Post by: Andy Simpson on September 15, 2005, 04:22:13 PM
Ok guys, how about this perspective:
In spatial terms, if 20k has a wavelength of 1.7 cm, then perhaps higher sampling rates can help us better represent the spatial timing differences in sounds.
Or, specifically, a recording made at 44.1 will 'quantize' the spatial timing aspects of a recording into chunks of 1.7cm.
When you think of the scale of a drumkit, cello or guitar, 1.7cm sampling seems rather crude.....Imagine a cello made of sugar cubes....pretty low resolution in my book.
Or imagine the round edge of a snare drum, quantized into 1.7cm blocks....not very round!
The strings on a guitar are often closer together than 1.7cm!
I'm not just talking left-right here, but front-back too.
In these terms, analogue tape really kills digital, but one can see in very real terms just how higher sampling rates might improve this aspect of imaging (ie. realness - and I would expect 192 to sound much more REAL than 44.1, where 1.7cm becomes 0.17cm - but I'd reckon on needing MUCH higher rates for really real sound).
Also (a small digression), these sorts of measurements start to make sense of the anti-NFB argument, where negative feedback can actually start to make spatially measureable distortions.
Andy
Post by: jimmyjazz on September 15, 2005, 04:27:43 PM
| andy_simpson wrote on Thu, 15 September 2005 16:22 |
In these terms, analogue tape really kills digital |
How so? Analog tape has the same ~ 20 kHz bandwidth, on a good day.
I think your reasoning is flawed, too. You're mapping the wavelength of sounds in air to structural "quanta" of instruments, but the wavelengths of those vibrations within the instruments themselves are far different.
Post by: Andy Simpson on September 15, 2005, 04:46:19 PM
| jimmyjazz wrote on Thu, 15 September 2005 21:27 | ||
How so? Analog tape has the same ~ 20 kHz bandwidth, on a good day. I think your reasoning is flawed, too. You're mapping the wavelength of sounds in air to structural "quanta" of instruments, but the wavelengths of those vibrations within the instruments themselves are far different. |
I think that you're missing the point.
It's not bandwidth that digital lacks over tape.
It's the spatial quantization that digital enforces, which can easily be measured in spatial terms (ie. 1.7cm).
To explain further; when you have a stereo signal, the time differences between left/right are limited to steps of 1.7cm by the sampling rate.
I am trying to illustrate that the sound made by acoustic instruments is as complex as their phyiscal shape and that 1.7cm quantization of this is very poor indeed.
This applies to reverb and all other aspects of acoustic sound.
Andy
Post by: Ged Leitch on September 15, 2005, 04:51:28 PM
The point about the crude quantizing of the spatial aspects of waveforms at 44.1khz is interesting though, but remember even though at 192khz this is improved in your theory, then a lot of it is thrown away when we reduce to 44.1khz again for CD.
So do you think it makes THAT much of a difference?
Post by: Andy Simpson on September 15, 2005, 04:59:48 PM
I'm saying that I would expect to need massively high sampling rates to provide the same spatial information that good tape has provided for years.
Even 192, at a quantization of 0.17cm, is still inadequate in my view. Probably my ears would be more impressed with something that approaches atomic scale.
Anyway, I would like to submit that this spatial quantization could explain why people can hear the difference and prefer 192 to 96 or 44.1 or whatever, despite a hearing bandwidth of 20-20k......
....but that tape still wins, as the quantization IS atomic level!
Andy
Post by: dcollins on September 15, 2005, 05:00:09 PM
| andy_simpson wrote on Thu, 15 September 2005 13:22 |
Or, specifically, a recording made at 44.1 will 'quantize' the spatial timing aspects of a recording into chunks of 1.7cm. |
Not true. The interchannel accuracy comes from the word-length, not the sample rate. With dither, there is essentially no limit to the "spatial resoultion."
DC
Post by: Andy Simpson on September 15, 2005, 05:21:05 PM
| dcollins wrote on Thu, 15 September 2005 22:00 | ||
Not true. The interchannel accuracy comes from the word-length, not the sample rate. With dither, there is essentially no limit to the "spatial resoultion." DC |
I disagree.
In terms of the human auditory system, you appear to be talking about interaural _level_ differences.
I am talking about interaural _timing_ differences.
Andy
Post by: Ged Leitch on September 15, 2005, 05:41:20 PM
| andy_simpson wrote on Thu, 15 September 2005 13:22 |
I disagree. In terms of the human auditory system, you appear to be talking about interaural _level_ differences. I am talking about interaural _timing_ differences. Andy |
And can you equate these "interaural timing differences" to something we can HEAR and understand?
Post by: Andy Simpson on September 15, 2005, 05:58:28 PM
| dcollins wrote on Thu, 15 September 2005 22:00 | ||
And can you equate these "interaural timing differences" to something we can HEAR and understand? |
As far as understanding is concerned, my explanations are as simple as I can manage.
In terms of hearing, good ears will probably help, and having an all-analogue recording chain to compare to digital would be useful.
I'm just trying to explain why I think we should be looking towards higher sampling rates, why we might hear a benefit and why tape (& vinyl) will continue to beat digital until this area is adressed specifically.
Andy
Post by: Norwood on September 15, 2005, 05:58:43 PM
| Quote: |
I am trying to illustrate that the sound made by acoustic instruments is as complex as their phyiscal shape and that 1.7cm quantization of this is very poor indeed. This applies to reverb and all other aspects of acoustic sound. |
This is absurd.
Post by: tom eaton on September 15, 2005, 06:00:16 PM
But before you do: How many samples represent a cycle at Nyquist? (Hint: it's not one)
And don't forget to think about oversampling at both converters.
-tom
Post by: Ged Leitch on September 15, 2005, 06:02:09 PM
| andy_simpson wrote on Thu, 15 September 2005 22:21 | ||||
As far as understanding is concerned, my explanations are as simple as I can manage. In terms of hearing, good ears will probably help, and having an all-analogue recording chain to compare to digital would be useful. I'm just trying to explain why I think we should be looking towards higher sampling rates, why we might hear a benefit and why tape (& vinyl) will continue to beat digital until this area is adressed specifically. Andy |
Too much maths and science and not enough real world comparisons and expanations to relate to for me so sorry.
cheers.
Post by: Andy Simpson on September 15, 2005, 06:10:37 PM
| TER wrote on Thu, 15 September 2005 23:00 |
What instruments do you record where the fundamental frequency is 20kHz? That's the only frequency represented by your math. Don't forget that 90% of the musical information in most cases is below 10k. Now try your math again. But before you do: How many samples represent a cycle at Nyquist? (Hint: it's not one) And don't forget to think about oversampling at both converters. -tom |
The fundamental frequency of a sound does not affect it's origin.
My taking of 20k is just a round number for the sample rate.
My whole point is that a 1k sound can be made at any distance from a mic (or two).
This distance affects the time of arrival.
According to the sampling rate, the time of arrival will be quantized.
I can't make this any simpler folks.
Andy
Post by: dcollins on September 15, 2005, 06:11:25 PM
| andy_simpson wrote on Thu, 15 September 2005 14:21 |
I disagree. In terms of the human auditory system, you appear to be talking about interaural _level_ differences. I am talking about interaural _timing_ differences. |
So am I. And it's as non-intuitive as they come!
DC
Post by: Ged Leitch on September 15, 2005, 06:16:58 PM
I can't make this any simpler folks.
Andy[/quote]
Ok so if you can't make it simpler can you at least tell us how this equates to audible differences we are supposed to be hearing?
e.g...
we all know that reducing a 24 bit file to 8 bit will sound terrible as the bitdepth used to represent the audio is reduced and therefore we lose information and get all sorts of unpleasant artefacts.
So, that said can you give us an example of how your theory relates to audible differences we can hear?
Post by: Andy Simpson on September 15, 2005, 06:48:33 PM
| andy_simpson wrote on Thu, 15 September 2005 23:10 I can't make this any simpler folks. Andy[/quote |
Ok so if you can't make it simpler can you at least tell us how this equates to audible differences we are supposed to be hearing? e.g... we all know that reducing a 24 bit file to 8 bit will sound terrible as the bitdepth used to represent the audio is reduced and therefore we lose information and get all sorts of unpleasant artefacts. So, that said can you give us an example of how your theory relates to audible differences we can hear? |
It basically relates to spatial resolution.
We know that we prefer 24bit dynamic range, but to the uneducated ear it can sound the same as 16bit.
I think the same is true of higher sampling rates. The higher sampling rate gives us greater resolution of the spatial aspects of sound - where it's coming from, how far it's travelling to us from the source.
How about this:
Take a 24/44.1 file and move the left channel forward by one sample.
Now do the same for a 24/192 file.
There is a difference in that one sample shift.
Andy
Post by: Eric Bridenbaker on September 15, 2005, 07:32:24 PM
Also, I had heard at one point that a good analog deck can go far higher than 20K. Is this true?
My own case for higher rates (apart from a reduction in audible aliasing during processing) involves the effects of harmonics happening at twelve times the frequencies we're actually perceiving (ie. 20KHz * 12 = 240 Khz, which would require a 480Khz sampling rate for accurate 20K reproduction). A full explanation can be found in the "Beat Harmonics" thread... which received a mixed reaction, and with any luck will continue to tick some people off.
Andy, it's nice to see you come up with and defend an idea like this. Someone has got to do it. We owe all of our current technology to those who had the guts to ask the questions, whether or not they were considered to be fashionable at the time. Sine wave tests in an audiology lab can only get one so far.
...to the universe, and beyond!!
Cheers,
Eric
Post by: Loco on September 15, 2005, 08:08:29 PM
| andy_simpson wrote on Thu, 15 September 2005 16:22 |
Or, specifically, a recording made at 44.1 will 'quantize' the spatial timing aspects of a recording into chunks of 1.7cm. |
What if you recorded the event under water? Or in the rain? What about the distance between your speaker's tweeter and woofer? what about the distance between your ears? What about what's in between?
| Quote: |
The strings on a guitar are often closer together than 1.7cm! |
Can you hit at least 20.000 of them in less than a second?
| Quote: |
In these terms, analogue tape really kills digital |
Can you get some useable dynamic range in analog tape beyond 22K after 20 passes?
Keep in mind that large diaphragm mics are 2.5 cm wide and your ears are about 5 cm. your ear canal is about 2 cm long and under one wide. A guitar pick is over 2 cm long and it's still irrelevant if you're recording analog or digital at any sample rate. It just doesn't make any sense what you said.
Post by: tom eaton on September 15, 2005, 08:22:03 PM
Interchannel Differential Phase (Stereo Pair) +/- 0.5 degrees at 20 kHz
Interchannel Differential Phase (Between DAC1 Units) +/- 0.5 degrees at 20 kHz
The DAC1 upsamples its input. This spec is REGARDLESS OF INPUT SAMPLE RATE.
Moving on...
Let's dig into your idea here a little. Let's "forget" about oversampling at the converter, as that would simply prove your argument void. Let's think about that 20K sine wave. One cycle of that sine wave is 1.7cm as a pressure wave, as you stated. At a sampling rate of 40kHz one cycle of that wave could be represented by two samples, in which case the zero crossings would be between the high and low points. But let's look at three or four consecutive samples as we see what happens if we shift that 20kHz wave LESS than one sample in time. Hey...it still recontructs fine!
Where the samples fall in time relative to the waveform has nothing to do with the system frequency response. The point in time where the samples are taken IS NOT the only place a wave (even at Nyquist) can crest!
-tom
Post by: timrob on September 15, 2005, 08:26:51 PM
| andy_simpson wrote on Thu, 15 September 2005 17:10 | ||
The fundamental frequency of a sound does not affect it's origin. My taking of 20k is just a round number for the sample rate. My whole point is that a 1k sound can be made at any distance from a mic (or two). This distance affects the time of arrival. According to the sampling rate, the time of arrival will be quantized. I can't make this any simpler folks. Andy |
You seem to be suggesting that A-D conversion is somehow nonlinear with regards to time. Of course, that is not the case.
I would say that the time of arrival cues suffer more from phase shift at higher frequencies due to Nyquist Filter. Higher sample rates can help, but are not a solution on their own.
Time of arrival will remain the same regardless of sample rate.
Can you show some actual data to support your claims?
Tim Roberts
Waterknot Music
Nashville
Post by: dcollins on September 15, 2005, 09:25:50 PM
| Eric Bridenbaker wrote on Thu, 15 September 2005 16:32 |
This is a very interesting way of looking at the issue, never heard that one before. 1.7cm is definitely crude. |
Even without dither 44/16 has "phase quantization" of
2pi/44100/2^16 or about 2ns from channel to channel.
With dither, there is essentially no limit. And, just like analog, whatever noise is at the zero crossing determines the phase resolution.
Some studies show that people can hear about 6us ITD, so I think we're safe.
I might also add that if PCM was as bad as all that, people would have noticed it way before it arrived in audio.
The inter-channel response is a particularly hard one to visualize, it took me forever, but remember that digital is really a continuous time system as we use it. Nothing can fall between the samples and be missed, as no matter how fast the input, the Nyqvist filters will always see a signal "smeared" over one or more samples.
The "smearing" question is slightly more interesting......
DC
Post by: jimmyjazz on September 16, 2005, 12:05:29 AM
| andy_simpson wrote on Thu, 15 September 2005 16:46 |
I think that you're missing the point. It's not bandwidth that digital lacks over tape. It's the spatial quantization that digital enforces, which can easily be measured in spatial terms (ie. 1.7cm). To explain further; when you have a stereo signal, the time differences between left/right are limited to steps of 1.7cm by the sampling rate. I am trying to illustrate that the sound made by acoustic instruments is as complex as their phyiscal shape and that 1.7cm quantization of this is very poor indeed. This applies to reverb and all other aspects of acoustic sound. |
You are trying to illustrate a hypothesis, and your hypothesis is flawed.
The wavelength of a 20 kHz signal is 1.7 cm, no doubt. It doesn't make any difference if it is captured "analog" or "digital". In air, the spatial separation of two pressure peaks of that 20 kHz signal will be (approximately) 1.7 cm. Space a pair of microphones 1.7 cm apart, and at some instant in time, both will sense an amplitude peak (in the absence of reflections), and then one micro-second later, they will both sense an amplitude ever so slightly less than that peak value. Approximately 12 micro-seconds later, they will both sense zero. And so on, and so on, and so on. There is no lack of "spatial information".
Look, Nyquist works. There may have been various implementations of digital audio that were less than spectacular, but the reasons weren't because Nyquist doesn't work. You can't really get around the fact that, properly implemented, digital audio is more accurate (not necessarily "better") than analog, at sampling frequencies far below 192 kHz. Maybe 44.1 kHz is cutting it a bit close, but this "spatial resolution" red herring is just that . . . a red herring.
Post by: Eric Bridenbaker on September 16, 2005, 01:37:05 AM
| dcollins wrote on Thu, 15 September 2005 21:25 |
Even without dither 44/16 has "phase quantization" of 2pi/44100/2^16 or about 2ns from channel to channel. |
Which might be just enough to screw with the mono mix. Phase offset can be very unforgiving in this case.
Best Regards,
Eric
Post by: J.J. Blair on September 16, 2005, 01:40:14 AM
Post by: Eric Bridenbaker on September 16, 2005, 01:42:22 AM
| J.J. Blair wrote on Fri, 16 September 2005 01:40 |
I know this guy who reads a bunch of shit on a particular topic, then tries to impress experts on the subject by butchering the nomenclature. I remember a converstaion he had with a web designer at a party years ago: "I was thinking about trying to have an intranet network over ethernet ... blah, blah, blah..." True story. If I didn't already know his whole family, I would swear he and Andy are related. |
Ad Hominum, No Dice!
Post by: maxdimario on September 16, 2005, 02:12:23 AM
as far as spatial and timing, I hear this as well, and I agree.
I never thought about the air distance.
in analog, frequency response is not a limit to timing resolution and is not tied to micro-timing differences -- as it is a continuous recording.
a slow slew rate does delay the initial impulse but does not shift it in time as digital does
in digital the timing differences, are quantized by the sampling rate and are therefore related to the frequency response...mathematically..because of the sampling rate.
Mix the above with jitter, and you have the 'digital' sound.
I have to say I agree in part with andy .
Post by: dcollins on September 16, 2005, 02:36:22 AM
| Eric Bridenbaker wrote on Thu, 15 September 2005 22:37 | ||
Which might be just enough to screw with the mono mix. Phase offset can be very unforgiving in this case. Best Regards, Eric |
That's two nanoseconds. You can forgive them, if combined.....
DC
Post by: Andy Simpson on September 16, 2005, 05:12:39 AM
| maxdimario wrote on Fri, 16 September 2005 07:12 |
andy, there is a problem with what you say, because the flaws that are associated with digital can be heard from mono sources as well. as far as spatial and timing, I hear this as well, and I agree. I never thought about the air distance. in analog, frequency response is not a limit to timing resolution and is not tied to micro-timing differences -- as it is a continuous recording. a slow slew rate does delay the initial impulse but does not shift it in time as digital does in digital the timing differences, are quantized by the sampling rate and are therefore related to the frequency response...mathematically..because of the sampling rate. Mix the above with jitter, and you have the 'digital' sound. I have to say I agree in part with andy . |
That's pretty much what I've been getting at, for better or worse.
In mono, those differences in air are as relevant to one mic as to a pair - for that front/back depth mono experience.
Andy
Post by: Bill Mueller on September 16, 2005, 06:35:18 AM
| andy_simpson wrote on Fri, 16 September 2005 05:12 | ||
That's pretty much what I've been getting at, for better or worse. In mono, those differences in air are as relevant to one mic as to a pair - for that front/back depth mono experience. Andy |
OH MY GOD,
This crap is catching.
This thread is giving me a headache. I try not to read it but I just can't turn away.
Best Regards,
Bill
Post by: Loco on September 16, 2005, 09:17:18 AM
I was gonna write an entire physics essay about this to illustrate how wrong they are, but when you don't want to hear there's no point on screaming.
Please, lock this thread.
Post by: zmix on September 16, 2005, 09:30:03 AM
| J.J. Blair wrote on Fri, 16 September 2005 01:40 |
I know this guy who reads a bunch of shit on a particular topic, then tries to impress experts on the subject by butchering the nomenclature. I remember a converstaion he had with a web designer at a party years ago: "I was thinking about trying to have an intranet network over ethernet ... blah, blah, blah..." True story. If I didn't already know his whole family, I would swear he and Andy are related. |
Really? You know George Bush? Shit....
Oh, wait, you did say 'reads'...
My bad...my little goat.
-CZ
Post by: Andy Simpson on September 16, 2005, 10:14:19 AM
If you take a 'really nice' 192 recording of, say, a drumkit.
Low-pass filter it at 20k or whatever.
Copy the file and downsample it to 44.1.
Now up-sample it back to 192.
Invert one of them and try to null them.
Can you get them to cancel?
My guess is that you end up with something left over....
Andy
Post by: maxdimario on September 16, 2005, 11:03:34 AM
| Quote: |
That's two nanobrains. You can't forgive them when combined... I was gonna write an entire physics essay about this to illustrate how wrong they are, but when you don't want to hear there's no point on screaming. Please, lock this thread. |
no use in getting upset kiddies..
keep your insults to yourself and your family please.
your reaction was inappropriate and shows that you have no valid SIMPLE argument to make.
It makes sense to me, because timing even in the range of a fraction of a millisecond influences the way we feel music and localize sound. you should study nature a little more.
digital is less involving than analog regardless of frequency bandwidth.
If I were to list all the people who have said the same, and the people who have denied it, you'd have a who's who of musicians and talented engineers for analog, and on the digital side a who's who of technicians who seemingly don't get it.
loco, you can list me all the physics theories you want (and probably have not studied in depth) but it don't change the fact that I hear the problem and so do many of the best musicians around.
of course we all know that digital is in, and analog is out...but some of us are actually interested in finding out why digital is more boring to listen to.
the fact is that many 'engineers' around are not really capable of listening, use crappy equipment that filters out the detail, and are mostly preoccupied with being right, based on some facts that they dont fully understand NOR DO THEY REALLY CARE TO because it is detrimental to their pocketbook and they are not so much interested in AUDIO REPRODUCTION as they are in mixing and mashing it all up with their lovely and cheap digital equipment.
if there was no digital, most modern production houses and pop records wouldn't exist, cause it would be impossible to polish a turd to the extent possible now for ...so cheap.
all these arguments only make sense if who is reading them recognizes the deficiency. To correct it, you have to come up with something that hasn't been said before.
All I see here is people reading out from textbooks.
How many of you have actually tried to say something new or bring the argument foreward?
doesn't sound very big minded to me.
actually sounds like a bunch of guys who really don't give a shit about music, but are in love with their gear.
Post by: jimmyjazz on September 16, 2005, 11:14:44 AM
You are exaggerating the demographics of the two camps. I'm sure it helps you sleep at night to tell yourself you're in bed with the brilliant musos as opposed to the gear-happy technicians, but that doesn't mean you've correctly assessed the situation. Not by a long shot.
Post by: maxdimario on September 16, 2005, 11:20:46 AM
some people actually KNOW what they like, and it takes them minutes, or seconds to choose.
instead of being generic, next time actually see if you can explain what you are talking about, instead of giving me your academic history.
Post by: tom eaton on September 16, 2005, 11:37:44 AM
And then go read a book on digital audio!
I might suggest Pohlman's Principles of Digital Audio, maybe starting with Chapter 2: Fundamentals of Digital Audio.
Geez...
-tom
Post by: Eric Bridenbaker on September 16, 2005, 11:38:17 AM
On the upside, whenever a topic comes up, that seems to strike a nerve, get people thinking, and passionately asserting this or that, there is most likely something to it.
Nice work Andy!!
Best Regards,
Eric
Post by: maxdimario on September 16, 2005, 11:54:55 AM
| Quote: |
Let's dig into your idea here a little. Let's "forget" about oversampling at the converter, as that would simply prove your argument void. Let's think about that 20K sine wave. One cycle of that sine wave is 1.7cm as a pressure wave, as you stated. At a sampling rate of 40kHz one cycle of that wave could be represented by two samples, in which case the zero crossings would be between the high and low points. But let's look at three or four consecutive samples as we see what happens if we shift that 20kHz wave LESS than one sample in time. Hey...it still recontructs fine! Where the samples fall in time relative to the waveform has nothing to do with the system frequency response. The point in time where the samples are taken IS NOT the only place a wave (even at Nyquist) can crest! |
what does that have to do with minute timing differences, please.
forget sine waves, I got that.
let's say a perfect pulse wave for theory's sake with an 'on' time of 10 ms or more
if the rise of the pulse wave does not corrispond exactly with the sample clock, will it not be shifted in time to the next sample clock division in some way, regardless how small.
wow and flutter are slow and predictable compared to jitter and this kind of distortion.
what do you think?
Post by: Eric Bridenbaker on September 16, 2005, 12:16:39 PM
Q: "If analog tape sounds so much better than digital, what improvements should be made in A/D, D/A converters?"
A: "First of all, the frequency response should extend from 3 Hz to 50 kHz, because we experience those frequency limits. We are able to detect audio up to 50 kHz. We don't hear it, but we experience it in other ways. I can give you tinnitus very quickly if I run an ultrasonic cleaner at 45 kHz. You are aware that it's on, and your ears ring when it's shut off. On the low end, we detect mechanical vibrations down to 3 Hz. When a marching band walks past you, you feel the drums in your stomach and bones. And that's all part of the sound.
Ten years ago in Stereophile, I said that digital was never going to work well in the chosen format. Digital should use a 400 kHz sampling rate and 24-bit words. Then it will satisfy the hearing mechanism and won't have a digital sound. Digital has a "sound" purely because it is based on lousy mathematics. The manufacturers presuppose too simplistic a view of our hearing mechanism."
The full interview is on the EAR site here:
http://www.ear-usa.com/timdeparavicini.htm
Best Regards,
Eric
Post by: Jack Schitt on September 16, 2005, 12:22:03 PM
If all the big names hear such a fatal flaw in digital why are they still using it? It doesn't make any sense.
Post by: timrob on September 16, 2005, 02:13:33 PM
| Eric Bridenbaker wrote on Fri, 16 September 2005 11:16 |
FWIW, Here's an excerpt from an interview with Tim de Paravicini (He designs the E. A. R. gear, very top end stuff, Tony Faulkner uses it for his classical recordings). Q: "If analog tape sounds so much better than digital, what improvements should be made in A/D, D/A converters?" A: "First of all, the frequency response should extend from 3 Hz to 50 kHz, because we experience those frequency limits. We are able to detect audio up to 50 kHz. We don't hear it, but we experience it in other ways. I can give you tinnitus very quickly if I run an ultrasonic cleaner at 45 kHz. You are aware that it's on, and your ears ring when it's shut off. On the low end, we detect mechanical vibrations down to 3 Hz. When a marching band walks past you, you feel the drums in your stomach and bones. And that's all part of the sound. Ten years ago in Stereophile, I said that digital was never going to work well in the chosen format. Digital should use a 400 kHz sampling rate and 24-bit words. Then it will satisfy the hearing mechanism and won't have a digital sound. Digital has a "sound" purely because it is based on lousy mathematics. The manufacturers presuppose too simplistic a view of our hearing mechanism." The full interview is on the EAR site here: http://www.ear-usa.com/timdeparavicini.htm Best Regards, Eric |
Yes, I'm sure many have read Tim de Paravicini's views. The article you quote here is from 1995 when 24-bit was just starting to proliferate to studios. I have heard many of the recordings by Kavi Alexander and they all sound beautiful. They sound beautiful on Compact Disc. Do they sound better on analog? Probably.
Few can afford the luxury of recording with ultra custom Analog Decks and Rectangle capsule Microphones using mechanically isolated stands. The recordings on CD are as good as it gets, so Big deal. None of this supports the theories laid out here.
Actually, it shows that digital can sound very good IMO.
It appears that some feel that some sort of time shift takes place in digital. I tell you what. Take a sine wave of any frequency in the passband and record it at any normal sample rate and play back. Hell, look at it on an Oscilloscope. Compare the input to the output. You'll see that the sine wave will be perfectly reconstructed.
Look I love analog as much as the next guy. But, you guys are grabbing at straws and trying to keep an argument going that is going nowhere.
Tim Roberts
Waterknot Music
Nashville
Post by: bblackwood on September 16, 2005, 02:15:32 PM
But wait, there's more!
If people are going to resort to such attacks, they will removed from George's forum. Keep it on topic and off one another.
Post by: tom eaton on September 16, 2005, 02:51:27 PM
There is a filter in front of any converter. That filter removes out of band information that would be incorrectly processed by the converter. So the highest frequency component of your square wave is limited by the sampling rate, which yes, to some extent does change the "timing" of the signal--the leading edge, if you will. But what you all are arguing is that change in a digital recording is only represented by the next sample...that's just not so! When a signal is reconstructed it can be CONSTANTLY changing. Two samples represent a continuum (not just instantaneous change), three samples, four samples represent a continuous motion. Samples don't come one by one by one and get rebuilt step by step...the result of reconstruction is an analog signal. And then... you have to listen to it somehow...which will likely cause more transient damage to the signal than the conversion process did. Unless your speakers are entirely phase accurate and flat from dc to 50k...
What are our options? Analog tape? No way. Digital is FAR closer to an accurate representation of the source than analog. Which sounds better to you or me is not (as far as I understand it) what we're talking about.
This really is covered in books...I was not being flip. There are many, many brilliant people who have worried about all these things before us, and solved them to the best extent of available technology. Nyquist knew more about this stuff years ago than most of us still know. What gets some of us (or just me) wound up is the fact that people post to Massenburg's forum without a basic grounding in the basics. The AES has papers available, there are great books out there, but you have to be willing to go do the homework.
-tom
Post by: bobkatz on September 16, 2005, 04:14:34 PM
| andy_simpson wrote on Thu, 15 September 2005 17:21 | ||||
I disagree. In terms of the human auditory system, you appear to be talking about interaural _level_ differences. I am talking about interaural _timing_ differences. Andy |
Andy, Dave Collins is right. You've been caught by an "urban myth". While it may seem counterintuitive to you, interaural timing issues are not affected by the sample rate, nor improved by a higher sample rate.
Mike Story has some issues regarding transient response and pre-echos with filters but this is peripheral to the issue of interaural timing. The two channels stay "in sync" with eachother down to pico seconds... in ANY current sample rate digital system.
BK
Post by: Andy Simpson on September 16, 2005, 05:41:28 PM
| bobkatz wrote on Fri, 16 September 2005 21:14 | ||||||
Andy, Dave Collins is right. You've been caught by an "urban myth". While it may seem counterintuitive to you, interaural timing issues are not affected by the sample rate, nor improved by a higher sample rate. Mike Story has some issues regarding transient response and pre-echos with filters but this is peripheral to the issue of interaural timing. The two channels stay "in sync" with eachother down to pico seconds... in ANY current sample rate digital system. BK |
But I'm not talking about whether the channels are in sync with eachother.
I'm talking about when a sound is 10 metres from the Left mic and 10 metres and 0.6cm from the Right mic.
What happens here?
Does the sound arrive at the same time when converted to 44.1?
It certainly doesn't arrive at the same time at the mics.
That is what I mean by interaural.
If anyone would be so kind as to try my down-sample/low-pass null test, this can be laid to rest I think, either way.
Andy
Post by: Ronny on September 16, 2005, 05:46:18 PM
| bblackwood wrote on Fri, 16 September 2005 14:15 |
If we are unable to discuss things, even things that seem rudimentary, without stooping to personal attacks, this thread will be locked. But wait, there's more! If people are going to resort to such attacks, they will removed from George's forum. Keep it on topic and off one another. |
I can't believe that there are still people debating annie and digi. It's been going on for over 20 years now.
They sound different and require different approaches. Good annie sounds good, bad annie sounds bad, good digi sounds good, bad digi sounds bad. What's the mystery folks? No debate, they are separate entities, each with there own unique benefits and drawbacks. The bottom line is, in todays industry, you are seldom going to have one without the other and there isn't any sense dickering about it.
Post by: Norwood on September 17, 2005, 01:48:40 AM
| Quote: |
If anyone would be so kind as to try my down-sample/low-pass null test, this can be laid to rest I think, either way. Andy |
All this would do is show the inaccuracies in the up/downsampling process.
Post by: maxdimario on September 17, 2005, 03:52:46 AM
insults are the equivalent to physical force to try and win an argument.
Andy's concerns are mostly intuitive, and may or may not be right, but to shrug them off by half-quoting texts does no good.
It is better to give a clear explanation of why not, than to say that the poster is incapable, period.
...anyway nobody answered my seemingly naive question about inter-sample pulses.
I am not talking about steady-state sine waves, but pulse waves.
if the rise of fall of the pulse signal falls between samples, will it not be shifted in time to the next sample?and can you explain why not.
Post by: maxdimario on September 17, 2005, 04:25:01 AM
higher sampling rates would create less of an error percentage.
think NON-steady-state.
anyone explain why this would not happen?
added: I am concerned only with distortion in the time domain and use a pulse wave to represent a distinct event in time.
Post by: timrob on September 17, 2005, 08:09:42 AM
| maxdimario wrote on Sat, 17 September 2005 03:25 |
here's a drawing to illustrate what I just posted. higher sampling rates would create less of an error percentage. think NON-steady-state. anyone explain why this would not happen? |
Square waves are a somewhat of a problem for digital systems.
But not in the way you describe. In most cases you'll see overshoot of the Leading edge of the waveform. Some call this ringing. It is third harmonic distortion caused by the anti-aliasing filter. As always the amount of this effect will depend on the quality of the filters.
Higher sample rates reduce this problem by using a less steep slope and higher cutoff frequency.
Tim Roberts
Waterknot Music
Nashville
Post by: Loco on September 17, 2005, 08:38:55 AM
| maxdimario wrote on Sat, 17 September 2005 03:52 |
if the rise of fall of the pulse signal falls between samples, will it not be shifted in time to the next sample?and can you explain why not. |
If you actually sample a 441 Hz square wave, you won't get full on/full off waves on your DAW. Instead, you will see ripples because the AD converter is not measuring an instant voltage value at 44100 Hz but instead a difference at a much higher rate (in the order of MHz). then it will decimate it to get the actul sample rate you need. this will happen after an analog low pass filter that will create an analog ripple as well. It will look mopre or less like Bart Simpson's hair.
Post by: tom eaton on September 17, 2005, 08:51:04 AM
A square wave coming off analog tape will likewise NOT BE SQUARE.
I'm not sure you're understanding that transient resolution and frequency response are tied together. If you take a snare drum hit and filter it down so that the highest frequency is within the capture range of the digital system (assuming the system is working properly) it will come back as it went in. By talking about the rise time of a wave you are talking about FREQUENCY. A square wave has high frequency components that create the very steep leading edge. That's to say that if a square wave is filtered down to 20kHz, it WILL NOT RISE faster than the system can resolve. Destructive, yes. Will higher sample rates help? Absolutley.
Does digital have flaws, sure. Does analog tape sound better? For many things, yes. Its flaws are pretty good sounding. Should we throw digital technology out the window because it has flaws, or should we let it evolve and become a mature technology?
I have to say that the worst digital recordings sound bad because the person using the technology isn't using it well. We're at the point now where one of the huge benfits of digital (dynamic range) has basically been forgotten by the recording community as it pushes to make things more loud more of the time. By the way...do you know what you get as things get louder and louder...more and more clipped? Square waves!
And why is pointing someone to a textbook a bad idea? All this stuff is out there in writing for people who want to know. Didn't your mom ever tell you to go look it up? Mine sure did.
-tom
Post by: bobkatz on September 17, 2005, 09:31:12 AM
| andy_simpson wrote on Fri, 16 September 2005 17:41 |
But I'm not talking about whether the channels are in sync with eachother. I'm talking about when a sound is 10 metres from the Left mic and 10 metres and 0.6cm from the Right mic. What happens here? Does the sound arrive at the same time when converted to 44.1? |
Yes, the errors are as low as picoseconds interaurally. The timing differences between channels are preserved because both left and right channels are sampled at identical times. At each frame, the next sample along represents identical sampling of left and right channel. So any delays between channels are exactly preserved, down to the noise limit of the system.
Imagine that your head is in a vise. Your left ear remains at the same spot in time for the duration of the recording as does your right ear. The A/D converter is in a similar vise.
BK
Post by: J.J. Blair on September 17, 2005, 11:18:07 AM
Christ, already.
Post by: Andy Simpson on September 17, 2005, 11:24:18 AM
| bobkatz wrote on Sat, 17 September 2005 14:31 | ||
Yes, the errors are as low as picoseconds interaurally. The timing differences between channels are preserved because both left and right channels are sampled at identical times. At each frame, the next sample along represents identical sampling of left and right channel. So any delays between channels are exactly preserved, down to the noise limit of the system. Imagine that your head is in a vise. Your left ear remains at the same spot in time for the duration of the recording as does your right ear. The A/D converter is in a similar vise. BK |
Bob, what is your take on the re-sample & null idea?
Andy
Post by: Eric Bridenbaker on September 17, 2005, 11:31:55 AM
| J.J. Blair wrote on Sat, 17 September 2005 11:18 |
Not to insult anybody (yeah, right) ... but if you don't listen to Dave Collins' years of experience and expertise on this issue, and choose to disagree and argue with him instead, then you are a fucking moron. Unless you are on the technical advisory board of AES or have similar credentials, listen to the man and learn something. He's giving you the answer. Christ, already. |
Not sure what's worse, bickering with the experts, or the kind of behavior in the above post, which is a great example of what NOT to do on the forum!! I dont think Brad or George is going to like it very much. I sure don't.
Post by: maxdimario on September 17, 2005, 11:38:38 AM
| Quote: |
Yes, it takes time to go from sample to sample...according to my memory it's about .02ms. In your drawing (of a non-periodic dc pulse--waves have positive and negative excursions from zero) you STILL are thinking that digital does not change until the next sample comes along. Reconstruction will CONNECT the dots. There will not be a right angle anywhere. There will be motion BETWEEN the samples. A square wave coming off analog tape will likewise NOT BE SQUARE. |
I didn't ask if the waveform wasn't gonna be square or not.
I am aware that there are filters involved.
I used a pulse wave as a theoretical example to indicate a distinct event in time.
of course analog tape doesn't reproduce square waves, in fact a lot of equipment doesn't.
Guys, try and focus out of the usual issues.
I am not talking about sine waves, or pulse waves because I want to record perfect pulse waves.
even if the pulse had a ramp-up time, there will have to be a moment in time when the ramp-up BEGINS.
quantization will make this precise moment fit into the sampling grid.
on tape there IS NO SUCH TIME RESTRICTION.
higher sampling rates will make the distortion in time less apparent because the resolution in time will increase.
it doesn't seem so complicated to me, please explain otherwise, related strictly to the time-distortion issue.
forget steady-state signals for now, PLEASE.
Post by: Eric Bridenbaker on September 17, 2005, 12:09:48 PM
However, I can also see that this is not what maxdimario is talking about...
I remember from my engineering classes a science concept called "significant digits". Basically it means that you can only get results as precise as the least precise measurement in the entire process. It will be the limiting factor with respect to the precision of the final result.
For instance, if you are measuring the floor area of a room, and the length is precise down to the millimeter, but the data for the width is only precise to the centimeter, then the final results can only be expressed in centimeters, and all of the "millimeter" decimal places in the final calculation are irrelevant and must be disregarded by rounding off to the least "significant digit", in this case centimeters.
So, back to our digital scenario, forgetting about amplitude for the sake of argument and focusing only on the timing of the digital audio conversion process, if indeed the data is reduced to a single value 44 thousand times every second, this IS the limiting precision of the process with respect to time, regardless of the smoothness achieved in reconstruction. "Filling" in the gaps IMHO is not the same as actually capturing the information with a higher degree of precision (ie, higher rate = smaller gaps in the resultant data).
I realize that oversampling will increase the accuracy of the data achieved for each sample value, but it does nothing for the precision with respect to "absolute" time (the concept is simple enough that we don't need to get into relativity on this one, and there is a difference between precision and accuracy). I also am aware that statistical principles do come into play with converter design, but is it really possible to restore timing information that is more precise than the sampling rate?
What Andy proposes about "spatial" precision being limited by the sampling rate is correct in my opinion, and has nothing to do with the smoothness of reconstruction, as any timing information is limited in precision to that of the sample rate in the A/D stage.
This argument is based more on scientific method than a working knowledge of converter design so I could be wrong about this.
Please correct me if so, and explain where the "missing" timing information is collected and stored, while keeping the resultant file size commeasurate with a 44Khz rate.
Best Regards,
Eric
Post by: bobkatz on September 17, 2005, 05:48:54 PM
| andy_simpson wrote on Sat, 17 September 2005 11:24 |
BK |
Bob, what is your take on the re-sample & null idea?
Andy
[/quote]
I'm lost. Can you point me to the full detailed test?
Bob
Post by: maxdimario on September 17, 2005, 05:56:15 PM
my point exactly.
I'm sure that digital engineers have been aware of this since the 70's since it is such a fundamental part of the process.
steady-state signals (test tones-sine waves) being repetitive do not exhibit this defect because of the convenient interplay between ratios of sampling rate (which is steady-state) and said signals.
on the other hand take a hi-hat recording and you have a very high degree of distortion in the time domain since there is nothing repetitive about such a signal, and the waveform oscillates at high frequencies in a very complex fashion.
not mentioning that the actual attack of the hihat hit will be quantized by the grid..
I wonder if anyone else gets it, or has something to say do disprove this.
frequency bandwidth is not really the problem...it's distortion in time. that's why higher sample rates sound more real.
Post by: howlback on September 17, 2005, 06:12:52 PM
| Eric Bridenbaker wrote on Sat, 17 September 2005 12:09 |
Please correct me if so, and explain where the "missing" timing information is collected and stored, while keeping the resultant file size concurrent with a 44Khz rate. |
Not to mention, if what you and Max are hypothesizing were so (and it is NOT), we would need a mic with "some kind of impulse response" to "catch" such small time differences.
-k.walker
Post by: maxdimario on September 17, 2005, 06:30:24 PM
| Quote: |
Not to mention, if what you and Max are hypothesizing were so (and it is NOT), we would need a mic with "some kind of impulse response" to "catch" such small time differences. |
no you don't need a mic.
it's very simple.
digital quantizes time in a grid.
if you have a hard time grasping this imagine a low sample rate.
at a low enough sample rate, the beginning of any percussive musical event is quantized to the next step, just like a sequencer quantizes musical midi events.
at a higher sample rate the beginning of the event becomes more accurately reproduced in time, but never 100%.
it will NEVER be 100% since it is against the very mathematical process of quantization.
the ear is very sensitive to time shifting, and can interpret and understand phase shift, as it occurs in nature, but quantization-time-distortion only happens in digital audio.
it is not the equivalent of phase shift, but something which confuses the hearing system.
Post by: howlback on September 17, 2005, 06:38:18 PM
| maxdimario wrote on Sat, 17 September 2005 18:30 |
digital quantizes time in a grid. |
-k. walker
Post by: maxdimario on September 17, 2005, 06:51:41 PM
If I have a sound event that starts at a specific point in time from rest to an excited state, and this point in time is in-between samples it will be registered in the next sample.
oversampling resolves some issues but not this one.
I would like an explanation of why not...it should be simple to state in a post.
Post by: J.J. Blair on September 17, 2005, 07:07:54 PM
| Eric Bridenbaker wrote on Sat, 17 September 2005 08:31 |
Not sure what's worse, bickering with the experts, or the kind of behavior in the above post, which is a great example of what NOT to do on the forum!! I dont think Brad or George is going to like it very much. I sure don't. |
Eric, I didn't single anybody out. You just have to understand my frustration with certain people. It just makes me lose my composure, especially when I get e-mails and phone calls saying "You're not going to believe what So-And-So is up to again on the forum." I don't think some people in here realize how heavy the credentials are of some of these guys, and Dave is one of those guys. You might as well be arguing with Paul Frindle about digital distortion. Certain people in this thread are not interested in learning anything, because they ignore everything that the experts say and insist on arguing with them, because for them learning isn't as important as trying to beat people down to get them to agree with their inept, uninformed concepts.
In that case, I don't mind being rude and calling them a moron. They have defined themselves by their behavior, wouldn't you say? If that makes me an asshole and pisses of George and Brad, so be it. Somebody needs to speak up. It's madness, I tell you. The lunatics are running the asylum.
"There is a principal which is a bar against all information, which is proof against all arguments, and which cannot fail to keep a man in everlasting ignorance -- that principal is contempt prior to investigation." - Herbert Spencer
Post by: Andy Simpson on September 17, 2005, 08:15:42 PM
| bobkatz wrote on Sat, 17 September 2005 22:48 | ||
Bob, what is your take on the re-sample & null idea? Andy |
I'm lost. Can you point me to the full detailed test?
Bob[/quote]
Sorry, it was earlier in the thread....
Take a 192 recording (of a drumkit or tamb or whatever).
Low-pass at say 20k.
Take the low-pass'd file and resample to 44.1.
Resample it back up to 192 again.
Invert one of the two and attempt to null them against each other (at 192).
Will they null (in theory or otherwise)?
Andy
Post by: Eric Bridenbaker on September 17, 2005, 08:20:18 PM
| J.J. Blair wrote on Sat, 17 September 2005 19:07 | ||
Eric, I didn't single anybody out. You just have to understand my frustration with certain people. It just makes me lose my composure, especially when I get e-mails and phone calls saying "You're not going to believe what So-And-So is up to again on the forum." I don't think some people in here realize how heavy the credentials are of some of these guys, and Dave is one of those guys. You might as well be arguing with Paul Frindle about digital distortion. Certain people in this thread are not interested in learning anything, because they ignore everything that the experts say and insist on arguing with them, because for them learning isn't as important as trying to beat people down to get them to agree with their inept, uninformed concepts. In that case, I don't mind being rude and calling them a moron. They have defined themselves by their behavior, wouldn't you say? If that makes me an asshole and pisses of George and Brad, so be it. Somebody needs to speak up. It's madness, I tell you. The lunatics are running the asylum. "There is a principal which is a bar against all information, which is proof against all arguments, and which cannot fail to keep a man in everlasting ignorance -- that principal is contempt prior to investigation." - Herbert Spencer |
Hey J.J.,
I do see your point here. Some of these people have devoted their lives to professional audio, and are willing to take the time to freely share the knowledge they've acquired over decades. That's what keeps a lot of the cats coming here, myself included, and why I felt so strongly about giving this forum in particular the respect it deserves.
My bad was singling out on your post, J.J., which was a loss of composure on my part.
I too started to get frustrated with exactly what you were mentioning. Some people would rather be right than learn something new. While I do try to keep the expletives to a minimum in writing, what's in my head sometimes can be quite a lot worse...
This particular thread seems to be hitting some nerves, which I like too see, if dealt with in a civil fashion.
Anyway, not to get OT here... but I do hope that some of the people reading this seriously try to answer the question of why they participate in the forum.
BTW: Great Spencer Quote!
Best Regards,
Eric
Post by: dcollins on September 17, 2005, 09:47:36 PM
| maxdimario wrote on Sat, 17 September 2005 15:30 |
digital quantizes time in a grid. |
Max, you must understand that the "graph paper" only exists in your mind.
All digital systems input and output in continuous time.
You must get this first, or all else is wasted.
Or perhaps you and Eric have a date with the King of Sverige....
The home of Nyquist, who clearly had no idea what he was doing.
As history has shown.
| Quote: |
the ear is very sensitive to time shifting, and can interpret and understand phase shift, as it occurs in nature, but quantization-time-distortion only happens in digital audio. |
And there you have it...
DC
Post by: dcollins on September 17, 2005, 09:53:18 PM
| maxdimario wrote on Sat, 17 September 2005 08:38 |
forget steady-state signals for now, PLEASE. |
Here's a quiz question:
What waveform is the _least_ "steady state?"
Why?
DC
Post by: Eric Bridenbaker on September 17, 2005, 10:07:41 PM
Specifically, I'm considering a short duration peak, sitting right in the middle of two sample points and overlapping neither, (this is obviously above the nyquist frequency). From my current understanding of the process his peak clearly cannot be reconstucted, as there is no data with which to form a sinc function. Now these peaks are of such short duration that they would not be heard anyway, however:
Instead of one peak, imagine a series of regular peaks, occuring at every third intersample space.
The peaks are now occuring at a regular interval which is below below the nyquist frequency.
The question: Is this pattern going to be represented in the sampled waveform?
My supposition is that if the temporal resolution were truly infinite below nyquist, these peaks would have to be represented in the resultant waveform, as they occur at a frequency below nyquist. If they are not appearing, then something is missing, right? Or am I missing something?
Someone help!
Best,
Eric
Post by: tom eaton on September 17, 2005, 10:52:12 PM
To have a rise time that's faster than the sample rate can resolve means having a frequency higher than the system can resolve. And because the system filters out those higher frequency signals, you will NEVER have situation like the one you're describing. Take the FILTER to task, not the digital conversion and reconstruction process.
And 24kHz is a pretty respectable frequency response window, especially when the filter is implemented well. BK has written about this EXTENSIVELY.
Eric-as to your specific question, a peak that appears between the sample points will be represented by the system, because:
1. in order to get into the system it must be below Nyquist
2. therefore it will exhibit a rise and fall time that the system can resolve
3. and therefore it will be sampled and reconstucted correctly
NOW...if you're talking about a continuous wave in which the crests and valleys of the wave are exactly aligned so that they fall in between each sample point, you are talking about a wave AT Nyquist. If you shifted this wave 1/2 sample to the right or left that would be obvious. This would not get past the filter on the way in.
If you're talking about a wave in which the valley is two samples away from the peak it will be obvious that the samples between those points will capture intermediate values that will allow the wave, again, to be reconstructed properly.
-tom
Post by: Ronny on September 17, 2005, 11:13:31 PM
| dcollins wrote on Sat, 17 September 2005 21:53 | ||
Here's a quiz question: What waveform is the _least_ "steady state?" |
Let me take a wild guess, DC.
Sawtooth
| Quote: |
Why? DC |
Infinite harmonics resulting in aliasing. Gibbs phenomenon.
I hope I don't have to go to summer school for this.
Post by: tom eaton on September 17, 2005, 11:20:54 PM
I'll bring the cookies if you bring the milk, Ronny. I'm hoping there's nap time.
-tom
Post by: Duardo on September 18, 2005, 12:22:43 AM
| Quote: |
But I'm not talking about whether the channels are in sync with eachother. I'm talking about when a sound is 10 metres from the Left mic and 10 metres and 0.6cm from the Right mic. What happens here? Does the sound arrive at the same time when converted to 44.1? It certainly doesn't arrive at the same time at the mics. |
The sound doesn't "arrive". Think of sound as a continuous wave. At 44.1 kHz the position of that wave is captured every forty four thousand, one hundredth of a second. If you picture it as a sine wave (which you can do, since all sound can be broken down into individual sine waves) and think of frequency as the x axis then you can see that the wave will be somewhere on that axis every time a sample is taken, right? A sample is taken every time it crosses the x axis so it doesn't need to be quantized as far as frequency is concerned. That is what you're talking about, whether you realize it or not.
The place where quantization takes place is when the sample crosses the y axis. Since it's captured every time it crosses the x axis it will almost always fall between two points on the y axis, so that's where quantization has to occur. And the y axis represents bit depth. At any given sampling rate, the smaller the "steps" in the y axis are the more accurately the position of the wave will be captured. This quantization error is thus pushed further and further down in terms of level, which is why more bits gives us more dynamic range. Again, since all sound can be broken down into individual sine waves, you can think of this quantization noise as being separate from the signal we're trying to capture, just as you can think of all the components of that signal...say, each instrument if we're recording an ensemble...as separate components. When all are combined together you have one complex waveform, but you can still hear each instrument, right? Same thing goes for the quantization noise, except for you typically can't hear it because it's so low in level compared to everything else. So what you have is the signal you recorded, with all of the spatial information intact. If you move one microphone that information will change, whether we're talking about an analog or a digital recording/
| Quote: |
if the rise of fall of the pulse signal falls between samples, will it not be shifted in time to the next sample?and can you explain why not. |
Sure, because if the rise and fall of a pulse signal falls between samples then it's above the Nyquist frequency and won't be captured at all. You can't say to leave sine waves out of the equation because all sound is composed of sine waves, so if you're looking at things in the context of two samples you can have nothing smaller than or equal to one half of one complete sine wave between those two samples. And regardless of whether those samples fall when that wave is at its peak, its null or anywhere in between it can be captured and reproduced with perfect accuracy.
| Quote: |
What Andy proposes about "spatial" precision being limited by the sampling rate is correct in my opinion, and has nothing to do with the smoothness of reconstruction, as any timing information is limited in precision to that of the sample rate in the A/D stage. |
Hopefully I've explained things clearly enough that you understand now that this is not the case.
| Quote: |
on the other hand take a hi-hat recording and you have a very high degree of distortion in the time domain since there is nothing repetitive about such a signal, and the waveform oscillates at high frequencies in a very complex fashion. not mentioning that the actual attack of the hihat hit will be quantized by the grid.. I wonder if anyone else gets it, or has something to say do disprove this. frequency bandwidth is not really the problem...it's distortion in time. that's why higher sample rates sound more real. |
There's nothing any more complex about the way a hi-hat waveform oscillates than anything else. Sure, there's more random high-frequency content than there is with something like a guitar string or a reed, but it's still just a bunch of sine waves as far as a sampling system is concerned. The higher frequencies are filtered out and what we can hear is captured. The attack of the hihat isn't quantized by any grid. The only distortion that will be induced will be the quantization noise (which really won't be perceived as distortion) and, more likely than not, distortion that occurs in the analog domain...at the microphone, in the speakers, and so on.
| Quote: |
if you have a hard time grasping this imagine a low sample rate. at a low enough sample rate, the beginning of any percussive musical event is quantized to the next step, just like a sequencer quantizes musical midi events. at a higher sample rate the beginning of the event becomes more accurately reproduced in time, but never 100%. it will NEVER be 100% since it is against the very mathematical process of quantization. |
That's not the way it works. The sampling rate will never be low enough where any perceptible event will be quantized to the next step. And remember, the quantization occurs in the amplitude domain, not the frequency domain. Within the Nyquist frequency of the lower sampling rate, which is what we're concerned with here, the signal will be no more or less accurately captured at the higher sampling rate than the lower one. "Accuracy" will improve with increased bit depth because the quantization is pushed down in level. Sure, it will never be 100% mathematically since that quantization noise will always be there, but frequency-wise it will be 100% with some added low-level noise.
Of course, in reality many other factors come into play...inaccuracies of the transducers on both ends, the effects of all of the other electronics the signal passes through...but for what we're talking about here digital audio works just fine.
-Duardo
Post by: JamSync on September 18, 2005, 02:10:48 AM
| dcollins wrote on Sun, 18 September 2005 02:53 | ||
Here's a quiz question: What waveform is the _least_ "steady state?" Why? DC |
Fractal? 'cuz it's aperiodic
Post by: maxdimario on September 18, 2005, 05:03:05 AM
| Quote: |
Sure, because if the rise and fall of a pulse signal falls between samples then it's above the Nyquist frequency and won't be captured at all. |
I can see we still aren't talking about the same thing.
I don't want to record anything above nyquist.
I don't want to record a 50 KHz pulse wave..
let's make a simple example.
I have a recording of a triangle in an anechoic room.
I begin to play a slow rhythm.
the signal is at rest until I strike the triangle.
if I strike the triangle between samples, the attack of the triangle won't be complete because It only picks up from the next sample. I am not calculating ramp-up time because that is a constant.
this is a minute difference but it still exists, enough to do something to the imaging and rhythm in a track.
now record a very dense rhythm with that triangle and the effect will more pronounced.
forget sine-waves or harmonic theory.
| Quote: |
There's nothing any more complex about the way a hi-hat waveform oscillates than anything else. Sure, there's more random high-frequency content than there is with something like a guitar string or a reed, but it's still just a bunch of sine waves as far as a sampling system is concerned. |
you take an irregular waveform, which by it's complexity and chaotic nature does not lend itself to the harmonic, or 'bunch of sine waves'theory, more or less applicable in different cases, and it will not sound as good on digital.
why? perhaps because in ADDITION to the fact that every sound can be broken up into sine waves, there is something ELSE going on, which has to do with the fact that the digital system is by definition a mathematical approximation of a real waveform.
the least signigicant digit in this case is the quantization rate, and bit depth.
the reason why bit depth is easier to hear than sample rate doesn't have as much to do with frequency extension per se, but because timing and spatial differences are less obvious to less people, but they are still there.
percussion, string attack etc. have always been iffy in digital, and this is perhaps why.
analog has no such time restrictions, and high sample rates make the errors less noticeable.
Post by: maxdimario on September 18, 2005, 06:16:54 AM
| Quote: |
Max, you must understand that the "graph paper" only exists in your mind. All digital systems input and output in continuous time. |
because of reconstruction filters?
how well do digital systems reproduce noise?
Post by: Eric Bridenbaker on September 18, 2005, 07:19:21 AM
| dcollins wrote on Sat, 17 September 2005 21:53 |
What waveform is the _least_ "steady state?" Why? DC |
Thanks Dave, You're making us think again!
Noise? because it's random, completely aperiodic. (If that qualifies as a wave "form").
If not, then KK's answer seems to be correct. A fractal generates a form over repeated iterations, though individual instances can be seemingly random.
Does anyone have an official definition of "Steady State" as it applies to audio theory?
Examples, From Wikipedia:
In biochemistry, steady state is a central term in osmosis.
In chemistry, steady state is a central term in chemical kinetics.
In cosmology, steady state is a non-standard cosmological view developed in 1949 by Fred Hoyle and others as an alternative to the Big Bang theory.
In physiology, steady state is a system in which a particular variable is not changing but energy must be continuously added to maintain this variable constant.
In macroeconomics, a time series variable reaches the steady state when its growth rate in time remains constant.
Waveforms with rapidly changing amplitude slopes, such as the sawtooth are useful for testing behavior in audio systems, and I can see how the sawtooth would be less steady than the square wave. Pulse seems even less steady, but I'm unsure about the definition...
Anyways, being Sunday and all, it might be time for some Sunday School. I'm going to read up a bit, starting with some of Dan Lavry's papers, and more Nyquist material. Here's hoping the answer to this specific question is in there somewhere. I have no problem coming back with an apology if it is, as I sure dont want to waste anyone's time here. I'm already appreciative of the fact that the experts here on the forum have been so patient. Will need to get more rigorous before making any further observations.
Best,
Eric
Post by: tom eaton on September 18, 2005, 08:09:43 AM
Your triangle example shows that you are still not understanding that transient response is tied to frequency response. The highest frequencies (above Nyquist) will be filtered out and this will "slow" the rise time to one which the system can resolve...but it WILL preserve the absolute timing of the signal relative to other elements of the recording due to the fact that the signal is now entirely within the capture range. The only way a signal would get "lost" is if it has NO content below Nyquist. Like a triangle where the fundamental is 30k going into a 44.1 system. It just isn't going to happen.
AND, even in an anechoic chamber, you have noise from the mic and preamp...you can only go from nothing (digital black) to signal in a computer. In a properly made digital recording this noise will be captured along with the signal, in an analog recording the noise would be swamped by the noise of the media.
The transfer function of tape is flawed, too! There are so many examples of this it's not really even worth explaining.
-tom
Post by: maxdimario on September 18, 2005, 08:18:40 AM
I'd like to keep steady state waveforms out of the discussion, because what I'm trying to get at is not heard in constant, regular waveforms, but random ones with high frequency attacks or noise-like components etc.
I mentioned that to try and get away from test-tone mania, and see if anyone could come up with some kind of explanation for the kind of timing distortion I hear.
with the anti-aliasing filter at the AD converter there can be no such thing as a perfect pulse wave because the rise time is limited by the filter, so it ends up ramping up at a speed determined by the slew-limiting of the said filter.
I thought this was obvious, so I drew a perfect pulse which was conceptually clearer and easy to draw.
even if the pulse ramps up, there is a point when ramp-up begins and ends.
there is still a definite point in time that the pulse wave starts.
if this point is inbetween samples, a small fraction of the rising wave will be truncated, no?
the sloping or rising part of the wave has no problem, as the filters re-construct the steady waveform no prob.
during DA conversion the 'truncated' (of the inter-sample attack) wave will be further modified by the lowpass filter on the output.
what's wrong with that? can anyone explain without being too mysterious?
my concern is that some sounds should not really be considered waveforms, but a composition of events, where harmonic content is not as important as the overall 'structure' of a sound in time.
Post by: Eric Bridenbaker on September 18, 2005, 08:27:27 AM
| TER wrote on Sat, 17 September 2005 22:52 |
YOU CAN"T HAVE AN INSTANTANEOUS PEAK BETWEEN SAMPLES UNLESS THE FREQUENCY THAT PEAK REPRESENTS IS ABOVE NYQUIST. SO IT CAN'T HAPPEN! To have a rise time that's faster than the sample rate can resolve means having a frequency higher than the system can resolve. And because the system filters out those higher frequency signals, you will NEVER have situation like the one you're describing. Take the FILTER to task, not the digital conversion and reconstruction process. And 24kHz is a pretty respectable frequency response window, especially when the filter is implemented well. BK has written about this EXTENSIVELY. Eric-as to your specific question, a peak that appears between the sample points will be represented by the system, because: 1. in order to get into the system it must be below Nyquist 2. therefore it will exhibit a rise and fall time that the system can resolve 3. and therefore it will be sampled and reconstucted correctly NOW...if you're talking about a continuous wave in which the crests and valleys of the wave are exactly aligned so that they fall in between each sample point, you are talking about a wave AT Nyquist. If you shifted this wave 1/2 sample to the right or left that would be obvious. This would not get past the filter on the way in. If you're talking about a wave in which the valley is two samples away from the peak it will be obvious that the samples between those points will capture intermediate values that will allow the wave, again, to be reconstructed properly. -tom |
Thanks TER, this does reiterate exactly what has been said with regards to this thread. I do understand in a very basic way why Nyquist works, if the samples obtained intersect with a theoretical sine wave at the sampling frequency, then sinc functions can be used to reproduce the original waveforms. Any waveform composed of samples that do not intersect with this sine wave MUST be above nyquist (half the sampling rate) and will be bandlimited by the filters anyway. Is this right??
I'll be reading up on more of this today... need to be more informed to come up with a satisfacory answer. the main issue I have with all of this is with the assertation that the temporal resolution is infinite, but only when we limit our discussion to rise times that the system can resolve... is this not a contradiction?
Anyway, I already promised these guys that I'd shut up until I did some homework...
Thanks,
Eric
Post by: tom eaton on September 18, 2005, 08:32:18 AM
We can't discuss sound without discussing waveform. Sound is a pressure wave, and harmonic content IS the structure of sound in time.
-tom
Post by: maxdimario on September 18, 2005, 08:32:19 AM
| Quote: |
...but it WILL preserve the absolute timing of the signal relative to other elements of the recording due to the fact that the signal is now entirely within the capture range |
what does that have to do with it?.
even if it's a 100 Hz triangle wave, it's got to start somewhere.
as far as analog tape's noise, slew limiting, I am aware of the headaches..
what I am saying is that tape, frequency response apart, has infinite 'event' resolution.
think of it this way: I have a full track of audio and I pass it through a perfect gate, which switches the signal on and off at a precise but unrelated time in relation to the sample clock.
so we have segments of audio that have a precise beginning and ending time.
once recorded those same segments of audio are going to be a fraction of a millisecond shorter or longer that the input waveform, because of the quantization.
this is only an example to try and explain....I don't need to actually do this because I like to record audio segments, you know..
Post by: tom eaton on September 18, 2005, 08:36:46 AM
-tom
Post by: maxdimario on September 18, 2005, 08:45:58 AM
| TER wrote on Sun, 18 September 2005 14:36 |
Nope, not infinite "event" resolution, because as you agreed, analog slew limits the signal. So which system more accurately places high frequency information in time? -tom |
analog..
just because they are shifted in time (phase shift)doesn't mean they don't have the resolution between events.
if the events are EXACTLY 10 ms apart, they will be 10ms apart. or if they shift it will be gradually over the course of the whole event, not suddenly at the moment of sampling.
of course you could say that wow and flutter will shift the events in time, true, but the minimal wow and flutter of pro analog machines does not distort time in the damaging way that digital does (don't forget jitter).
phase shift is one thing, distortion in time another.
this isn't really an analog vs. digital debate anyway.
it's a sampling rate issue.
Post by: maxdimario on September 18, 2005, 08:52:22 AM
| Quote: |
We can't discuss sound without discussing waveform. Sound is a pressure wave, and harmonic content IS the structure of sound in time |
again I don't agree that fourier or any of the theories traditionally employed to resolve these issues can really capture what is going on 100%.
people have been saying digital is perfect from day one, but it keeps improving.. why?
sound is also a composite of events in time which are chaotic when approached from a harmonic point of view.
Post by: Bob Olhsson on September 18, 2005, 09:01:20 AM
| maxdimario wrote on Sun, 18 September 2005 07:32 |
...what I am saying is that tape, frequency response apart, has infinite 'event' resolution... |
Tape is far worse at this than digital audio because of manufacturing variations between magnetic head gaps. Gap scatter meant microphones sent to individual tracks could never be recombined accurately. Tape also involves the polarization of discrete iron oxide particles. The sample rate is somewhat random but this idea of infinite time resolution has no basis at all in fact.
I'm not trying to be a digital apologist and I just mastered a project a couple days ago where a 1/2" tape absolutely slaughtered the sound of a very popular high-end A to D converter. More open and effortless, more dynamic, deeper image, the whole enchelada. Still I don't buy that it sounded any better because of bandwidth or timing. A 44.1 Pacific Microsonics converter had no problem at all revealing how wonderful this tape sounded.
Post by: maxdimario on September 18, 2005, 09:32:46 AM
as far as the iron particles I also agree, but magnetized iron particles are much smaller than the typical sampling rate, and there is a soft focus effect of analog tape that tends to smear events because of the head gap, but it doesn't shift them in time erratically.
I don't see how digital is better than analog, when dealing in the time resolution of events, multitrack aside.
by the way if anyone has any theories that explain what I'm hearing I'd appreciate reading them...hoping it doesn't create too much of a stir.
Post by: Bob Olhsson on September 18, 2005, 11:58:09 AM
All of the good things about digital audio are only potentially good if the product happens to be well implemented and then interfaced advantageously with other gear.
My point is that there is a LOT that manufacturers routinely screw up. Getting digital audio right isn't nearly as simple as buying the best chips.
Post by: Johnny B on September 18, 2005, 01:32:34 PM
"Time Smear" *may* be one of the dirty little secrets in the digital capture and reproduction process; however, Julian Dunn (RIP) was attempting to understand, describe, and explain it before his unfortunate demise. IIRC, some of Julian Dunn's white papers discuss "Time Smear."
Here's a link to some of Julian Dunn's white papers, that may be a good place for one to begin to understand the many "Timing Issues."
http://www.nanophon.com
see also:
http://www.cco.caltech.edu/~boyk/spectra/spectra.htm
And David Blackmer's article:
http://www.drtmastering.com/blackmer.htm
"Digital? Is that the thing where they take a good old sine wave and they chop it up into little bits?" --- Rupert Neve
"Why do you record any better than MP3 quality? The answer is: Because you are putting your work, your labor of love, and your artist's vision, to a format that has life for a longer period than 'right now.'" --- George Massenburg
According to 3-D Audio's Lynn Fuston,
George Massenburg said he thought that "384K is when PCM catches up sonically with DSD."
Lynn goes on to say that GM implied that he'd be happy cutting on that.
Hope this helps.
Post by: bobkatz on September 18, 2005, 03:25:28 PM
| maxdimario wrote on Sat, 17 September 2005 18:51 |
sorry, I don't get it and to tell you the truth I don't think it should be kept a mistery, and there is no point in making a critical comment without backing it up with an understandable explanation. If I have a sound event that starts at a specific point in time from rest to an excited state, and this point in time is in-between samples it will be registered in the next sample. |
Actually, no it won't "register in the next sample". Stop thinking about samples. The low pass filtering smooths out the transitions and it ends up one continuous flow.
This IS the post that explains your answer as simple as it can be.
BK
Post by: AndreasN on September 18, 2005, 04:00:32 PM
| maxdimario |
even if it's a 100 Hz triangle wave, it's got to start somewhere. think of it this way: I have a full track of audio and I pass it through a perfect gate, which switches the signal on and off at a precise but unrelated time in relation to the sample clock. so we have segments of audio that have a precise beginning and ending time. once recorded those same segments of audio are going to be a fraction of a millisecond shorter or longer that the input waveform, because of the quantization. |
(I'm only a n00b in this company
, but I'll give it a try.)The sampling rate does not quantize the continuous waveform. If an analog spurious event starts between digital sample points, it will be sampled mid-wave, as you assume. All sampling is actually "off timing" 99.9999+% of the time, if you like to see it that way. The good news is that it doesn't matter.
Even if the first sample of the waveform comes after the slope of the analog wave starts rising, it will still be recreated with the right timing. The sampling points represents a transition spot between the previous and coming samples, where only one waveform can exist. The only way the bandwith limited waveform can be recreated is with a smooth slope from zero to the sample point. Waveform reconstructed.
Example, your triangle. I assume the waveform starts with a pulse-like impact as it is hit. The start/pulse will fall between sample points. When filtered, the pulse will calm down and assume a more relaxed shape. Down to sine wave, if the fundamental frequency falls within the upper octave of the sampled bandwidth. (there was that S-word again.
)This will be represented by two or more sample points with X bit depth on the vertical axis. Regardless of the bit depth, or quantization, the outcome on the horizontal time axis will be a continuous waveform which is an exact reconstruction of the input, including the trailing slope before the actual sample point.
Irregularities would come from too few bits or timing problems with the clock, not the actual sampling as such.
Hope this clears matters a little bit!
..AFAIK, IMHO, etc.
Cheers,
Andreas Nordenstam
Post by: Sam Lord on September 18, 2005, 04:23:07 PM
| Eric Bridenbaker wrote on Thu, 15 September 2005 19:32 |
Andy, it's nice to see you come up with and defend an idea like this. Someone has got to do it... |
I beat the drum about timing errors long ago, no doubt echoing what many had said before me, with the same response. It certainly appears that improved bit resolution, around 20 bits (~121dB) in very good ADCs (now peaking at about 130dB with the best ones) at 48kHz actually finishes the spatial and tonal resolution job, unlike 16 bits at 2x or 4x FS. I certainly didn't expect this, but will accept what countless trustworthy golden ears have reported. Even the terrific constraints of 44.1 kHz appear surmountable when truly optimized ADCs are at work. (Bob Katz has written extensively on this.) Now, when lots of processing is to occur, higher sample rates may show validity. Many experienced MEs have said this. As of now, the emerging consensus is that 88.2k or 96k are the best recording choices (of those now available) for those with high-quality ADCs. This is the result of the performance conflicts between anti-aliasing, circuit-settling, and jitter. And equipment compatibility and money.
I still wonder what our ears can discern. Why do all of my 16/44.1k recordings of complex music lack that outstanding honesty which my best SACDs and 24/96 recordings have? Is this the "time smear" that Dave Collins speaks of, and is this the effect of the LP filter chain or quantization error? And do the dCS up- and over-sampling units ($$$$) mostly eliminate this by their careful interpolation? I expect that the most revealing test is very complex music, performed well and recorded with no effects straight to digital. Bruno Putzeys of Phillips, along with Tim de P. and GM are convinced that we will yet benefit from more than 4 FS a la DXD; we shall see. Best wishes, Sam
Post by: Johnny B on September 18, 2005, 06:50:24 PM
| Sam Lord wrote on Sun, 18 September 2005 21:23 |
Bruno Putzeys of Phillips, along with Tim de P. and GM are convinced that we will yet benefit from more than 4 FS a la DXD... |
Apparently, Walter Sear has recently expressed his feeling that even 384kHz speeds will have to be increased five-fold to approach the sound quality of analogue. I have not verified whether he actually said this or not, but to comtemplate that kind of increase in horsepower should not be discounted out-of-hand, after all, we've been seeing geometric increases in digital speeds and the number-of-X-isters-on-a-chip for years. With proper implementation, 5 x 384kHz just might work and sound as good as analogue. It could take years to get there, then again, it *could* happen overnight. I think I'll try to keep an open mind and reserve judgement until I can actually hear the new devices which are headed our way. However, there's just no way I'll ever trust lab bench tests or propaganda published by the chip makers et al., I'll have to actually hear how these Next Gen chips perform in real world situations to make up my mind.
For me, the "Ear Test" is the most important test of all. That, and the feelings I get in my brain and in my body.
Post by: dcollins on September 19, 2005, 12:09:53 AM
| JamSync wrote on Sat, 17 September 2005 23:10 | ||||
Fractal? 'cuz it's aperiodic |
Ok, ok, what periodic waveform?
DC
Post by: mark fassett on September 19, 2005, 01:08:45 AM
However, a sound wave can be represented at any point in the cycle by a sample, it doesn't matter whether it begins at the zero crossing or not... the sample isn't delayed relative to the original audio to the next zero crossing point.
Post by: dcollins on September 19, 2005, 01:21:41 AM
| Ronny wrote on Sat, 17 September 2005 20:13 |
Let me take a wild guess, DC. Sawtooth |
Nej.
Think of something that changes all-the-time.
DC
Post by: JamSync on September 19, 2005, 03:53:20 AM
| dcollins wrote on Mon, 19 September 2005 06:21 | ||
Nej. Think of something that changes all-the-time. DC |
a sweep? nah...probably noise since you can generate it with pseudo-random numbers
Post by: Eric Rudd on September 19, 2005, 09:48:04 AM
| Ronny wrote on Fri, 16 September 2005 22:46 |
I can't believe that there are still people debating annie and digi. It's been going on for over 20 years now. They sound different and require different approaches. Good annie sounds good, bad annie sounds bad, good digi sounds good, bad digi sounds bad. What's the mystery folks? No debate, they are separate entities, each with there own unique benefits and drawbacks. The bottom line is, in todays industry, you are seldom going to have one without the other and there isn't any sense dickering about it. |
The debate continues for a variety of reasons...
1) People have their beliefs and feel the need to share them because they are excited about the work they are doing. There are good number of people, many on this forum, who are doing GREAT work towards the advancement of digital.
2) They share because they are competitive and the think they are "right."
3) People subconsiously feel the need to justify/defend their choices or investments (i.e., personal investments of time, gear purchases, etc)
4) Nostalgic reasons are a part of this discussion (I know this is a big one for me....I miss the smell of tape.)
5) And finally, the internet forum is ultimately a rather poor place to communicate. Too often people read between the lines and hear a "tone" in the post that often times isn't there, or misinterpreted, etc.
Did I miss anything??
My digital "coming of age" happened when I heard one particular all-digital recording by a colleague of mine and I realized that I could no longer use the excuse of "digital bad....analog good."
For me, I can fully respect the people on this forum...as long as I see them providing the same respect of others. I can disagree with someone and still not make it personal. That's where a few on this forum cross the line.
Best,
Eric
Post by: Eric Bridenbaker on September 19, 2005, 09:49:01 AM
| JamSync wrote on Mon, 19 September 2005 03:53 | ||||
a sweep? nah...probably noise since you can generate it with pseudo-random numbers |
Pseudo Random... that's good!! Fits the bill. Here's to thinking outside the box, KK!
Some other possibilities:
Pulse wave? (because of the uneven slope of the rise time, combined with the instanteneous jump in amplitude).
As for the square wave, it will force the audio circuit to try and do the amplitude jump twice as often as pulse or sawtooth. So, from a certain point of view this could be considered to be even less steady state but this is where I start to realize that I'm not quite clear about what is meant by "Steady State".
Of course, combining any two tones that are relatively unrelated harmonically will generate a periodic, yet crazy looking waveform. Detuning one of the tones by a few cents should do the trick. Musically speaking, my vote is the tritone, which cuts the octave in half, the logarithmic equivalent of the most harmonically unrelated combination.
I had thought for a while also about those sample/hold modulation effects, like on the moog, you know, that "computer" sound, but it does have random behavior and should be probably be considered aperiodic.
The other possibility I'm thinking of is a ring modulated signal.
Best Regards,
Eric
Post by: Sam Lord on September 19, 2005, 10:13:56 AM
Post by: StudioRhythm on September 19, 2005, 12:04:54 PM
| dcollins wrote on Sun, 18 September 2005 21:09 |
Here's a quiz question: What waveform is the _least_ "steady state?" Why? DC |
Music. (oops -- not periodic, though!)
Post by: four on September 19, 2005, 12:09:40 PM
| Quote: |
Ronny wrote on Sat, 17 September 2005 20:13 Let me take a wild guess, DC. Sawtooth Nej. Think of something that changes all-the-time. DC |
The least steady-state periodic waveform... changes all the time... Ooh! Ooh! Sine Wave!
-Tom
Post by: JamSync on September 19, 2005, 01:43:06 PM
| Eric Bridenbaker wrote on Mon, 19 September 2005 14:49 |
As for the square wave, it will force the audio circuit to try and do the amplitude jump twice as often as pulse or sawtooth. So, from a certain point of view this could be considered to be even less steady state but this is where I start to realize that I'm not quite clear about what is meant by "Steady State". Best Regards, Eric |
I've always thought the opposite was "chaos" or a system leading to chaos which is often discussed with period doubling.
You could think of "transient" vs "steady state".
Not sure what he's looking for here.
Post by: Johnny B on September 19, 2005, 01:48:06 PM
I could be wrong tho', I often am as "She Who Must Be Obeyed" routinely points out.
Post by: Tomas Danko on September 19, 2005, 03:58:42 PM
| andy_simpson wrote on Thu, 15 September 2005 21:22 |
In spatial terms, if 20k has a wavelength of 1.7 cm, then perhaps higher sampling rates can help us better represent the spatial timing differences in sounds. Or, specifically, a recording made at 44.1 will 'quantize' the spatial timing aspects of a recording into chunks of 1.7cm. Also (a small digression), these sorts of measurements start to make sense of the anti-NFB argument, where negative feedback can actually start to make spatially measureable distortions. Andy |
This sounds like a statement not considering the reconstruction filters in the modern DAC's of today. It's not completely intuitive, but the DAC will reproduce phase differences and such much smaller than those "1.7 cm" aspects you mention.
Sincerely,
Tomas Danko
Post by: 12345 on September 20, 2005, 04:08:09 AM
So now I can post some effort onto this thread. There is a very good reason why higher sampling rates are beneficial, and this is when the transducer must be called on to accurately transfer audio along with additional information. In order to do this correctly, the transducer's pickup flux must be known and compensated for to create an optimized pressure-receiving system, and the microphone must be located near either an interferometric source or a cyclic, directed known source.
For a bit of background, let's assume that 1) the microphone is used to pickup only the frequency and amplitude of a given signal, and let's also assume that 2) the microphone has the inherent capability to be part of an audio capturing system which is capable of recording more than simply the frequency and amplitude of a given signal. Let's further assume that 3) the microphone we are using has a diaphragm (ribbon, etc.) which is torqued differently about its holding structure and hence subject to differences in how the pressure creates an electrical pulse on the other end of the transducer, i.e. if the wave hits the microphone from the left side the signal will be different than if it hits the microphone on the right-hand side. Let's further assume that 4) this diaphragm "imperfection" can be determined such that its total "imperfection" can be calculated out to create an optimized system such that after the correction there will be no difference whether the signal is incident from the left of the right-hand.
Following these assumptions, there is now a method to create a "perfect" microphone response across the extent of the diaphragm's (ribbon's, etc.) surface area.
Bringing in assumption (2), we can now devise a way to extract more information from the microphone than we gather from assumption (1).
I present such an approach in the attached .jpeg image, and I propose that the following information can now be extracted using a permutation of this approach:
1) the sound’s position in space at all times;
2) the sound’s origin in space at all times;
3) the sound’s frequencies at all times;
4) any pressure changes not directly attributable to the sound source which occur in the room while recording is taking place.
I further propose that this method can be applied "in reverse" so that the audio event can be recreated by the speaker.
I further propose that the sinc function approach is a single design methodology which has benefits and disadvantages compared to other existing techniques such as:
1) cardinal splines
2) nth-order polynomials
3) bitrate sampling
So what are the design methodology assumptions present in the sinc function approach:
1) that the transform of the audio event be restricted to linear ordinary differential equations (l.o.d.e.);
2) that the convolution of the sinc function be specified mathematically
3) that the interaction of the Laplace and Fourier transforms be resolved either in the s-plane or through poles and zeroes.
Design methodologies of cardinal splines:
1) the end boundaries of each spline must be driven to continuity
2) the nodes of each spline must lie within the specified set of equations
3) the splines must be driven to sinusoidal functions.
4) the splines must be limited in their order.
Design methodologies of nth-order polynomials:
1) the order of the polynomials must be limited
2) the polynomials must be driven to sinusoidal functions
3) the end boundaries of each polynomial must be driven to continuity
Design methodologies of bitrate:
1) chaotic architectures must be stabilized
2) cascading coverters must be stabilized
3) frequencies must be separated by element
Please consider the above techniques along with .jpeg image.
Sincerely,
MW
Post by: 12345 on September 20, 2005, 04:18:59 AM
My guess as to which type of wave is the least steady-state would be feedback because its propogation is not governed by an impulse, sweep, or predicable event--feedback is a byproduct of chaos and system anomalies.
MW
Post by: 12345 on September 20, 2005, 04:34:46 AM
the beginning and end of the waveform
MW
Post by: J.J. Blair on September 20, 2005, 12:28:14 PM
| dcollins wrote on Sat, 17 September 2005 18:53 |
What waveform is the _least_ "steady state?" |
Tsunami! It destroys everything in it's path.
Post by: maxdimario on September 20, 2005, 01:28:00 PM
for argument's sake..
A lot of the bad sound of digital comes from it's being engineered at a lower standard than should be as Bob O. says, and there is plenty of improvement to do there.
since so much of the way a digital system works depends on filters, any defects in the engineering and production of the filters will result in bad audio.
as someone who always opens up a new piece of gear I was surprised to see very expensive state-of-the art convertes with no internal RF shielding between circuits, one PC board that housed most of the cicuits, along with dubious placement of analog circuitry near digital, and power supply distribution etc.
getting back to the theoretical.
I realize that if a wave starts mid sample, and it is a test tone, or a regular repeating waveform, it will be reconstructed by the filters if it is below the brickwall frequency.
what about if the wave does not repeat itself?
how about percussion, where the sound begins with a sharp attack and may be more of a noise than a tone?
Yes, what I am talking about could be called 'time smear'.
and to make the argument even more basic let's say we sample at 10 khz, which would make the brickwall 5 KHz
what would be the delay between the samples? and wouldn't that delay have an effect on how, for example, a hihat pattern was reproduced in time? (I don't have any intention of recording hihats at 10 KHz sampling frequency, i'm just using it as a worse-case scenario)
curious.
Post by: lord on September 20, 2005, 01:59:51 PM
| maxdimario wrote on Tue, 20 September 2005 13:28 |
and to make the argument even more basic let's say we sample at 10 khz, which would make the brickwall 5 KHz what would be the delay between the samples? and wouldn't that delay have an effect on how, for example, a hihat pattern was reproduced in time? |
Yes, you are right. The hi-hat is delayed.
It is further delayed by the time it takes for you to rewind and play back the file a minute later.
Do you understand?
It would be delayed another ms if you moved the microphone back a foot.
THIS IS THE MOST RETARDED THREAD EVER.
Post by: timrob on September 20, 2005, 02:01:53 PM
| maxdimario wrote on Tue, 20 September 2005 12:28 |
Ok, for argument's sake.. A lot of the bad sound of digital comes from it's being engineered at a lower standard than should be as Bob O. says, and there is plenty of improvement to do there. since so much of the way a digital system works depends on filters, any defects in the engineering and production of the filters will result in bad audio. as someone who always opens up a new piece of gear I was surprised to see very expensive state-of-the art convertes with no internal RF shielding between circuits, one PC board that housed most of the cicuits, along with dubious placement of analog circuitry near digital, and power supply distribution etc. getting back to the theoretical. I realize that if a wave starts mid sample, and it is a test tone, or a regular repeating waveform, it will be reconstructed by the filters if it is below the brickwall frequency. what about if the wave does not repeat itself? how about percussion, where the sound begins with a sharp attack and may be more of a noise than a tone? Yes, what I am talking about could be called 'time smear'. and to make the argument even more basic let's say we sample at 10 khz, which would make the brickwall 5 KHz what would be the delay between the samples? and wouldn't that delay have an effect on how, for example, a hihat pattern was reproduced in time? (I don't have any intention of recording hihats at 10 KHz sampling frequency, i'm just using it as a worse-case scenario) curious. |
Well, the reason 44.1k was chosen as the standard originally was because it seemed to be the best compromise. By placing the cutoff frequency above the known limits of average human hearing.
It is the same reason that 20-20k has become the standard for measurement.
Your example places the cutoff frequency smack in the middle of one hearings most sensitive areas. A brick wall at 5k at any sample rate will sound like crap. Beyond that, most of the sharp attack you mention will be completely filtered out by that 5k filter. I just don't find this a very useful way of looking at things. All it does is show the reason we don't use a 10k sample rate. What you are really trying to do is reason that if a low sample rate is crap then a higher sample rate must be better. Where your average run of the mill converter is concerned, I'd have to agree with that. On the other hand, when it comes to the higher end, I think it has been shown that even 44.1k can compete with 96k when really high quality filters are used.
My understanding is that the Time Smear that keeps getting referred to is really an effect of the filters ringing.
Tim Roberts
Waterknot Music
Nashville
Post by: Ronny on September 20, 2005, 03:40:19 PM
| timrob wrote on Tue, 20 September 2005 14:01 | ||
Well, the reason 44.1k was chosen as the standard originally was because it seemed to be the best compromise. By placing the cutoff frequency above the known limits of average human hearing. It is the same reason that 20-20k has become the standard for measurement. Your example places the cutoff frequency smack in the middle of one hearings most sensitive areas. A brick wall at 5k at any sample rate will sound like crap. Beyond that, most of the sharp attack you mention will be completely filtered out by that 5k filter. I just don't find this a very useful way of looking at things. All it does is show the reason we don't use a 10k sample rate. What you are really trying to do is reason that if a low sample rate is crap then a higher sample rate must be better. Where your average run of the mill converter is concerned, I'd have to agree with that. On the other hand, when it comes to the higher end, I think it has been shown that even 44.1k can compete with 96k when really high quality filters are used. My understanding is that the Time Smear that keeps getting referred to is really an effect of the filters ringing. Tim Roberts Waterknot Music Nashville |
On a related note, 78 rpm records seldom had a frequency above 7k. AM radio 8-10k and some FM radio uses 32k sampling rates on their digital broadcasts. Reason being that FM maxes at 15k, so there is no need to go above the Nyquist of 32k. The bottom line is this: You can only reproduce what the microphone is capable of capturing, which for most large condensers is 20k. 44.1k will effectively capture and reproduce all freq's that the human ear can perceive and will accommodate everything that a 20k mic can throw at it.
Post by: crm0922 on September 20, 2005, 05:34:08 PM
| Quote: |
Instead of one peak, imagine a series of regular peaks, occuring at every third intersample space. The peaks are now occuring at a regular interval which is below below the nyquist frequency. The question: Is this pattern going to be represented in the sampled waveform? |
Your peaks are above Nyquist. Try to remember, if the signal is above Nyquist, it won't be captured. That makes this whole thread irrelevant, more or less.
| Quote: |
even if the pulse had a ramp-up time, there will have to be a moment in time when the ramp-up BEGINS. quantization will make this precise moment fit into the sampling grid. on tape there IS NO SUCH TIME RESTRICTION |
Also wrong. The ability of tape to accurately represent rise times is directly related to the bandwidth. Same as digital
As far as the "between-sample" pickup, the first
"lost" sample will be deluged in noise anyways, so it doesn't matter. The waveform will be faithfully reconstructed according to Nyquist, with the proper rise-time. There is no questioning this. It is mathematical fact.
The signal component realized by collecting the "missing" sample would be above nyquist, were it to be in a different spot than where the reconstruction filter puts it.
Frequency and time and completely related and cannot be separated just for fun. If a peak appears between samples, it is above Nyquist in frequency. This is because of its rise time, not because of how often it repeats.
Chris
Post by: tom eaton on September 20, 2005, 06:37:04 PM
I don't think he's listening. You just restated five of my posts from earlier in the thread.
Oh well.
-tom
Post by: crm0922 on September 20, 2005, 07:53:52 PM
| TER wrote on Tue, 20 September 2005 18:37 |
Chris- I don't think he's listening. You just restated five of my posts from earlier in the thread. Oh well. -tom |
I know you've been right all along, Tom. It's just so irritating seeing the same (incorrect) theory reposed in 9 million different ways. I felt compelled to restate the (correct) reasoning yet again. And I'm hardly an expert, despite having taken some graduate signals courses in college.
Here's another good one:
| Quote: |
in analog, frequency response is not a limit to timing resolution and is not tied to micro-timing differences -- as it is a continuous recording. |
Yikes.
Chris
Post by: Duardo on September 20, 2005, 08:41:18 PM
| Quote: |
to make the argument even more basic let's say we sample at 10 khz, which would make the brickwall 5 KHz what would be the delay between the samples? and wouldn't that delay have an effect on how, for example, a hihat pattern was reproduced in time? |
To say the same thing that's been said over and over again in this thread yet another slightly different way...
It doesn't matter what the sampling rate is. To oversimplify, a microphone picks up whatever SPL is present at its capsule, that signal is amplified and send on to a converter, which filters out everything below the Nyquist frequency and converts it to digital. It doesn't matter if the hihat is struck between samples. I suppose it will almost always be struck between samples. The thing is, since we're dealing only with frequencies below the Nyquist frequency, even though the initial attack of the sound "begins" between samples, between the sample before the hihat is struck (where there will still be some noise) and and the sample after it is struck, there is one and only one "curve" that will fit between those two samples, which will recreate the "beginning" of the sound of the hihat being struck exactly as it occurred.
-Duardo
Post by: Bill Mueller on September 20, 2005, 09:19:34 PM
Can I ask that you accept the multiple and clear explanations of the experts on this forum and give up this sad circus? Your premise was, and is flawed and no amount of uninformed arguing will bring it to life. I taught audio for 25 years and never met anyone so unwilling to attempt to understand what was being taught to them. When you have it so wrong, it is no longer a reasonable discussion, but a sad, futile and childish enterprise.
I am asking this so that you can maintain your standing here, so that when you post a question in the future, everyone does not groan and say, "Oh no, here we go again". Your tenacity shows that you are passionate about audio, and I laud you on that. However, this one is lost. Please give it up so we all can move on to a more productive exchange of ideas.
Best Regards,
Bill
Post by: crm0922 on September 20, 2005, 10:48:40 PM
Chris
Post by: echotp on September 20, 2005, 11:55:05 PM
Its been determined that time delay differences of 15 microseconds between left and right ears are easily discernible by nearly anyone. Thats less than the time difference between two samples at 48kHz (about 20 microseconds). Using a single pulse, one microsecond in length as a source, some listeners can perceive time delay differences of as little as five microseconds between left and right. It is therefore, indicated that, in order to provide a system with exact accuracy concerning imaging and positioning, the individual samples should be less than five microseconds apart. At 96kHz (a popularly preferred sample rate) there is a 10.417-microsecond space between samples. At 192kHz sample rate there is a 5.208-microsecond space between samples. This reasoning suggests that a sample rate of 192kHz is probably a good choice. As processors increase in speed and efficiency and as storage capacity expands high sample rates, long word length will become an insignificant concern and we'll be able to focus on the next audio catastrop
Post by: dcollins on September 21, 2005, 12:25:31 AM
| echotp wrote on Tue, 20 September 2005 20:55 |
a interesting thought from Bill Gibson: Its been determined that time delay differences of 15 microseconds between left and right ears are easily discernible by nearly anyone. Thats less than the time difference between two samples at 48kHz (about 20 microseconds). Using a single pulse, one microsecond in length as a source, some listeners can perceive time delay differences of as little as five microseconds between left and right. It is therefore, indicated that, in order to provide a system with exact accuracy concerning imaging and positioning, the individual samples should be less than five microseconds apart. At 96kHz (a popularly preferred sample rate) there is a 10.417-microsecond space between samples. At 192kHz sample rate there is a 5.208-microsecond space between samples. This reasoning suggests that a sample rate of 192kHz is probably a good choice. As processors increase in speed and efficiency and as storage capacity expands high sample rates, long word length will become an insignificant concern and we'll be able to focus on the next audio catastrop |
Have you or Bill read the rest of this thread?
DC
Post by: dcollins on September 21, 2005, 12:26:40 AM
| four wrote on Mon, 19 September 2005 09:09 |
The least steady-state periodic waveform... changes all the time... Ooh! Ooh! Sine Wave! |
Winner!
DC
Post by: Bob Olhsson on September 21, 2005, 12:31:25 AM
| timrob wrote on Tue, 20 September 2005 13:01 |
...the reason 44.1k was chosen as the standard originally was because it seemed to be the best compromise... |
Everybody knew 44.1kHz. was a sonic compromise but it was necessary because almost nobody could afford to edit digital audio on mainframe computers.
Post by: Johnny B on September 21, 2005, 12:55:13 AM
First, Nyquist has been dead for a long, long time. Does "above-Nyquist" mean standing above his cold, dead corpse?
Next, the SOB was not God.
And third, even if the SOB had the math right, it don't mean shit.
Why? Because if all the lemmings following him implemented it all wrong or made some bad assumptions, then it all still sucks.
Further, even if it looks perfect on a damn scope does not mean it sounds worth a shit. Lots of things can spec out great, look good on a scope, and still sound like shit. There's only one way to "prove" something sounds good, and that's to listen to it in real world situations.
Since people have been raging about digital vs. analougue for 25 years, I cannot worship at the alter of Nyquist, or at least, not yet.
If you really want to capture that 104-plus kHz that Boyk measured, that means a minimum of 208kHz, does it not? Is that the new Nyquist figure? 208kHz?
And has it ever occurred to some of the luddites stuck in 1980's tech, that the R&D labs at the chip makers just might know something that others don't...emmm...say like "trade secrets"...and that they are not involved in a huge right-wing world-domination conspiracy to sell everyone faster chips only because they can make some more money...
Did Intel, Motorola, IBM, and AMD push those CPU speeds ever higher in line with this same conspiracy theory?
Let's say it was a big monopolistic conspiracy for the moment, did you benefit from those speed increases?
Does not chip tech always advance forward? That's the only conspiracy, it's a conspiracy to move things forward and advance the technology. Some people might call it "competition."
And what about that 384Khz that will be needed for PCM to catch up with the sonics of DSD that GM discussed with Lynn?
Regardless of the side that one comes down on re: the speed issue, the conspiracy issue, the mic and speaker issue, things will change whether we like it or not.
I suspect new formats with new increased speeds and with new methodogies will be upon us very, very soon. My only hope is that it all sounds better than what came before, but if other areas of tech innovation provide a road map, it *should* sound better. Whether it will or not, is an open question.
I wish all of you good luck when the new tech hits.
Post by: crm0922 on September 21, 2005, 02:11:38 AM
If "the math don't mean shit", I don't know what anyone can say to convince you of anything.
There are mathematically quantifiable errors and distortion in digital recordings, minimizing these and removing artifacts is what has been improving year by year. Improving clock performance, noise specs (not that it matters all that much), and other meaningful improvements have brought digital audio to quite a respectable level. Some would say, anyways.
It is not being questioned whether the Nyquist theorem works or not. This is established fact. There isn't much more to say.
"Above Nyquist" is a frequency that is higher than the sample rate divided by two. That is 48000Hz/2 = 24000Hz, so higher than 24kHz frequencies.
"Below Nyquist" are frequencies below the above-mentioned number.
Any piece of audio signal that peaks in between two sample points, by definition, is of a frequency is higher than the Nyquist limit (sample rate/2).
I don't know how many more times this can be stated.
Please read:
http://www.lavryengineering.com/documents/Sampling_Theory.pd f
Explains sampling theory (a must read for Andy and Max) and makes a case for why 192kHz is excessive and *less* optimal at capturing audio frequencies for accurate reproduction.
http://www.lavryengineering.com/white_papers/sample.pdf
Explains some of the problems with digital audio using Nyquist roll-offs close to 20k.
Please read these papers and try to internalize the Nyquist theory fully. It is an absolute necessity before suggesting (to very experienced engineers) how digital audio might be improved.
Chris
Post by: mark fassett on September 21, 2005, 02:42:47 AM
| My World wrote on Tue, 20 September 2005 01:08 |
I am personally glad to see this topic on GM's forum because I have all but been kicked off of Dan Lavry's forum. |
You could be kicked off this board too, but for a different reason. See the sticky thread above.
Post by: Johnny B on September 21, 2005, 03:20:12 AM
And digital just does not sound as good as analogue...not yet anyway.
Why do you think people have been complaining for so long about the sound quality of digital?
Cause it's mathematically perfect....I don't think soooo.
At best, math can give one only a meager approximation of reality, it's not reality itself. Math is nothing more than a concept, it exists only in people's heads.
Maybe GM or Walter Sear are right...384kHz or maybe five times that fast...
We'll have to wait and see what the Next Gen Chips and New Formats bring...one thing's sure...they are headed our way...no doubt about that...
And when they hit, I wish all of you the very best of luck. I really do.
Looking forward to GM's return. I miss him, I'll bet we all do.
Post by: Tim Gilles on September 21, 2005, 03:28:22 AM
| lord wrote on Tue, 20 September 2005 13:59 |
THIS IS THE MOST RETARDED THREAD EVER. |
Not fer nuthin.... but if I see Lord @ AES, I am gonna RUN LIKE HELL in the opposite direction.
Tim "Rumblefish" Gilles
Post by: crm0922 on September 21, 2005, 04:41:24 AM
That is all I will contribute to this thread. The brainless grasping at technical topics by people without the background or intelligence to comprehend them is too much for me.
Enjoy your lives of eating up marketing hype and truly having no idea what any of it means.
Just for the record, I like analog better as well. This does not make Nyquist incorrect, nor does it mean that higher and higher sampling rates are the answer. They aren't. Read Dan's paper.
Chris
24K is arbitrary???....for the love of god man, read a book.
Post by: maxdimario on September 21, 2005, 05:27:43 AM
| Quote: |
On a related note, 78 rpm records seldom had a frequency above 7k. |
this is actually a very interesting point!
the medium which seems to capture best what I am talking about is the 78 record..
I have a Caterina Valente 78 from the late 50's on polydor which is an example of exactly what is lacking in digital.
the voice and the snare drum come out at you in a way that is so rock-stable sounding if feels as if the sound is in the room.
even if they are bandwidth limited, they sound so fixed in time and space that the sound has a presence in the room.
lp's don't exhibit as much of the presence effect, but the right LP's are more present than CD for sure (cd as an example...)
Post by: maxdimario on September 21, 2005, 05:30:37 AM
| Quote: |
Also wrong. The ability of tape to accurately represent rise times is directly related to the bandwidth. Same as digital |
I am not talking about rise time, it's obvious that rise time is a function of slew rate. I am talking about the exact moment a sound begins and ends.. how can I make this more understandable?
Post by: maxdimario on September 21, 2005, 05:42:39 AM
| Quote: |
Yes, you are right. The hi-hat is delayed. It is further delayed by the time it takes for you to rewind and play back the file a minute later. |
I am so glad that some of you take that extra little bit of time to express your observations, thanks.
if the time-window between two samples is X period of time long, then any event that begins at a given voltage will begin to charge the filter at any fraction of X.
so if an event begins right after the last sample, or between samples or right next to the following sample the same waveform should charge the filter a different amount because of the limited slew rate...no?
Post by: crm0922 on September 21, 2005, 06:05:20 AM
| maxdimario wrote on Wed, 21 September 2005 05:30 |
I am not talking about rise time, it's obvious that rise time is a function of slew rate. I am talking about the exact moment a sound begins and ends.. how can I make this more understandable? |
Here's your original quote:
| Quote: |
even if the pulse had a ramp-up time, there will have to be a moment in time when the ramp-up BEGINS. quantization will make this precise moment fit into the sampling grid. on tape there IS NO SUCH TIME RESTRICTION |
You are saying that the ramp-up happens, but it's "accuracy in time" is limited somehow by being "sampled into a grid" (your wording). This is saying the rise time is not able to be represented accurately by Nyquist. If there is information that is between Nyquist samples, this information is in the above-Nyquist realm.
Your hi-hat or whatever has a rise time that is represented within Nyquist unless it contains higher than 20kHz frequencies and you wish to capture them for some strange reason. The first rise is slow enough that the position of the first sample is not important, it will be reconstructed from either 0 or the noise floor. Nor does it affect timing, phase, or anything else you have attribted to this flawed theory.
I can't believe I just said it again. What is going on in here?
Read Dan Lavry's papers. Please....pretty please with sugar on top. Even if some of it is over your head. Please.
You did not just come upon something here that has been missed for 80 years. Sorry. At first I was trying to help, now I am just so baffled I can feel my IQ decreasing just posting with these people.
Chris
Post by: maxdimario on September 21, 2005, 06:23:24 AM
| Quote: |
You are saying that the ramp-up happens, but it's "accuracy in time" is limited somehow by being "sampled into a grid" (your wording). This is saying the rise time is not able to be represented accurately by Nyquist. If there is information that is between Nyquist samples, this information is in the above-Nyquist realm. |
I wasn't talking about the accuracy of the ramp angle, or maximum slew rate, I was talking about the precise moment that an impulse begins.
you could have a 100hz triangle wave which has a ramp/slew rate much lower than nyquist, coming out of a sinth, but if it begins between samples it will affect the filter differently, albeit in a very small way. Who's to say you can't hear it? if it does exist, of course.
I don't want to lower anyone's IQ level, but at least I'd like to be understood for what I have said, right or wrong.
Post by: maxdimario on September 21, 2005, 06:59:48 AM
Maybe he's wrong as well...
http://www.digitalprosound.com/2002/12_dec/editorials/hear_t his.htm
my fave part!
| Quote: |
But the problem is the true reproduction of the higher frequencies in that 3-18K region that makes all the difference, and it has very little to do with distortion measurements. The Nyquist theory is correct by stating that any sine wave sampled at twice its frequency can be reproduced accurately. The problem is it does not guarantee that it will be reproduced in the correct location in time as it was originally — with relation to other present waveforms. |
Post by: Eric Bridenbaker on September 21, 2005, 08:43:49 AM
| maxdimario wrote on Wed, 21 September 2005 06:59 | ||
I've found this article on the net by Paul Wolff, director of engineering at API, which seems to basically say what I am trying to say.. Maybe he's wrong as well... http://www.digitalprosound.com/2002/12_dec/editorials/hear_t his.htm my fave part!
|
Yep, he's onto it. This thread is getting really long over what seemed to be a simple idea. Ask the questions, get the answers, right?
The problem must be that we have been given the answers, and some of us just don't want to hear them.
But this should come as no suprise. The jist of what I keep hearing here is:
- If it can't be expressed as a sine wave, then it doesn't exist.
- If it's above 20K, you can't hear it so don't worry about it.
- The system works. How dare you question the system? You must be a fool.
Still reading, but have not found a satisfactory answer to the time smear question within the Nyquist documentation so far. But then again, I'm a little slow.
Cheers,
Eric
Post by: Johnny B on September 21, 2005, 09:45:41 AM
Without trying to cast any undue aspersions, it should be pointed out that not everyone agrees with Dan. In fact, many people disagree with Dan. I believe GM expressed his concerns about Dan not making 192 boxes...I guess one might say that some of you are calling GM names too.
Why not disagree with people, but still show some respect.
I can't wait for GM to get back, then people can ask him what his current feelings are about new SRC's and new formats.
Post by: timrob on September 21, 2005, 10:53:04 AM
| maxdimario wrote on Wed, 21 September 2005 05:59 | ||
I've found this article on the net by Paul Wolff, director of engineering at API, which seems to basically say what I am trying to say.. Maybe he's wrong as well... http://www.digitalprosound.com/2002/12_dec/editorials/hear_t his.htm my fave part!
|
I met Paul a few years ago, when he was doing some mods to the Sony console at Sony/ATV. I'd be careful not to assume that he totally agrees with your hypothesis as stated. You've been talking about impulse response and such not phase vs. frequency.
It is what he didn't say that is important and that is where the time difference comes from. You extrapolated from what he said that sampling frequency is the only factor. If that were the case it would effect all frequencies within the Nyquist range. The Filters cause phase distortion at higher frequencies.
BTW, this phenomenon exists in the analog realm. Capacitors and Trasformers both exhibit phase distortion to some degree.
ELI the ICE man anyone?
The truth is that no audio system around reproduces time perfectly. Not analog, not digital. Once the sound is recorded, It will never be experienced the same way again. Analog may sound better, but it is not necessarily because of its time resolution.
Tim Roberts
Waterknot Music
Nashville
Post by: JamSync on September 21, 2005, 10:56:01 AM
| dcollins wrote on Wed, 21 September 2005 05:26 | ||
Winner! DC |
Ah...you mean the *slope* changes all the time.
Post by: Johnny B on September 21, 2005, 11:18:31 AM
Bring on those faster chips and new formats, let's see how they perform in real world stressed out situations. More importantly, let's "hear" with our own ears how they sound.
Let's also see if the new chips and new formats can provide listeners the ear/brain/body experience of hours of good feelings instead of giving people more headaches.
Post by: timrob on September 21, 2005, 11:19:10 AM
from
http://www.sfu.ca/~truax/bourges3.html
"From a psychoacoustic point of view, the auditory system can be understood as needing to balance the perception of temporal variation with frequency or spectral variation. Again, it is a matter of the linkage between the time and frequency domains. If we were able to track the time behaviour of sound precisely - the waveform level being the most ridiculous example - we could not perceive anything in the frequency domain. And likewise, more precise spectral analysis than that of the critical bandwidth would have to occur at the expense of tracking temporal variations. Speech perception probably defines the normal balance between temporal and spectral tracking on the part of the auditory system where phonemes occur at a speed of around five per second. This aspect of perception is sometimes termed "demodulation" because it is not primarily the sound "carrier" that is most important, but rather the changes that have been "encoded" into it in both the time and frequency domains."
Tim Roberts
Waterknot Music
Nashville
Post by: Johnny B on September 21, 2005, 11:29:51 AM
It will be interesting to compare male and female populations in this respect.
Bottom line is that more research is needed as there is still much to learn about the ear/brain/body interaction.
Post by: Johnny B on September 21, 2005, 11:43:03 AM
It is often the case that there is an accidental discovery or lucky scientific break-through which is then later followed by some attempt to explain and help us understand. Not altogether unlike a decision to act based on compulsive-obsessive behavior disorder followed by one's excuses or rationalisations. More simply---act first...explain later.
New science and new technology often follows this path...it is not infrequent that you may hear someone say, "We don't know yet why it works."
But to really understand what the hell is really going on with that ear/brain/body interaction, it will undoubtedly require years of effort by entire research teams which properly utilise a multi-disciplinary approach.
Math alone will never give all the important answers.
Let us not fool ourselves that everything "important" is fully understood.
Let us not be limted and constrained by old belief systems or fooled by old mythology like 20Hz-to-20kHz are the only "important" frequencies.
Our ears, our brain reactions, and the feelings in our bodies strongly suggest that digital sound quality and the underlying technology have a long way to go before it meets the "Gold Standard" set by Analogue.
However, I for one, will attempt to keep an open mind and am perfectly willing to experience what the Next Gen in digital has to offer.
Post by: StudioRhythm on September 21, 2005, 11:56:30 AM
| Johnny B wrote on Wed, 21 September 2005 08:29 |
Bottom line is that more research is needed as there is still much to learn about the ear/brain/body interaction. |
-- does that mean you just finished your thesis...?I'm no expert on Nyquist, but let me try to explain some of the things you've been missing during this thread in a different way.
First, there's no such thing as an "instantaneous" beginning to an event like an attack -- an instant attack has an infinite frequency response, and we only hear to 20kHz (or so; I like Dan Lavry's approach of doubling it for safety, so call it 40kHz, OK?)
NOTHING can reproduce an instantaneous attack -- no microphone, no mic preamp, no equalizer, no compressor, not even your favorite tape machine. In fact, especially your favorite tape machine!
EVERYTHING "delays" the attack of something with super-high frequency content if it doesn't reproduce all of the (inaudible) higher frequencies. Does this mean every piece of gear ever invented sounds bad?? No, it means that the gear isn't reproducing things we don't hear. Of course, for analog gear, it's useful to have the gear extend its bandwidth to very high frequencies, but that's because it affects those frequencies we do hear. Using digital filters and oversampling, digital doesn't need to reproduce anything over what we hear.
The point is that if the ear can't perceive whether an event started with 1/100th of a millisecond accuracy (which is what we're talking about here, or more with double sample rates), it doesn't matter if it's represented with exact precision, which is good, because exact precision doesn't exist in any piece of gear that has ever been made or ever will be made. Instead, we use equipment that has exact precision within the bandwidth of our hearing.
If you still think it matters, try getting your drummer to play his hihat 1/100th of a millisecond earlier -- then he'll be "right on the grid" and it won't matter....!
| Johnny B wrote on Wed, 21 September 2005 at 08:18 |
Since math is constrained and limited to being purely conceptual, math can take us only so far, it's always real world "stress tests" that provide us with what really counts. In this case, whether or not it sounds as good as the time-tested and proven "Gold Standard"---Analogue. Bring on those faster chips and new formats, let's see how they perform in real world stressed out situations. More importantly, let's "hear" with our own ears how they sound. Let's also see if the new chips and new formats can provide listeners the ear/brain/body experience of hours of good feelings instead of giving people more headaches. |
Can you see the basic flaw in your argument here? What do you think the people who make the new, faster chips use in designing them? I'll give you a hint -- it isn't chemistry, history, industrial arts or home economics! Give up? IT'S MATH!!!! The horror!
What are we supposed to do, just throw every combination of possible circuits into a box and then do listening tests on all of them? I guess that considering 99.9999999999999% of them won't pass any audio whatsoever, that'll speed up the listening process somewhat....
Post by: Johnny B on September 21, 2005, 12:57:54 PM
| Dan Feiszli wrote on Wed, 21 September 2005 16:56 |
we only hear to 20kHz (or so; I like Dan Lavry's approach of doubling it for safety, so call it 40kHz, OK?) |
I disagree with the myth that "ONLY" 20Hz to 20kHz contains "ALL" the "IMPORTANT" frequencies. I think this is total bullshit. Please take the time to read what David Blackmer had to say about this, ok?
CalTech Prof. James Boyk has already measured frequencies up to 104kHz and he did not even measure "everything" there is to measure! We also "know" that pipe organs produce lows of 8Hz...I can "feel" those lows in my body, can you?
Oh wait a minute, following blindly along behind the now discredited 20-to-20 myth, that's below 20Hz, so let's just say those frequencies are unimportant. We don't have any good evidence that people don't feel or experience these lows, that's Ok, we will just perpetuate the lie. We'll just try to ignore it.
Let's switch the topic and talk about mics and speakers, that'll keep people from noticing that the lows are completely missing and screwed up beyond belief.
The lows are far too neglected in all this increased speed discussion, who will speak out for the lows? Oh, and BTW, there are a lot of problems in the low-end in digital, do a little reading and you may learn something about it.
The very first thing that must go is the abandonment of old mythology, like the crusty old myth that says "ONLY 20-to-20 is IMPORTANT." What utter nonsense!
Post by: maxdimario on September 21, 2005, 01:10:43 PM
phase shift is a form of time distortion found in analog, but it is linear, predictable and occurs in nature.
in digital the kind of time distortion that exists is erratic, mathematically derived.
depending on the frequency and the waveform in general time distortion can occur even for events that are within the slew rate limit.
the fact that there is no microphone that has an instantaneous attack has nothing to do with the issue, since AD conversion is slew-limited and since we are talking about time distortion as it exists in digital systems: the reproduction of the waveforms is dependent on a fixed quantization rate and the errors are not linear such as in normal phase lag, but mathematically derived errors which produce distortion which is not predictable and not relative to what is going on sonically.
it's no use in insisting on textbook definitions of issues that have to do with reconstruction of test tones.
there is a difference, and it can be heard.. especially at low sampling rates.
my personal dislike for digital has always had to do with some kind of phase distortion that I was hearing, which manifests itself in poor imaging, lack of a solid image etc.
people must realize how important it is for the ear to have the sound time-aligned, if not in phase... at least in relationship of an absolute timeline and the waveform as a whole.
Post by: StudioRhythm on September 21, 2005, 01:16:22 PM
| Johnny B wrote on Wed, 21 September 2005 09:57 | ||
I disagree with the myth that "ONLY" 20Hz to 20kHz contains all the "Important" frequencies. I think this is total bullshit. Please take the time to read what David Blackmer had to say about this, ok? CalTech Prof. James Boyk has already measured frequencies up to 104kHz and he did not even measure "everything" there is to measure! We also "know" that pipe organs produce lows of 8Hz...I can "feel" those lows in my body, can you? Oh wait a minute, folllowing blindly along along behind the 20-to-20 myth, that's below 20Hz, so let's just say thoe frequencies are unimportant. We don't have any good evidence that people don't feel or experience these lows, that's Ok, we will just perpetuate the lie. We'll just try to ignore it. Let's switch the topic and talk about mics and speakers, that'll keep people from noticing that the lows are completely missing and screwed up beyond belief. The lows are far too neglected in all this increased speed discussion, who will speak out for the lows? Oh, and btw, there are a lot of problems in the low-end in digital, do a little reading and you may learn something about it. The very first thing that must go is the abandonment of old mythology, like the crusty old myth that says "ONLY 20-to-20 is IMPORTANT." What utter nonsense! |
Like I said, double it, do whatever you want -- there's a point where we just can't hear it. Why don't you tell me what you can hear and we'll go from there.
You know, there's a huge difference between measuring that instruments produce frequencies and saying we can hear them. My cell phone produces frequencies too -- should we include those frequencies in our audio recording equipment? Your argument so far has been claiming that tape is better than digital, and you're pointing to frequency response as the answer, when tape starts rolling off well before high-rate digital.
http://www.endino.com/graphs/
There are clear, obvious, measureable differences in tape vs. digital -- if you like the sound of tape better, just say "I like it better" and use it. Why try and use really flawed arguments when there's clear differences already there? No need to use voodoo, how about just accept the obvious proven differences?
As for your claim that digital doesn't reproduce the lows, my Lavy A/D converter is flat to +-0.05 dB down at 10Hz (that's 1/20th of 1dB.) They go down farther than that, as well; it's my understanding that it basically goes down to around 1Hz with a filter to prevent DC offset. Again, look at those tape machine response graphs -- many machines are as much as 6dB down at 30Hz! Not much comparison there, I'm afraid.
Post by: maxdimario on September 21, 2005, 01:27:54 PM
| Quote: |
The point is that if the ear can't perceive whether an event started with 1/100th of a millisecond accuracy (which is what we're talking about here, or more with double sample rates), it doesn't matter if it's represented with exact precision, which is good, because exact precision doesn't exist in any piece of gear that has ever been made or ever will be made. Instead, we use equipment that has exact precision within the bandwidth of our hearing. |
you never know what the ear can hear or not technically. Timing is crucial for the interpretation of music and for realism's sake. If that error occurs once every two seconds it's one thing, but when the error continues constantly it's another.
my point was to identify that digital has some shortcomings, and they are derived from quantization.
the shortcomings are not the same as analog shorcomings, they are worse for feel and imaging, especially at low sampling rates.
I am not concerned with sounds above 20 Khz, the artifacts I hear on 44.1 go down into the 3.5 Khz area and above and they create non-linear phase-distortion.
Get to a point where you can't hear it and digital will be o.k.
I have a feeling the sample rate needs to be pretty high indeed to compete with a lower bandwidth analog recorder.
Post by: StudioRhythm on September 21, 2005, 01:31:53 PM
| maxdimario wrote on Wed, 21 September 2005 10:27 | ||
you never know what the ear can hear or not technically. Timing is crucial for the interpretation of music and for realism's sake. If that error occurs once every two seconds it's one thing, but when the error continues constantly it's another. my point was to identify that digital has some shortcomings, and they are derived from quantization. the shortcomings are not the same as analog shorcomings, they are worse for feel and imaging, especially at low sampling rates. I am not concerned with sounds above 20 Khz, the artifacts I hear on 44.1 go down into the 3.5 Khz area and above and they create non-linear phase-distortion. |
But did you understand my point that a band-limited analog recorder (i.e. every analog recorder ever made) suffers from those same "problems"? I'm not saying digital is perfect, but good digital sounds really good to me, and this particular argument seems deeply flawed.
Post by: maxdimario on September 21, 2005, 01:41:00 PM
you cannot hear this if you record through equipment that has already smeared the sound enough to filter out some of that precious information.
A lot of high end amplifiers reduce negative feedback, or remove it precisely for this reason. Playing with the time-alignment of the sound-components confuses the ear and makes them progressively dead sounding.
Post by: Johnny B on September 21, 2005, 01:48:45 PM
I am not, however, going to be sucked into a hysterical, fear-based diatribe against technologial progress which can include greater bit-depths and vastly increased speeds. If people want to stay stuck in the past, with all the time smear, strange phase issues, weird imagining, truncations, math errors, and all the other digital anomalies, I'm also Ok with that.
Me? I'm curious about whether the Next Gen chips and new formats will improve digital sound quality. I'm also curious whether or not entirely new methods will be needed in the future to get digital sound quality up to world class analogue sound quality standards.
I can say this, it won't be scope tests, spec sheets, or atempts to convince me with propaganda based on math by any of those making devices and indirectly trying to sell them to me which will do the trick for me, only a good listening test by me will do.
I'm willing to give the "New and Improved" digital stuff a good listening test, that's the only way I'll know if I really like it or not.
As always, some people will normally be leaders and find their own path, others will be followers.
Post by: StudioRhythm on September 21, 2005, 01:49:06 PM
| maxdimario wrote on Wed, 21 September 2005 10:41 |
Analog &digital..they both have problems, but digital screws with the time erratically (changes continuously, illogicaly to the ear), therefore screwing up the imaging, depth, feel and everything else which depends on a good reproduction of the elements of sound that make a recording 'real' sounding and emotionally captivating. you cannot hear this if you record through equipment that has already smeared the sound enough to filter out some of that precious information. A lot of high end amplifiers reduce negative feedback, or remove it precisely for this reason. Playing with the time-alignment of the sound-components confuses the ear and makes them progressively dead sounding. |
But digital doesn't "screw with the time erratically"! It's predictable and "screwed with" by the same amount as an analog recorder with the same bandwidth -- if an attack has higher frequency content than the digital system can handle, it starts "late" -- by the same amount as an analog recorder with the same bandwidth. Within the bandwidth, it is reproduced exactly as it entered (plus a little noise, around 120dB down.) Frequency and time are the same thing, right?
Post by: StudioRhythm on September 21, 2005, 02:07:32 PM
| Johnny B wrote on Wed, 21 September 2005 10:48 |
Dan, if you cannot hear it, that's ok with me. I will only add that there are many tone deaf people involved with digital. |
Dude, that's priceless! That's it, I must be tone deaf. All the mastering engineers who've told you that you were wrong probably are too. But you have golden ears!
| Quote: |
I am not, however, going to be sucked into a hysterical, fear-based diatribe against technologial progress which can include greater bit-depths and vastly increased speeds. If people want to stay stuck in the past, with all the time smear, strange phase issues, weird imagining, truncations, math errors, and all the other digital anomalies, I'm also Ok with that. Me? I'm curious about whether the Next Gen chips and new formats will improve digital sound quality. I'm also curious whether or not entirely new methods will be needed in the future to get digital sound quality up to world class analogue sound quality standards. I can say this, it won't be scope tests, spec sheets, or atempts to convince me with propaganda based on math by any of those making devices and indirectly trying to sell them to me which will do the trick for me, only a good listening test by me will do. I'm willing to give the "New and Improved" digital stuff a good listening test, that's the only way I'll know if I really like it or not. As always, some people will normally be leaders and find their own path, others will be followers. |
Rock on! I'm not trying to say what we have now is perfect and can't be improved upon, all I'm saying is that the arguments you're making against it don't make sense and have no basis in reality. If you can use your skills as a natural born leader to lead us into a real discussion that makes sense, be my guest. Until then, I'll just follow the path of rational thought.
Post by: Johnny B on September 21, 2005, 02:21:39 PM
Wrong! The arguments I make are based upon sound science and human experience.
You, OTOH, are relying entirely on math. Math is not real, at best, it's only an aproximation of something that IS real.
But I'm glad to hear you claim that you do not want to stand in the way of technologial progress. I suspect there are those around the digital field who would like nothing better than to freeze things just as they are...but we all know in our hearts that ain't gonna happen, don't we?
Technological progress will march forward despite the best efforts of corporations and nations to thwart it, technological progess stands still and waits for no man, woman, or child.
Bring it on! Let's all keep an open mind and hear how it sounds.
Post by: StudioRhythm on September 21, 2005, 02:26:39 PM
| Johnny B wrote on Wed, 21 September 2005 11:21 |
Dan, Wrong! The arguments I make are based upon sound science and human experience. You, OTOH, are relying entirely on math. Math is not real, at best, it's only an aproximation of something that IS real. |
Actually, I'm a musician first and an engineer second, and the math that I have learned (which is not all that much) has been to understand what I hear. But, I'm willing to learn, which is something we don't appear to share.
| Quote: |
But I'm glad to hear you claim that you do not want to stand in the way of technologial progress. I suspect there are those around the digital field who would like nothing better than to freeze things just as they are...but we all know in our hearts that ain't gonna happen, don't we? Technological progress will march forward despite the best efforts of corporations and nations to thrawt it, technological progess stands still and waits for no man, woman, or child. |
Actually, those corporations are the ones who are trying to feed you bigger, better, faster, cheaper -- so you buy their stuff! And the nations want their corporations to do well so they have money...etc. Progress is only progress when it makes things better, you know....and I don't mean when the corporate bottom line gets better.
Post by: Johnny B on September 21, 2005, 02:46:58 PM
Sometimes, things that do not look good on paper sound really good. The math people have claimed that digital is great the way it is, they have been saying that for a long time, I don't agree. It still does not sound right to me.
Maybe the Next Gen will improve, maybe not. It might take several Gens, it might take entirely new tech to get it right. I dunno, but I will at least give whatever comes along a good listen. I will not dismiss it out of hand, and I certainly won't dismiss it based on hysterical, fear-based arguments.
I will give the new tech a fair chance to fail or succeed on its own merits or demerits.
I will listen for the sound quality, I will hold it up against analogue sound quality.
For me, that's the "Acid Test."
Post by: blairl on September 21, 2005, 02:49:26 PM
George's Old Forum
Post by: StudioRhythm on September 21, 2005, 02:57:52 PM
| Johnny B wrote on Wed, 21 September 2005 11:46 |
The difference between us is I do not automatically buy the argument of "better, faster, cheaper," although we have plenty of digital examples where that's been true. I am, however, willing to give the new tech a chance to prove itself by actually keeping an open mind and listening to it myself. |
What do you think I'm doing? I use it and listen, and the good stuff sounds good to me, better than some analog for some things. I'm not automatically buying anything anybody says, including flawed arguments that ignore things that we do know about sound, electronics and digital audio!
| Quote: |
Sometimes, things that do not look good on paper sound really good. The math people have claimed that digital is great the way it is, they have been saying that for a long time, I don't agree. It still does not sound right to me. Maybe the Next Gen will improve, maybe not. It might take several Gens, it might take entirely new tech to get it right. I dunno, but I will at least give whatever comes along a good listen. I will not dismiss it out of hand, and I certainly won't dismiss it based on hysterical, fear-based arguments. I will give the new tech a fair chance to fail or succeed on its own merits or demerits. I will listen for the sound quality, I will hold it up against analogue sound quality. For me, that's the "Acid Test." |
And, what exactly have you listened to, how did you test and compare it, and what specifically are your problems with it?
Post by: Johnny B on September 21, 2005, 03:01:20 PM
Good idea, providing a link to GM's old forum, I believe that's where GM may have said he had some problems with the narrow approach advocated by certain vendors of ADDA boxes who resist increasing speeds.
IIRC, GM also came down in favor of more R&D and more advanced and applied scientific research into the brain and body. More research just might hold the key to unlocking some of the puzzle.
Ok, I'm outta this thread. My best wishes to all.
Post by: timrob on September 21, 2005, 03:17:15 PM
| Johnny B wrote on Wed, 21 September 2005 14:01 |
blairl, Good idea, providing a link to GM's old forum, I believe that's where GM may have said he had some problems with the narrow approach advocated by certain vendors of ADDA boxes who resist increasing speeds. IIRC, GM also came down in favor of more R&D and more advanced and applied scientific research into the brain and body. More research just might hold the key to unlocking some of the puzzle. Ok, I'm outta this thread. My best wishes to all. |
I Love it. A guy jumps into a discussion, makes a bunch of inflammatory statements, then bails when someone asks him back it up. It would have been ok if any of his posts actually contained any information.
Post by: Ronny on September 21, 2005, 04:14:24 PM
| timrob wrote on Wed, 21 September 2005 15:17 | ||
I Love it. A guy jumps into a discussion, makes a bunch of inflammatory statements, then bails when someone asks him back it up. It would have been ok if any of his posts actually contained any information. |
That's ok Tim, others "have" gotten something out of this discussion, so our efforts have not fallen on deaf ears.
Post by: timrob on September 21, 2005, 04:35:28 PM
| Ronny wrote on Wed, 21 September 2005 15:14 |
That's ok Tim, others "have" gotten something out of this discussion, so our efforts have not fallen on deaf ears. |
Ronny, I sincerely hope someone gained some insight. It is certainly hard to tell through the noise.
Peace.
Post by: Eric Bridenbaker on September 21, 2005, 05:36:49 PM
| timrob wrote on Wed, 21 September 2005 16:35 | ||
Ronny, I sincerely hope someone gained some insight. It is certainly hard to tell through the noise. Peace. |
This discussion has definitely driven me to take my understanding of digital farther, brushing up on sampling theory, thinking about it etc... While I may not "get" the case in regards to the timing issue, it's still been of great benefit, so thanks for that.
I'm sure that this has fallen on some deaf ears, though (everyone knows most sound engineers are deaf anyway...)
Best,
Eric
Post by: tom eaton on September 21, 2005, 05:40:32 PM
If I play the snare between two kick samples will it play back?
-tom
Post by: bobkatz on September 21, 2005, 07:06:53 PM
Its been determined that time delay differences of 15 microseconds between left and right ears are easily discernible by nearly anyone. Thats less than the time difference between two samples at 48kHz (about 20 microseconds).
Which is completely irrelevant... the interchannel time resolution of 48 kHz 24 bit digital audio is much finer than the intersample timing. As DC said, it is limited by the noise of the system, and I suspect, also the jitter. Correct me if I'm wrong on the latter
.BK
Post by: Ronny on September 21, 2005, 07:40:09 PM
| TER wrote on Wed, 21 September 2005 17:40 |
If I play the snare between two kick samples will it play back? -tom |
Sure it will play back, those are called syncopated samples. I see it all of the time when I zoom in on reggae music.
Post by: Andy Simpson on September 21, 2005, 07:44:52 PM
| echotp wrote on Tue, 20 September 2005 23:55 |
a interesting thought from Bill Gibson: Its been determined that time delay differences of 15 microseconds between left and right ears are easily discernible by nearly anyone. Thats less than the time difference between two samples at 48kHz (about 20 microseconds). Which is completely irrelevant... the interchannel time resolution of 48 kHz 24 bit digital audio is much finer than the intersample timing. As DC said, it is limited by the noise of the system, and I suspect, also the jitter. Correct me if I'm wrong on the latter .BK |
Just out of interest and to help me get a different angled-handle on this question, how would one implement a 1 microsecond advance (or delay) of one channel of a 44/16 recording?
In the physical realm, one can simply move the mic closer (a teeny bit).
Presumably it can be done in theory?
Could it be done without upsampling?
Andy
Post by: Duardo on September 21, 2005, 10:07:39 PM
| Quote: |
some listeners can perceive time delay differences of as little as five microseconds between left and right. It is therefore, indicated that, in order to provide a system with exact accuracy concerning imaging and positioning, the individual samples should be less than five microseconds apart. |
That's not corrrect. While some can perceive time delay differences of as little as five microseconds, it does not follow that individual samples need to be taken less than five microseconds apart. You seem to think that what you're "hearing" is the individual samples. You're not.
| Quote: |
I am not talking about rise time, it's obvious that rise time is a function of slew rate. I am talking about the exact moment a sound begins and ends.. how can I make this more understandable? |
I think most people here to understand what you're talking about. What you don't seem to understand is that the "beginning" of the sound is still captured, even if it falls between two samples.
| Quote: |
I've found this article on the net by Paul Wolff, director of engineering at API, which seems to basically say what I am trying to say.. Maybe he's wrong as well... The Nyquist theory is correct by stating that any sine wave sampled at twice its frequency can be reproduced accurately. The problem is it does not guarantee that it will be reproduced in the correct location in time as it was originally ? with relation to other present waveforms. |
He is wrong as well, at least with the second sentence. The Nyquist theorem (it's not a theory) does indeed state that the any sine wave samled at twice its frequency can be reproduced accurately...in the correct location in time. If it's not, its a shortcoming somewhere else...like the filters...
| Quote: |
I find this term "above-" or "below- Nyquist" to be confusing the issue. |
It's fairly obvious that you find it confusing. As far as this discussion is concerned, it's important to distinguish what's below the Nyquist frequency in terms of waveforms as opposed to phase. People want to equate phase differences that are "smaller" than the amount of time between two samples to being able to capture sounds above the Nyquist frequency (regardless of sampling rate...forget 20-20kHz for a second). They are two separate issues. We don't need to capture those frequencies to be able to capture timing differences that are the same as the wavelength of those frequencies.
| Quote: |
Next, the SOB was not God. And third, even if the SOB had the math right, it don't mean shit. |
The "SOB" did have the math right. It was proven, and it does mean everything for digital audio. Without it it wouldn't work, at any sampling rate. You seem to have more of a problem with the whole 20-20 kHz thing, which has nothing to do with Nyquist himself other than his theorem dictates what sampling rate we'd need to use to be able to capture and recreate those waveforms. So it absolutely does mean shit.
| Quote: |
If you really want to capture that 104-plus kHz that Boyk measured, that means a minimum of 208kHz, does it not? Is that the new Nyquist figure? 208kHz? Is that the new Nyquist figure? 208kHz? |
Well, sure, if you really want to capture those frequencies, then 208 kHz would be the minimum. Nobody's arging that, are they? I for one don't want to capture those frequencies. I haven't heard a convincing reason why I'd want to and there are plenty of reasons why I don't.
And what do you mean by "the" new Nyquist figure? There is no single Nyquist frequency.
| Quote: |
Does not chip tech always advance forward? That's the only conspiracy, it's a conspiracy to move things forward and advance the technology |
Hopefully it will. But you seem to equate technology advancing forward with sampling rates going up. Would you not consider chips that are more accurate at the existing samling rates to be advancing forward? I certainly would.
| Quote: |
Let's also see if the new chips and new formats can provide listeners the ear/brain/body experience of hours of good feelings instead of giving people more headaches. |
Funny, I listen to CD's for hours and the only ones that give me headaches are the ones that have been over-compressed and -limited, and I've stopped listening to them.
| Quote: |
However, I for one, will attempt to keep an open mind and am perfectly willing to experience what the Next Gen in digital has to offer. |
Your mind is far from open. You are totally unwilling to accept the possibility that the principles upon which digital audio is based are actually correct.
I'm not saying that digital audio is perfect as it is now. I don't think it ever will be. And I don't think that analog recording was either.
| Quote: |
NOTHING can reproduce an instantaneous attack -- no microphone, no mic preamp, no equalizer, no compressor, not even your favorite tape machine. In fact, especially your favorite tape machine! |
Don't forget your ears!
| Quote: |
I disagree with the myth that "ONLY" 20Hz to 20kHz contains "ALL" the "IMPORTANT" frequencies. I think this is total bullshit. Please take the time to read what David Blackmer had to say about this, ok? |
| Quote: |
CalTech Prof. James Boyk has already measured frequencies up to 104kHz and he did not even measure "everything" there is to measure! We also "know" that pipe organs produce lows of 8Hz...I can "feel" those lows in my body, can you? |
As has already been mentioned (in this thread and others you've participated in) nobody is arguing the fact that those frequencies are present. Just that they don't need to be captured.
As for the lows...sure, I can feel them. The shortcomings there aren't with digital audio nearly as much as they are with the transducers used to capture and play back those sounds (or with analog recording, for that matter).
| Quote: |
I am not, however, going to be sucked into a hysterical, fear-based diatribe against technologial progress which can include greater bit-depths and vastly increased speeds. If people want to stay stuck in the past, with all the time smear, strange phase issues, weird imagining, truncations, math errors, and all the other digital anomalies, I'm also Ok with that. |
Again, this is where you're being closed-minded. Are you not willing to accept that 44.1 kHz, or even 96 kHz, may continue to sound better and better as time marches on and technology improves? It seems that you think that "technological progress" is limited to greater bit depths (which nobody here is arguing about anyhow) and faster sampling rates. As for weird "imagining"...I think that's only going on in your head.
-Duardo
Post by: tom eaton on September 21, 2005, 10:13:23 PM
Just for our edification, what converters, clock, monitors and amps are you using to monitor your digital recordings?
Thanks-
tom
Post by: Ronny on September 21, 2005, 11:04:38 PM
Post by: Terry Demol on September 21, 2005, 11:04:53 PM
| maxdimario wrote on Wed, 21 September 2005 18:10 |
the whole issue has to do with time resolution. my personal dislike for digital has always had to do with some kind of phase distortion that I was hearing, which manifests itself in poor imaging, lack of a solid image etc. people must realize how important it is for the ear to have the sound time-aligned, if not in phase... at least in relationship of an absolute timeline and the waveform as a whole. |
Max,
I have followed a lot of your posts and really like your
passionate approach to things.
However, I definately think you need to hear *great* digital
playback before judging it so harshly, WRT it's lack of image
solidity, apparent phase problems, HF hardness etc etc.
Out of curiosity, what is your dig playback setup?
Cheers,
Terry
Post by: Johnny B on September 21, 2005, 11:54:37 PM
| Duardo wrote on Thu, 22 September 2005 03:07 | ||
Well, sure, if you really want to capture those frequencies, then 208 kHz would be the minimum. Nobody's arging that, are they? I'm not saying that digital audio is perfect as it is now. I don't think it ever will be. -Duardo |
It appears we have some basis for agreement, perhaps not on every point tho' and that's OK by me.
Alright, now I'm really out.
Cheers.
Post by: J.J. Blair on September 22, 2005, 03:25:47 AM
Post by: dcollins on September 22, 2005, 04:08:52 AM
| J.J. Blair wrote on Thu, 22 September 2005 00:25 |
If I slit my wrists after reading this thread and record it at 44.1 khz, will I die slower than if I record it at 192? |
As we have "learned" from this thread, you can slit your wrists by a mere 1.7 cm at 44.1, so it may take a while.....
DC
Post by: maxdimario on September 22, 2005, 05:17:08 AM
| Quote: |
Out of curiosity, what is your dig playback setup? Cheers, Terry |
I have a modified RME, but I don't use my playback setup as an absolute reference, i use records that have been released commercially, as well as work done in studios that have better converters than mine.
there is a definite digital sound. People know this and they have been saying it from day one.
I sold all of my vinyl in the early 90's because I thought digital was a superior medium (yes, even 16/44.1).
After a while I noticed I stopped listening regularly to records like I used to.
a few years ago I got back into records and re-discovered the magic element in performance, that had been partially lost on cd.
a lot of performance-based records are just not as effective without that vibe getting through 100%.. just because they 'SOUND' good (balanced, aesthetically pleasing)doesn't mean that they are done correctly, the records also must fool you into feeling as if the artist was performing in front you, just for you.
to do that the sound has to be real enough, image and depth-wise, to convey the performance.
when I started working with early digital multitrack I noticed it was hard work getting a feel going, although things have improved.
I still hear this diminished resolution of feel, depth..especially ITB and/or with a lot of tracks going.
even a pre straight into an ampex 440 does not exhibit this high end time distortion that makes sounds less 'there', although it does distort the sound aesthetically more than most digital.
It's obvious I'd rather have digital that was capable of doing this as well..being more convenient and edit-able.
maybe with very high sample rates the time-resolution is good enough at high frequencies that it can be as convincing as analog regarding feel and rock-solid imaging, depth etc..
but then again, like I said above, the most stable (mono) imaging I have heard so far is late 50's 78's..noisy, distorted, band-limited but real-sounding.
I'm not saying we should go back to recording on lathes, but we should be learning from the comparisons, which are often surprising.
Post by: tom eaton on September 22, 2005, 08:41:01 AM
And you're saying the best mono imaging you've heard is from a mono recording, right? That seems logical to me. What mono digital recordings have you heard for comparison?
You keep adding variables to your argument.
It's hard to believe that you're now stating "high end time distortion" as some kind of given in your arguments. At or above what frequency do you hear this distortion?
I'm really trying to understand what you're getting at that hasn't been addressed here.
-tom
Post by: StudioRhythm on September 22, 2005, 11:38:12 AM
| Quote: |
I have a modified RME, but I don't use my playback setup as an absolute reference, i use records that have been released commercially, as well as work done in studios that have better converters than mine. |
If you're making all your claims about digital sound based on listening through an RME converter, now it all starts to make sense. I also have an RME converter (ADI-8DS,) which sounds OK for what it is (an older mid-priced unit,) but I certainly wouldn't put it in the same league as a Lavry or Benchmark or Genex (and probably plenty of others, but those are what I use/have used.) The biggest thing I notice with the RME is that things sound smeared and the low-mids get rounded in a false-sounding way compared with the Lavry or Benchmark; there's no comparison -- once I got my DAC-1, I was amazed at the night and day difference in clarity, imaging, and "life" in what I was hearing as opposed to the RME. You should try getting yourself a nice D/A and start listening to CDs through that -- I have a feeling we won't be seeing as many posts about this "high-end time distortion"....
| Quote: |
there is a definite digital sound. People know this and they have been saying it from day one. |
And they were right about it on day one when converters were really weak and had serious (and measureable) problems, but things have changed a lot since then.
| Quote: |
maybe with very high sample rates the time-resolution is good enough at high frequencies that it can be as convincing as analog regarding feel and rock-solid imaging, depth etc.. |
how about with very high quality equipment, instead?
Post by: bobkatz on September 22, 2005, 11:55:02 AM
| andy_simpson wrote on Wed, 21 September 2005 19:44 |
Just out of interest and to help me get a different angled-handle on this question, how would one implement a 1 microsecond advance (or delay) of one channel of a 44/16 recording? In the physical realm, one can simply move the mic closer (a teeny bit). Presumably it can be done in theory? Could it be done without upsampling? Andy |
No.
BK
Post by: Andy Simpson on September 22, 2005, 12:32:45 PM
| bobkatz wrote on Thu, 22 September 2005 16:55 | ||
No. BK |
But we think that if we upsample high enough, shift it a sample or two & down-sample back to 44.1, we can do it?
Andy
Post by: Norwood on September 22, 2005, 12:58:42 PM
| andy_simpson wrote on Thu, 22 September 2005 09:32 |
But we think that if we upsample high enough, shift it a sample or two & down-sample back to 44.1, we can do it? Andy |
We...?
Don't listen to the voices Andy. Just kidding. Using your initial argument wouldn't it just get "quantized" back to the same sample, since 44.1k can't represent differences in time that small?
And if it can represent differences that small wouldn't your argument be disproven? And also if it can accurately portray differences as small as a microsecond by up/downsampling why do you not believe that it can do this with oversampling/reconstruction at the a/d?
Post by: Johnny B on September 22, 2005, 01:04:17 PM
And while we wait for them to improve the chips, we can also wait for them to improve the digital formats as well.
With digital, people are in a constant "Wait State!"
Post by: Andy Simpson on September 22, 2005, 01:04:57 PM
| Norwood wrote on Thu, 22 September 2005 17:58 | ||
We...? Don't listen to the voices Andy. Just kidding. Using your initial argument wouldn't it just get "quantized" back to the same sample, since 44.1k can't represent differences in time that small? And if it can represent differences that small wouldn't your argument be disproven? And also if it can accurately portray differences as small as a microsecond by up/downsampling why do you not believe that it can do this with oversampling/reconstruction at the a/d? |
That's what I'm trying to get at......it's the exact practicalities of how much information is lost on the down-sample....
Andy
Post by: Eric Bridenbaker on September 22, 2005, 02:01:06 PM
Compare the total magnetic surface area of a 15/30 minute two inch reel to that of a typical SCSI of IDE hard drive, which stores many hours of information, even when the audio files are at high sample rates.
Which do you figure would have a higher information density?
Best,
Eric
Post by: Chrunchy on September 22, 2005, 02:10:00 PM
Post by: Malcolm Boyce on September 22, 2005, 03:31:47 PM
| Quote: |
The 96kHz Alligator Ok, so you now are riding on the bleeding edge of technology. You are even transferring your cassettes to 96kHz/24-bit. Life is good… almost. With an upper frequency response limit of 40kHz you can make excellent dog whistle recordings. I’m just finishing up my first DVD entitled “Studies in Parallel Harmonies Above 20kHz”. You could play it to entertain your pets without disturbing any humans. Another project idea is to record everything an octave below where I really want it, then pitch shift every instrument up an octave so I will have plenty of supersonic overtones. Even if you recorded everything at 48kHz on your ADATs, the act of mixing, EQing and adding reverb and effects will generate program content above 20kHz. That is the good news. The bad news is that you can’t check to see what is up there because you can’t hear it. You can look at it with a spectrum analyzer, or zoom in and look at the waveform, but you can’t really tell by looking, what effect it will have on your hearing. Audio equipment is designed for a fairly flat response from 20Hz to 20kHz. We have known for a long time that there are problems with recording information below the 20Hz limit. DC components must be filtered out, capacitor noise and power supply ripple must be eliminated, and does anybody remember “turntable rumble?” With 96kHz recordings I have run into some supersonic problems that you should watch out for. Remember I said that supersonic material is generated during the mixing process. Sometimes these can be pretty healthy transients generated by the music. Other times they can be harmonic impulses that are caused by the console EQ. In the 20Hz to 20kHz world there is no problem because if something causes a click you usually hear it and fix it. Supersonic transients go undetected. I have some mixes that contain some of these supersonic transients. When I play back the mix in the studio, or on my studio quality gear at home, everything is fine. When I play it back through a consumer power amp and speakers there is a giant click. When I listen on headphones powered by very expensive amp, everything is fine, but when I plug the headphones into a $500 receiver, the click made my nose bleed. Low price amplifiers contain circuits that can not change the voltage fast enough for the high frequencies coming from the 96kHz material. It is kind of like the click you hear when you have the bass turned up too loud and your speakers hit the stops. (My daughter Ashlee actually likes the extra click added to the kick drum attack. Oh well, she’s out of my will.) The Answer. I talked to one mastering engineer about the problem and he said he just rolls off the stuff above about 22kHz so he won’t have that problem, and as long as the end product says 96kHz, who will know? Wait a minute. Doesn’t this negate the need for 96kHz? Have I wasted all of my money again? I guess a good comparison would be owning a Dodge Viper. You can’t find many places to drive 180mph, but it sure impresses people who see it in your driveway. I guess the Apogees and Genex and Mytek and Alesis and TC 96k stuff looks good in my rack, so maybe everything will turn out ok. I’ll have to think about this and get back to you. |
That Roger... always good for a laugh!
If we haven't heard of these artifacts before, doesn't it stand to reason that anything before 96K digital wasn't reproducing anything up there, including the best analog?
Just something else... sorry for taking up more space in this thread.
Malcolm
Post by: dcollins on September 22, 2005, 03:47:02 PM
DC
Post by: Johnny B on September 22, 2005, 04:05:28 PM
Man, that 0100110100101100 sure sounds warm...NOT!
Who were the deaf bastard math guru's who came up with that one?
Oh, and don't we all just love the detail provided by MPfreakin3's....
More moronic math guru's with tin ears...
Gimme some good analogue, it blows this digital crap out of the water
Post by: StudioRhythm on September 22, 2005, 05:18:43 PM
| Johnny B wrote on Thu, 22 September 2005 13:05 |
Yeah, but many people simply cannot love the sound of 16 wimpy bits at 44 freakin'k! Man, that 0100110100101100 sure sounds warm...NOT! Who were the deaf bastard math guru's who came up with that one? Oh, and don't we all just love the detail provided by MPfreakin3's.... More moronic math guru's with tin ears... Gimme some good analogue, it blows this digital crap out of the water |
OK, one more time -- what digital equipment have you listened to exactly and what are your specifc problems with it? And no, your consumer CD player doesn't count....
Post by: Duardo on September 22, 2005, 05:24:59 PM
| Quote: |
Well, it's now been established that the "New and Improved Nyquist Figure" is a minimum of 208kHz |
Did I miss something? When was that established?
| Quote: |
OK, one more time -- what digital equipment have you listened to exactly and what are your specifc problems with it? And no, your consumer CD player doesn't count.... |
His problems are all related to his interpretations of things that he's read. He won't say what his actual experience is and he won't leave, even though he says he will.
-Duardo
Post by: StudioRhythm on September 22, 2005, 05:26:17 PM
| Johnny B wrote on Thu, 22 September 2005 10:04 |
Well, it's now been established that the "New and Improved Nyquist Figure" is a minimum of 208kHz, so we will have to wait for those faster chips, wait for the implementation and design improvements to take their course, and wait for them to reach the consumers before these comparisons with King Analogue can be set up properly. And while we wait for them to improve the chips, we can also wait for them to improve the digital formats as well. With digital, people are in a constant "Wait State!" |
You're kidding, right? It's established because you said it once? Yes, I know that instruments produce frequencies up to 104 kHz in some cases, but that doesn't mean we hear them!
Like I said before, my cell phone produces lots of super-high frequencies too, maybe those are an important part of the sound I hear on a call. My microwave produces lots of high frequencies too, maybe if I want to record the true sound of my microwave I'll need a sampling rate in the hundreds of millions of Hz.
Every light source, including the sun, produces a much broader spectrum of light than the eye can see -- nobody argues with this because people are generally more aware of our visual sense than sound. You don't see people suddenly suggesting that digital video sucks because it's not recording infared, do you?
What part of basic logic don't you understand? Or is this all a joke?
Post by: StudioRhythm on September 22, 2005, 05:29:06 PM
| Duardo wrote on Thu, 22 September 2005 14:24 |
His problems are all related to his interpretations of things that he's read. He won't say what his actual experience is and he won't leave, even though he says he will. -Duardo |
As they say, a little knowlege is a dangerous thing. In this case, I'm not sure it's even a little....
Post by: Johnny B on September 22, 2005, 06:11:24 PM
First, I'm not on trial, digital sound quality is on trial.
My damn opinions are only my own. I think digital sounds bad, OK? With digital it comes out weak, thin, and ice cold, dead sounding.
Digital still gets blown away when compared to King Analogue.
Do your own listening tests, if you like the way formats like CDs and MP3's sound, that's fine with me.
Second, it's not ok to attempt to muddy the waters or attempt to use misdirection by bringing up issues with mics and speakers, the issues are:
1. What does all that slicing and dicing do to the sound?
and
2. Why do people keep complaining about digital's sound quality?
I, for one, blame:
A) the ADDA chips and the Codecs;
B) the poor digital formats; and
C) the moron element who came up with this digital crap
Just think about this too, if there were better digital formats it just might be possible to remove many of those big nasty digital anomalies that the chip makers publish long lists about...Why you might even begin to get rid of all those big nasty truncation problems...
Like I said above,
"With digital, people are in a constant "Wait State"
Here's my real take---Junk all of digital and start over.
In the meantime, use analogue until the digital nerds get it right.
Post by: StudioRhythm on September 22, 2005, 06:48:40 PM
| Johnny B wrote on Thu, 22 September 2005 15:11 |
Alright, one last post on this thread. First, I'm not on trial, digital sound quality is on trial. |
says who?
| Quote: |
My damn opinions are only my own. I think digital sounds bad, OK? With digital it comes out weak, thin, and ice cold, dead sounding. |
And when you won't tell us how these opinions are formed, it makes you look like somebody who forms his opinions based on absolutely no knowlege or experience, and therefore, hold opinions that don't make any kind of sense and are worth nothing.
| Quote: |
Digital still gets blown away when compared to King Analogue. |
because analog tape reproduces 104kHz??
| Quote: |
Do your own listening tests, if you like the way formats like CDs and MP3's sound, that's fine with me. |
I think a well-recorded CD, played back through a good quality D/A, sounds good. MP3s can also sound good, although they can also sound really bad depending on how they're encoded and at what bit rates.
| Quote: |
Second, it's not ok to attempt to muddy the waters or attempt to use misdirection by bringing up issues with mics and speakers, the issues are: |
Who's muddying the waters here? Hmm....
| Quote: |
1. What does all that slicing and dicing do to the sound? |
you're right, it doesn't sound good, but it makes a nice salad.
| Quote: |
and 2. Why do people keep complaining about digital's sound quality? |
Oh, as you've proved quite conclusively, people say all kinds of B.S, whether or not they know what they're talking about.
| Quote: |
Just think about this too, if there were better digital formats it just might be possible to remove many of those big nasty digital anomalies that the chip makers publish long lists about...Why you might even begin to get rid of all those big nasty truncation problems... |
Where can I find these lists? Do you even know what truncation means?
| Quote: |
Like I said above, "With digital, people are in a constant "Wait State" |
Hey, that's catchy. You know, maybe you should be in advertising. It doesn't make sense, but it has a nice ring to it.
| Quote: |
Here's my real take---Junk all of digital and start over. In the meantime, use analogue until the digital nerds get it right. |
I'm sure everybody here who has been waiting anxiously to hear your take on it can now breathe a sigh of relief to know that the answers to our imaginary problems are that simple -- just stop everything and start over. Whew -- good thinkin' there, chief!
Post by: bobkatz on September 22, 2005, 06:51:22 PM
| andy_simpson wrote on Thu, 22 September 2005 12:32 |
BK |
But we think that if we upsample high enough, shift it a sample or two & down-sample back to 44.1, we can do it?
Andy
[/quote]
And that's exactly how the Cedar Azimuth fixer works. If you want to do some subsample slipping, just put it through the Cedar.
BK
Post by: maxdimario on September 22, 2005, 07:01:59 PM
| TER wrote on Thu, 22 September 2005 14:41 |
So you've compared analog playback to what digital chain to determine that digital is lacking? It's a simple question which might help us understand if you've heard or used good, properly set up digital gear. From your post it would seem you were listening to cds with a stock a/d converter built into a consumer player. Is that incorrect? And you're saying the best mono imaging you've heard is from a mono recording, right? That seems logical to me. What mono digital recordings have you heard for comparison? You keep adding variables to your argument. It's hard to believe that you're now stating "high end time distortion" as some kind of given in your arguments. At or above what frequency do you hear this distortion? I'm really trying to understand what you're getting at that hasn't been addressed here. -tom |
HI
the problem with some of you guys is that you never question, you never read in-between the lines, you simply seem to elaborate what has been established as a rule without trying to understand the underlying trends or patterns.
it has been my experience that when I talk with people who have had hits, the conversation is simple and to the point, but always a bit on the 'loose' side: one phrase conjures up another which is related but with a deeper meaning. One doesn't feel the need to be diplomatic and orderly, methodical, correct and politically just when those involved really love what they are talking about and would do anything just to improve it that bit further and raise the bar a notch. It all fixes itself.
what does this all mean?
it means that you read the worst into my post.. Why?
isn't it obvious from the beginning that I am trying to improve what I hear?
didn't you read the part where it says I listen to cd's AND stuff coming out of studios?
Apogee and Prism . they're not weiss etc. but they are the MCI or maybe even AMPEX, as far as esteem. i dunno and I don't care. tried them out, took 'em apart, listened to the results.
what does it sound like in THE HOME.
and the people who don't have a clue of why they like music?
they are waiting for the music to entertain them, why pay the money they earned working getting weiss converters for their cd player etc.. when they can hardly tell the difference anyway ...and nobody informs them?.
What about if they are turned off of music because although it sounded perfect! it was BORING to listen to?
as any good performer will show, the difference between very good and being totally convincing is actually very subtle, and difficult to capture 100% with recording equipment.
aesthetic-based music is not so much of a performance-related music. make arrangements with sequenced keyboards and the recording system is greatly alleviated of it's expectations.
when sounds become synthetic, the musical transparency of a sound is not as important. The synthesised sound is created from scratch, filtered, distorted, eq'ed to fit into the space of an arrangement to make a lovely perfect sound, but with no real identity.
this kind of music actually does not need an excellent mixer, that focuses on detail. but of course a nice one helps, sometimes.
live musicians generate a human feel which can be represented both in timing and in quality of sound.
Most people can't differentiate between a live rhythm section and a sequenced one, but they will be drawn to both for different reasons and in different ways.
6 month singles are mostly sequenced or have been cut up ...so to speak.
feel is mostly timing from the musicians to the recorder, sound wise.
so the first thing you want to keep intact in music reproduction is time integrity.
the high-end, hi-fi amp community has already supported this issue thouroughly through simple minimal, low-feedback designs.
as did early recording equipment.
Post by: StudioRhythm on September 22, 2005, 08:06:37 PM
| maxdimario wrote on Thu, 22 September 2005 16:01 |
the problem with some of you guys is that you never question, you never read in-between the lines, you simply seem to elaborate what has been established as a rule without trying to understand the underlying trends or patterns. |
Hey, we're questioning you, aren't we?
| Quote: |
it has been my experience that when I talk with people who have had hits, the conversation is simple and to the point, but always a bit on the 'loose' side: one phrase conjures up another which is related but with a deeper meaning. One doesn't feel the need to be diplomatic and orderly, methodical, correct and politically just when those involved really love what they are talking about and would do anything just to improve it that bit further and raise the bar a notch. It all fixes itself. |
Well, first of all, this is an engineering forum, not the hitmakers forum, that's across the hall. I do know what you're talking about, though -- it's what I think of as the artist personality. And sure, that's a very useful thing to have when you're trying to be creative, I agree 100%. But, we're not talking about the actual creative part here, we're talking about the technology that people can use to be creative, and the technology does need to be designed precisely and carefully.
| Quote: |
what does this all mean? it means that you read the worst into my post.. Why? isn't it obvious from the beginning that I am trying to improve what I hear? |
I'm not doubting that at all; what I'm doubting is whether what you're claiming is correct and makes sense, that's all. No offense intended, and no attempt to question your intentions. I do question whether or not you're misguided, though.
| Quote: |
didn't you read the part where it says I listen to cd's AND stuff coming out of studios? |
Yes, but that's not real specific, you know.
| Quote: |
Apogee and Prism . they're not weiss etc. but they are the MCI or maybe even AMPEX, as far as esteem. i dunno and I don't care. tried them out, took 'em apart, listened to the results. what does it sound like in THE HOME. and the people who don't have a clue of why they like music? they are waiting for the music to entertain them, why pay the money they earned working getting weiss converters for their cd player etc.. when they can hardly tell the difference anyway ...and nobody informs them?. What about if they are turned off of music because although it sounded perfect! it was BORING to listen to? |
I've heard great performances that were recorded terribly, and I wasn't bored in the least. And I've heard "audiophile quality" recordings that bored me to tears. I don't think the excitement comes from the recording technology in most cases.
| Quote: |
as any good performer will show, the difference between very good and being totally convincing is actually very subtle, and difficult to capture 100% with recording equipment. aesthetic-based music is not so much of a performance-related music. make arrangements with sequenced keyboards and the recording system is greatly alleviated of it's expectations. when sounds become synthetic, the musical transparency of a sound is not as important. The synthesised sound is created from scratch, filtered, distorted, eq'ed to fit into the space of an arrangement to make a lovely perfect sound, but with no real identity. this kind of music actually does not need an excellent mixer, that focuses on detail. but of course a nice one helps, sometimes. live musicians generate a human feel which can be represented both in timing and in quality of sound. Most people can't differentiate between a live rhythm section and a sequenced one, but they will be drawn to both for different reasons and in different ways. 6 month singles are mostly sequenced or have been cut up ...so to speak. feel is mostly timing from the musicians to the recorder, sound wise. so the first thing you want to keep intact in music reproduction is time integrity. the high-end, hi-fi amp community has already supported this issue thouroughly through simple minimal, low-feedback designs. as did early recording equipment. |
Max, if you just said "I don't like the way digital sounds, I like analog better. Analog keeps a more natural feel and I like it, and I wish digital could sound as good." I wouldn't have a problem with anything you said. It's when you make up psuedo-scientific terms that don't really mean anything and pretend to be diagnosing "the problem" that I will challenge your statements. This is engineering, not voodoo -- if you don't think digital sounds as good as analog, that's fine, but unless you know what you're talking about technically and can back it up, keep it at that. There's plenty of people here who really know what they're talking about technically, and I've learned a lot from many of them; I respect their posts very much, in part because they don't claim to be able to explain things they can't.
Post by: Duardo on September 22, 2005, 08:40:07 PM
| Quote: |
Alright, one last post on this thread. |
You've said that before.
| Quote: |
First, I'm not on trial, digital sound quality is on trial. |
I don't think either you or digital sound quality is on trial, but at this point I'd say that more people participating in this thread have a problem with your statements than they do with "digital" as an entity.
| Quote: |
Digital still gets blown away when compared to King Analogue. |
You're not British, are you?
| Quote: |
Do your own listening tests, if you like the way formats like CDs and MP3's sound, that's fine with me. |
It doesn't seem to be...
| Quote: |
1. What does all that slicing and dicing do to the sound? |
Nothing. Sound isn't "sliced and diced".
| Quote: |
2. Why do people keep complaining about digital's sound quality? |
For the same reasons people comlain about analog's sound quality. These reasons include, but aren't limited to, the fact that it doesn't sound exactly like being there. Becuase there are some poor implementations of the technology out there. Because they've read things that get them all excited about something they really don't understand.
And some people complain about "digital" because it doesn't sound the same as "analog". Some people complain about "analog" because it doesn't sound the same as "digital". People have preferences.
-Duardo
Post by: dcollins on September 23, 2005, 02:39:11 AM
| Johnny B wrote on Thu, 22 September 2005 15:11 |
I, for one, blame: A) the ADDA chips and the Codecs; B) the poor digital formats; and C) the moron element who came up with this digital crap |
"When you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot measure it, when you cannot express it in numbers, your knowledge is of a meager and unsatisfactory kind: it may be the beginning of knowledge, but you have scarcely, in your thoughts, advanced to the stage of science."—William Thomson, Lord Kelvin, 1824-1907
Post by: Johnny B on September 23, 2005, 04:33:24 AM
http://www.analog.com/UploadedFiles/REDESIGN_IC_Anomalies/19 5504460ts101_anomaly52605.pdf
http://search.analog.com/search/default.aspx?query=anomalies &local=en
BTW it's not my intent to single out AD...all the chip makers have them, the above are but examples.
Alright, I'm really out this time. I think I may be more on GM's page in regard to SRC's and sample rates, so I will be in a "wait state" until he gets back.
Best wishes to all of you, I hope you all do well when the new formats hit.
Really, I do wish you all the best.
Post by: maxdimario on September 23, 2005, 04:50:27 AM
| Quote: |
"When you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot measure it, when you cannot express it in numbers, your knowledge is of a meager and unsatisfactory kind: it may be the beginning of knowledge, but you have scarcely, in your thoughts, advanced to the stage of science."—William Thomson, Lord Kelvin, 1824-1907 |
first you have to know what you are expressing in numbers.
here is a re-quote from the Paul Wolff article:
| Quote: |
The Nyquist theory is correct by stating that any sine wave sampled at twice its frequency can be reproduced accurately. The problem is it does not guarantee that it will be reproduced in the correct location in time as it was originally — with relation to other present waveforms. Since the sample rate is constant, as you slowly increase the frequency that is being sampled, the lower frequencies get more samples than the higher ones — and it slowly ends up at 22.05KHz, which has only two samples. As the frequencies slide up, like the subtle harmonics of just about anything, they are not all getting the same treatment as far as sampling. |
the problem with science and math is that is far too simple to make models which approximate reality but do not represent it fully.
I have a friend who works with econometrics in a reasearch institute, and they are always trying to figure out what mathematical model to use on what.
Post by: Duardo on September 23, 2005, 10:19:00 AM
| Quote: |
here is a re-quote from the Paul Wolff article: |
Just because Paul Wolff wrote it doesn't mean that it's correct. It is not.
-Duardo
Post by: timrob on September 23, 2005, 10:24:45 AM
| maxdimario wrote on Thu, 22 September 2005 18:01 |
didn't you read the part where it says I listen to cd's AND stuff coming out of studios? Apogee and Prism . they're not weiss etc. but they are the MCI or maybe even AMPEX, as far as esteem. i dunno and I don't care. tried them out, took 'em apart, listened to the results. what does it sound like in THE HOME. and the people who don't have a clue of why they like music? they are waiting for the music to entertain them, why pay the money they earned working getting weiss converters for their cd player etc.. when they can hardly tell the difference anyway ...and nobody informs them?. What about if they are turned off of music because although it sounded perfect! it was BORING to listen to? as any good performer will show, the difference between very good and being totally convincing is actually very subtle, and difficult to capture 100% with recording equipment. aesthetic-based music is not so much of a performance-related music. make arrangements with sequenced keyboards and the recording system is greatly alleviated of it's expectations. when sounds become synthetic, the musical transparency of a sound is not as important. The synthesised sound is created from scratch, filtered, distorted, eq'ed to fit into the space of an arrangement to make a lovely perfect sound, but with no real identity. this kind of music actually does not need an excellent mixer, that focuses on detail. but of course a nice one helps, sometimes. live musicians generate a human feel which can be represented both in timing and in quality of sound. Most people can't differentiate between a live rhythm section and a sequenced one, but they will be drawn to both for different reasons and in different ways. 6 month singles are mostly sequenced or have been cut up ...so to speak. feel is mostly timing from the musicians to the recorder, sound wise. so the first thing you want to keep intact in music reproduction is time integrity. the high-end, hi-fi amp community has already supported this issue thouroughly through simple minimal, low-feedback designs. as did early recording equipment. |
Bad music is bad music. Whether its recorded digitally or analog.
I have no doubt that you have trouble enjoying listening to a lot of music these days. I do, too. BUT, it has NOTHING to do with analog vs. digital or high sample rates vs. low sample rates.
It has to do with Production Values. Values that say everything must be quantized to a grid. There's the timing weirdness you are hearing, especially in popular music. Hyperlimiting and clipping to achieve loudness. A phenomenon that was not a huge problem when the final product was LP. The medium couldn't physically handle it. You are trying to kill the messenger because you don't like the message.
Unfortunately, with Digital we do have the power to screw things up in rather ridiculous ways. That has nothing to do with capturing and playing back music.
These kinds of discussions always seem to descend into the Analog VS. Digital debate. You might as well be debating Creation vs. Evolution. In the end it doesn't matter cuz we're all here and we have to deal and work with what we have at hand.
Post by: Andy Simpson on September 23, 2005, 10:34:55 AM
| maxdimario wrote on Fri, 23 September 2005 09:50 | ||||
first you have to know what you are expressing in numbers. here is a re-quote from the Paul Wolff article:
|
This is interesting, as it relates the increasing error of the harmonics of a sound relative to its fundamental.
Do we agree that error increases with frequency?
Andy
Post by: maarvold on September 23, 2005, 10:54:22 AM
BTW, Andy: I think your original premise represents GREAT 'outside-the-box' thinking.
Post by: Jon Hodgson on September 23, 2005, 11:13:33 AM
| Johnny B wrote on Fri, 23 September 2005 09:33 |
By popular request, examples of digital anomaly lists: http://www.analog.com/UploadedFiles/REDESIGN_IC_Anomalies/19 5504460ts101_anomaly52605.pdf http://search.analog.com/search/default.aspx?query=anomalies &local=en BTW it's not my intent to single out AD...all the chip makers have them, the above are but examples. Alright, I'm really out this time. I think I may be more on GM's page in regard to SRC's and sample rates, so I will be in a "wait state" until he gets back. Best wishes to all of you, I hope you all do well when the new formats hit. Really, I do wish you all the best. |
You really should understand what you are looking at before you post it as support for your theories.
The ICs you've linked to aren't converters, they're DSPs. As for the "anomalies" they are what most manufacturers refer to as "errata", they are hardware equivalent of software bugs, and indicate circumstances in which the results might not be what you expect, you'll also notice that they all have workarounds, it means that so long as the programmer is aware of these anomolies, and implements the workarounds when needed, he will get the expected results.
It's no different to having a label stuck on your analogue 24 track that says "track 22 distorts really badly, so don't use it", so long as the engineer reads the label and doesn't use that track his recordings will sound just as good.
Post by: maxdimario on September 23, 2005, 11:17:47 AM
| Quote: |
Just because Paul Wolff wrote it doesn't mean that it's correct. It is not. -Duardo |
Why not?
please don't post affirmations like that without giving an explanation of why, as it does not add to the discussion.
otherwise your comments appear to be self-serving, period.
Post by: Norwood on September 23, 2005, 11:26:01 AM
| timrob wrote on Fri, 23 September 2005 07:24 |
Bad music is bad music. Whether its recorded digitally or analog. I have no doubt that you have trouble enjoying listening to a lot of music these days. I do, too. BUT, it has NOTHING to do with analog vs. digital or high sample rates vs. low sample rates. It has to do with Production Values. Values that say everything must be quantized to a grid. There's the timing weirdness you are hearing, especially in popular music. Hyperlimiting and clipping to achieve loudness. A phenomenon that was not a huge problem when the final product was LP. The medium couldn't physically handle it. You are trying to kill the messenger because you don't like the message. Unfortunately, with Digital we do have the power to screw things up in rather ridiculous ways. That has nothing to do with capturing and playing back music. These kinds of discussions always seem to descend into the Analog VS. Digital debate. You might as well be debating Creation vs. Evolution. In the end it doesn't matter cuz we're all here and we have to deal and work with what we have at hand. |
In my best black preacher voice... "CAN I GET A WITNESS!"
Post by: jimmyjazz on September 23, 2005, 11:42:28 AM
| Johnny B wrote on Wed, 21 September 2005 00:55 |
If you really want to capture that 104-plus kHz that Boyk measured, that means a minimum of 208kHz, does it not? Is that the new Nyquist figure? 208kHz? |
First of all, I would take anything James Boyk says with an enormous lump of salt. I have had numerous exchanges with him in the past, and he is dead wrong as often as he is right.
Regardless, it seems you are accepting the fact that Nyquist was right when you yourself double this magic "104 kHz" number to come up with what you perceive to be the "correct" sampling rate. Interesting.
Post by: maxdimario on September 23, 2005, 11:44:29 AM
| Quote: |
I have no doubt that you have trouble enjoying listening to a lot of music these days. I do, too. BUT, it has NOTHING to do with analog vs. digital or high sample rates vs. low sample rates. It has to do with Production Values. |
Let's not get into that! that is a long discussion as well.
I am focusing in strictly on the recording technology, and am aware of the fact that 90% of a good record is what's going on in the room, musician, arranging, instruments, the feel, groove, playing together, recording working live bands who know how to play effectively etc.
The feel of the music comes through worse with digital.
the feel of the music also comes through worse with a lot of poorly designed analog gear.
some analog gear can have a frequency response of 100 KHz but still screw up the time-alignment of the various components in the sound in a way that it sounds flat and lifeless.
the more lively the sound is to begin with, the more of a possibility that as a result of repeated listenings the listener will 'get it', assuming there is something to be gotten.
the whole point of these discussions is to raise the bar, not lower it.
If digital as it is now is good enough, why does 60 year old technology beat it at imaging, depth and feel?
Making a choice in what recording medium to use should be based on (predominantly)artistic considerations as much as technical ones.
you need to get the artistic point across, and to do that with performances, you need to keep the subtleties of time intact.
Post by: Jon Hodgson on September 23, 2005, 11:48:35 AM
| andy_simpson wrote on Fri, 23 September 2005 15:34 | ||||||
This is interesting, as it relates the increasing error of the harmonics of a sound relative to its fundamental. Do we agree that error increases with frequency? Andy |
No we don't.. because it doesn't.
I'll try to explain this without diagrams.
Imagine you have a band limited signal, which contains no frequency above y Hz, you draw this as a line on your paper.
Now you sample it at regular intervals, at ANY frequency which is GREATER than 2y Hz, for now you do it with infinite level resolution (so you have no level quantisation, only time quantization).
You now have a series of dots, which you need to join up with a smooth line. You might think there are an infinite number of ways to do this, but in fact that's not true if you apply a restriction to the output using a reconstruction filter.
If that line is restricted to containing only frequencies which are BELOW half the sample rate (and the can mean a fraction of a Hz below) there is ONLY ONE LINE that can be drawn, and that is THE SAME SHAPE AS YOUR ORIGINAL LINE.
That is the purpose of the reconstruction filter.
Not only are frequencies maintained, but so is phase.
Now it's true that you don't have infinite level resolution when sampling, but even in a perfectly clean system the error would sometimes round up, sometimes down, so you wouldn't get a shift in phase, you would get the original signal plus some additional low level artifacts. In a real system however the step size is almost certainly swamped by the noise floor (either through intentional dithering or with 20 or 24 bit converters simply because the step size is less than the circuit noise). The net result is that you get your original signal (frequency and phase content intact) plus some white noise, frequency elements are not shifted relative to each other.
Of course in the real world the filter is less than perfect, so level and phase of the frequency components will be altered to some degree (as it is when the signal passes through any circuit more complex than a simple wire), but that is an implementation issue, not a problem with sampling theory.
(edited for typo)
Post by: maxdimario on September 23, 2005, 11:59:45 AM
| Quote: |
If that line is restricted to containing only frequencies which are BELOW half the sample rate (and the can mean a fraction of a Hz below) there is ONLY ONE LINE that can be drawn, and that is THE SAME SHAPE AS YOUR ORIGINAL LINE. That is the purpose of the reconstruction filter. |
there is no sense in quoting and re-quoting the nyquist theory and the purpose of reconstruction filters. It has been quoted identically above many times.
as you go higher near the 1/2 of sampling rate, the sample rate is disproportional in resolution to frequency.
22.5 KHz has two samples, 11.25 has 4, and so on.
can you disprove specifically that a lower amount of samples in the high frequency range does not create distortion in complex signals, without re-quoting nyquist?
Post by: Jon Hodgson on September 23, 2005, 12:06:40 PM
| maxdimario wrote on Fri, 23 September 2005 16:59 | ||
there is no sense in quoting and re-quoting the nyquist theory and the purpose of reconstruction filters. It has been quoted identically above many times. as you go higher near the 1/2 of sampling rate, the sample rate is disproportional in resolution to frequency. 22.5 KHz has two samples, 11.25 has 4, and so on. can you disprove specifically that a lower amount of samples in the high frequency range does not create distortion in complex signals, without re-quoting nyquist? |
You can't do it mathematically without "requoting" nyquist, because if you show the mathematics of it, you are simply showing what nyquist has already shown.
When I have more time I can try to show things diagrammatically, wherein they may seem more intuitive.
But note one thing, Nyquist states that if your sample frequency is 44.1, then you cannot sample 22.05, you can only sample below that, this makes a big difference.
Post by: maxdimario on September 23, 2005, 12:12:08 PM
how about 22.04?
O.K. I'd be interested in seeing something like that.
no sine waves though please.
Post by: Jon Hodgson on September 23, 2005, 12:27:13 PM
| maxdimario wrote on Fri, 23 September 2005 17:12 |
eek. no 22.05? how about 22.04? O.K. I'd be interested in seeing something like that. no sine waves though please. |
I thought that the orginal subject of this thread was whether signal components near the nyquist frequency where phase shifted by sampling?
I can't think of any clearer way of showing this than with a sine wave.
What signal would you consider acceptable?
Post by: J.J. Blair on September 23, 2005, 12:29:43 PM
| maarvold wrote on Fri, 23 September 2005 07:54 |
BTW, Andy: I think your original premise represents GREAT 'outside-the-box' thinking. |
Michael, what's so great about thinking outside the box if it's wrong? The whole premise of digital quantizing the soundwave is absurd. I don't mind somebody throwing out an idea, but I for one don't think we should encourage expressing a thought to only later ignore the experience of people who know what they are talking about.
I had a very interesting discussion with Hutch over at Manley Labs yesterday. I ran Andy's quantization premise by him to make sure that DC and I are not the only ones who think it's ridiculous. (We aren't.) It was interesting, because he mentioned that he 'knows', not 'thinks he knows', why digital sounds different from analog, and not a single person in here who has tried to explain why they sound different has even touched on his explanation. I've used the gear that Hutch builds and designs and I'm going to trust him. Manley builds George's boxes, so I assume George thinks trusts him, too.
That not withstanding, to say that "a recording made at 44.1 will 'quantize' the spatial timing aspects of a recording into chunks of 1.7cm" is patently absurd. I think Andy is a really talented engineer from the stuff I've heard him post, but I honestly don't think he knows what the hell he's talking about theoretically, from every single theoretical discussion I've seen him start. I'd like to see people encouraging him to listen to other people's experience to learn something, rather than encouraging his iconoclastic, rebellious, 'ignore the experts' posting style. That people nod their heads in agreement with his farcical conclusions boggles my mind.
I'm sorry, but somebody has to say it and call a spade a spade. And this is nothing against Andy personally. Hey, I even compliment his recording. This behavior has to be attenuated though, IMO.
And for some of you other guys suggesting that digital sounds like shit: I'm an analog head. My preferred sampling rate is 499 @ 15ips +6. But to suggest that you can't make a good recording on digital is laughable. I guess I'll have to throw out a bunch of my record collection, then. Whoever is saying this has very limited experience with digital, and should save their biases for when they've worked on a few different formats with some high end gear.
As my friend David Palmer says, "Eventually, your experience catches up with your opinions."
Post by: jimmyjazz on September 23, 2005, 12:30:15 PM
| maxdimario wrote on Fri, 23 September 2005 04:50 | ||||
first you have to know what you are expressing in numbers. here is a re-quote from the Paul Wolff article:
the problem with science and math is that is far too simple to make models which approximate reality but do not represent it fully. I have a friend who works with econometrics in a reasearch institute, and they are always trying to figure out what mathematical model to use on what. |
The problem with Wolff's quote is that it neglects the fact that a sampled signal is expressed as two spectra, not one: the amplitude spectra (which is what most people think of as the signal's "frequency response") and the phase spectra (which provides the "timing cues", if you will). The timing information is all there when a signal is (correctly) sampled.
The fact that lower frequencies are sampled more than twice per period but the highest frequencies aren't is a non-issue. Nyquist was correct when he postulated that one must sample at a frequency higher than twice that of the signal of interest . . . sampling at, say, three times the signal frequency doesn't make the result any more or less "perfect". So, for a signal that is band-limited to below 22.05 kHz, we are safe in sampling at 44.1 kHz, at least as far as sampling theory goes. The high frequencies will be sampled perfectly. The low frequencies will not be sampled any MORE perfectly. Perfect is perfect.
Post by: maxdimario on September 23, 2005, 12:31:16 PM
or a violin section.
pretty hard to draw..
Post by: Jon Hodgson on September 23, 2005, 12:34:56 PM
| maxdimario wrote on Fri, 23 September 2005 17:31 |
how about white noise at -3 db? or a violin section. pretty hard to draw.. |
Exactly...
I could draw it, or rather I could write a program to draw it.
But you'd have no way of knowing if I was making it up, hence a simple signal is best.
What is your problem with sine waves, do you doubt Fourier?
Post by: Andy Simpson on September 23, 2005, 12:35:39 PM
| Jon Hodgson wrote on Fri, 23 September 2005 16:48 | ||||||||
No we don't.. because it doesn't. I'll try to explain this without diagrams. Imagine you have a band limited signal, which contains no frequency above y Hz, you draw this as a line on your paper. Now you sample it at regular intervals, at ANY frequency which is GREATER than 2y Hz, for now you do it with infinite level resolution (so you have no level quantisation, only time quantization). You now have a series of dots, which you need to join up with a smooth line. You might think there are an infinite number of ways to do this, but in fact that's not true if you apply a restriction to the output using a reconstruction filter. If that line is restricted to containing only frequencies which are BELOW half the sample rate (and the can mean a fraction of a Hz below) there is ONLY ONE LINE that can be drawn, and that is THE SAME SHAPE AS YOUR ORIGINAL LINE. (edited for typo) |
Thankyou kindly for your (polite) explanation - maybe it's the one that finally lights a bulb.....
So, level quantization error = spatial timing error?
Does this error become more significant with frequency?
Andy
Post by: Jon Hodgson on September 23, 2005, 12:43:52 PM
| andy_simpson wrote on Fri, 23 September 2005 17:35 |
Thankyou kindly for your (polite) explanation - maybe it's the one that finally lights a bulb..... So, level quantization error = spatial timing error? Does this error become more significant with frequency? Andy |
You're welcome
level quantisation error has the same effect as a spacial timing error OF THAT POINT.
However the signal consists of a series of points, joined together by a band limited line, and the error of those points is not contant, so you don't get a phase shift of that frequency forward or backwards in time, what you end up with is the correct signal (because ON AVERAGE the points are correctly placed) plus a random error, a random error is white noise.
Post by: maxdimario on September 23, 2005, 12:45:18 PM
| Quote: |
That not withstanding, to say that "a recording made at 44.1 will 'quantize' the spatial timing aspects of a recording into chunks of 1.7cm" |
yes, if you take it literally it is absurd. Andy's post was interesting because he made an association between speed of sound and sampling frequency.
in itself it means nothing, but it gives an indication of the relationship between physical distance and sampling frequency.
Post by: Andy Simpson on September 23, 2005, 12:46:12 PM
| J.J. Blair wrote on Fri, 23 September 2005 17:29 | ||
Michael, what's so great about thinking outside the box if it's wrong? The whole premise of digital quantizing the soundwave is absurd. I don't mind somebody throwing out an idea, but I for one don't think we should encourage expressing a thought to only later ignore the experience of people who know what they are talking about. I had a very interesting discussion with Hutch over at Manley Labs yesterday. I ran Andy's quantization premise by him to make sure that DC and I are not the only ones who think it's ridiculous. (We aren't.) It was interesting, because he mentioned that he 'knows', not 'thinks he knows', why digital sounds different from analog, and not a single person in here who has tried to explain why they sound different has even touched on his explanation. I've used the gear that Hutch builds and designs and I'm going to trust him. Manley builds George's boxes, so I assume George thinks trusts him, too. That not withstanding, to say that "a recording made at 44.1 will 'quantize' the spatial timing aspects of a recording into chunks of 1.7cm" is patently absurd. I think Andy is a really talented engineer from the stuff I've heard him post, but I honestly don't think he knows what the hell he's talking about theoretically, from every single theoretical discussion I've seen him start. I'd like to see people encouraging him to listen to other people's experience to learn something, rather than encouraging his iconoclastic, rebellious, 'ignore the experts' posting style. That people nod their heads in agreement with his farcical conclusions boggles my mind. I'm sorry, but somebody has to say it and call a spade a spade. And this is nothing against Andy personally. Hey, I even compliment his recording. This behavior has to be attenuated though, IMO. And for some of you other guys suggesting that digital sounds like shit: I'm an analog head. My preferred sampling rate is 499 @ 15ips +6. But to suggest that you can't make a good recording on digital is laughable. I guess I'll have to throw out a bunch of my record collection, then. Whoever is saying this has very limited experience with digital, and should save their biases for when they've worked on a few different formats with some high end gear. As my friend David Palmer says, "Eventually, your experience catches up with your opinions." |
Maybe the only good thing to come out of this thread will be that some of us are thinking about digital error in terms of spatial timing error, rather than amplitude error.....and that is 'outside the box'.
Heck, maybe some of us have never thought about sound in terms of spatial timing at all.....
In any case, I appreciate the restraint shown in your post.
Thanks.
Andy
Post by: jimmyjazz on September 23, 2005, 01:04:47 PM
I mean, I don't really know all that much about chemistry and thermodynamics, but I'm also not coming up with alternate theories on how my car works.
Post by: Johnny B on September 23, 2005, 02:17:00 PM
| Jon Hodgson wrote on Fri, 23 September 2005 16:13 | ||
You really should understand what you are looking at before you post it as support for your theories. The ICs you've linked to aren't converters, they're DSPs. As for the "anomalies" they are what most manufacturers refer to as "errata", they are hardware equivalent of software bugs, and indicate circumstances in which the results might not be what you expect, you'll also notice that they all have workarounds, it means that so long as the programmer is aware of these anomolies, and implements the workarounds when needed, he will get the expected results. |
First, I guess no one doing digital uses DSP, aye?
Second, I said they are "but examples," guess you may have missed that part.
Third, I do agree with your point of calling the many digital anomalies "hardware bugs."
What a pity that digital has so many hardware *and* software "bugs."
What a pity there were so many initial errors, esp. in the format selections.
Hindsight? Yes, but some dumb mistakes were made. Things like emasculated little shit chips are far too often the result of bean counters sticking their cheap ass noses into the process.
Here's always been my underlying theme: Digital is in need of more R&D and applied science that uses a multi-disciplinary approach which will look much more closely at the "ear/brain/body interactions" in the continuing effort to move the technology forward.
If it turns out we need much more speed and greater bit-depth or entirely new methods and new technology, I'm cool with that so long as it approaches the best in analogue.
Actually, I'm very excited about what the future may bring and the technological innovations we will see.
'Nuf said. This time I'm out for sure.
Best wishes to everyone.
Post by: dcollins on September 23, 2005, 04:12:15 PM
| maxdimario wrote on Fri, 23 September 2005 01:50 |
here is a re-quote from the Paul Wolff article: The Nyquist theory is correct by stating that any sine wave sampled at twice its frequency can be reproduced accurately. The problem is it does not guarantee that it will be reproduced in the correct location in time as it was originally — with relation to other present waveforms. Since the sample rate is constant, as you slowly increase the frequency that is being sampled, the lower frequencies get more samples than the higher ones — and it slowly ends up at 22.05KHz, which has only two samples. As the frequencies slide up, like the subtle harmonics of just about anything, they are not all getting the same treatment as far as sampling. |
Paul and Rupert have made some great analog gear, but they seem to have gone to the same school regarding digital.....
It's not like this was fully understood by everyone else for, oh, seventy years!
Max; Johnny B. Take a week off posting nonsense, get a couple books, and read them.
Then you can come back and make some "reasonable" criticisms.
Or not.
DC
Post by: maxdimario on September 23, 2005, 05:22:24 PM
all analog engineers that deal with audio sooner or later understand that the best circuit is the simplest one with the largest bandwidth within reason(which can be limited in a controlled manner afterwards), because nobody worth their salt dares to mess with all the information that is available in an audio signal.
..granted doing this with digital becomes quite the ordeal. and so the theories abound.
Dc,
From what I understand, and has been quoted repeatedly, the filters will not let any signal above the cutoff frequency through, so anything between samples will not be reproduced if it is above cutoff.. like little spurious HF glitches etc between samples.. in reconstructing the waveform, since the original signal is band limited, by limiting the band of the DA converter the waveform is accurately reproduced, since any information 'between the dots' is above the permissible slew rate, and since it all relates mathematically to the sampling frequency/filter and the slope from one 'dot' to the next.
Seemingly Nyquist himself did not promise that all of the waveform's information would be captured by the digitization process, furthermore, it seems he stated that his theory could only work if the filters were theoretically perfect (perfect impulse response, distortion, phase issues).
You tell me if he claimed otherwise.
And since I am admittedly more of an analog guy, I would like to ask you a question dc, given your expertise in the matter: Can you tell me what types of signal are the most difficult to recreate perfectly on a digital system?
this might give some interesting insights.
Post by: Jon Hodgson on September 23, 2005, 05:32:21 PM
| Johnny B wrote on Fri, 23 September 2005 19:17 |
First, I guess no one doing digital uses DSP, aye? |
Yes, I do, for a living, all day every day.
You've missed the point, these errata HAVE NO EFFECT ON THE FINAL RESULT THE PROGRAMMER KNOWS ABOUT THEM.
| Johnny B wrote on Fri, 23 September 2005 19:17 |
Second, I said they are "but examples," guess you may have missed that part. |
They are pointless examples, since they don't affect the sound in any way whatsoever so long as the programmer knows about them and implements the workarounds in his code.
Johnny, you may be able to hear problems with the current implementations of digital audio... but you are looking in the wrong place for the causes of what you hear and for the solutions. You don't actually know enough about sampling and signal processing to start making claims about what the problems are with it.
Post by: maxdimario on September 23, 2005, 07:17:17 PM
| Quote: |
What is your problem with sine waves, do you doubt Fourier? |
Hi Jon,
I doubt the efficiency of a sine wave to represent the worst-case scenario for any audio system.
fourier is ok to explain waveforms mathematically, but as I mentioned before, for a mathematical model to be correct it not only has to work, but it has to produce the desired result, which usually means that it has to encompass every possible variable.
as someone who works with dsp's you are probably aware of how hard it is to model the behaviour of an analog circuit perfectly through a mathematical model. I use this as an example, I know that recorded audio and synthesised audio are not the same thing, I am talking about the relationship between a math model and the reality it needs to represent.
to me Nyquist and Fourier do not cover all the bases.
resolution in digital is not simply an issue of the audio frequency/sampling frequency ratio, it is a total issue of waveform resolution.
the higher the sampling frequency, the more samples OVERALL the more accurate the reproduction of the signal even below nyquist.
and as an analog guy, let me say that no self-respecting analog system functions on the border of it's limitations, not because theory says so but because experience dictates.
44.1 digital does.
Post by: Malcolm Boyce on September 23, 2005, 07:56:24 PM
| Quote: |
the higher the sampling frequency, the more samples OVERALL the more accurate the reproduction of the signal even below nyquist. |
I think that's what's been proven to be incorrect
| Quote: |
and as an analog guy, let me say that no self-respecting analog system functions on the border of it's limitations, not because theory says so but because experience dictates. 44.1 digital does. |
Don't the microphones and the loudspeakers we use, work pretty much to the extremes of their usable bandwidth? That's pretty much both ends of the signal path, analog, OR digital. I'd say that's bordering on their limitations.
Post by: compasspnt on September 23, 2005, 08:14:49 PM
| Johnny B wrote on Fri, 23 September 2005 14:17 |
What a pity that digital has so many hardware *and* software "bugs." What a pity there were so many initial errors, esp. in the format selections. |
Neve V series consoles are famous for a poorly designed heat tolerance/capacitor issue. Caps had to be replaced in many, if not all, desks. I know we spent an entire summer pulling modules one or two at a time, and in total made 96,000 desolder-resolder connections that year. Then the Neve supplied replacement caps WERE ALSO BAD for a different reason, and the next summer, we had to do it all over again. Hardware bug? Analogue? A big, well known company like Neve? Pretty expensive piece of gear? Analogue? It happens.
| Quote: |
Hindsight? Yes, but some dumb mistakes were made. Things like emasculated little shit chips are far too often the result of bean counters sticking their cheap ass noses into the process. |
In the analogue world, this also happens. Where do you think we get Behringer from? Focusrite (new stuff)? There's a long list of terrible analogue gear, reduced in quality by the "bean counters" in order to make a profit while ignoring quality. It's all in how good a company they want to be, not whether they make digital v. analogue equipment.
| Quote: |
'Nuf said. This time I'm out for sure. |
Then I wasted this post, because you're not around to read it.
Post by: dcollins on September 23, 2005, 08:34:23 PM
| maxdimario wrote on Fri, 23 September 2005 16:17 |
to me Nyquist and Fourier do not cover all the bases. . |
Max, get ready for an all-expense paid trip to Sweden!
Dinner is included.
DC
Post by: Johnny B on September 23, 2005, 09:38:47 PM
| Jon Hodgson wrote on Fri, 23 September 2005 22:32 |
You've missed the point, these errata HAVE NO EFFECT ON THE FINAL RESULT THE PROGRAMMER KNOWS ABOUT THEM...They are pointless examples, since they don't affect the sound in any way whatsoever so long as the programmer knows about them and implements the workarounds in his code. Johnny, you may be able to hear problems with the current implementations of digital audio... but you are looking in the wrong place for the causes of what you hear and for the solutions. |
Workarounds? Trust the coders to get it right? Emmm, I think you may be barking up the wrong tree with that one. C'mon man, just face the fact that digital is buggy. You have buggy digital iron and you have buggy digital code. That's no secret.
As for me looking in the right or wrong places for solutions, at least my mind is open enough to still be looking, and I will not discount out of hand even something that might seem bizarre, out of the mainstream, or completely revolutionary.
Frankly, I half expect a couple of bright young minds working out of some garage to crack the digital nut wide open and blow all the existing technology completely out of the water. Maybe I'm dreaming, maybe not. But if such young bright minds were able to come up with some revolutionary and patentable technology, they could rapidly become "Billionaires." (As a side note: When Texas Instruments bought out Mr. Burr's little company, they paid in excess of 4 Billion Dollars for it.) The possibility to make such vast sums of money may provide an incentive for some of these youngsters, but I also think some are more motivated by assuming the role of "innovative gunslinger" and all the money that rolls in is a secondary issue.
This time I really, really, want to be out.
Please don't quote me or take unnecessary pot shots at me and I will not feel compelled to respond.
The future will be interesting, I hope you all enjoy the ride.
Once again, I wish the best to everyone.
Cheers.
Post by: dcollins on September 23, 2005, 09:57:46 PM
| Johnny B wrote on Fri, 23 September 2005 18:38 |
This time I really, really, want to be out. |
Buh-bye!
DC
Post by: bobkatz on September 23, 2005, 11:42:05 PM
| Quote: |
The Nyquist theory is correct by stating that any sine wave sampled at twice its frequency can be reproduced accurately. The problem is it does not |
snip from Paul Wolf.
Was that supposed to be in support of or not of Digital Audio and sampling? If it's "against", then please realize that Wolf was speaking from complete ignorance. The entire thrust of the sampling frequency theorem is that ALL the information in the source up to the Nyquist frequency can be completely and accurately captured. There is NOTHING in the samplng theorem that says that "there will be less accuracy as the frequency increases". In fact, the theorem makes it clear in very simple terms that even though only two points may (in extreme cases) define an extremely high frequency, all the information can be restored.
THIS IS THE SCIENCE. ACCEPT IT. UNFORTUNATELY, FOR THOSE WHO LIKE TO PICK LOOPHOLES, IT'S CALLED A "THEORY", BUT IT IS A FACT.
What YOU HAVED TO REALIZE IS THAT THE SONIC ARTIFACTS YOU HEAR DO NOT COME FROM THE SAMPLING! Many of the artifacts that we complain about come from the filters that are necessary to prevent aliasing. A little phase shift, a little preecho, a little ripple in the passband, a little distortion... All that we can hear, and it can sound a lot like something you don't like. You're not hearing any steps or glitches or breaks in the sound, trust me, that's not the problem. There are other anomalies to converters such as quantization distortion, non-linear amplitude conversion, etc.; but most of these have been reduced considerably with current-day converters. The last frontier is the filtering, and it's one of the reasons we had to raise the sample rate. NOT because "there are only two samples to describe the 20 kHz wave."
You don't have to sample to learn to find that filters (can) change the sound. You can hear problems if you try to construct a sharp filter, even in the analog domain.
A smart man (Bob Olhsson) once said, "The issues of the audibility of bandwidth and the audibility of artifacts caused by limiting bandwidth must be treated separately. Blurring these issues can only lead to endless arguments."
BK
Post by: dcollins on September 23, 2005, 11:48:21 PM
| bobkatz wrote on Fri, 23 September 2005 20:42 |
You can hear problems if you try to construct a sharp filter, even in the analog domain. |
Make that especially in the analog domain!
DC
Post by: Duardo on September 24, 2005, 12:06:24 AM
| Quote: |
This is interesting, as it relates the increasing error of the harmonics of a sound relative to its fundamental. Do we agree that error increases with frequency? |
No, we don't agree on that. Sure, harmonics are missing as frequency increases,
but they're harmonics we don't hear. As far as the information that's actually
captured they don't.
| Quote: |
How can one explain the fact that film and its 24 Hz/sec 'sampling rate'--more than 3 orders of magnitude less than cd quality digital audio--has any apparent depth at all? |
Even a picture has apparent depth, doesn't it? 24 Hz is fine for film as that's
fast enough for our eyes to register what happens onscreen as motion.
| Quote: |
please don't post affirmations like that without giving an explanation of why, as it does not add to the discussion. |
I didn't say why as it's already been said many many times in this discussion.
| Quote: |
there is no sense in quoting and re-quoting the nyquist theory and the purpose of reconstruction filters. It has been quoted identically above many times. as you go higher near the 1/2 of sampling rate, the sample rate is disproportional in resolution to frequency. 22.5 KHz has two samples, 11.25 has 4, and so on. |
As has already been mentioned, you have to go back to the Nyquist theorem because that's what digital audio is based on. There is no decrease in resolution as you approach the Nyqusist frequency. Just because 11 kHz is represented by twice as many samples as 22 kHz (as mentioned earlier, only frequencies below...not at or below, just below...the Nyquist frequency will be represented) does not mean that it will be any less accurately reproduced than 22 kHz. In fact, it won't.
| Quote: |
BTW, Andy: I think your original premise represents GREAT 'outside-the-box' thinking. Michael, what's so great about thinking outside the box if it's wrong ? The whole premise of digital quantizing the soundwave is absurd. I don't mind somebody throwing out an idea, but I for one don't think we should encourage expressing a thought to only later ignore the experience of people who know what they are talking about. |
I think he was referring to the original premise. There's really nothing intuitive about the way digital audio works.
| Quote: |
how about white noise at -3 db? or a violin section. pretty hard to draw.. |
That's the thinking that you have to get away from. Sure, it's hard for us to "draw", but for a digital sampling system to convert and reconstruct it's no more or less difficult to "draw" than a pure sine wave since we're operating within the limits of the Nyquist frequency.
| Quote: |
can you disprove specifically that a lower amount of samples in the high frequency range does not create distortion in complex signals, without re-quoting nyquist? |
The Nyquist theorem is the one that proved it. Why would you not want to quote it?
| Quote: |
Emmmm, I believe the Manley Slam uses 192kHz 'verters which is much closer to the "New and Improved" Nyquist Figure of 208kHz. Slams, Weiss, and Lynx sound the best to me and, given the constraints, are doing the best they can do |
All converters sample at rates much higher than 192 kHz initially. The converters you mention sound great at 44.1 kHz. Have you actually heard them anywhere other than on the 3d CD? That's more of a rhetorical question as I know you won't answer.
| Quote: |
Seemingly Nyquist himself did not promise that all of the waveform's information would be captured by the digitization process, furthermore, it seems he stated that his theory could only work if the filters were theoretically perfect (perfect impulse response, distortion, phase issues). You tell me if he claimed otherwise. |
No, he didn't promise that all of the waveform's information would be captured by the digitization process...just all the information below the Nyquist frequency. And of course it only works perfectly if the filters are perfect, which none are.
The point is, that the filters have gotten better and better over the years, so the answer to higher quality isn't necessarily raising sampling rates. Likewise, higher sampling rates are not synonymous with higher quality. Any deficiencies in the sound of digital audio are due to flaws in the implementation, not because Nyquist was "wrong" or because two or three samples of a waveform aren't enough to accurately capture and reproduce it.
| Quote: |
the higher the sampling frequency, the more samples OVERALL the more accurate the reproduction of the signal even below nyquist. |
No. That is simply wrong. Not the frequencies below Nyquist. If they aren't reproduced as accurately it's because of the filters or some other flaw in the system, not because there aren't enough samples to reproduce it. Because there are. It's not a theory or an opinion. It's been proven.
| Quote: |
As for me looking in the right or wrong places for solutions, at least my mind is open enough to still be looking, and I will not discount out of hand even something that might seem bizarre, out of the mainstream, or completely revolutionary. |
Again, your mind is not open. I have no doubt, however, that you would embrace the bizarre.
It is so obvious that you ignore most of the facts that are presented and are so stubborn in your theories that I'm halfway convinced that you actually know exactly what you're talking about and just post the far-fetched nonsense that you do for pure entertainment.
| Quote: |
THIS IS THE SCIENCE. ACCEPT IT. UNFORTUNATELY, FOR THOSE WHO LIKE TO PICK LOOPHOLES, IT'S CALLED A "THEORY", BUT IT IS A FACT. |
It's not actually a theory, but a theorem, right? Because it was proven correct?
In any case, I'm all for the advancement of science and technology when it comes to converters. But personally, I'd much rather see the technology for existing sampling rates improve than see the necessary storage and processing power needed be quadrupled or octupled. Why do people so strongly want higher sampling rates to be the answer?
-Duardo
Post by: compasspnt on September 24, 2005, 02:09:45 AM
Post by: Johnny B on September 24, 2005, 04:10:51 AM
| Duardo wrote on Sat, 24 September 2005 05:06 |
Why do people so strongly want higher sampling rates to be the answer?-Duardo |
Five answers off the top:
The first point is that there are dents in the old 20-to-20 thing which have been brought into finer focus during the digital vs. analogue debate. Some of the measurement tools have been refined which makes it easier and faster to do tests. People are not in total agreement anymore as to how much of the frequency spectrum is really important to capture and reproduce. The entire 20-to-20 thing could very well crumble when subjected to more applied science using a broader multi-disciplinary approach. There are heated arguments about bones and brain scans and the thumps felt by your body. There are studies which show that the brain chemistry changes, so a simplistic counter-argument based upon mics and speakers is unsatisfying and can be viewed by many as an attempt to thrawt further study. And while 20-to-20 may have served people well in the analogue world to some limited degree, it may not serve people well in regard to digital. My sense is that the entire 20-to-20 thing is now legitimately open for debate. Further tests and study could show that David Blackmer's upper figures or Boyk's 104kHz are extremely important for digital. How much of the frequency spectrum is really important is now very much an open issue.
Second, just as we see people are not persuaded by the mic and speaker counter-argument (which BTW should be totally separated from the SRC Sample Rate Debates) we also see they are equally unpersuaded by the "using vast resources" counter-argument. This "using vast resources" is very easy for them to set aside because they have the actual experence of having geometric increases in speed, capacity, and horsepower delivered into their hands for ever cheaper prices. They have lived with "better, faster, cheaper" for so long now that everyone counts on it. So this counter-argument about using vast resources goes against their own experience of what they know is true.
Third, some people intuitively sense that higher sample rates may help with decreasing or eliminating phase and 'time smear" issues. And while it may be true that the higher rates, in and of themselves, may not do this, the supporting technology surrounding the higher rate converters could very well deliver the result they desire.
Fourth, the filters operate better when used in conjunction with higher sample rates which results in people liking the higher sample rate converters.
Fifth point, and the bean-counters and certain types of "double e minds' love this last point, it's cheaper to do the filters this way.
So those are but a few of the reasons one could assign for the push for higher sample rates.
Besides any monetary issues, either in the form of rapidly depreciated gear or making money from increased sales of new gear, I think both sides of the argument have a good deal of emotion bound up in the "slow vs. faster rate" debate.
Funny thing about people's emotions, it can spur action.
Personally, I'd simply like digital to progress to where it sounds better.
By using the term "sounds better" I mean to take in the entire chain, from performer thru delivery, all the way until it plays on the consumers' system. In this sense, Digital should progress and be made to sound better from top-to-bottom of the chain.
I wish I had all the answers to revolutionise and improve digital sound quality because I might like a shot at that brass ring and walk off with a Billion Dollars. Undoutedly, it'll be some twenty-year-old kids who get the billion-dollar-jackpot.
Ok, this time, I'm really, really, really, really, out. Way, way, out ...and going to get some sleep.
My best wishes to all.
Post by: Jon Hodgson on September 24, 2005, 04:12:39 AM
| maxdimario wrote on Sat, 24 September 2005 00:17 |
as someone who works with dsp's you are probably aware of how hard it is to model the behaviour of an analog circuit perfectly through a mathematical model. I use this as an example, I know that recorded audio and synthesised audio are not the same thing, I am talking about the relationship between a math model and the reality it needs to represent. to me Nyquist and Fourier do not cover all the bases. |
Hi Max,
Actually one of the DSP tasks I have done was emulation of an analogue synthesizer, and generally the response I've had is that I did a pretty good job (though I won't even hint that it's perfect). So I am painfully aware of how difficult it is to model that behaviour.
But here's the thing, Nyquist and Fourier PREDICT THE PROBLEMS PERFECTLY. When it comes to issues of sample rate Nyquist and Fourier DO cover all the bases. And they tell us...
1) For SAMPLING and PLAYBACK you need a sample rate which gives a nyquist frequency high enough that the required filter does not affect audio in the audible band (Whatever that band may actually be, though the evidence in favour of 20kHz looks pretty good to me)
2) For PROCESSING that involves non linear elements (such as emulating the distortions in an analogue filter) you're almost certainly going to have to use an oversampled stage if you're going to avoid the harmonics that are generated by those non linearities from folding back into the audible band and making everything sound really nasty.
Now from the above you might conclude that since you're likely to have to process at a higher sample rate, you might as well do your sampling and playback at that rate, since nyquist doesn't say it would get any worse, what's the problem? But pretty much everything I see from people who's knowledge of converters I respect (such as Dan Lavry), says that as the sample rate goes up you bump more and more into the limitations of the physical world in your circuits and you end up with more samples each of which is LESS accurate.
| maxdimario wrote on Sat, 24 September 2005 00:17 |
as resolution in digital is not simply an issue of the audio frequency/sampling frequency ratio, it is a total issue of waveform resolution. the higher the sampling frequency, the more samples OVERALL the more accurate the reproduction of the signal even below nyquist. |
As several other people have stated here, this is not true. Nyquist and Fourier are proven as fact. They are the basis of communications, if they didn't work you wouldn't be reading this right now.
| maxdimario wrote on Sat, 24 September 2005 00:17 |
and as an analog guy, let me say that no self-respecting analog system functions on the border of it's limitations, not because theory says so but because experience dictates. 44.1 digital does. |
I doubt you'll get much disagreement on this one. But (unless you believe that you can hear anything over 20kHz) the problem is not digital, it is analogue. Putting the nyquist that close to the audible band makes producing a sampling and reconstruction filter extremely difficult, it has to be really steep, whilst simultaneously having zero effect in the audible band.
But as increasing the sample rate reduces those issues in the analogue circuitry, it brings in others, again in the analogue circuitry of the converter. Dan Lavry knows infinitely more about converter design that I do, and if I recall correctly he says that for an audible band of 20-20kHz a sampling rate of 60kHz would be enough to get the filter artifacts well out of the audible range, 96kHz is already overkill, and 192kHz gives you LESS accurate sampling.
Post by: maxdimario on September 24, 2005, 05:30:50 AM
| Quote: |
But here's the thing, Nyquist and Fourier PREDICT THE PROBLEMS PERFECTLY. When it comes to issues of sample rate Nyquist and Fourier DO cover all the bases. And they tell us... |
yes, I get it.
perfectly with perfect filters (and does that exist either?)
when I start hearing digi recordings that have retained that specific element that is missing compared to analog, i'll be happy to know that someone has got it down perfect or imperceptibly defective, and then i'll probably buy one.
to me it's not a distortion it's a DEFECT, it sounds like a defect. well...that's because digital distortion is different than analog distortion, and the ears have only ever analized analog signals, for the last million years? ..what's that stuff happening there in the middle of the music...it's the quality of the converters? great, if the converter fixes the problem then every household player needs to be at least that good.. cheap analog still had that true quality at low speed though it was band-limited...
dc you ignored my question above, what is the hardest waveform to reproduce in digital, there must be a 'hardest' one, come on, what kind of signal is most difficult to capture in digital.
well it's time for some quote hunting and text reading very soon. This interesting boxcar of a discussion will perhaps get somewhere past 'the brickwall filter' (or filters present).
Post by: Jon Hodgson on September 24, 2005, 06:26:06 AM
| maxdimario wrote on Sat, 24 September 2005 10:30 | ||
Analog has it's defects too...I remember connectors....grrr..
yes, I get it. perfectly with perfect filters (and does that exist either?) when I start hearing digi recordings that have retained that specific element that is missing compared to analog, i'll be happy to know that someone has got it down perfect or imperceptibly defective, and then i'll probably buy one. |
No, I'm afraid that going by the part of what I said you chose to quote that you don't get it. When I process in the digital domain nyquist and fourier and various other calculations predict the results (both good and bad) correctly. I'm not making any claim about what those results need to be for it to sound good, I'm saying that the maths tells us EXACTLY what the results will be.
Nyquist and Fourier don't say anything about what sounds good, they simply state what you need to do if you want to store or generate and replay a signal of a given bandwidth, whether that bandwidth is 10Hz or 10 Billion Hz.
When I then convert those signals to analogue nyquist once again predicts what will happen with imperfect filters (which are the only kind which exist, as you already know). The problem is in predicting what the performance of those imperfect filters will be. That requires analysis of the analogue circuitry and here things become very complex, because transistors are not linear, and above 0 Kelvin you have electrons moving around randonly and creating noise, and you can have physical interaction between circuit components. Basically there are a huge number of variables, so you can never calculate what will happen with the precision you can in the digital domain, but you can do so to a high level of precision.
| maxdimario wrote on Sat, 24 September 2005 10:30 |
to me it's not a distortion it's a DEFECT, it sounds like a defect. |
It is still distortion, it is just distortion that you find unpleasant. Similarly in the analogue world you get distortions which are pleasant or unpleasant... they are all distortions. Trying to change the terminoligy only muddies the water.
| maxdimario wrote on Sat, 24 September 2005 10:30 |
well...that's because digital distortion is different than analog distortion, |
No it is not different. The distortion you are referring to IS ANALOGUE DISTORTION, it is happening in the analogue domain.
For example an amplifier can have clipping distortion or crossover distortion, they both sound completely different (crossover is generally more unpleasant), but the are both analogue distortions.
The distortions you are hearing (assuming we are talking purely about recording and playback, processing is another discussion) are happening in the analogue domain, and it is in the analogue domain that the solutions must be found, because that is where the compromises are to be made. As previously stated increasing sampling frequency reduces the stress on your filter design by moving artifacts out of the audible band, so that is good... but then it also makes the actual samples less accurate, so that is bad, so you have to make a compromise.
| maxdimario wrote on Sat, 24 September 2005 10:30 |
dc you ignored my question above, what is the hardest waveform to reproduce in digital, there must be a 'hardest' one, come on, what kind of signal is most difficult to capture in digital. |
I'll answer this one for him, though there seems little point since you appear to refuse to believe it.
In the case of sampling there isn't a worse case signal for digital. As long as your input signal is banlimited to less than half your sampling frequency, any waveshape is captured to the limitations of your quantisation steps (how many bits per sample).
There may however be worst case signals for the ANALOGUE circuitry on the input and output of the system.
Post by: Andy Simpson on September 24, 2005, 07:26:45 AM
| Jon Hodgson wrote on Fri, 23 September 2005 17:43 | ||
You're welcome level quantisation error has the same effect as a spacial timing error OF THAT POINT. However the signal consists of a series of points, joined together by a band limited line, and the error of those points is not contant, so you don't get a phase shift of that frequency forward or backwards in time, what you end up with is the correct signal (because ON AVERAGE the points are correctly placed) plus a random error, a random error is white noise. |
If level quantisation error has the same effect as a spatial timing error OF THAT POINT, then as frequency rises and samples per cycle reduce, the probability that this spatial timing quantisation error combines to produce consistent timing error over the entire period of the cycle increases, no?
Andy
Post by: Jon Hodgson on September 24, 2005, 07:55:50 AM
| andy_simpson wrote on Sat, 24 September 2005 12:26 | ||||
If level quantisation error has the same effect as a spatial timing error OF THAT POINT, then as frequency rises and samples per cycle reduce, the probability that this spatial timing quantisation error combines to produce consistent timing error over the entire period of the cycle increases, no? Andy |
No, because sometimes that error puts the point slightly ahead and sometimes slightly behind, amd on average in exactly the right spot.
So what you get after the reconstruction filter is not a time shifted freqency component, instead you get the original component in the correct phase, PLUS a random error element (noise).
Post by: Andy Simpson on September 24, 2005, 09:16:34 AM
| Jon Hodgson wrote on Sat, 24 September 2005 12:55 | ||||||
No, because sometimes that error puts the point slightly ahead and sometimes slightly behind, amd on average in exactly the right spot. So what you get after the reconstruction filter is not a time shifted freqency component, instead you get the original component in the correct phase, PLUS a random error element (noise). |
It is this 'sometimes' that I'm interested in.....
'On average' would be fine if we considered the entire waveform at once, and used the reconstruction filter on the entire waveform, but we know that this is not how it works.
Also, 'on average' surely is related to how many samples are taken to make the average?
Presumably, 20k sampled at 40k, for the period of a cycle has less information for this average to be derived from, than say 20hz at 40k?
Perhaps the most 'important' sample is the first of a waveform, since from that point onwards the reconstruction filter will force all subsequent points to line-up in time with this, so that the reconstructed waveform falls below the filter limit?
Andy
Post by: Jon Hodgson on September 24, 2005, 10:02:00 AM
| andy_simpson wrote on Sat, 24 September 2005 14:16 |
It is this 'sometimes' that I'm interested in..... 'On average' would be fine if we considered the entire waveform at once, and used the reconstruction filter on the entire waveform, but we know that this is not how it works. |
What do you mean by this?
| andy_simpson wrote on Sat, 24 September 2005 14:16 |
Also, 'on average' surely is related to how many samples are taken to make the average? Presumably, 20k sampled at 40k, for the period of a cycle has less information for this average to be derived from, than say 20hz at 40k? |
Well as previously mentioned, you can't successfully sample 20KHz at 40 kHz... but 19.999999 is ok
More data (sample points) does not neccessarily equal more information.
For example you have a factory that manufactures blue cars, yellow busses and green vans. You have two production lists for the month, the first one reads
buses - 7
cars - 3
vans - 5
the second one reads
buses - 7
yellow vehicles - 7
cars - 3
blue vehicles - 3
vans - 5
green vehicles - 5
Which one has more data (samples)? The second one obviously, which one contains more INFORMATION? Neither, they are exactly the same, if you know the number of cars you also know the number of blue vehicles, if you know the number of blue vehicles you also know the number of cars.
| andy_simpson wrote on Sat, 24 September 2005 14:16 |
Perhaps the most 'important' sample is the first of a waveform, since from that point onwards the reconstruction filter will force all subsequent points to line-up in time with this, so that the reconstructed waveform falls below the filter limit? Andy |
No, that's not how it works.
Also bear in mind that in a real sampling situation the quantisation error is well below the noise floor, which means to all intents and purposes, there is no error. (This is a strange one the first time you encounter it, but you can actually improve the accuracy of a sampling system by adding noise to the input.... or more precisely you spread the same error across the whole audio spectrum rather than having a bigger error in fewer frequencies).
I don't know how to say this any more clearly, if you are hearing problems due to phase then that problem is not caused by sampling frequency, because this does not affect phase, it must therefore be caused by some form of processing, whether that is the input filter, the reconstruction filter, or some internal processing that is going on.
Now some increase in sampling frequency above 44.1 may well help this if the phase problems are caused by the steepness required of the anti-aliasing filters, but this does not mean that there is anything wrong with nyquist's theorem, rather that building a less than optimal digital system (optimal in the case of using the minimum resources to get 100% of what we want) makes it easier to build a more optimal analogue part, and therefore a better overall system.
Post by: bobkatz on September 24, 2005, 10:08:28 AM
| andy_simpson wrote on Sat, 24 September 2005 09:16 |
Perhaps the most 'important' sample is the first of a waveform, since from that point onwards the reconstruction filter will force all subsequent points to line-up in time with this, so that the reconstructed waveform falls below the filter limit? Andy |
Andy keeps jumping on that "first of a waveform" issue. I am not a mathematician, but it seems to me Andy's conclusion is wrong, the first sample does NOT determine the accuracy of the rest. Andy's worried that the first sample has been "missed" because it may not land in "synchronization" with his high frequency wave of interest. Well, it does not matter where on the waveform the sample lands, the reconstruction filter will recreate the original (filtered) waveform. The truth is that the output waveform will be exactly correct, plus the addition of random noise, as John mentioned. That's it.
In other words, even if you produce a sine wave that is seemingly exactly synchronized with the sample rate, and sample it, you will get a correct output. The offset of this wave to the sample rate is completely irrelevant and I urge Andy to get off that bandwagon.
I think there is a pathological case of exactly the Nyquist frequency if it is exactly in sync with the sample rate. But Nyquist didn't say that you could reproduce the Nyquist frequency, did he?
BK
Post by: J.J. Blair on September 24, 2005, 12:23:48 PM
I do have a serious question here though. Who here has built there own AD converters and uses them? I mean, some of us have some great book knowledge about this theoretical crap, but who here has actual working knowledge?
Post by: J.J. Blair on September 24, 2005, 12:38:12 PM
Here's an example: I mix down to 1/2" @ 30ips on my ATR. I could go buy a Lavry Blue or a Weiss or a UA 192, and I'm sure that my mix would be recorded more accurately than with the 1/2" machine. The 1/2" adds a very subtle amount of tape compression (if any at +6), and a head bump at 50hz. It is certainly less accurate. I just like the sound. Does that mean that analog is better? No. In fact, it might be inferior in many ways. I just like the sound. But I don't go around referring to it as "King Analog" and dismissing the validity of digital. That would sound downright ignorant, don't you think?
Post by: compasspnt on September 24, 2005, 01:49:43 PM
JB is not around here any more to read your post.
But I agree, that digital probably (if fully implemented properly, with highest grade converters, etc.) does capture more accurately (compared to the original live signal) than does analogue. I know (as I've posted before) that when I mix I will usually mix to both 1/2" analogue on 499, and to 96/24 through a UA 192 converter. Almost always, upon immediate playback, I think that the digital reproduction is closest to what I was actually hearing "live" as it went down. But then very often I will like the 1/2" version better when listening the next day, or upon choosing between the two during mastering (but not always; on some occasions, especially for softer, quieter music, I will still choose the 96/24).
*Analogue has flaws, but I (we) love most of the changes it makes to the sound.
*Digital has flaws, but if properly employed, has a tremendous capability to reproduce a live signal accurately.
*At the end of the day, it will all end up in an iPod.
Post by: timrob on September 24, 2005, 02:04:08 PM
| compasspnt wrote on Sat, 24 September 2005 12:49 |
JJ, *At the end of the day, it will all end up in an iPod. |
Just wondering. Do you say that because you think the iPod sucks...
Or, because you believe it only supports lossy formats. I only ask because it seems to be a common misperception that it doesn't support non compressed formats like AIFF and WAV. Its not the greatest sounding piece of gear in the world, but it sounds as good or better than most Portable personal music devices on the market.
Of course, none of us can help it if a kid wants 50,000 songs on his iPod and encodes everything at really low quality.
Post by: CCC on September 24, 2005, 02:10:21 PM
| J.J. Blair wrote on Sat, 24 September 2005 17:23 |
I do have a serious question here though. Who here has built there own AD converters and uses them? I mean, some of us have some great book knowledge about this theoretical crap, but who here has actual working knowledge? |
Well, GM has that knowledge, but he doesn't participate in this forum anymore. I don't know him well enough to make a presumption as to why that is, although it is proposed that he's busy building a new room. In any case, I sometimes wonder if threads like this proliferate because he's not here participating - or if he's not participating because threads like this proliferate.
I mean really - analog vs. digital.
We are still talking about analog vs. digital and the qualitative and quantitative differences?
Bloody hell.
Hey, I discovered the font size and color features. Cool, huh?
Post by: compasspnt on September 24, 2005, 03:05:42 PM
| John Sorensen wrote on Sat, 24 September 2005 14:10 |
I sometimes wonder if threads like this proliferate because [GM]'s not here participating - or if he's not participating because threads like this proliferate. |
It's one or the other, isn't it?
http://recforums.prosoundweb.com/index.php/t/7211/6490/?SQ=f 47f0bb0bf251397965108e00bccdbf7
Post by: TotalSonic on September 24, 2005, 03:26:36 PM
* poster with essential misunderstanding of the basics of PCM digital audio posts assertions based on fallacies
* other posters try to explain where initial poster is completely wrong - various other posters still insist on going with the fallacies
* gee willickers, analog sounds different from digital
* golly gee, analog and digital recording both have limitations
Can we all go home now?
Best regards,
Steve Berson
Post by: Ged Leitch on September 24, 2005, 03:29:09 PM
| J.J. Blair wrote on Sat, 24 September 2005 17:38 |
It just might sound more musical to some of our ears, including mine. Here's an example: I mix down to 1/2" @ 30ips on my ATR. I could go buy a Lavry Blue or a Weiss or a UA 192, and I'm sure that my mix would be recorded more accurately than with the 1/2" machine. The 1/2" adds a very subtle amount of tape compression (if any at +6), and a head bump at 50hz. It is certainly less accurate. I just like the sound. Does that mean that analog is better? No. In fact, it might be inferior in many ways. I just like the sound. But I don't go around referring to it as "King Analog" and dismissing the validity of digital. That would sound downright ignorant, don't you think? |
And I would second that J.J, I think thats whats been missing from this thread all along.
Post by: Johnny B on September 24, 2005, 03:50:36 PM
After all, whether or not analogue has some technical flaws is not the point of comparison, that's plain silly to compare the two on that level.
Where the rubber hits the road is when you actually listen to the products produced by both technologies.
I don't think there would be any debate at all if digital behaved and sounded exactly like analogue. That's where the bone of contention is, is it not?
I do wish GM would chime in to clarify his comments to Lynn Fuston about PCM having to go to 384kHz to "catch up" with the sonics of DSD.
I think that would be helpful or advance people's understanding and move the discussion forward.
Who knows, maybe that statement "I feel the need for speed" will turn out to be more accurate than many now realise.
Cheers.
Post by: StudioRhythm on September 24, 2005, 05:02:58 PM
| Johnny B wrote on Sat, 24 September 2005 12:50 |
It seems to be to be just as ignorant to try and discount all the complaints about digital sound quality or take unnecessary pot shots at analogue. |
Johnny, if someone actually would bring up specific complaints, I, and probably everyone else would be happy to discuss them. But saying, "digital sucks and is inferior to King Analog" hardly counts as a specific complaint. Unless you've personally compared the same mix recorded to both digital and analog, as Terry Manning just told you he has and does practically daily, you don't really have much of an argument -- how can you separate out the production decisions from the format exactly at the end?
| Quote: |
After all, whether or not analogue has some technical flaws is not the point of comparaison, that's plain silly to compare the two on that level. Where the rubber hits the road is when you actually listen to the products produced by both technologies. |
Such as? Show me two comparable examples, one recorded digitally and one recorded analog, and then we'll talk. Generalizations don't cut it.
| Quote: |
I don't think there would be any debate at all if digital behaved and sounded exactly like analogue. That's where the bone of contention is, is it not? |
The bone of contention is when people make uninformed sweepingly incorrect statements and then argue endlessly with people who clearly have way more knowlege than they do on the points being addressed (not speaking of myself here, although I think I would qualify in this particular case....)
| Quote: |
I do wish GM would chime in to clarify his comments to Lynn Fuston about PCM having to go to 384kHz to "catch up" with the sonics of DSD. |
I'd certainly be curious to hear anything George has to say on the subject, but don't forget that he has been using digital almost exclusively for quite some time....
| Quote: |
I think that would be helpful or advance people's understanding and move the discussion forward. Who knows, maybe that statement "I feel the need for speed" will turn out to be more accurate than many now realise. |
Well, the scientific evidence is leaning strongly towards the opposite -- it appears to most people who have done intelligent objective testing that it's all about the implementation of the filters (both digital and analog), and when the highest quality filters are used, 48 sounds almost (if not completely) identical to 96. Of course, things keep getting better, but it appears to be a waste to just keep upping the sampling rate, especially considering the toll it takes on system requirements -- which would you rather have, 48 tracks or 96? There's no such thing as infinite computing power, you know, no matter how fast processors get.
Post by: crm0922 on September 24, 2005, 06:47:21 PM
| Dan Feiszli wrote on Sat, 24 September 2005 17:02 | ||
I'd certainly be curious to hear anything George has to say on the subject, but don't forget that he has been using digital almost exclusively for quite some time.... |
I find it confusing why GM would suggest the use of 384k. If you have a working knowledge of sampling theory, and implementation, the idea that higher than 50k-60k sampling rates could ever be necessary is patently absurd.
Someone should just build a FATSO type distortion circuit and a fake head bump into a digital converter, not tell anyone it's in there, and push it as "sounding as good as analog". Maybe include a little wow and flutter, etc. I bet it would sell like hotcakes.
Chris
Post by: Eric Bridenbaker on September 24, 2005, 07:27:32 PM
| crm0922 wrote on Sat, 24 September 2005 18:47 |
I find it confusing why GM would suggest the use of 384k. If you have a working knowledge of sampling theory, and implementation, the idea that higher than 50k-60k sampling rates could ever be necessary is patently absurd. |
Reduced latency would be my guess.
Best,
Eric
Post by: compasspnt on September 24, 2005, 07:33:37 PM
| crm0922 wrote on Sat, 24 September 2005 18:47 |
Someone should just build a ... distortion circuit and a fake head bump into a digital converter, not tell anyone it's in there, and push it as "sounding as good as analog". Maybe include a little wow and flutter, etc. |
Ssshhh Chris!
I have actually been working on just that very thing.
Post by: thedoc on September 24, 2005, 08:57:11 PM
Post by: Andy Simpson on September 24, 2005, 09:00:18 PM
| Jon Hodgson wrote on Sat, 24 September 2005 15:02 | ||
The problem with explaining all this stuff is it is not intuitive, though it has been proven as fact again and again both mathematically and experimentally.
What do you mean by this? |
Let's consider 2bit 44.1, for the sake of intuition....
Imagine a full scale 20k wave represented in samples.
Could this wave not move forward or back in time significantly without changing the resultant samples?
What is the significance of this at reconstruction?
Perhaps I am arguing for greater bit depths......
Andy
Post by: crm0922 on September 24, 2005, 09:22:31 PM
You see, the time component (sample rate) is integral to the frequency in question. If the next sample is less or more "time away", the frequency will be different when heard as audio.
frequency = 1 / time
The sample rate and the bandlimiting filter are the non-intuitive "missing information" that is required for 2 measly samples to accurately reproduce a high frequency waveform.
It makes perfect sense when you understand the underlying theory. The closer you get to the Nyquist cutoff, the more the cutoff becomes part of the "trick" that only requires 2x highest frequency required. This is not exactly true, per se, but might help you visualize what I'll mention next.
Without proper bandlimiting, the waveform in question will *not* be faithfully reproduced. The requirements of Nyquist must be met as completely as possible for an "as complete as possible" reproduction of the input waveform.
Chris
Post by: maxdimario on September 24, 2005, 09:43:42 PM
| Quote: |
But I agree, that digital probably (if fully implemented properly, with highest grade converters, etc.) does capture more accurately (compared to the original live signal) than does analogue. I know (as I've posted before) that when I mix I will usually mix to both 1/2" analogue on 499, and to 96/24 through a UA 192 converter. Almost always, upon immediate playback, I think that the digital reproduction is closest to what I was actually hearing "live" as it went down. But then very often I will like the 1/2" version better when listening the next day, or upon choosing between the two during mastering (but not always; on some occasions, especially for softer, quieter music, I will still choose the 96/24). |
ooohhh!
I don't agree..
It may sound identical aesthetically, but does it have the same musical effect as live?
it's TIME-distortion.
come on! are you saying that the digital recording gives you a better idea of what was going on emotionally in the studio than analog?
who cares about perfect aesthetic reproduction, it's the overall effect on the listener that counts, and that has a lot to do with TIME resolution.
forget frequency response, THD etc. .. you know what? most people at home can't tell the difference about these things, otherwise they'd complain.
better put a smiley on..
Post by: maxdimario on September 24, 2005, 09:46:45 PM
I don't have too much faith in your monitoring chain, guys.
maybe that's the thing that's keeping the discussion from coming to a point.
Post by: maxdimario on September 24, 2005, 09:49:49 PM
| Quote: |
How about a 15ips/30 ips headbump switch |
We've had the plug-ins around for ages, and they sound like plug-ins, don't they.
listen to some live music once in a while, then you'll get it.
Post by: JamSync on September 24, 2005, 09:51:12 PM
| bobkatz wrote on Thu, 22 September 2005 23:51 | ||
But we think that if we upsample high enough, shift it a sample or two & down-sample back to 44.1, we can do it? Andy |
And that's exactly how the Cedar Azimuth fixer works. If you want to do some subsample slipping, just put it through the Cedar.
BK[/quote]
Just for more useless info: you can subsample slip on a Lex 480 or 300 <yawn>.
Post by: maxdimario on September 24, 2005, 09:59:26 PM
| Quote: |
It is still distortion, it is just distortion that you find unpleasant. Similarly in the analogue world you get distortions which are pleasant or unpleasant... they are all distortions. Trying to change the terminoligy only muddies the water. |
Hi,
I hear you.
what I'm saying above is that analog distortion is something the ear understands and can actually make a recorded sound MORE intelligible.
digital distortion is alien to nature, and the hearing system, being human and therefore natural, considers it as a noise.
if you're going to replace analog for digital because it is so much closer to the original sound and therefore better, but the overall effect on the listener is negative and cold SOMETHING IS WRONG isn't it?
maybe the key elements in music are not what you are consciously listening to.
I don't hate digital,
I want digital to work, but FOR REAL, just like analog did.
in the sense that from the studio to the home player, you must be able to 'get it'
BECAUSE PERFECT AESTHETICS DOES NOT SELL RECORDS, FEEL AND THE HUMAN ELEMENT DOES.
Post by: CCC on September 24, 2005, 10:30:17 PM
| maxdimario wrote on Sun, 25 September 2005 02:46 |
you know what? I don't have too much faith in your monitoring chain, guys. maybe that's the thing that's keeping the discussion from coming to a point. |
So let me get this straight - you don't have faith in the monitoring rigs of Dave Collins, Bob Katz, JJ Blair, Brad Blackwood, Ronny Morris. KK Proffitt. Bob Olhsson or Terry Manning - among others?
Tell you what, bubba, there's one thing that's keeping this discussion (to use that term very loosely) from coming to a point. See if you can guess what it is.
.
.
.
Think hard.
.
.
.
Somebody in this discussion has absolutely positively no interest whatsoever in either genuinely contributing or learning. See if you can guess who.
So tell you what - let's give you the front and centre stage here to show us how dumb we all are - please post an analog recording of your own that you think 'captures what was going on emotionally in the studio' and explain precisely how your choices of equipment were made with the intention of preserving analog emotion - as you like to phrase it. I'm interested in the whole chain, but I am particularly interested in; the make and model of the emotional tape recorder, the brand of emotional audio tape, and whether it was 1/4", 1/2", 1", 2" or whatever. I'm interested in the speed you prefer and why, as well as the alignment (I like to think in terms of 185nW/m so if you could give us your preferred emotional level referenced against 185 that would be very gratifying). If there was any emotionally disturbing noise reduction in the chain it would be nice to document, as well as perhaps a review of your preferred practices with respect to cleaning and demagging. If you keep the machine in very good shape so as to preserve it's perfect aesthetics there may be a loss of emotion, I fear.
I'd be interested in clips of your work that exhibit "analog distortion" effects that made something more intelligible in your opinion. If there could be an A/B comparison between the undistorted version and the more intelligible distorted version that would be hot.
Do you have a recording to show that represents, in your opinion, an example of an analog recording you made that preserved the sense of "getting it" from the studio to the home listener.
Mr. DeMille, he's ready for his close-up.....
(edited for typo)
Post by: tom eaton on September 24, 2005, 10:35:30 PM
And, more to the point, if digital sounded EXACTLY like analog it would be broken. If you have any contact with folks that record classical for a living see what they think about digital recording. You'll find that many PREFER it to analog.
Maybe their monitors suck, though.
-tom
Post by: bobkatz on September 24, 2005, 11:22:48 PM
| J.J. Blair wrote on Sat, 24 September 2005 12:23 |
Quick note to Andy re my post testerday: I genuinely admire your enthusiasm and I honestly respect the recordings you've posted. That's not me tempering my post. I just figure I'd rather change my tactics and try to encourage you to raise the level of your posting rather than tell you to shut up. Who am I to do that? (Unless you're alpha jerk! LOL.) I do have a serious question here though. Who here has built there own AD converters and uses them? I mean, some of us have some great book knowledge about this theoretical crap, but who here has actual working knowledge? |
I can't say "uses" anymore. I can only say I "built" (past tense) and "used" (past tense). Both units based on eval boards and customized. The first from a DBX eval board, the second from an Ultra Analog. I knew a lot about the inner architecture of the UA board so my mods were significant. However, building your own A/D from scratch including designing and laying out your own 5 layer PC board? That's a lot heavier task and the simple question of "has anyone built their own A/D?" becomes a moot point.
There are some "kits" around, I believe. Does that count, either?
BK
Post by: compasspnt on September 24, 2005, 11:28:10 PM
| maxdimario wrote on Sat, 24 September 2005 21:43 |
come on! are you saying that the digital recording gives you a better idea of what was going on emotionally in the studio than analog? who cares about perfect aesthetic reproduction, it's the overall effect on the listener that counts, and that has a lot to do with TIME resolution. |
Oh, excuse me. I forgot all about the emotion. I guess I was just all caught up in the gear/format/aesthetics argument. Thanks for reminding me.
Post by: maxim on September 24, 2005, 11:39:25 PM
fwiw, i'm typing this as i'm listening to some classic french pop (leo ferre, jacques brel, serge lama etc) as a bunch of mp3 through the apple I speakers via pismo converters, and they don't fail to evoke an emotional response by any means
max dm wrote:
"digital distortion is alien to nature"
now you're starting to sound like a fundamenalist christian
Post by: Johnny B on September 25, 2005, 01:36:26 AM
Why do people fear even testing Next Gen higher rate converters?
Have any of you designed chips that work at Next Gen high speeds?
Have any of you ever listened to prototype units which use very high speed Next Gen converters?
If you cannot say "Yes" to those kinds of questions, based upon your own personal experience, then you are condemning Next Gen high speed technology based upon your own ignorance and mere speculation. Some may have some one-sided mathematical theory to condemn the Next Gen higher rates, but they don't have any hard physical evidence. And what if the Next Gen high speed design folks were bright enough to overcome any challenges or had adequate work-arounds?...I think many here would simply condemn the Next Gen high speed technology to a death sentence without even providing the Next Gen high speed tech a "Fair Trial."
Why do some people fear *even* looking at a more holistic approach including more study and greater understanding of how the frequency spectrum operates on the entire human system, the ear, the brain, and *the* body?
I sense the underlying fear of those who would seemingly want to hold the technology back is so great that you could bottle it and sell it to the masses.
There is apparently an emotion underlying all this resistance here, that emotion has a name, it's called "Fear."
No guts?---No Air Medal.
And no billion dollars.
Post by: StudioRhythm on September 25, 2005, 03:30:40 AM
| maxdimario wrote on Sat, 24 September 2005 18:46 |
you know what? I don't have too much faith in your monitoring chain, guys. maybe that's the thing that's keeping the discussion from coming to a point. |
That's funny, Max. So, you're questioning the monitoring chain of everyone here, and when I questioned you on your monitoring chain, you revealed that one component of it is a mediocre DA converter (of which I have personal experience with, in my room, in comparison with Lavry and Benchmark units, which both smoke it.) After I told you that I wasn't surprised you thought digital didn't sound great if you were listening through that unit, you changed the subject as quick as you could -- now you're attacking everyone else's monitoring chain? So, why don't you explain your chain, in detail?
Post by: StudioRhythm on September 25, 2005, 03:40:50 AM
| maxdimario wrote on Sat, 24 September 2005 18:43 |
ooohhh! I don't agree.. It may sound identical aesthetically, but does it have the same musical effect as live? it's TIME-distortion. |
Max, you're talking complete B.S. here. Random error in the digital signal DOES NOT create some magic "time distortion" that could never exist in the analog world. The random error is WHITE NOISE. Don't go making up terms to try and explain things you don't understand.
| Quote: |
come on! are you saying that the digital recording gives you a better idea of what was going on emotionally in the studio than analog? who cares about perfect aesthetic reproduction, it's the overall effect on the listener that counts, and that has a lot to do with TIME resolution. |
Many people care about perfect aesthetic reproduction, and since you've been ranting and raving like a madman about how digital is flawed in this regard, it's hardly appropriate to start changing your argument now just because it keeps getting pointed out how ridiculous the things you say are.
The overall effect on the listener can be manipulated so many ways that digital vs. analog is about the least important decision anyone can make in the overall scope of things at this point (assuming good quality equipment, which it sounds like you have never used....)
Post by: StudioRhythm on September 25, 2005, 03:42:59 AM
| maxdimario wrote on Sat, 24 September 2005 18:49 | ||
We've had the plug-ins around for ages, and they sound like plug-ins, don't they. listen to some live music once in a while, then you'll get it. |
I make most of my living playing live music, and have for the past 10+ years. The rest comes from playing on sessions and engineering. I know very well what live music sounds like, thank you very much, and I think I "get it."
Post by: StudioRhythm on September 25, 2005, 03:53:34 AM
| Johnny B wrote on Sat, 24 September 2005 22:36 |
It seems to me that those who advocate more research, more study, and moving the technology forward are often met with unnecessary name-calling and emotion-based resistance. Why do people fear even testing Next Gen higher rate converters? |
Johnny, nobody "fears" testing anything -- what I fear is that someone might read the ridiculous things you posted and think that there is something in there that bears some slight resemblance to fact or reality. That frightens me very much. Listening to a super-high rate converter? Not so much....
| Quote: |
Have any of you designed chips that work at Next Gen high speeds? Have any of you ever listened to prototype units which use very high speed Next Gen converters? If you cannot say "Yes" to those kinds of questions, based upon your own personal experience, then you are condemning Next Gen high speed technology based upon your own ignorance and mere speculation. Some may have some one-sided mathematical theory to condemn the Next Gen higher rates, but they don't have any hard physical evidence. And what if the Next Gen high speed design folks were bright enough to overcome any challenges or had adequate work-arounds?...I think many here would simply condemn the Next Gen high speed technology to a death sentence without even providing the Next Gen high speed tech a "Fair Trial." |
Johnny, you don't just throw parts into a box and listen to it. Things are designed based on theory! There are no theories to support using sample rates higher than what are out there currently (no, your made-up drivel and gross misunderstanding of just about everything you've written about doesn't count as a theory.) And, if the "one-sided mathematical theory" you speak of were to be wrong (which it has been proven not to be,) the converters just wouldn't work. You can't have it both ways.
| Quote: |
Why do some people fear *even* looking at a more holistic approach including more study and greater understanding of how the frequency spectrum operates on the entire human system, the ear, the brain, and *the* body? I sense the underlying fear of those who would seemingly want to hold the technology back is so great that you could bottle it and sell it to the masses. There is apparently an emotion underlying all this resistance here, that emotion has a name, it's called "Fear." |
What's all this fear garbage about? I don't think anybody posting here has shown any fear about anything; we've just been correcting the endless amounts of B.S. that have been spewing forth from your keyboard.
Post by: J.J. Blair on September 25, 2005, 04:25:09 AM
For such a fucking know-it-all, how come I've never heard of this guy outside of internet forums? Where does he get the temerity to question Terry Manning's listening abilities? Christ, don't people's experience and body of work mean ANYTHING to some of you idiots? max should be kissing Bob Ohlsson's feet for imparting his experience with us so freely, not questioning his monitoring chain.
Sofa king retarded.
Let me put it this way, you were probably conceived to music that Bob Ohlsson recorded (assuming that your parents have taste in music). I'd say he has seniority on you. Fermes ta gueule, already.
Post by: Ronny on September 25, 2005, 04:38:58 AM
| Dan Feiszli wrote on Sun, 25 September 2005 03:53 |
What's all this fear garbage about? I don't think anybody posting here has shown any fear about anything; we've just been correcting the endless amounts of B.S. that have been spewing forth from your keyboard. |
Dan, ask him how many frequencies his high sample rate captures with a 20-20k U87 or a 40-15k SM 57.
Post by: gehrke111 on September 25, 2005, 06:09:16 AM
Post by: maxdimario on September 25, 2005, 09:34:07 AM
| Quote: |
Or max, who apparently doesn't even know enough to know that his converters, upon which he had chosen to judge all digital audio, suck balls. |
prism and apogee suck balls?
tried those too jj.
what converters do you use? digidesign??
gees you guys are beginning to get a little hot under the hood eh.
guess what J.J. I DON'T EVEN NEED CONVERTERS!
All I have to do is to listen to SOMEBODY ELSES, and listen to commercial releases.
Post by: maxdimario on September 25, 2005, 09:39:56 AM
I've never put any doubt into anything Bob O. says, including the part where he says that in the end analog still sounds better than digital.
Bob O. has actually worked with real artists, and the best equipment, and I trust anything he says as a probable truth.
can't say the same about you, or anybody who uses aggression to get his point across.
threads die off when people stop responding.
And J.J.
stop your insults, if only for the reason that even though you DO work a lot more than I even care to you still have your own opinions that seem to contrast with a lot of the more experienced crowd.
DO I EVEN BOTHER TO POINT THAT OUT TO OTHERS?
I am not the first to put the perfect theory of Nyquist in question, and won't be the last.
you seemingly take it for granted.
Post by: maxdimario on September 25, 2005, 09:46:54 AM
| Quote: |
Many people care about perfect aesthetic reproduction, and since you've been ranting and raving like a madman about how digital is flawed in this regard, it's hardly appropriate to start changing your argument now just because it keeps getting pointed out how ridiculous the things you say are. |
Do you actually read what I write?
Digital is aesthetically superior to analog, but the depth and feel of the music gets lost.
I don't really care about aesthetic perfection, since there is no such thing in recorded music such as rock'n'roll and blues, or even classical.
I listen to recordings from the 40's ..far from aesthetically perfect.
Post by: maxdimario on September 25, 2005, 09:51:03 AM
| Quote: |
The random error is WHITE NOISE. Don't go making up terms to try and explain things you don't understand. |
is that superimposed white noise?
Post by: Jon Hodgson on September 25, 2005, 09:54:34 AM
| maxdimario wrote on Sun, 25 September 2005 14:51 | ||
is that superimposed white noise? |
Well it's not on its own channel, what do you mean by "superimposed"?
Post by: maxdimario on September 25, 2005, 10:00:46 AM
what I meant to say is since noise does not exist in digital like in tape etc.. on analog it's always there regardless if a signal is being fed into it or not ...Then the noise is related to the recorded signal?
could this be correct?
Post by: maxdimario on September 25, 2005, 10:14:31 AM
| Quote: |
That's funny, Max. So, you're questioning the monitoring chain of everyone here, and when I questioned you on your monitoring chain, you revealed that one component of it is a mediocre DA converter (of which I have personal experience with, in my room, in comparison with Lavry and Benchmark units, which both smoke it.) After I told you that I wasn't surprised you thought digital didn't sound great if you were listening through that unit, you changed the subject as quick as you could -- now you're attacking everyone else's monitoring chain? So, why don't you explain your chain, in detail? |
Those are the converters I personally have at home, and I am not stupid or ignorant enough to think they are superior to the high end modern ones, sorry.
as I mentioned above to J.J. there is such a thing as going to a recording studio and using someone else's gear.
I am not going to buy anything else until I hear something that makes me believe that it's going to be obsolete in a couple of years, I let others do that because I don't need to.
sure I can explain about the monitoring chain.
if you monitor your sound through an IC mixer a lot of the fine detail is lost, if the signal goes out of the control room out into an amp and then ns-10's as the primary reference you lose more.
I check out converters with a tube mic or ribbon into a tube pre live and compare what is going into the converter and what is going out through sennheiser headphones.
that is a very quick way of checking out the sound of a converter or recorder or any piece of gear.
once it's in-line with the monitor section of the average mixer you don't hear it as easily, though it's still there.
peace.
Post by: Jon Hodgson on September 25, 2005, 10:34:57 AM
| maxdimario wrote on Sun, 25 September 2005 15:00 |
Hi Jon. what I meant to say is since noise does not exist in digital like in tape etc.. on analog it's always there regardless if a signal is being fed into it or not ...Then the noise is related to the recorded signal? could this be correct? |
Well firstly white noise is not related to the recorded signal in either case, it is random noise spread equally throughout the spectrum. In the case of analogue circuitry it is the result of heat making electrons move around randomly, this random element is superimposed on the electron movement due to the audio signal.
If you have silence input into a digital system, then you can record it without noise, since it is easy to quantize a value of zero without error.
But as soon as you have a signal you have errors due to the quantization steps. That error is noise, or distortion, depending on whether it is signal dependent.
If you quantize a signal where the noise floor is below the quantization steps then some of the errors may be signal dependent, so they come out as distortion rather than noise.
If you want an example of this, take a really low level sine wave, which is only one quantization step high, it will be quantized into a square wave, which means that it has odd harmonics added to it.. that's distortion, part of the reason for those "gritty" sounding early samplers.
So what you want to do is spread that error out across the frequency spectrum, if you have an 8 or 12 bit converter then you do this by adding some analogue white noise before you sample. I'm not sure when the crossover point occurs, but by the time you reach 20 bits you don't need to add any noise, since it is already there due to the limitations of the analogue circuitry (you can't escape the laws of physics).
Your sampled signal now contains white noise over your desired signal, just as it would on analogue tape.
You have the same issue when you reduce sample length in the digital domain, say from 24 to 16 bit. That's why when you do this you use digital dither, which is basically adding noise in the digital domain. In the digital domain you can get cleverer about it though, and instead of simply spreading the error out equally over the whole spectrum, you move the noise into areas where it is less audible, systems like Apogees UV22 do this.
Post by: maxdimario on September 25, 2005, 10:50:19 AM
since you're being very clear,
does sampling rate affect this issue, or is it only bit-depth?
do certain frequencies generate more 'noise' than others?
Post by: maxdimario on September 25, 2005, 11:08:53 AM
| Quote: |
Oh, excuse me. I forgot all about the emotion. I guess I was just all caught up in the gear/format/aesthetics argument. Thanks for reminding me. |
..yeah posted that a little too late in the evening and perhaps was not completely cohesive in expressing myself, and was writing with the wrong tone, sorry.
anyway, I do agree that tape sounds more distorted, as analog is apt to be, whereas digital is more of a 'mirror-image'.
my point is that there are certain minute elements in the music which give the ear time-related clues that seem to get lost along the way with digital, especially 44.1.
to me on a rock'n'roll record, for example, the time and feel issue is more important than the 'mirror-image' issue.
I don't believe that it's an issue of analog distortion, but what doesn't get lost.
Post by: bobkatz on September 25, 2005, 12:03:06 PM
| Johnny B wrote on Sun, 25 September 2005 01:36 |
It seems to me that those who advocate more research, more study, and moving the technology forward are often met with unnecessary name-calling and emotion-based resistance. Why do people fear even testing Next Gen higher rate converters? |
You talk about fear and speak about higher rates as if they were nirvana. Scientific testing requires isolation of variables. If a 192 K converter sounds better than a 96K it is not enough to say, "yes it does". It is necessary to find out why.
Testing "next gen higher rate converters" is only part of the research that is needed. I suggest that you test all the sample rate converters you can find. They all sound different. Try upsampling 44.1K to 96 and beyond and then back to 44.1. Then, of course, the DAC you are listening to evaluate has its effect. But you will find that a surprising amount of information remains after that up/downsampling if you use an excellent SRC. That's a data point that we must not ignore.
Or, listen to that upsampled material at 192K or 96K. It sounds better! Did more information get produced? Of course not. So, what about the upsampled reproduction makes it sound better? That's where the research lies. I say it's in the filter design.
Unfortunately, no one except DCS or Prism or Meridian has the knowledge to construct a DAC and ADC with a custom filter set to prove my hypothesis that a properly-constructed filter will make 96K indistinguishable from any higher rate and the highest rate that will be needed.
Most manufacturers use off the shelf IC filters that just don't have the resolution, sufficient coefficient accuracy, ripple, slop and phase response to qualify as a candidate. My point being that we may have to "suffer" with 192K converters because no one knows how to make a really good 96K version.
I know enough now about what it takes to make a transparent-sounding low pass filter to say that current pre-masked silicon chips don't cut it.
A good friend who is the designer for Algorithmix (Christoph Musialik) visited this weekend. His linear phase Algorithmix Red uses 80 bit floating point arithmetic and he says that to minimize the accumulation of errors and distortions you need those 80 bits. Certainly more than 64 float. I nominate the Algorithmix Red as the most transparent (also warm and pure) sounding digital equalizer on the planet.
And yes, you need a digital filter to do the anti-imaging.
BK
Post by: J.J. Blair on September 25, 2005, 12:16:05 PM
| maxdimario wrote on Sun, 25 September 2005 06:39 |
What coverters do you use? Digidesign? |
I guess you never bothered to check my gear list. Actually I've been using Apogees exclusively starting with the AD-500 eleven years ago.
However, I'll retract my remark about your converters sucking balls, since you don't have any. (Converters, that is.)
| Quote: |
you still have your own opinions that seem to contrast with a lot of the more experienced crowd. DO I EVEN BOTHER TO POINT THAT OUT TO OTHERS? |
Hmmmm. What that Bob Ohlsson likes RE15s better than D19s and I disagree with him? Find a single time that I've stated something technical where one of these guys has disagreed with me or corrected me that I have not deferred to them. Just once. I even have several posts where I reconsidered my initial statement and admit that I was wrong. If you could learn to do that yourself, I probably wouldn't lose my patience.
| Quote: |
I am not the first to put the perfect theory of Nyquist in question, and won't be the last. you seemingly take it for granted. |
I do? I'm sorry, I missed that part.
Post by: lord on September 25, 2005, 12:27:58 PM
| maxdimario wrote on Sun, 25 September 2005 09:39 |
I am not the first to put the perfect theory of Nyquist in question, and won't be the last. you seemingly take it for granted. |
It's a THEOREM not a theory.
There is a rather large difference. Taking it for granted is not such a bad idea, since it's been.... proven!
Can you get it into your head that your completely random predilection for 78s over CDs has absolutely no bearing on the soundness of the fundamental principles of digital audio? You are entitled to your preference. But your refusal to acknowledge the validity of basic known physical properties of the universe continues to discredit your position.
The following is an example of a THEORY:
maxdimario is a computer virus sitting on a server somewhere endlessly regurgitating the same madness day after day until the whole world goes insane.
The comical thing is that I agree with you in many ways. In most cases, I prefer tape and vinyl to digital audio. But that preference has nothing to do with anything mentioned in this thread.
Post by: CCC on September 25, 2005, 12:32:09 PM
I see you have posted a lot of messages since I last queried you. I'm still waiting for your dissertation on;
-your preferred analog recorder, and why it is your preferred analog recorder. Bonus points for discussing your preferred multitrack and 2 track deck.
-your preferred brand of tape. It would be ok to discuss a brand no longer available.
-your preferred tape speed, and why it is your preferred tape speed.
-your preferred alignment, preferably referenced against 185nW/m. If you must use another reference number, like say 250nW/m as the baseline for your level then that's fine, we can work the numbers out from there.
-how you feel about noise reduction and whether you have a preference with respect to this?
-and do you have anything you'd like to offer about very routine/basic maintenance on analog recorders?
Please, at your convenience, tell me about any experiences you have with respect to A/B comparisons of the same program material printed simultaneously on analog and digital decks. I mean, let's face it, this is where the 'rubber meets the road', right? I would say that at some point in our lives those of us who, you know, actually record things, have run a 2" side by side with some flavor of digital deck - and/or have run 2 track analog machines side by side with various flavors of digital 2 tracks (DATs, hi res PCM, even DSD). So you must have similar experience to draw on since you have such strong views on the matter.
Just to 'confuse' matters more I am additionally interested in your views on the relative merits of using a PPM style meter vs a VU meter when cutting to analog. You may not be surprised to hear that there is a school of thought that suggests that misreading your meters when tracking on analog leads to overmodulation with all the attendant side-effects. Your views please.
There's another question that's been bothering me. Once upon a time someone told me that their analog edits sounded better the day after they did the editing. I thought this was quite a queer thing. Seems that some folks believe that due to issues of retentivity/coercivity the magnetic domains around an edit will re-orient slightly after a day or so, making the transition across the splice smoother. I think that's interesting, but I'm not sure. What do you think?
That point ties into another question or two - like, the nature of aging in analog tape. Do you think that some of the itsy-bitsy-teeny-weenie little magnetic particles on tape lose their orientation over time? I believe that this is a known phenomenon, but could stand to be corrected. If some of the itsy-bitsy-teeny-weenie little magnetic particles do reorient over time then what is the audible effect? Perhaps this could manifest as a loss of transient snap....I mean, the + and - orientation of the itsy-bitsy-teeny-weenie little magnetic particles will be most dramatic around the onset of any given note. If a tape eventually loses it's 'magnetic value' over time does it sound softer/smoother/more "rolled-off"? I don't know. Can you give me some insight?
And on that subject of the impact of age and wear on sonics do you have views on HF loss on worn tapes and whether this impacts on the impression of emotion in performance or if you would categorize this as simply being an irrelevant issue of aesthetics? Do I understand your views on categorization correctly or am I misrepresenting your position?
Additionally I must confess considerable ignorance on my part as to the manufacture of vinyl. Being something of an analog-guy you might have distinct views on approaches to cutting vinyl so as to preserve....uh, emotion? Care to comment? I eagerly await learning about this, since, as I said, I know virtually nothing about how vinyl gets made.
I'm sorry to, you know, "put you to your proof" again .... twice in two days....but hey, if you actually know what you are talking about then at least some of these questions are quite trivial. Thanks, JS
Post by: compasspnt on September 25, 2005, 12:58:27 PM
Regarding your recent post concerning filter design/sample freq, have you found that it is the QUALITY of the filter design alone which improves overall sound reproduction, or is it the filter design in conjunction with a higher sample rate?
That is, in a converter with the very best possible filters, would there be an improvement in quality at 192 over 96 or 48, or would the improvment already have occurred because of the better filters?
Post by: StudioRhythm on September 25, 2005, 01:13:39 PM
| Ronny wrote on Sun, 25 September 2005 01:38 | ||
Dan, ask him how many frequencies his high sample rate captures with a 20-20k U87 or a 40-15k SM 57. |
Post by: StudioRhythm on September 25, 2005, 01:30:02 PM
| maxdimario wrote on Sun, 25 September 2005 06:46 | ||
Do you actually read what I write? |
Max, I read what you write -- I've been doing it all week. I think the question is more whether or not you read what you write! When it comes to making broad generalizations with no supporting evidence, or even any description of how you came to the extreme positions you hold, you're quite the Dostoevsky.
| Quote: |
Digital is aesthetically superior to analog, but the depth and feel of the music gets lost. |
Hey, there's another one! Would you mind explaining that one a little bit for us?
| Quote: |
I don't really care about aesthetic perfection, since there is no such thing in recorded music such as rock'n'roll and blues, or even classical. |
That's something I agree with; however, not everyone likes all the added noise, distortion, and altered frequency response of tape, especially for very dynamic music. The bulk of what I work on as an engineer is jazz, and most of the musicians I work with (who know nothing about technical specs, and don't know anything much about digital vs. analog) like the way digital sounds! And, to be honest, I think it sounds pretty good myself! (and I appear to be in good company in that regard on this thread, to say the least)
| Quote: |
I listen to recordings from the 40's ..far from aesthetically perfect. |
That's not really a valid comparison for several reasons:
1) the only recordings people still listen to from that period are the best recordings, of the best performances. All the bad and mediocre has been forgotten or lost. Actually, back then, most of the mediocre wasn't even recorded! (Hmmm -- there's an idea....
)2) For better or for worse (OK, for better...,) that was a time where musicians had nothing better to do than practice, perform, and create. Life was somewhat less complicated then; musicians weren't being fed the line that everyone has to be a recording engineer and own their own studio, be their own PR person, webmaster, tour manager, record label. The end result, in my opinion (although I guess I could claim that it's absolute truth if that would help you relate a little better...) was that musicians became better musicians, worked better as a group (since there was more geographic isolation,) and the end result was better performances. Better performances make better recordings, period, with almost any technology.
3) If you took away the great performance, the nice big-sounding hall they were recorded in, I wouldn't care to listen to anything from that era at all -- I'm not really fond of the way those records sound at all, sonically.
4) "The nostalgia factor": Genius + time = a classic
Genius right now = yeah, that's good.
Ever since the second caveman started pounding on a log, his critics were saying, "That's cool, Zog, but you should have heard how Zak used to do it..."
Post by: StudioRhythm on September 25, 2005, 01:35:33 PM
| maxdimario wrote on Sun, 25 September 2005 06:34 | ||
prism and apogee suck balls? tried those too jj. what converters do you use? digidesign?? gees you guys are beginning to get a little hot under the hood eh. guess what J.J. I DON'T EVEN NEED CONVERTERS! All I have to do is to listen to SOMEBODY ELSES, and listen to commercial releases. |
Max, you're not understanding. If you're listening to digital through lousy converters, you are not listening to digital on a level playing field to analog, even if you're listening to the most perfectly engineered recording ever made, recorded with the best eqiupment -- it's all about the weakest link (or weakest converter.) Is anyone here saying it's a good thing that most consumer DAs sound lousy? No, but that's how it is. With digital you need good quality converters to compare it to analog, period. Is that a good thing or a bad thing? Well, it's probably bad, but whatever, that's the way it is right now. Hopefully that will change; I'm sure in a few years things will be at least a little better.
Post by: maarvold on September 25, 2005, 01:58:27 PM
| J.J. Blair wrote on Fri, 23 September 2005 09:29 | ||
Michael, what's so great about thinking outside the box if it's wrong? |
Isn't this kind of like saying "What's so great about Freedom Of Speech?" The reason I barely skim these types of topics anymore is because it just takes too much energy. I can't believe it, I come back 2 days later and there's 5 more pages. It's too much like a homework assignment.
BTW, the idea that 2 samples/cycle can accurately describe a waveform still strikes me as pretty 'outside the box'.
Maybe we should be concentrating all this energy on a more far-reaching topic, like: Why do broadcast journalists, and their superiors, think it's acceptable to editorialize?
Post by: J.J. Blair on September 25, 2005, 01:58:47 PM
Post by: CCC on September 25, 2005, 02:05:53 PM
| J.J. Blair wrote on Sun, 25 September 2005 18:58 |
Can we please stop saying "where the rubber hits the road"? I'm trying to forget about that god forsaken song. |
bukkake!
Post by: thedoc on September 25, 2005, 02:33:53 PM
hmmm, I wasn't aware that you were aware of my listening habits.
Once again, hmmmmmm...
Post by: timrob on September 25, 2005, 02:42:06 PM
| maarvold wrote on Sun, 25 September 2005 12:58 | ||||
Isn't this kind of like saying "What's so great about Freedom Of Speech?" The reason I barely skim these types of topics anymore is because it just takes too much energy. I can't believe it, I come back 2 days later and there's 5 more pages. It's too much like a homework assignment. BTW, the idea that 2 samples/cycle can accurately describe a waveform still strikes me as pretty 'outside the box'. Maybe we should be concentrating all this energy on a more far-reaching topic, like: Why do broadcast journalists, and their superiors, think it's acceptable to editorialize? |
Perhaps it could be considered "outside the box" if the question had never been raised. This stuff has been hashed out many times.
It is beating the decaying corpse of a dead horse. It is the kind of thinking that first year engineering students come up with.
I appreciate someone putting in the effort to understand things. But when they won't accept what is already known to be true then it just becomes infantile.
I knew a kid in high school who's father was a physics researcher. His father explained to him a few things about anti-matter and black holes. He then used this information to try to refute every subject in every class. For him 2+2=4, except when it encountered anit-matter.
These guys aren't interested in actually learning anything. They are just sitting around getting their jollies trying to get peoples hackles raised. I joined this forum because I thought the signal to noise ratio would be very high. After all, the George's header reads "Reason in Audio". It is just kind of silly when someone won't use some.
This thread is stuck in a loop. So there's not much point in reading any more. It's all been said over and over again.
Post by: Andy Simpson on September 25, 2005, 03:10:38 PM
| crm0922 wrote on Sun, 25 September 2005 02:22 |
Andy, please rephrase the question. Maybe I misunderstood. ... Chris |
Let's say we have a mic 10ft away from the source.
The source is making a 20k sinewave.
The converter is running at 2bit 44.1.
As the samples are taken and quantized we see the results.
I think that if we move the mic 1cm closer to the sound (or further), the results will not change.
Andy
Post by: Jon Hodgson on September 25, 2005, 03:26:18 PM
| andy_simpson wrote on Sun, 25 September 2005 20:10 | ||
Let's say we have a mic 10ft away from the source. The source is making a 20k sinewave. The converter is running at 2bit 44.1. As the samples are taken and quantized we see the results. I think that if we move the mic 1cm closer to the sound (or further), the results will not change. Andy |
2 bit?
Actually the results will change, even at 2 bits
Post by: dcollins on September 25, 2005, 03:39:24 PM
| andy_simpson wrote on Sun, 25 September 2005 12:10 |
Let's say we have a mic 10ft away from the source. The source is making a 20k sinewave. The converter is running at 2bit 44.1. As the samples are taken and quantized we see the results. I think that if we move the mic 1cm closer to the sound (or further), the results will not change. |
Good to see you've been paying attention here, Andy. You seem to really be getting it.
DC
Post by: dcollins on September 25, 2005, 03:42:59 PM
| maxdimario wrote on Sun, 25 September 2005 07:14 |
sure I can explain about the monitoring chain. |
Tell us about your system.
Max, are you familiar with them term "plonk?"
DC
Post by: Jon Hodgson on September 25, 2005, 03:49:18 PM
| dcollins wrote on Sun, 25 September 2005 20:39 | ||
Good to see you've been paying attention here, Andy. You seem to really be getting it. DC |
I think one of us may have misunderstood here...
I believe Andy is implying that the system will not be able to pick up a phase difference caused by moving the source 1cm, whereas analogue would
Post by: Bill Mueller on September 25, 2005, 05:29:31 PM
| Jon Hodgson wrote on Sun, 25 September 2005 15:49 | ||||
I think one of us may have misunderstood here... I believe Andy is implying that the system will not be able to pick up a phase difference caused by moving the source 1cm, whereas analogue would |
I'm starting to find this thread entertaining! I usually pride myself on not looking at the dead bodies when I pass an accident on the highway. But somehow I can't look away this time.
At one time, I had some sympathy for Max, Andy and Johnny, but this has been going on long enough that they have had time to read any number of books on digital audio and find that they are pissing in the wind. It also is getting interesting how normally genteel professionals, intent on sharing their hard fought experience and educating others, have begun to come apart at the seams with frustration.
As entertaining as it has become, I still suggest that this has gone on long enough. There are other, more rewarding topics to pursue. It is ok if Andy, Max or Johnny have the last word. It just means that they are talking to themselves. Everyone has done an exemplary job attempting to resolve this in a civilized way. Bravo.
Best Regards,
Bill
Post by: Phil on September 25, 2005, 06:02:13 PM
| Bill Mueller wrote on Sun, 25 September 2005 14:29 |
I'm starting to find this thread entertaining! I usually pride myself on not looking at the dead bodies when I pass an accident on the highway. But somehow I can't look away this time. |
Great analogy Bill -- thread gridlock! Is that a KCAL news chopper I hear overhead?
Phil
Post by: J.J. Blair on September 25, 2005, 06:32:46 PM
Post by: dcollins on September 25, 2005, 07:06:56 PM
| J.J. Blair wrote on Sun, 25 September 2005 15:32 |
Oh, the humanity. |
http://www.cartoonbank.com/product_details.asp?mscssid=HEKPN 8U2DMQP9HXJWWLR0E3M8FQS8E31&sitetype=1&did=4&sid =22230&whichpage=1&sortBy=popular&keyword=intern et§ion=cartoons
Post by: Andy Simpson on September 25, 2005, 07:46:47 PM
| Jon Hodgson wrote on Sun, 25 September 2005 20:49 | ||||
I think one of us may have misunderstood here... I believe Andy is implying that the system will not be able to pick up a phase difference caused by moving the source 1cm, whereas analogue would |
That is what I'm implying....thanks.
I picked 2bit to make it intuitive....for me!
Andy
Post by: tom eaton on September 25, 2005, 10:15:07 PM
-tom
Post by: Bob Olhsson on September 25, 2005, 10:53:02 PM
| maxdimario wrote on Sun, 25 September 2005 08:39 |
...I've never put any doubt into anything Bob O. says, including the part where he says that in the end analog still sounds better than digital. |
Post by: Ronny on September 26, 2005, 12:47:02 AM
I can't wait until they come out with analog cds. They'll sound so much better.
Post by: Johnny B on September 26, 2005, 12:47:20 AM
| Dan Feiszli wrote on Sun, 25 September 2005 08:53 |
What's all this fear garbage about? I don't think anybody posting here has shown any fear about anything... |
Of course not, everyone here is very brave, very open-minded, willing to look under every rock in order to advance the tecnology and improve the sound quality. But that kind of bravery may also pass for cowardice and really mask the state of denial.
The kind of bravery you suggest is here probably would not be the kind of bravery I'd want around in a good barroom brawl. I think some people are really chicken, chicken to even look at Next GEN very high speed chips, let alone listen to them or give them a fair chance to fight for market share.
I say, "Bring 'em on." Let's see how they actually perform when stressed in real world situations.
Post by: Ronny on September 26, 2005, 12:59:54 AM
| Johnny B wrote on Mon, 26 September 2005 00:47 | ||
Of course not, everyone here is very brave, very open-minded, willing to look under every rock in order to advance the tecnology and improve the sound quality. But that kind of bravery may also pass for cowardice and really mask the state of denial. The kind of bravery you suggest is here probably would not be the kind of bravery I'd want around in a good barroom brawl. I think some people are really chicken, chicken to even look at Next GEN very high speed chips, let alone listen to them or give them a fair chance to fight for market share. I say, "Bring 'em on." Let's see how they actually perform when stressed in real world situations. |
There are people on this forum that are beta'ing the highest speed chips that are in pre-market development, and you aren't one of them. Inscinuating that people at psw are chicken to try anything new is preposterous, many designers on these forums that are not only at the leading edge of technology, but they are writing it.
Generalizations and opinions from uneducated minds abound these days.
Post by: StudioRhythm on September 26, 2005, 01:26:08 AM
| Johnny B wrote on Sun, 25 September 2005 21:47 | ||
Of course not, everyone here is very brave, very open-minded, willing to look under every rock in order to advance the tecnology and improve the sound quality. But that kind of bravery may also pass for cowardice and really mask the state of denial. The kind of bravery you suggest is here probably would not be the kind of bravery I'd want around in a good barroom brawl. I think some people are really chicken, chicken to even look at Next GEN very high speed chips, let alone listen to them or give them a fair chance to fight for market share. I say, "Bring 'em on." Let's see how they actually perform when stressed in real world situations. |
So now you're calling people names? Wow, this is impressive.
At the risk of appearing chicken, I think this will be my last response to you. Do yourself a favor -- if you're serious about improving sound quality, learn something about digital audio first, and then you'll have some ideas about what to improve. Stop blindly following the marketing B.S. that says that all we need is faster sampling rates (oh, and by the way, you'll need a faster computer, and more storage, and new and improved software, new interfaces, and new converters -- and did we mention we can sell you those...?)Post by: Johnny B on September 26, 2005, 01:27:33 AM
| bobkatz wrote on Sun, 25 September 2005 17:03 |
So, what about the upsampled reproduction makes it sound better? That's where the research lies. I say it's in the filter design... And yes, you need a digital filter to do the anti-imaging. A good friend who is the designer for Algorithmix (Christoph Musialik) visited this weekend. His linear phase Algorithmix Red uses 80 bit floating point arithmetic and he says that to minimize the accumulation of errors and distortions you need those 80 bits. Certainly more than 64 float. I nominate the Algorithmix Red as the most transparent (also warm and pure) sounding digital equalizer on the planet. I know enough now about what it takes to make a transparent-sounding low pass filter to say that current pre-masked silicon chips don't cut it. |
Ok, now we are getting somewhere! 80-bit float. And better chips.
All this foward movement and having people show they have an open mind and a "seeking spirit" is how things like technology get advanced forward. That's all anyone can ask for, move the tech forward, and make it sound better. Perhaps one can also ask for it to be easier to integrate or for easier, seamless operation. I'd settle for the forward movement and better sound quality.
I talked about "fear" above, I have two fears or concerns:
1) That video will take center stage at the expense of sound quality
2) That sound quality will not advance as it should.
Personally, I'd much rather have a bad picture if it meant I could have better sound.
Pictures are cool, but without great sound, who cares? Not me anyway. Others may differ and assign more importance to pictures, I'm not one of those people tho'
Now let me ask a rhetorical question or two:
Do you advance the technology by defending things as they already exist?
Do you protect or defend sound from video by saying everything in digital sound as it now exists is just fine?
Post by: natpub on September 26, 2005, 01:33:05 AM
Post by: J.J. Blair on September 26, 2005, 01:52:41 AM
I actually think that not even analog could help this mess.
Post by: J.J. Blair on September 26, 2005, 01:56:36 AM
| Johnny B wrote on Sun, 25 September 2005 21:47 |
I say, "Bring 'em on." |
There's another person known for this statement. Simple minds think alike?
Post by: Johnny B on September 26, 2005, 01:58:20 AM
Ok, back on topic.
Emmmm. I don't think GM is afraid to look at or use high speed chips.
If anyone is going to AES, maybe someone will be kind enough to pull him aside and find out if he really meant to tell Lynn Fuston that PCM would have to be bumped up in speed to 384kHz to "catch up" with the sonics of DSD.
Perhaps a few probing questions could be posed if GM has the time to answer. The interviewer could get some details and a little clarification about that 384kHz and come back here and share it with others. I'm assuming there are no I.P., trade secrets, or NDA difficulties that would bar GM from helping all of us put in proper context or better understand his statement about 384kHz.
Post by: compasspnt on September 26, 2005, 02:10:32 AM
Post by: Johnny B on September 26, 2005, 02:27:37 AM
GM is a man with a lot of integrity. If he is not legally obligated to not share with us, I sense he will share some enlightening info with us, esp. if someone else will make it easy and do the post for him. I'm also assuming that the person who goes to AES and asks him about his statement, which seemingly calls for the adoption of 384kHz and apparently has some sound basis for doing so, will not be afraid of what GM might say or what they might learn.
After hearing from GM about 384kHz, maybe we really do want to "bring 'em on."
And Bob K. just said he likes 80-bit float and that the current chips are not really that hot in some respects.
Current chips not that hot, 80-bit float, GM calling for 384kHz PCM, looks to me like there just might be some significant room for digital improvement.
Post by: maxdimario on September 26, 2005, 03:53:01 AM
I'll try to finish this off.
I don't believe that there is a major difference between brand names of tape, other than the effect on frequency/saturation levels, for what I'm looking for as an ideal audio recorder.
I do believe that discrete machines such as the old ampex with the deep head gap or the early a80 machines do more of what I like to hear.
my opinions were formed on ampex 440 and studer a80 machines, as well as critical listening to records made in different periods of recording history.
It's probably agreed by most people that digital 'sounds like' the source more than analog, as I used to believe as well.
there is an element in music that I have grown sensitive to over the years of modifying and designing electronic equipment which I call time distortion.
this would be in analog network-induced phase distortion, NFB.
the erratic shifting and ringing of the harmonics to the point that they are SLIGHTLY misaligned creates a dullness which makes sounds less real.
this kind of distortion affects feel.
hence the best equipment to record rock'n'roll is the simple discrete stuff, because it is the most open-sounding, time-aligned (phase shift is not as much of a problem as phase distortion).
digital errors, whatever they may be, sound like this to me, and therefore I imagine that something is happening in the time domain.
96K sounds a little more stable, maybe because the filters work better, maybe because the higher you go the more of a chance there is that the imperfections of digital reproduction (the real ones not the theoretical ones).
I imagine that the quality of a system depends on the potential resolution in conjunction with the quality of engineering.
a poorly engineered hi-res converter sounds worse than a great lo-res.
what bothers me is when people start quoting texts, and theory as if they were the be-all and end-all of designing a good audio system, when it is obvious that what is lacking is the end result, which I perhaps erroneously put down to a 'balanced' system.
what I mean by balanced is that you find an equilibrium between the theoretical, what components you have at your disposition, and the end result.
Is it the filters? is any filter perfect enough to work 100% at the limit?
is it the sampling rate? could it be that there is some kind of distortion that is not covered by nyquist? or that comes as a side-effect of an overlooked aspect, which could be alleviated by a hi-res sampling rate?
I used to think you didn't need higher than 44.1 (edit), before I started listening.
THANK GOD THIS IS ONLY A POST...phew!
Post by: Jon Hodgson on September 26, 2005, 04:12:02 AM
| Johnny B wrote on Mon, 26 September 2005 07:27 |
Current chips not that hot, 80-bit float, GM calling for 384kHz PCM, looks to me like there just might be some significant room for digital improvement. |
you're like some kind of soundbyte king aren't you? Trawl through everything that's said and take the bits you can quote out of context to seemingly support your position.
1) Current chips not that hot, that's referring to the filter stage, not sampling rates, and yes people are already working on that. Why do you think Dan Lavry's converters sound better than a soundblaster? It's not sample rate or bit depth, and with each generation the ICs are generally getting better.
2) 80 bit float is for PROCESSING, not conversion, at conversion once you get to 24 bits you are so far down into the noise floor that the lower bits are random, 4 random bits or 40 random bits, it makes no difference. I'll have to look into whether 80 bits is really neccessary in most cases, I know I've run into problems at 32 bit float with some filters, but adjusting the topology improved things enormously. I also know from talking with one of them that when the Sony Oxford guys were porting their algorithms to the OXF R3 they had some major challenges since that uses 32 bit float... they managed it though (could be they're using double precision in places, I didn't ask for details).
Anyway, this doesn't require new chips, though new chips could let you process more channels for less money. Also, even if it is neccessary in some processing, it is not neccessary in all processing, using it for a simple volume control gains you nothing.
3) 384Khz? I'll tell you what, you show me proof that keeping the signal content up to 160 kHz (let's give those filters room to work) gives us more audible benefit than the negative effect of the extra noise created by clocking at those sorts of speeds (it's those pesky laws of physics getting in the way, what a shame we can't tell electrons to sit exactly still until we want them to move, and then to jump instantaneously to where we want them), then I'll be there, but I don't believe you can, so far the weight of evidence is against you... for a start if those frequencies were essential then your analogue recordings would be inadequate, and you've already stated how great they are.
Actually I don't know why I'm addressing this to you, this post is for those with open minds who may have been confused by what you posted.
Post by: Norwood on September 26, 2005, 04:37:30 AM
| maxdimario wrote on Mon, 26 September 2005 00:53 |
a poorly engineered hi-res converter sounds worse than a great hi-res. |
...good point.
Post by: maxdimario on September 26, 2005, 04:48:40 AM
Post by: tom eaton on September 26, 2005, 08:06:10 AM
"I used to think you didn't need higher than niquist, before I started listening."
Nyquist is a theorem. Nyquist is a statement about how digital audio works. What Nyquist do you need higher than?
This is really starting to hurt.
-tom
Post by: Bill Mueller on September 26, 2005, 08:38:22 AM
| Jon Hodgson wrote on Mon, 26 September 2005 04:12 | ||
you're like some kind of soundbyte king aren't you? Trawl through everything that's said and take the bits you can quote out of context to seemingly support your position. 1) Current chips not that hot, that's referring to the filter stage, not sampling rates, and yes people are already working on that. Why do you think Dan Lavry's converters sound better than a soundblaster? It's not sample rate or bit depth, and with each generation the ICs are generally getting better. 2) 80 bit float is for PROCESSING, not conversion, at conversion once you get to 24 bits you are so far down into the noise floor that the lower bits are random, 4 random bits or 40 random bits, it makes no difference. I'll have to look into whether 80 bits is really neccessary in most cases, I know I've run into problems at 32 bit float with some filters, but adjusting the topology improved things enormously. I also know from talking with one of them that when the Sony Oxford guys were porting their algorithms to the OXF R3 they had some major challenges since that uses 32 bit float... they managed it though (could be they're using double precision in places, I didn't ask for details). Anyway, this doesn't require new chips, though new chips could let you process more channels for less money. Also, even if it is neccessary in some processing, it is not neccessary in all processing, using it for a simple volume control gains you nothing. 3) 384Khz? I'll tell you what, you show me proof that keeping the signal content up to 160 kHz (let's give those filters room to work) gives us more audible benefit than the negative effect of the extra noise created by clocking at those sorts of speeds (it's those pesky laws of physics getting in the way, what a shame we can't tell electrons to sit exactly still until we want them to move, and then to jump instantaneously to where we want them), then I'll be there, but I don't believe you can, so far the weight of evidence is against you... for a start if those frequencies were essential then your analogue recordings would be inadequate, and you've already stated how great they are. Actually I don't know why I'm addressing this to you, this post is for those with open minds who may have been confused by what you posted. |
As I was reading this I was wondering why you were sending it to Johnny. From his posts above, he clearly does not see the relationship between noise and bandwidth in an analog circuit.
Best Regards,
Bill
Post by: Ronny on September 26, 2005, 08:58:46 AM
| TER wrote on Mon, 26 September 2005 08:06 |
If you had been listening HERE you would never have written something like: "I used to think you didn't need higher than niquist, before I started listening." Nyquist is a theorem. Nyquist is a statement about how digital audio works. What Nyquist do you need higher than? This is really starting to hurt. -tom |
It's as simple as what Nyquist captures everything that a 20k mic can capture? Any sampling rate above that simply does not apply, as these higher frequencies aren't recorded by the mic to any significant degree to begin with.
The confusion I think lies in them thinking that because you sample more times you get a more accurate representation. That's true from a very low res standpoint, but only until you reach Nyquist of the microphone max, the limit that humans can hear and the limitations of the playback speakers. Anything above that does not apply at the converter. They also forget that most 24 bit AD's, use 128x delta-sigma oversampling, which oversamples the signal to over 5 megahertz, so the initial capture before the decimation stage is very, very high oversampling, much higher than 384k which would capture 192k, but only if you had a mic that went that high and you still wouldn't be able to play back or hear those frequencies. A 400Hz fundamental will be at the 100th harmonic at 40k and the energy above 20k, way under .01%, that's using a ref mic that will go above the higher harmonics of an instrument, but the best tracking mics seldom go above 20k. They also don't seem to be aware of relative octave structure, for example a Les Paul or Strat guitar, the low A is 110Hz, next octaves are 220Hz and 440Hz and the highest A note that you play on the high E string 17th fret is 880Hz, next octave and out of range of the 6 string guitar tuned to concert pitch is 1.7k, next octave 3.5k. Now we are getting close to the highest note on an 88 key piano or the typical highest instruments high note in an orchestra, the piccolo at 4.2k. so the next octave above 3.5k is 7k (almost twice as high as the high C on a piccolo), than 14k than 28k than 56.3k, than 112k than 225k. You aren't even covering a full octave going from 56 to 96k, nor are you gaining another octave when you go from 96 to 192k. Octave frequencies double, so that on paper 192k may look good but in reality it's not giving you any more significant notes anyway, that's if you could capture them in the first place and if instruments play above the fundamental of 4.2k to begin with. Couple that with the Fletcher-Munson curve that reflects frequency perception in humans, even in real world live audio, our ears highly attenuate frequencies above 15k, so we've reached what we are going to gain capturing and playing back frequencies at around 30k. 44.k will do it and is actually a little overkill for the 20k mics, what the ear can perceive and the play back systems that we have.
Post by: tom eaton on September 26, 2005, 09:11:24 AM
As you stated, the fundamental of the note has nothing to do with its harmonic structure or the reach of its overtones. If all we heard were fundamentals, all we'd be hearing was sine waves. The harmonics of many, many instruments extend well beyond the 22k brickwall point of many converters. At what point do those partials become inaudible? Who knows...varies from person to person.
My point was that Nyquist is not a frequency as Max suggested. Nyquist is the math which determines the highest frequency that a given system can capture...and that can be ANY frequency. It just depends on the sampling rate, as Nyquist stated.
Saying something like "higher than Nyquist" has no meaning, because Nyquist is a formula.
-tom
Post by: tom eaton on September 26, 2005, 09:26:15 AM
The original post was about "what happens in between the samples" and we've beat that point to death.
-tom
Post by: Duardo on September 26, 2005, 10:44:23 AM
| Quote: |
BTW, the idea that 2 samples/cycle can accurately describe a waveform still strikes me as pretty 'outside the box'. |
Of course it's outside the box. It's completely nonintuitive. But it's been proven. And since it's been proven true, the original theory that started off this thread has been proven false. So pushing it really isn't thinking outside the box.
| Quote: |
Do you advance the technology by defending things as they already exist? |
No. You push for them to get better. But there's more than one way to go about doing that. Raising sampling rates isn't the only way, and in this case it isn't the best way.
-Duardo
Post by: crm0922 on September 26, 2005, 10:50:55 AM
| TER wrote on Mon, 26 September 2005 09:11 |
Ronny- Saying something like "higher than Nyquist" has no meaning, because Nyquist is a formula. |
It's an abbreviation for "the Nyquist frequency". From that I assumed he meant 44,100 Hz. You know, us science geeks like to shorten things. There are plenty of things to dog on the "ear-plugging-la-la-la-i-can't-hear-you" guys around here about than their use of some shorthand.
Dan L and the other dudes down the hall refer to the Nyquist rate as "Nyquist" quite often.
Moving on...
I thought whoever was blabbering about other people "quoting texts" sounded remarkably ignorant.
We are quoting science, they are making biased judgements with no properly designed comparison structure, and trying to justify it with theories that are factually incorrect. And continuing to do so when proven wrong over and over again.
Oh, and JJ, re: that band in your live room. I feel your pain. Pretty much on a daily basis. Too bad a good analog 2" doesn't help make bands know how to play guitar or something.
Chris
Post by: maxdimario on September 26, 2005, 11:13:00 AM
| Quote: |
Nyquist is a theorem. Nyquist is a statement about how digital audio works. What Nyquist do you need higher than? |
You're right.
I've got a bad cold and I can't concentrate.
I meant 44.1 as opposed to high res.
Post by: Johnny B on September 26, 2005, 01:17:56 PM
| Johnny B wrote on Mon, 26 September 2005 06:58 |
Emmmm. I don't think GM is afraid to look at or use high speed chips. If anyone is going to AES, maybe someone will be kind enough to pull him aside and find out if he really meant to tell Lynn Fuston that PCM would have to be bumped up in speed to 384kHz to "catch up" with the sonics of DSD. Perhaps a few probing questions could be posed if GM has the time to answer. The interviewer could get some details and a little clarification about that 384kHz and come back here and share it with others. I'm assuming there are no I.P., trade secrets, or NDA difficulties that would bar GM from helping all of us put in proper context or better understand his statement about 384kHz. |
Guess a few of you missed the part about trying to get some claifification and attempting to put the 384k into the proper conext and help us understand why that might be.
God, are there any open minds here?
What's wrong with getting a different view or a taking a fresh look at the problems?
That's not a crime in my book. Asking questions, even seemingly dumb questions is never a crime, and challenging past technology is also always Ok. Without those challenenges, nothing would ever move forward. Sheeesh.
Here, we just want to know what GM might have meant. Maybe Lynn could help clarify the 384k statement some.
Post by: J.J. Blair on September 26, 2005, 01:37:59 PM
| maxdimario wrote on Mon, 26 September 2005 00:53 |
I don't believe that there is a major difference between brand names of tape, other than the effect on frequency/saturation levels, for what I'm looking for as an ideal audio recorder. |
Frequency/saturation levels are not major differences? How about inherent noise levels? Max, if you can;t hear the difference between 1.5db over bias and 2.5 db over bias, you have no business judging the quality of digital audio.
Since nobody asked me what Hutch mentioned was the reason why digital doesn't sound like analog, I'm going to throw it out there. If true, and if I am quoting him correctly, this explains my perceptions that analog sounds more '3 dimensional' to me than digital. These are my terms not his, but what I understand as his idea, and it has nothing to do with sample rates, quantizing or any of that shit. Basically, analog has the ability to capture the immediate impact of transients, like it can go from 0-60 immediately with the hit of a stick on a snare or a pick striking a string. For whatever technical reasons, digital actually has to ramp up from that 0-60, so the leading edge of the transient has lost its immediacy.
Now, can somebody who really knows what the fuck they are talking about(ie. not Jonny, Max or Andy), who has built AD converters, etc. please comment on this?
Post by: J.J. Blair on September 26, 2005, 01:49:49 PM
Post by: Jon Hodgson on September 26, 2005, 02:12:20 PM
| J.J. Blair wrote on Mon, 26 September 2005 18:37 | ||
Frequency/saturation levels are not major differences? How about inherent noise levels? Max, if you can;t hear the difference between 1.5db over bias and 2.5 db over bias, you have no business judging the quality of digital audio. Since nobody asked me what Hutch mentioned was the reason why digital doesn't sound like analog, I'm going to throw it out there. If true, and if I am quoting him correctly, this explains my perceptions that analog sounds more '3 dimensional' to me than digital. These are my terms not his, but what I understand as his idea, and it has nothing to do with sample rates, quantizing or any of that shit. Basically, analog has the ability to capture the immediate impact of transients, like it can go from 0-60 immediately with the hit of a stick on a snare or a pick striking a string. For whatever technical reasons, digital actually has to ramp up from that 0-60, so the leading edge of the transient has lost its immediacy. Now, can somebody who really knows what the fuck they are talking about(ie. not Jonny, Max or Andy), who has built AD converters, etc. please comment on this? |
Actually what you appear to be describing has everything to do with sample rate. A vertical step in the signal requires infinite bandwidth, and therefore infinite sample frequency. But analogue circuits and tape don't have infinite bandwidth, so they cannot step instantaneously as you appear to be describing (please correct me if I misunderstood you).
However this theory does raise one interesting question, which is whether tape responds differently to the initial transient than the steady state response might otherwise indicate (this could be either on record or playback). Any electronics engineers around here who really know analogue tape machines?
Post by: StudioRhythm on September 26, 2005, 02:14:31 PM
| J.J. Blair wrote on Mon, 26 September 2005 10:37 |
Since nobody asked me what Hutch mentioned was the reason why digital doesn't sound like analog, I'm going to throw it out there. If true, and if I am quoting him correctly, this explains my perceptions that analog sounds more '3 dimensional' to me than digital. These are my terms not his, but what I understand as his idea, and it has nothing to do with sample rates, quantizing or any of that shit. Basically, analog has the ability to capture the immediate impact of transients, like it can go from 0-60 immediately with the hit of a stick on a snare or a pick striking a string. For whatever technical reasons, digital actually has to ramp up from that 0-60, so the leading edge of the transient has lost its immediacy. |
JJ, an "immediate impact" means something with infinite bandwidth, which neither analog nor digital have, and thankfully neither do our ears. If a transient occurs immediately, that would look like a straight vertical line on a frequency response graph (or close to a straight line, anyway.) Anyway, such a straight line would contain frequency content well above 20kHz, well above 96kHz, well above 192 kHz, well above, uh, 1000kHz... you get the picture.
If one were to take that "immediate impact" of this transient and apply some kind of filter to it -- for sake of argument let's use a perfect filter that creates no phase distortion, and has complete attenuation at its stop point, and let's set that at 40kHz for fun, so the people claiming we hear well above 20kHz are satisfied.
After that filter is applied, the straight vertical line is gone, since that immediate impact is purely the product of super high-frequency content, and what you're left with "ramps up."
This is what both digital and analog recording do, albeit in different ways, but the end result is that neither gives you that immediate impact (aka super-high frequency content.)
Is it non-intuitive? Yes. Does it take some serious thought to grasp? Yes. Does this mean that time and frequency are basically the same thing? Yes. Will someone who has actually built converters and understands digital audio more completely than I do agree with and, if necessary, elaborate on my description here? Hopefully...
Post by: four on September 26, 2005, 02:16:46 PM
| Quote: |
Since nobody asked me what Hutch mentioned was the reason why digital doesn't sound like analog, I'm going to throw it out there. If true, and if I am quoting him correctly, this explains my perceptions that analog sounds more '3 dimensional' to me than digital. These are my terms not his, but what I understand as his idea, and it has nothing to do with sample rates, quantizing or any of that shit. Basically, analog has the ability to capture the immediate impact of transients, like it can go from 0-60 immediately with the hit of a stick on a snare or a pick striking a string. For whatever technical reasons, digital actually has to ramp up from that 0-60, so the leading edge of the transient has lost its immediacy. |
JJ-
What that does have to do with is frequency response. The leading edge of the transient represents a high frequency component to the signal. If it happened "immediately", then no band-limited system on any sort could capture the transient exactly. Of course, it doesn't really happen immediately - the object that creates the sound can only begin to vibrate so fast, and the medium that carries the wave can only change so fast - and this limitation is the frequency response of the system. A digital system will round this signal off to a shape that is limited to it's bandwidth... and so will an analog system.
Think of it like this: take the first part of that transient waveform - from 0-60 - and stop where is reaches it's peak. Now make a periodic waveform of that. If the frequency of this signal is above the range of the system, then that transient cannot be captured, but the lower frequencies (of the original signal) still will, and that will result in a band-limited transient.
-Tom
Post by: four on September 26, 2005, 02:19:46 PM
-Tom
Post by: Johnny B on September 26, 2005, 02:21:28 PM
You are talking about filters, speeds, bit depth, more efficient thruput, possibly new methods and new formats. Those are the issues.
If GM thinks there are some good reasons to go with 384kHz, there are many who would like to know why he feels that way. Please don't try to shut those questions off because failing to ask questions, failing to examine other views, keeps us all in the dark and operating blind. We could speculate why GM may have said that that he feels PCM has to have its speed bumped 384k to "catch up" with the sonics of DSD, but that may not answer some of people's questions in a satisfying manner.
I'm very curious about GM's purported statement about 384kHz, aren't you?
Post by: dcollins on September 26, 2005, 03:36:32 PM
| J.J. Blair wrote on Mon, 26 September 2005 10:37 |
Now, can somebody who really knows what the fuck they are talking about(ie. not Jonny, Max or Andy), who has built AD converters, etc. please comment on this? |
I assume he's talking about "pre-ringing" that comes from FIR filter structures. It rings before the impulse but "How Do it Know?"
This has been mildly controversial as it never occurs in nature, but I'm not sure that's why Max et. al. don't like digital......
There is a paper at the next AES from Wolfson that investigates this topic and compares minimum phase (rings only after the impulse) to linear phase.
Personally, I don't have a single piece of linear phase processing gear in the studio, although guys that live with digital EQ's seem to like them.
Causally,
DC
Post by: maxdimario on September 26, 2005, 04:53:20 PM
Post by: Johnny B on September 26, 2005, 05:59:02 PM
Do you have a link to the Wolfson paper, or maybe you could give us the gist of it.
Post by: bblackwood on September 26, 2005, 06:21:36 PM
| Johnny B wrote on Mon, 26 September 2005 16:59 |
Do you have a link to the Wolfson paper, or maybe you could give us the gist of it. |
Ahem, the next AES...
Post by: JamSync on September 26, 2005, 06:25:31 PM
| Johnny B wrote on Mon, 26 September 2005 22:59 |
Dave, Do you have a link to the Wolfson paper, or maybe you could give us the gist of it. |
After AES, you can order the CD with the papers from AES.org. I'm going to pick one up, since I don't have time to go to the papers this year.
Post by: Johnny B on September 26, 2005, 06:31:51 PM
Oh, so does Dave have some pre-release inside dope?
Do we have to wait for clues, or can some further light be shed right now without too much difficulty.
Probably a couple of sentences would be enough...just enough to get an idea of what Wolfson might be up to...
Post by: JamSync on September 26, 2005, 07:02:55 PM
| Johnny B wrote on Mon, 26 September 2005 23:31 |
Oh, so does Dave have some pre-release inside dope? Do we have to wait for clues, or can some further light be shed right now without too much difficulty. Probably a couple of sentences would be enough...just enough to get an idea of what Wolfson might be up to... |
AES sent out a brochure to members about the upcoming papers. I tossed it already. It may be somewhere on the website.
Hmm...here it is. I love Google.
http://www.aes.org/events/119/papers/session.cfm?code=P10
"P10-6 An Ultra High Performance DAC with Controlled Time-Domain Response?Paul Lesso, Anthony Magrath, Wolfson Microelectronics - Edinburgh, Scotland, UK
This paper describes the design of an ultra-high performance stereo digital-to-analog converter (DAC) employing advanced digital filtering techniques. Recently there has been a renewed interest in the time-domain properties of digital filters used for interpolation and decimation. Linear phase FIR filters, which have proliferated digital filter design for the last two decades, have the undesirable properties of pre-ringing and high group delay. Conversely, minimum phase filters, which offer lower levels of pre-ringing, do not have a uniform phase response. This paper describes the trade-offs in the design of filters with controlled pre-ringing, coupled with desirable phase and magnitude characteristics. The paper also describes architectural choices in the implementation of the DAC signal processing chain, required to achieve commensurate analog performance.
Convention Paper 6577"
Note: it's a DAC, not an ADC. I think most of the brouhaha here has been about the capture side, not the repro side.
Post by: JamSync on September 26, 2005, 07:23:21 PM
| J.J. Blair wrote on Mon, 26 September 2005 18:49 |
All this talk about higher sampling rates has missed a very important point: Storage. Where are you going to come up with the drive space and drive speed? I've heard complaints of massive dropouts at 192. I can only imagine the problems at higher sampling rates. Personally I'd rather have a lower res recording with less errors. |
J.J,
The bitrate and storage of uncompressed video makes audio look like a breeze. Yes, there are issues for multitrack, but frankly, I haven't seen many insurmountable ones and I've worked with 192 on three different converters with several different DAWs (and I'm scheduled to take delivery on a fourth converter for review).
By far the 192 kHz archival format is sought after by classical and jazz recordists who are most interested in stereo, surround, or modified surround with accent mikes, not 128-track pop and rock types or movie soundtracks, so at the moment, we're not really addressing problems in that area.
Post by: Johnny B on September 26, 2005, 07:25:36 PM
I'm not sure you might not get even better implementions of filters in the future with the higher speed 'verters...I'm never sure about anything tho'...Ha Ha
Here's a link to the Wolfson ADC product page which shows at least 3 192kHz products. Naturally, they have slower ones too.
http://www.wolfson.co.uk/products/digital_audio/adcs/
And here Wolfson has 7 or so Codec products at 192kHz.
http://www.wolfson.co.uk/products/digital_audio/codecs/
There's a different page for the DACs.
When I've googled the "384kHz," most of the stuff that comes back appears to be about upsampling.
I'm very curious about the 384kHz, very curious.
Post by: Johnny B on September 26, 2005, 07:42:56 PM
"Recent personal research of Dr. Ferreira on audio compression has lead to innovative methods of signal analysis, decomposition and synthesis in the frequency domain, paving the way to dedicated and efficient coding of separate audio objects such as harmonic sounds, transients and stationary noise. In addition to audio compression application, this approach makes possible new functionalities such as semantic scalability and access to compressed audio material. The new approach also allows for "low-delay" encoding and lower/symmetric encoder-decoder complexity, providing new ways for real-time, two-way, high-quality audio communication."
A cursory exam of some of the material makes it seem like the sound quality is almost as good as CD...However, some of that research might lead to something else...something better than CD.
Post by: J.J. Blair on September 26, 2005, 07:56:27 PM
As far as KK mentioning video storage, video has the benefit of using RAID, which audio doesn't. Video doesn't have to deal with drop outs in the same manner either, does it? I was speaking of the feasibilty of multitracking in particluar. I realize that we can do whatever for 2 track, but I don't see a point in recording the basic tracks at a lower rate than the final 2 track product.
Post by: Jon Hodgson on September 26, 2005, 08:24:06 PM
| J.J. Blair wrote on Tue, 27 September 2005 00:56 |
Dave, if I receall, Hutch mentioned the term 'pre-ringing'. You hit the nail on the head. But can somebody tell me how pre-ringing can be mitigated by infinite bandwidth or does bandwidth even have anything to do with pre-ringing? Or is it merely a filter dilemma? If analog also suffers from finite bandwidth, why do we perceive less of the ramp up from analog than from digital? I guess what I'm saying is, is pre-ringing purely a bandwidth issue or a filter issue? If somebody could clear this up for me I'd appreciate it. As far as KK mentioning video storage, video has the benefit of using RAID, which audio doesn't. I was speaking of the feasibilty of multitracking in particluar. I realize that we can do whatever for 2 track, but I don't see a point in recording the basic tracks at a lower rate than the final 2 track product. |
ok, in simplified terms...
any filter with more than one pole is going to have some ringing (even if it is so low as to be imperceptable). In analogue filtering the filter cannot predict what is going to happen, so the ringing always happens after the step that causes it.
However in digital filtering you can "predict" what is going to happen because you're working with a delay line. There are advantages to doing this as regards the phase response of the filter, but one of the effects is to move some of the ringing before the transient that causes it.
The debate is whether the possible negative effects of this "pre ringing" are worse than the positive effects of the linear phase.
Post by: rankus on September 26, 2005, 08:35:39 PM
Briliant explanation Jon.! Thank you....
even I can understand now
Post by: JamSync on September 26, 2005, 08:55:40 PM
| J.J. Blair wrote on Tue, 27 September 2005 00:56 |
Dave, if I receall, Hutch mentioned the term 'pre-ringing'. You hit the nail on the head. But can somebody tell me how pre-ringing can be mitigated by infinite bandwidth or does bandwidth even have anything to do with pre-ringing? Or is it merely a filter dilemma? If analog also suffers from finite bandwidth, why do we perceive less of the ramp up from analog than from digital? I guess what I'm saying is, is pre-ringing purely a bandwidth issue or a filter issue? If somebody could clear this up for me I'd appreciate it. As far as KK mentioning video storage, video has the benefit of using RAID, which audio doesn't. Video doesn't have to deal with drop outs in the same manner either, does it? I was speaking of the feasibilty of multitracking in particluar. I realize that we can do whatever for 2 track, but I don't see a point in recording the basic tracks at a lower rate than the final 2 track product. |
I use Medea audio and video RAIDs here. I've been using a RAID for my audio for several years. Just because PT doesn't recognize RAIDs with ATTO cards doesn't mean that hardware RAIDs are not used every day for audio. I built the ATTO RAID for testing video with PT years ago, but it is put to shame by the Medea stuff.
Digital video and audio don't have to deal with dropouts (which to me means "tape formulation instability"), but both have to deal with error correction.
I think the most tracks I've done at 192 are 10 or 12, but as in earlier days, there's nothing stopping people from synchronizing multiple DAWs from a central clock. Like I said, it isn't insurmountable.
Unless you're dealing with live music that isn't overdubbed and you're capturing a real soundstage, either stereo or surround, I'd say the benefits of multitracking with huge numbers of tracks at 192 are nil since you will induce phase cancellation that will disrupt the soundstage. Your typical pop/rock/country album with massive auto-tune dosages is NOT going to show any benefit from being tracked at 192, even if this extraordinary thread were to resolve itself in favor of 192 kHz A/D converters.
Again, are you guys talking about ringing in an A/D or D/A situation? You can't just posit "filter" and "ringing" without knowing which side of "reality" you're on.
Post by: dcollins on September 26, 2005, 09:11:17 PM
| Johnny B wrote on Mon, 26 September 2005 16:42 |
A cursory exam of some of the material makes it seem like the sound quality is almost as good as CD...However, some of that research might lead to something else...something better than CD. |
Why don't you give Pohlmans "Principles of Digital Audio" a cursory exam?
We'll see you in a couple weeks...............
DC
Post by: dcollins on September 26, 2005, 09:16:37 PM
| Jon Hodgson wrote on Mon, 26 September 2005 17:24 |
The debate is whether the possible negative effects of this "pre ringing" are worse than the positive effects of the linear phase. |
In the olden days, before oversampling, where we had 9 or more pole Cauer filters and the HF group delay was off-the-map, we had the late great Deane Jensen model the LPF with his Comtran program and design an all-pass filter that would basically delay the LF to meet up with the HF.
I built one with 990's, circa 1986, and sure enough you could turn the tweaker and make the ringing "symetrical."
Iirc we used it on one record, then moved on to the real problems of the day......
Like sample/hold's, and nonlinear conversion.
DC
Post by: tom eaton on September 26, 2005, 09:22:28 PM
-tom
Post by: Johnny B on September 26, 2005, 09:40:34 PM
Some people claim that nothing is lost, but
people might be well advised to judge for themsleves by listening.
Apparently, Sony does a similar routine as well.
Some people feel that DSD retains good timing info, but at the cost of less HF
'equivalent bit' resolution and mega ultrasonic noise. This is based on both
theory and their extensive listening. Good PCM may have plenty of HF resolution but has a tendency to "time smear" at lower sample rates. And this is due to the FIR
filters which can be relaxed at higher rates. How high is 'good enough'?
That is something I doubt that anybody knows for sure at this time.
Can future FIR filters be 'fixed' and avoid the problem? Maybe/probably someday. And if that is the case, then the issue with sample rate may become less of a concern. Maybe not tho'...it's hard to tell the future.
Do any of you plan to support DXD? Buy some new gear and so on?
Oh, and one of my little birdies mentioned that it would not be surprising if Wolfson made an announcement at AES of some kind of "break-thru" technology.
No promises however. Have to wait and see.
Post by: dcollins on September 26, 2005, 09:56:23 PM
| TER wrote on Mon, 26 September 2005 18:22 |
He didn't go for "Principles" when I suggested it on page three of this thread...in fact I think someone got upset that I was pointing people towards books for information. This place can be strange. -tom |
I'm reminded of the beginning of "Animal House," long tracking shot up the lawn of university -- close-up on the founders statue -- tilt down to the plaque that reads:
"Knowledge is Good"
But who needs books when we have the Internet?
DC
Post by: Johnny B on September 26, 2005, 10:02:01 PM
Post by: Bill Mueller on September 26, 2005, 10:34:26 PM
| J.J. Blair wrote on Mon, 26 September 2005 13:37 | ||
Frequency/saturation levels are not major differences? How about inherent noise levels? Max, if you can;t hear the difference between 1.5db over bias and 2.5 db over bias, you have no business judging the quality of digital audio. Since nobody asked me what Hutch mentioned was the reason why digital doesn't sound like analog, I'm going to throw it out there. If true, and if I am quoting him correctly, this explains my perceptions that analog sounds more '3 dimensional' to me than digital. These are my terms not his, but what I understand as his idea, and it has nothing to do with sample rates, quantizing or any of that shit. Basically, analog has the ability to capture the immediate impact of transients, like it can go from 0-60 immediately with the hit of a stick on a snare or a pick striking a string. For whatever technical reasons, digital actually has to ramp up from that 0-60, so the leading edge of the transient has lost its immediacy. Now, can somebody who really knows what the fuck they are talking about(ie. not Jonny, Max or Andy), who has built AD converters, etc. please comment on this? |
JJ, it is interesting that you experience this difference between analog and digital. I have always experienced it the other way. I had to try and re-eq my analog mixes to bring out the "ping" on the cymbals two weeks after the session, where digi (Sony 3324) retained that transient quality permanently despite having hundreds of degrees of phase shift at 20K (filters). Today I don't even give transients a thought. If I want it sharp, I make it sharp and it stays sharp.
An easy way to illustrate your question (if I understand you correctly) is to record a high amplitude square wave on analog and digital and then view them both on a scope. I believe you will see a much "square-er" wave from the digital recorder, especially if the signal level is above -20db. It was George who taught me to record drums at -20 to retain transients. I had never seen anyone do that before him. Of course everyone has their own way. I engineered for Bob Ezrin once and he had a vintage U47 one inch from a snare drum with the meter pegging solid on the A80.
Max's inability to discern the difference between Scotch 250, Ampex 456, Quantegy and Agfa 468 is just sad. From that statement I don't believe he has actually recorded on analog multitrack. Personally, Scotch 250 was always my favorite for rock. It had a punch that I could never get out of either Agfa or Ampex tape. However Agfa was great tape for jazz and clasical music because of it's noise character, which always sounded less "hissy" to me. I was never a fan of Ampex/Quantegy even though I burned through truck loads of it. It was cheaper and more available but the quality control left lots to be desired. In the end the slitting was very bad. When Quantegy took over, I think they rebuilt their machines because the slitting improved alot. The other thing. All my Scotch and Agfa tapes are still in good shape, even after thirty years. The Ampex is all a mess.
Best Regards,
Bill
Post by: CCC on September 26, 2005, 10:49:35 PM
| Bill Mueller wrote on Tue, 27 September 2005 03:34 | ||
I don't believe he has actually recorded on analog multitrack. |
News flash.
Post by: Johnny B on September 26, 2005, 11:11:20 PM
| Johnny B wrote on Tue, 27 September 2005 02:40 |
Is it be possible that GM's statement about PCM having to have a speed increase to 384kHz to "catch up" with the sonics of DSD had to do with Pyramix DXD? Where they convert DSD to 8-bit 384k for editing and processing. Some people claim that nothing is lost, but people might be well advised to judge for themsleves by listening. Apparently, Sony does a similar routine as well. Some people feel that DSD retains good timing info, but at the cost of less HF 'equivalent bit' resolution and mega ultrasonic noise. This is based on both theory and their extensive listening. Good PCM may have plenty of HF resolution but has a tendency to "time smear" at lower sample rates. And this is due to the FIR filters which can be relaxed at higher rates. How high is 'good enough'? That is something I seriously doubt that anybody knows for sure at this time. Can future FIR filters be 'fixed' and avoid the problem? Maybe/probably someday. And if that is the case, then the issue with sample rate may become less of a concern. Maybe not tho'...it's hard to tell the future. Do any of you plan to support DXD? Buy some new gear and so on? Oh, and one of my little birdies mentioned that it would not be surprising if Wolfson made an announcement at AES of some kind of "break-thru" technology. No promises however. Have to wait and see. |
Thoughts on this? Not so much the part about Wolfson, more interested in comments about the other parts of the quote.
Post by: Ronny on September 26, 2005, 11:18:20 PM
Same here Bill.
If you think that annie retains fast transients better than annie, after you digitize it, just zoom in on a spot where the drummer accidentally hits the snare mic and compare it with a digi track where the same mistake is made. Tape saturation may have more to do with the transients being lessened, than attack being profound and the hotter you go to tape the more the natural compression from the tape. All snare hits, kick hits and cybmal hits ramp up naturally and I've never seen the natural ramp up, shorter than the ramp up from the medium being digital.
DC, didn't they whip audible pre-ringing about 15 years ago?
Post by: timrob on September 26, 2005, 11:27:04 PM
| Bill Mueller wrote on Mon, 26 September 2005 21:34 |
The other thing. All my Scotch and Agfa tapes are still in good shape, even after thirty years. The Ampex is all a mess. |
I always thought it was interesting that Scotch and Agfa exhibited the opposite problems from Ampex when it comes to long term storage.
Scotch and Agfa to some degree would dry out and the Oxide would literally fall off the backing. (I guess that would be pre 250 day.IIRC 206 was the worst) Whereas Ampex turns to goo.
Either way it is a PITA to get a decent transfer. Sometimes impossible.
But, I don't know how you could expect to get the feeling of the music without the goo.
Post by: compasspnt on September 26, 2005, 11:30:38 PM
| timrob wrote on Mon, 26 September 2005 23:27 |
But, I don't know how you could expect to get the feeling of the music without the goo. |
Not to mention the emotion.
Post by: timrob on September 26, 2005, 11:33:39 PM
| compasspnt wrote on Mon, 26 September 2005 22:30 | ||
Not to mention the emotion. |
Well, you know it really glues the track together.
Post by: JamSync on September 26, 2005, 11:57:17 PM
| dcollins wrote on Tue, 27 September 2005 02:56 | ||
I'm reminded of the beginning of "Animal House," long tracking shot up the lawn of university -- close-up on the founders statue -- tilt down to the plaque that reads: "Knowledge is Good" But who needs books when we have the Internet? DC |
Well, the internet *does* provide a fairly broad library...
http://www.digitalaudio.dk/technical_papers/Tradeoff%20of%20 192%20kHz.PDF
http://www.dcsltd.co.uk/technical_papers/effects.pdf
http://www.dcsltd.co.uk/technical_papers/aes97ny.pdf
Post by: J.J. Blair on September 27, 2005, 02:29:49 AM
Bill, it was particularly on upright bass and electric guitars that I noticed the 3D-ishness more with analog. I didn't notice it had been missing until returning to analog after a long period of digital.
I also tend to peg the needle when recording snare in analog, simply because I'm looking for tape compression, rather than a realistic response. I'll set my kick and snare settings while listening to the repro head. I'll try the -20 thing on snare next time for digital though. I take it you have to make up the gain?
BTW, I'm not concerned about altering a square wave with tape. Transformers and other types of circuits will alter the square wave on the oscilloscope, yet they sound better to my ears. I'm not after accurate reproduction, as much as capturing a musical sound that I can work with when it's time to mix.
Post by: maxdimario on September 27, 2005, 06:07:15 AM
| Quote: |
Max's inability to discern the difference between Scotch 250, Ampex 456, Quantegy and Agfa 468 is just sad. From that statement I don't believe he has actually recorded on analog multitrack |
jesus, I can't leave without somebody twisting words around, can I?
I said that they all are similar in the sense that they all have that analog effect.
and of course I know about bias, in fact I rebuilt and set up many tape recorders, beginning with cassetes when I was 13, a teac 3340 when I was 16, tons of studers and technics at the radio I used to work at, an m-10 telefunken 8-track that I sold when I almost broke my knee lifting it up (that machine had potential) an ampex 440, and studer a 80 machines.
I grew up before digital even existed at a low price.
now since I am back.......I guess those transients that ramp up quickly can define the attack of a drum hit PRECISELY in time...uh.. right?
Post by: maxdimario on September 27, 2005, 06:54:25 AM
It also seems like I'm not the only one around here that talks about things they don't know about for sure with conviction, huh? otherwise you wouldn't be making such claims about my past.
time=feel
Post by: Andy Simpson on September 27, 2005, 07:21:50 AM
| TER wrote on Mon, 26 September 2005 03:15 |
Ooh! Wait...I posted the phase accuracy specs for the Benchmark DAC-1 on PAGE 2 of this very thread. Come on! -tom |
Indeed you did.
But with my 2bit example, I'm trying to get at how the bit depth relates to the sample rate.
Could I increase the phase accuracy of my 2bit recording by sampling at 192 instead of 44.1?
Could I say that a 192/24bit recording is the equivalent of 44.1/96bit in terms of spatial timing (and in terms of pure information) - for frequencies below ~20k - ?
Maybe we like the sound of 192/24 better because it has the equivalent spatial accuracy of 44.1/96bit.........?
Does this make any sense to anyone else?
Andy
Post by: Jon Hodgson on September 27, 2005, 08:07:23 AM
| andy_simpson wrote on Tue, 27 September 2005 12:21 | ||
Indeed you did. But with my 2bit example, I'm trying to get at how the bit depth relates to the sample rate. Could I increase the phase accuracy of my 2bit recording by sampling at 192 instead of 44.1? Could I say that a 192/24bit recording is the equivalent of 44.1/96bit in terms of spatial timing (and in terms of pure information) - for frequencies below ~20k - ? Maybe we like the sound of 192/24 better because it has the equivalent spatial accuracy of 44.1/96bit.........? Does this make any sense to anyone else? Andy |
If your system was perfect in every way, and the only noise in the system was injected intentionally due to dithering, then by sampling at 192kHz and converting down to 48kHs you could gain 2 bits worth of resolution.
However your system is not perfect, and as you sample faster your samples become less accurate, so you don't get that gain, in fact you're almost certainly going to get a reduction in accuracy. In addition 24 bits or 26 bits of sampling accuracy makes no difference since you're so far down into the noise floor that those bits are random. You'd get the same audible effect with a random number generator.
Post by: Andy Simpson on September 27, 2005, 08:50:21 AM
| Jon Hodgson wrote on Tue, 27 September 2005 13:07 | ||||
If your system was perfect in every way, and the only noise in the system was injected intentionally due to dithering, then by sampling at 192kHz and converting down to 48kHs you could gain 2 bits worth of resolution. However your system is not perfect, and as you sample faster your samples become less accurate, so you don't get that gain, in fact you're almost certainly going to get a reduction in accuracy. In addition 24 bits or 26 bits of sampling accuracy makes no difference since you're so far down into the noise floor that those bits are random. You'd get the same audible effect with a random number generator. |
I'm glad what I said makes sense, in theory.....
What I'm saying is that IF you could do 96bit 44.1 with accuracy, it would have the equivalent spatial resolution of 192/24.
BUT, since we can't accurately do 44.1/96bit, we are better off doing 192/24 instead.
I just want to propose that 192/24bit has the spatial resolution of a theoretical 44.1/96bit - to explain why we might prefer 192/24 despite the fact that we can't hear above ~20k.
Andy
Post by: Bill Mueller on September 27, 2005, 09:00:25 AM
| J.J. Blair wrote on Tue, 27 September 2005 02:29 |
KK, we're talking A/D. Bill, it was particularly on upright bass and electric guitars that I noticed the 3D-ishness more with analog. I didn't notice it had been missing until returning to analog after a long period of digital. I also tend to peg the needle when recording snare in analog, simply because I'm looking for tape compression, rather than a realistic response. I'll set my kick and snare settings while listening to the repro head. I'll try the -20 thing on snare next time for digital though. I take it you have to make up the gain? BTW, I'm not concerned about altering a square wave with tape. Transformers and other types of circuits will alter the square wave on the oscilloscope, yet they sound better to my ears. I'm not after accurate reproduction, as much as capturing a musical sound that I can work with when it's time to mix. |
JJ,
Actually I was speaking of analog when recording at -20db, just the opposite of most folks. I noticed that George would record certain instruments at very different deflections of the VU meter. Guitars were up around OVU but drums were barely moving the meters. This was before I had any idea what a peak meter was and I thought it was strange. George was actually recording, according to his own understanding of peak levels no matter what the VU was showing. This was 1974-5 but his stuff was so much cleaner and hi fidelity than anything I had ever heard that it made an impression.
I have always wondered about the leading edge overshoot that you see on some square waves off tape. I wonder if those spikes are the "observed" transients that you are sensing? Anyone have an opinion?
Best Regards,
Bill
Post by: Jon Hodgson on September 27, 2005, 09:23:20 AM
| andy_simpson wrote on Tue, 27 September 2005 13:50 |
I'm glad what I said makes sense, in theory..... What I'm saying is that IF you could do 96bit 44.1 with accuracy, it would have the equivalent spatial resolution of 192/24. BUT, since we can't accurately do 44.1/96bit, we are better off doing 192/24 instead. I just want to propose that 192/24bit has the spatial resolution of a theoretical 44.1/96bit - to explain why we might prefer 192/24 despite the fact that we can't hear above ~20k. Andy |
Ok, I'll say it one more time...
if you had that perfect 192kHz system at 24 bits then you would have the equivalent of a 26 bit system at 48kHz, or 25 bits at 96kHz, if you bandimited the output to 24 or 48kHz respectively.
But in the real world you won't even get that, it will get worse, but even if you did get a more accurate sample, what would you gain?
a 24 bit signal has a dynamic range of 144dB, which means that if you set your levels so that your peak is the noise of a jet taking off at 200 feet (just below the pain threshold), the lowest signal you can capture (if the world and electronic circuitry were completely silent and flawless) is a gnat farting in the next room. Extending this by one bit would admitedly allow you to capture the quieter sound of that same gnat farting whilst sitting on a cushion, which you might consider an important difference artistically, however unless you're dead the sound of your own breathing will drown this out.
When you record to analogue tape the signal gets filtered, then because the tapes magnetic surface doesn't respond linearly you have to add a bias signal, that signal can interact with your audio signal so ideally you have a system like HX Pro to reduce this effect, then it gets written to this non linear medium which is moving at a non constant speed and has a less than perfectly even covering, and has probably had other things recorded on it before, on top of that you probably like the sound of tape saturation so you push it into that non linearity. Then when you playback, again at a non constant speed (not in sync with the speed variations on record) the playhead is non linear, so you have to filter the output from the playback head... and on top of all of this generally screwing around with the signal you may be running noise reduction, which further messes the signal.
What comes out of that tape machine is most definately not what you put into it... but it sounds good, especially if they recording engineer knows how to use its limitations to good effect (tape saturation mostly).
But after all this you listen to an analogue recording and a digital recording and you think "They don't sound the same, there must be something wrong with the digital recording"
Post by: compasspnt on September 27, 2005, 09:33:59 AM
Well said.
Post by: Bill Mueller on September 27, 2005, 09:51:53 AM
| Jon Hodgson wrote on Tue, 27 September 2005 09:23 | ||
Ok, I'll say it one more time... if you had that perfect 192kHz system at 24 bits then you would have the equivalent of a 26 bit system at 48kHz, or 25 bits at 96kHz, if you bandimited the output to 24 or 48kHz respectively. But in the real world you won't even get that, it will get worse, but even if you did get a more accurate sample, what would you gain? a 24 bit signal has a dynamic range of 144dB, which means that if you set your levels so that your peak is the noise of a jet taking off at 200 feet (just below the pain threshold), the lowest signal you can capture (if the world and electronic circuitry were completely silent and flawless) is a gnat farting in the next room. Extending this by one bit would admitedly allow you to capture the quieter sound of that same gnat farting whilst sitting on a cushion, which you might consider an important difference artistically, however unless you're dead the sound of your own breathing will drown this out. When you record to analogue tape the signal gets filtered, then because the tapes magnetic surface doesn't respond linearly you have to add a bias signal, that signal can interact with your audio signal so ideally you have a system like HX Pro to reduce this effect, then it gets written to this non linear medium which is moving at a non constant speed and has a less than perfectly even covering, and has probably had other things recorded on it before, on top of that you probably like the sound of tape saturation so you push it into that non linearity. Then when you playback, again at a non constant speed (not in sync with the speed variations on record) the playhead is non linear, so you have to filter the output from the playback head... and on top of all of this generally screwing around with the signal you may be running noise reduction, which further messes the signal. What comes out of that tape machine is most definately not what you put into it... but it sounds good, especially if they recording engineer knows how to use its limitations to good effect (tape saturation mostly). But after all this you listen to an analogue recording and a digital recording and you think "They don't sound the same, there must be something wrong with the digital recording" |
Excellent! Now we are getting somewhere.
I have wondered for a while, if we should have a thread dedicated to the actual limitations of both analog tape recording and analog disc mastering. There seem to be so many kids here who think there once were things called a record and analog recordings, that were perfect and somewhere we lost the technology, like the lost city of Atlantis.
Think about things like wow and flutter. Double the spec sheet because it wows and flutters when you record it and it wows and flutters when you play it back. The only thing that distorts more than your analog tape is your speaker. Digital, even old digital, is absolutely pristine in comparison.
I have mastered some of my own recordings in self defense and if these folks knew what you have to do to get a decent sounding track on the inner grooves on an LP, I think it would be a real wake up call.
Best Regards,
Bill
Post by: Andy Simpson on September 27, 2005, 09:54:40 AM
| Jon Hodgson wrote on Tue, 27 September 2005 14:23 | ||
Ok, I'll say it one more time... if you had that perfect 192kHz system at 24 bits then you would have the equivalent of a 26 bit system at 48kHz, or 25 bits at 96kHz, if you bandimited the output to 24 or 48kHz respectively. But in the real world you won't even get that, it will get worse, but even if you did get a more accurate sample, what would you gain? a 24 bit signal has a dynamic range of 144dB, ........ |
Oops, sorry my mistake in terms of the numbers.....192/24 = 48/26 in terms of spatial resolution.....if you say so!
I think you're missing my point still, though.
My point still remains that it's not the amplitude dynamic range of 144dB that we like about 24bit.
Clearly we never record anything with anywhere near that dynamic range IN TERMS OF AMPLITUDE.
BUT, this dynamic range has an impact on the spatial resolution.
Why indeed do we like 24bit recordings of music which never manages more than 90dB amplitude dynamics?
I say that higher sampling rates have benefits in these spatial resolution terms, which we can hear, if only because they increase the effective dynamic range of the system in a useable way.
To summarise: frequencies above 20k are not the ONLY benefit to 96 or 192/etc - which I think we can agree on, in theory at least?
To put it in context, in the real world - if we have reached the limit of bit-depth accuracy (where increase in bit-depth will not yield greater accuracy) then we can increase the effective bit-depth/spatial resolution of the system by increasing sampling rates.
Andy
Post by: Jon Hodgson on September 27, 2005, 10:02:54 AM
NO
Post by: Andy Simpson on September 27, 2005, 10:38:26 AM
| Jon Hodgson wrote on Tue, 27 September 2005 15:02 |
Andy, in a word NO |
Why?
Don't we all like 24bit better than 16?
But do we like 192/16 better than 48/24? I would expect not, all things being equal.....
Andy
Post by: maxdimario on September 27, 2005, 10:49:13 AM
| Quote: |
There seem to be so many kids here who think there once were things called a record and analog recordings, that were perfect and somewhere we lost the technology, like the lost city of Atlantis. Think about things like wow and flutter. ... |
some lp masters were indeed horrid, especially best-of type records which took long-playing to it's excess.
a lot of analog gear sucks, including some pro gear.
moderate wow and flutter is predictable and relatively natural sounding, because it modulates the entire track at low frequency.
the real test is placing the recorder in line and out of line from a live source, does it do what you want it to?
non linearity is inherent in every transducer, microphone, speaker etc.
Anyone know of a perfectly linear microphone?
nothing that makes music is perfectly linear, the whole idea is to get the ones that sound like they are natural enough, if that's what you seek, which maybe isn't everybody's objective?
Post by: Bob Olhsson on September 27, 2005, 10:50:39 AM
| andy_simpson wrote on Tue, 27 September 2005 08:54 |
...To put it in context, in the real world - if we have reached the limit of bit-depth accuracy (where increase in bit-depth will not yield greater accuracy) then we can increase the effective bit-depth/spatial resolution of the system by increasing sampling rates. |
Why not examine the REAL problems with common digital audio products, the ones manufacturers and software developers avoid discussions of like the plague?
Why just keep repeating meaningless Madison Ave. spin that's aimed only at making people think they'll look kewl?
Do we HAVE to define who we are with what gear we own?
Post by: Andy Simpson on September 27, 2005, 11:06:59 AM
| Bob Olhsson wrote on Tue, 27 September 2005 15:50 | ||
|
Are we at that point? If we are at that point with bit depth, then lets take sampling rate to the same point and hear the results....
Maybe these real-world issues mean that digital will never equal tape?
I'm just trying to explain why we might prefere 192/24 over 48/24, where previously we thought that there was no advantage other than the increase of bandwidth.....
Infact, I think I (we) have.
Andy
Post by: Jon Hodgson on September 27, 2005, 11:19:23 AM
Post by: J.J. Blair on September 27, 2005, 11:49:35 AM
BTW, can we give Jonny, Max and Andy their own forum? I think that would be a swell idea.
Post by: StudioRhythm on September 27, 2005, 12:03:07 PM
| andy_simpson wrote on Tue, 27 September 2005 08:06 | ||||
Are we at that point? If we are at that point with bit depth, then lets take sampling rate to the same point and hear the results.... Maybe these real-world issues mean that digital will never equal tape? |
Andy, the best designed A/D converters in the world have a dynamic range of around 120dB, maybe 126 or so tops. That's about 20 or 21 bit resolution. And, that appears to be about where it tops out due to the laws of physics, barring anything dramatic and unforeseen, as I understand it.
The dynamic range of the best analog tape recorder is far, far below that. When I said that the dynamic range of ADs tops out around 120-130dB, that same figure would also apply to every possible kind of recording ever invented or to be invented that relies on analog circuitry, since that's about as quiet as any kind of analog circuit can possibly be (such as those found on the front end of AD converters -- it's the ANALOG part of the converters that is restricting the dynamic range, not the digital part.)
| Quote: |
I'm just trying to explain why we might prefere 192/24 over 48/24, where previously we thought that there was no advantage other than the increase of bandwidth..... |
Andy, you're not making any kind of sense. Over the past few pages, many patient people have explained to you why there is not any increased resolution, and they have also explained to you over and over how the phase information that is within the bandwidth of the digital system is kept intact (that means the timing is all there!) Stop making yourself look stupid and read a book on digital audio. Then read this post and see how many times people explained your questions and you ignored them because you didn't understand what they were saying. You're grasping at straws, and they're all the wrong straws, too....
Post by: PaulyD on September 27, 2005, 02:09:29 PM
| dcollins wrote on Thu, 15 September 2005 18:25 |
Even without dither 44/16 has "phase quantization" of 2pi/44100/2^16 or about 2ns from channel to channel. With dither, there is essentially no limit. And, just like analog, whatever noise is at the zero crossing determines the phase resolution. Some studies show that people can hear about 6us ITD, so I think we're safe. I might also add that if PCM was as bad as all that, people would have noticed it way before it arrived in audio. The inter-channel response is a particularly hard one to visualize, it took me forever, but remember that digital is really a continuous time system as we use it. Nothing can fall between the samples and be missed, as no matter how fast the input, the Nyqvist filters will always see a signal "smeared" over one or more samples. The "smearing" question is slightly more interesting...... DC |
Somehow, I would have thought this would answer the question and end the debate about spatial information accuracy in a well-implemented digital system...
| andy_simpson wrote on Tue, 27 September 2005 13:50 |
BUT, this dynamic range has an impact on the spatial resolution. Why indeed do we like 24bit recordings of music which never manages more than 90dB amplitude dynamics? I say that higher sampling rates have benefits in these spatial resolution terms, which we can hear, if only because they increase the effective dynamic range of the system in a useable way. To summarise: frequencies above 20k are not the ONLY benefit to 96 or 192/etc - which I think we can agree on, in theory at least? To put it in context, in the real world - if we have reached the limit of bit-depth accuracy (where increase in bit-depth will not yield greater accuracy) then we can increase the effective bit-depth/spatial resolution of the system by increasing sampling rates. |
Andy, I don't mean to jump on you, brother, but are you understanding what other people here have already posted? Good digital systems have more dynamic range than analog systems. Good digital systems have more dynamic range than we can practically use right now. Digital systems, especially playback systems, also achieve far greater channel separation than analog systems. So...brace yourself...good digital systems preserve spatial information better than analog systems.
Think about this: when a sound reaches two mics spaced slightly apart, do those mic's react precisely the same way? No. Now consider cables...stray capacitance...mic pre's...more cables...other devices that may be used between the mic pre and the MTR...more cables....now the signal, which has just passed through a labrynth of slew rates, finally reaches the analog MTR. Revisit Jon Hodgson's last post on the previous page. Have you ever pulled the covers off a 2" 24-track MTR? They have card cages with separate circuits for each channel that are loaded with discrete electronic components. I'd bet dollars to doughnuts all those circuit cards will have less linear slew rates than the circuits in good digital converters.
Digital recordings are more accurate than analog recordings. So why do we like analog? Because it's less accurate. Analog can add distortion, noise, and harmonic emphasis to a signal that can give it that "larger than life" quality that many listeners find pleasing (myself included). Pop music needs makeup.
So why do higher-sample rate digital recordings sometimes sound better than 44.1? The most logical theory I have read is that higher sample rates cause aliasing that resolves itself in the audible range, and that the aliasing is mimicking the behavior of an analog recorder. I am guessing digital converter designers are experimenting right now with filter designs that can recreate this same behavior without having to resort to higher sampling rates. I'd love for my next set of converters to be usable with an ordinary computer setup. I'd also love it if it had an "Analog Character" section that could be switched on or off with adjustable settings for "tape speed," "saturation," etc. Dave Hill is already onto this (Crane Song HEDD 192) and I absolutely love its affect when I listen to something like Lynn Fuston's 3D ADC CD.
A friendly suggestion: get Nika's book and read it.
Paul
(apologies to everyone else for perpetuating this thread...)
Post by: Johnny B on September 27, 2005, 02:13:49 PM
"Taking into account the speed of sound, we can convert energy defocusing in the time domain to smear in distance estimation by the ears. Energy spread over
Post by: timrob on September 27, 2005, 02:30:38 PM
| PaulyD wrote on Tue, 27 September 2005 13:09 |
So why do higher-sample rate digital recordings sometimes sound better than 44.1? The most logical theory I have read is that higher sample rates cause aliasing that resolves itself in the audible range, and that the aliasing is mimicking the behavior of an analog recorder. I am guessing digital converter designers are experimenting right now with filter designs that can recreate this same behavior without having to resort to higher sampling rates. |
Paul,
Do you really mean to call it aliasing? Or are you referring to sum and difference frequencies(mostly difference in this case).
I can understand the latter, but isn't aliasing taken care of by the Filter.
IIRC, Nika addressed this as a viable explanation in the thread on the old GM forum.
Post by: Johnny B on September 27, 2005, 02:54:53 PM
Maybe a way for people to think about this would be the standard "straight wire" analogy. IOW, the less you mess with the signal, the better.
Ideally, there would be no intermediate conversion steps at all, the signal would retain the optimal bit-depth from recording all the way thru consumer playback. I know this is difficult to imagine this being implemented at this time, but we could see future systems that utilise this kind of thinking.
Post by: dcollins on September 27, 2005, 02:56:14 PM
| maxdimario wrote on Tue, 27 September 2005 07:49 |
moderate wow and flutter is predictable and relatively natural sounding, because it modulates the entire track at low frequency. |
Especially the flutter.
DC
Post by: Johnny B on September 27, 2005, 04:31:59 PM
Ok, do we have any thoughts or comments on DXD, etc. two more high-density disc formats - Blu Ray and HD-DVD?
I'd like to hear comments about sample rates, bit-depths, latency, thruput, truncation, and all the other ususal supsects, but most importantly, I'd like to hear comments about these formats' overall sound quality.
Post by: PaulyD on September 27, 2005, 05:10:30 PM
| timrob wrote on Tue, 27 September 2005 11:30 |
Paul, Do you really mean to call it aliasing? Or are you referring to sum and difference frequencies(mostly difference in this case). I can understand the latter, but isn't aliasing taken care of by the Filter. IIRC, Nika addressed this as a viable explanation in the thread on the old GM forum. |
Hi Tim,
this is a quote from chapter 21 of Nika's book:
"Myth Number One: Higher Sample Rate Recordings Inherently Sound Better Than Lower Sample Rate Recordings, All Things Being Equal
False...We cannot hear above 20KHz. We cannot hear the effect of anything above 20KHz. We cannot hear inter-modulation distortion or beat frequencies caused by material above 20KHz being mixed with material in our hearing range. The only way that we can hear the effect of material over 20KHz is if it is aliased back into the hearing range, or if it is combined with other frequencies in a non-linear environment, creating artificial tones within our hearing range that we can hear..."
None of those italics are my emphasis. That is how it appears in his book, verbatim. I highly recommend Nika's book. It is absolutely outstanding.
Cheers,
Paul
Post by: jimmyjazz on September 27, 2005, 05:12:42 PM
Post by: timrob on September 27, 2005, 05:26:45 PM
| PaulyD wrote on Tue, 27 September 2005 16:10 | ||
Hi Tim, this is a quote from chapter 21 of Nika's book: "Myth Number One: Higher Sample Rate Recordings Inherently Sound Better Than Lower Sample Rate Recordings, All Things Being Equal False...We cannot hear above 20KHz. We cannot hear the effect of anything above 20KHz. We cannot hear inter-modulation distortion or beat frequencies caused by material above 20KHz being mixed with material in our hearing range. The only way that we can hear the effect of material over 20KHz is if it is aliased back into the hearing range, or if it is combined with other frequencies in a non-linear environment, creating artificial tones within our hearing range that we can hear..." None of those italics are my emphasis. That is how it appears in his book, verbatim. I highly recommend Nika's book. It is absolutely outstanding. Cheers, Paul |
Oh, I agree with what this says, I just think that using the term alias in this manner creates confusion. I have read Nika's book and did not recall this particular wording. Honestly, with all of the debates surrounding this issue, Nika would have been a bit more careful with the wording.
Post by: Johnny B on September 27, 2005, 05:30:21 PM
You know, it's quite possible that these two self-proclaimed authority figures are indeed wrong about a few things...for that matter, more than a few things...
For example, the ear/brain/body interaction cannot be reduced to some simplistic theory of being a purely mechanical digital device...the ear/brain/body interaction is far more complex than simple "on-off" states...people are not either in a "zero or a one" state...there are far more complex and subtle factors at play...
And there is some good empirical evidence, which they always seem to ignore or discount, which strongly suggests they are both wrong about quite a few things.
Hell, if they were right, you'd never need anything more than 44k...44k looks great on paper, in practice, it's been well established that it's lacking
Post by: Bill Mueller on September 27, 2005, 05:33:33 PM
| dcollins wrote on Tue, 27 September 2005 14:56 | ||
Especially the flutter. DC |
LOL!
Now this is the spirit with which we should further conduct this discussion!
Best Regards,
Bill
Post by: jimmyjazz on September 27, 2005, 05:41:05 PM
| Johnny B wrote on Tue, 27 September 2005 17:30 |
Hell, if they were right, you'd never need anything more than 44k...44k looks great on paper, in practice, it's been well established that it's lacking |
Guh. How many times does it need to be said . . . it's not the theory, it's the implementation of the theory which has been lacking?
Post by: Johnny B on September 27, 2005, 05:44:29 PM
I can't help it if 44k cannot be made to work, even though the theory says it should.
There are a lot of complete failures in digital tech, get used to it.
Post by: Bill Mueller on September 27, 2005, 05:55:45 PM
| J.J. Blair wrote on Tue, 27 September 2005 11:49 |
Bill, that's really interesting re: the -20db thing. George's meters weren't averaging, were they? Perhaps they were just slow. Also, I'm assuming that if he was using dolby? What a nightmare of noise 456 must have been at those levels. I am going to take a guess at what was going on: Slow VUs not accurately capturing the snare peaks, combined with George not liking the tape compression or lack of headroom on 456? Just a theory. What do you think? BTW, can we give Jonny, Max and Andy their own forum? I think that would be a swell idea. |
JJ,
The sessions I was talking about were on the original ITI console (1st parametric eq). The meters were VU on the console and machine. I don't remember if GM had Dolby on the multi. But my response was the same as yours. What about the noise? However, the transient sparkle on the overheads were better than anything I had ever recorded and I don't remember hearing any noise problems. Of course, VU meters don't capture peaks and that's the point. George knew were the peaks were and were they weren't and set his levels accordingly.
This was my first clue that recording music involved much more than cranking the level up to O and kicking back to listen. That there was something in there that I did not know about, that I could not see, but that could make the sound much better if I controlled it.
Bets Regards,
Bill
Post by: compasspnt on September 27, 2005, 06:07:47 PM
| Johnny B wrote on Tue, 27 September 2005 17:30 |
For a variety of good reasons, there are many who strongly disagree with both Dan Lavry and Nika... You know, it's quite possible that these two self-proclaimed authority figures are indeed wrong about a few things...for that matter, more than a few things... |
Self proclaimed? OK, how many of us here would like to now add our proclamation to these gentlemen's credentials? I for one.
| Quote: |
For example, the ear/brain/body interaction cannot be reduced to some simplistic theory of being a purely mechanical digital device...the ear/brain/body interaction is far more complex than simple "on-off" states...people are not either in a "zero or a one" state...there are far more complex and subtle factors at play... |
If you're referring to not being able to record proper audio because of using only ones and zeros (or "offs" amd "ons"), as I recall this happens AT LEAST 44,100 times each second.
| Quote: |
And there is some good emphical evidence, which they always seem to ignore or discount, which strongly suggests they are both wrong about quite a few things. |
Just exactly what emphical [sic] evidence is that?
| Quote: |
Hell, if they were right, you'd never need anything more than 44k...44k looks great on paper, in practice, it's been well established that it's lacking |
A) We all use more than 44k already.
B) Practice can always be improved upon. That's just what they are doing this very minute.
Scientists have always been misunderstood by non-scientists. In fact, that will probably be the real societal conflagration of the next century, not lib v. cons, or black v. white, or rich v. poor. There are those scientists who do think "outside the you-know-what," and implement great change. And there are always some scientists, spurred on by those not in the know, who accept only the status quo, and aren't part of the upward movement of change. But to blindly state that Nika and Dan (not to mention others here who have so carefully made their case) are NOT TRYING TO IMPROVE THE STATE OF MODERN AUDIO is simply ludicrous.
I know you are trying to provoke people. And provoking for the good of viewing new or unseen horizons can be a good thing. But surely it can be done in a gentlemanly and reasonable manner, paying attention to absolute FACT at the same time. And surely those to whom you refer don't actually NEED the provocation to begin with.
Post by: Johnny B on September 27, 2005, 06:11:20 PM
from Webster's:
empirical: 1. experiment or experience; 2. Depending on observation alone, without regard to science or theory...
Empirical evidence is fully admissible in most courts of law around the world because it often can have a high degree of reliability.
Here's an example I just found.
Lawyer Smith: And what did you hear?
Witness Jones: I heard a gunshot.
Lawyer Smith: How do you know it was a gunshot and not a truck backfiring?
Witness Jones: I'm a well-seasoned hunter, I know a gunshot when I hear it.
Lawyer Smith: And what direction did the gunshot come from?
Witness Jones: It came from my right, about 50 feet away.
Opposing Counsel: Objection. Mr Jones is not a scientific expert, he cannot prove with math exactly what he heard.
Trial Judge: Objection overruled. Mr. Jones' direct obsevation and his direct experience are relevant and material evidence. Please continue giving your evidence Mr. Jones.
So we see that most courts of law will allow this kind of experience and observation into evidence. But there are people running around the digital debate who simply discount, deny, or ignore this kind of admissible evidence.
Post by: timrob on September 27, 2005, 06:23:47 PM
| Johnny B wrote on Tue, 27 September 2005 17:11 |
Maybe this will help: from Webster's: empirical: 1. experiment or experience; 2. Depending on observation alone, without regard to science or theory... Empirical evidence is fully admissible in most courts of law around the world because it often can have a high degree of reliability. Here's an example I just found. Lawyer Smith: And what did you hear? Witness Jones: I heard a gunshot. Lawyer Smith: How do you know it was a gunshot and not a truck backfiring? Witness Jones: I'm a well-seasoned hunter, I know a gunshot when I hear it. Lawyer Smith: And what direction did the gunshot come from? Witness Jones: It came from my right, about 50 feet away. Opposing Counsel: Objection. Mr Jones is not a scientific expert, he cannot prove with math exactly what he heard. Judge: Objection overruled. Mr. Jones' direct obsevation and his direct experience are relevant and material evidence. Please continue giving your evidence Mr. Jones. So we see that most courts of law will allow this kind of experience and obsevation into evidence. But there are people running around the digital debate who simply discount or ignore this kind of admissible evidence. |
I think I can state unequivocally that Nika Aldrich and Dan Lavry have put in more research and have more empirical evidence available to them than you. All your arguments are based on hearsay and conjecture not empirical evidence.
Post by: Johnny B on September 27, 2005, 06:26:14 PM
By the way I was not aware that both Nika and Dan Lavry were also highly trained biologists, licensed brain surgeons, or established medical researchers.
Are they? Or are they claiming to be experts in matters based purely on a layman's attempt at understanding, and giving opinions as to matters for which they are not really qualified.
Post by: Andy Simpson on September 27, 2005, 06:31:55 PM
| PaulyD wrote on Tue, 27 September 2005 19:09 |
Andy, I don't mean to jump on you, brother, but are you understanding what other people here have already posted? Good digital systems have more dynamic range than analog systems. Good digital systems have more dynamic range than we can practically use right now. Digital systems, especially playback systems, also achieve far greater channel separation than analog systems. So...brace yourself...good digital systems preserve spatial information better than analog systems. ...) |
If digital systems have more dynamic range than we can practically use, right now, why doesn't 44/16 sound good enough? It has 'enough' dynamic range, but alas 44/24 sounds better. Do we need 144dB of amplitude dynamic? Nope. But in terms of spatial resolution, maybe we need even more......
Anyway, I wanted to relate 192/16 to 48/24 to illustrate how a lower bandwidth system can have greater spatial resolution.
And to show how a very high sampling medium with low dynamic range can also have good spatial resolution.
And having done that we might imagine how tape, as a low dynamic range medium, might overcome this by sampling more often and giving a greater spatial resolution in this way.
If you know that tape has a dynamic range of 90dB and you know how often it samples, you should be able to work out the equivalent resolution - but wait - we don't know how often it samples....
.....unless you maybe take 1/2" 15ips, and work out how many magnetic particles there are per second passing the head......
my guess - quite a few.....maybe someone can chime in with a reasonable estimate?
Let's just say that in theory, greater spatial resolution can be had by increasing bit-depth OR sampling rate, and leave it there.
Andy
Post by: timrob on September 27, 2005, 06:33:43 PM
| Johnny B wrote on Tue, 27 September 2005 17:26 |
Nice try. No sale. |
Well, one can't ignore the truth. That is unless their handle is Johnny B.
Post by: compasspnt on September 27, 2005, 06:41:03 PM
| Johnny B wrote on Tue, 27 September 2005 18:26 |
...I was not aware that both Nika and Dan Lavry were also highly trained biologists, licensed brain surgeons, or established medical researchers. Are they? Or are they claiming to be experts in matters based purely on a layman's attempt at understanding, and giving opinions as to matters for which they are not really qualified. |
???
Where did this come from? A bit OT?
Post by: maxdimario on September 27, 2005, 06:52:18 PM
ears?
Post by: CCC on September 27, 2005, 07:10:30 PM
| Johnny B wrote on Tue, 27 September 2005 23:11 |
So we see that most courts of law will allow this kind of experience and observation into evidence. But there are people running around the digital debate who simply discount, deny, or ignore this kind of admissible evidence. |
Admissibility is one thing - finding evidence to be factual in the final analysis is something else altogether.
Post by: Johnny B on September 27, 2005, 07:12:02 PM
To be honest and fair, we really need far more research that relies on a proper multidisciplinary approach to fully understand how much more of the frequency spectrum is important to our entire human system, the ears, the brain, and the body.
You'll need such things as biologists, qualified medical researchers, brain specialists or surgeons and all manner of other disciplines to all come together to fully resolve the burning question: "How much of the frequency spectrum is important to digitally capture and digitally reproduce for humans to experience pleasure when listening to music?"
Post by: timrob on September 27, 2005, 07:27:12 PM
| Johnny B wrote on Tue, 27 September 2005 18:12 |
The biology and medical ref above comes from their claims about what amounts to their idle speculation about the frequency spectrum as far as the ear/brain/body interaction is concerned. They cannot prove their claims with sound medical science that only 20 to 20 is important, other contrary evidence suggests that the "important" frequency spectrum is far greater. To be honest and fair, we really need far more research that relies on a proper multidisciplinary approach to fully understand how much more of the frequency spectrum is important to our entire human system, the ears, the brain, and the body. You'll need such things as biologists, qualified medical researchers, brain specialists or surgeons and all manner of other disciplines to all come together to fully resolve the burning question: "How much of the frequency spectrum is important to digitally capture and digitally reproduce for humans to experience pleasure when listening to music?" |
And what makes you think that research in those areas hasn't been or isn't being done.
Post by: dcollins on September 27, 2005, 07:28:36 PM
| Johnny B wrote on Tue, 27 September 2005 16:12 |
The biology and medical ref above comes from their claims about what amounts to their idle speculation about the frequency spectrum as far as the ear/brain/body interaction is concerned. They cannot prove their claims with sound medical science that only 20 to 20 is important, other contrary evidence suggests that the "important" frequency spectrum is far greater. |
Sweetie, is it more likely that >20kHz hearing would be discovered by audiophiles, or by people that have actually studied physiology?
Since we don't have a hair-cell for much over 15k (and the HF cells are nearest the outside world and thus the first to "go") you would have to show us where these freqs are detected.
You must have finished Pohlman already!
Then move on to Fletcher (not that one) or even Brian C. J. Moore, if you are really serious.
If you weren't such a pud I'd offer to lend you my copies.......
DC
Post by: Johnny B on September 27, 2005, 07:36:00 PM
Dave, you bring up hair, but the fluid mechanics may not be as straight forward as people once believed. From what my "pud-like" brain can gather, you are looking at a fairly complex system that does more than a fair bit of multi-tasking with each hair. Contrast that with the old belief system that said each hair only responded to one, and only one, frequency. Anyway, the old belief system about hair seems to be falling by the wayside.
And let's not forget the brain and body reactions. There's a lot to this, simplistic answers just don't cut it anymore.
Those with only limited skill sets are no longer sufficient to adequately answer many of the newer questions or adequately resolve many of the complex issues raised.
Under the circumstances, a proper multidisciplinary approach is warranted.
Post by: electrical on September 27, 2005, 11:05:33 PM
| dcollins wrote on Tue, 27 September 2005 19:28 |
Sweetie, is it more likely that >20kHz hearing would be discovered by audiophiles, or by people that have actually studied physiology? Since we don't have a hair-cell for much over 15k (and the HF cells are nearest the outside world and thus the first to "go") you would have to show us where these freqs are detected... DC |
Dave, do you mean you can't hear the diference between a 7kHz sine wave and a 7kHz square wave? Its odd-order harmonics begin at 21kHz. For a professional listener, I'd be shocked if you couldn't. I know I can, and I don't believe I'm special.
I'll put money on it gentlemen. I'll bet that Dave Collins, professional listener, can hear the difference between a 7kHz sine wave and a 7kHz square wave.
Post by: tom eaton on September 27, 2005, 11:36:22 PM
Ummm...ask Apple. They've done the most relevant market research for you. It's not the answer you want, in fact quite the opposite.
-tom
Post by: minister on September 27, 2005, 11:45:33 PM
...amazing...
and YET, not having read the books mentioned, and not having a deep understanding of digital audio design, but being a pretty decent composer/engineer, i have learned a tremendous amount from the preposterous propositions and the elucidating emendations both textual and conceptual.
...amazing...
uh...THANKS
Post by: dcollins on September 28, 2005, 12:05:50 AM
| electrical wrote on Tue, 27 September 2005 20:05 |
I'll put money on it gentlemen. I'll bet that Dave Collins, professional listener, can hear the difference between a 7kHz sine wave and a 7kHz square wave. |
I've never tried it!
There is a potential gotcha here since most generators match the peak-to-peak value as you switch waveforms, making the square wave a couple three dB louder unless otherwise matched.
I cannot hear a 21k sine, that much I know for sure.
But I will try this test, with an attempt at matching, at some point.
I got a fin says you're hearing something else, though.
DC
Post by: dcollins on September 28, 2005, 12:10:00 AM
| Johnny B wrote on Tue, 27 September 2005 16:36 |
Emmmm...yes, yes....maybe I should also dig up my old edition of Handbook of Recording Engineering (Hardcover) by John M. Eargle and review that as well, aye? |
Hey, at this point it couldn't hurt.
| Quote: |
And let's not forget the brain and body reactions. There's a lot to this, simplistic answers just don't cut it anymore. |
Fortunately hearing has never been studied by Science as it was too "simplistic."
| Quote: |
Under the circumstances, a proper multidisciplinary approach is warranted. |
And you're just the guy to bring a rigorous, scholarly, non-sectarian approach to the subject, so off to library with you!
DC
Post by: Ronny on September 28, 2005, 12:41:46 AM
| andy_simpson wrote on Tue, 27 September 2005 18:31 | ||
If digital systems have more dynamic range than we can practically use, right now, why doesn't 44/16 sound good enough? It has 'enough' dynamic range, but alas 44/24 sounds better. Do we need 144dB of amplitude dynamic? Nope. But in terms of spatial resolution, maybe we need even more...... Anyway, I wanted to relate 192/16 to 48/24 to illustrate how a lower bandwidth system can have greater spatial resolution. And to show how a very high sampling medium with low dynamic range can also have good spatial resolution. And having done that we might imagine how tape, as a low dynamic range medium, might overcome this by sampling more often and giving a greater spatial resolution in this way. If you know that tape has a dynamic range of 90dB and you know how often it samples, you should be able to work out the equivalent resolution - but wait - we don't know how often it samples.... .....unless you maybe take 1/2" 15ips, and work out how many magnetic particles there are per second passing the head...... my guess - quite a few.....maybe someone can chime in with a reasonable estimate? Let's just say that in theory, greater spatial resolution can be had by increasing bit-depth OR sampling rate, and leave it there. Andy |
Only up to a certain point, when you pass human acuity it doesn't matter anymore. It's like cleaning your windshield more, when you've already cleaned all of the dirt off. On analog tape the magnetic particles overlap, analog is one continuous sample, it's not pieced together like digital audio that has finite samples.
Post by: Phil on September 28, 2005, 01:14:56 AM
You can bet your Fruit of the Looms that, while this subject is being talked to death here, there are people working theories and ideas, and they have the ability and resources to do something. And, when the ones who can do are finished, their products will be brought to market. Some will be good; some will sound like hammered shit.
When the products of those who can do are on the market, the talkers will post…
“Has anyone used the new Zizzwheel Purple yet…and how does it compare to a Neve?”
“I’m thinking about buying a matched pair of U-47s to use with my Zizzwheel Purple for overheads…will they sound better than a 57?”
“Will I need a Mackie Big Knob to connect my Zizzwheel Purple to my monitors?”
“Which end of the guitar cord goes in the guitar, and which end to the amp?”
We’ll have new tools and refined technology, and what will we do with this new methodology?
Why, we’ll record baldheaded thugs grunting bad rhymes about God and Peace and Hoes and Bootie, over a beat track containing at least 20% distortion. We’ll record a too-beautiful-for-words nubile who will whine 128 tracks of mindless lyrics that will need to be comped, auto-tuned, and mercilessly smashed. And the Zizzwheel Purple will faithfully reproduce every insult to creativity that is fed into it. Some, of course, will argue that the reproduction still isn’t all that faithful. Some will like it anyway. Many will pay money to hear it. Many more will not know they are hearing it, and will still pay money. Others will steal it.
At the end of the day, the baldheaded thug may be shot. The beautiful nubile won’t practice because she misses her boyfriend, and besides, she has an interview and a photo shoot. Politicians will still be lying rat bastards. Shit will still stink.
Mothers will bake apple pies, and their kitchens will smell good, and some of us will wish we could go home again.
Even if it’s just for a little while...
* * * * *
I humbly submit that our tools, as they exist now, are very good, and yet, they will still continue to evolve. Sadly, they are frequently better than what they are required to do. We need a balance.Now, back to our regularly scheduled program; currently in progress.
Post by: Eric Bridenbaker on September 28, 2005, 03:00:15 AM
Here's two:
#1 - A ninteenth century french poet (not sure which, be sure to chime in if you know who it was) was once asked...
"Where are we French going here, revolution after revolution, promise after promise, things look like they will improve but still we end up in trouble again. Why?"
To which he responded, "In any full revolution, you are bound to wind up in the same place. That is the definition!"
#2 - The farthest point on earth from you (and likely the universe for that matter) is the back of your head!
Best,
Eric
Post by: Ronny on September 28, 2005, 03:09:34 AM
| Phil wrote on Wed, 28 September 2005 01:14 |
I do believe this circle jerk will, like Hell’s torment, continue forever. You can bet your Fruit of the Looms that, while this subject is being talked to death here, there are people working theories and ideas, and they have the ability and resources to do something. And, when the ones who can do are finished, their products will be brought to market. Some will be good; some will sound like hammered shit. When the products of those who can do are on the market, the talkers will post… “Has anyone used the new Zizzwheel Purple yet…and how does it compare to a Neve?” |
Better than Neve Capricorn but not as good as Neve Campbell.
| Quote: |
“I’m thinking about buying a matched pair of U-47s to use with my Zizzwheel Purple for overheads…will they sound better than a 57?” |
I was working with a matched pair of 38's, I went for some bottom end, she crossed her legs and broke my glasses.
| Quote: |
“Will I need a Mackie Big Knob to connect my Zizzwheel Purple to my monitors?” |
I'd suggest that you use your own Big Knob instead of Mackies.
| Quote: |
“Which end of the guitar cord goes in the guitar, and which end to the amp?” |
Just follow the direction of the arrows for the best skin effect tone.
| Quote: |
We’ll have new tools and refined technology, and what will we do with this new methodology? |
Masters & Johnson and Doctor Ruth have some great books on the subject of tools.
| Quote: |
Why, we’ll record baldheaded thugs grunting bad rhymes about God and Peace and Hoes and Bootie, over a beat track containing at least 20% distortion. |
I ran FOH for Bootie and the Hoe Fish back in the 80's.
| Quote: |
We’ll record a too-beautiful-for-words nubile who will whine 128 tracks of mindless lyrics that will need to be comped, auto-tuned, and mercilessly smashed. And the Zizzwheel Purple will faithfully reproduce every insult to creativity that is fed into it. Some, of course, will argue that the reproduction still isn’t all that faithful. Some will like it anyway. Many will pay money to hear it. Many more will not know they are hearing it, and will still pay money. Others will steal it. At the end of the day, the baldheaded thug may be shot. The beautiful nubile won’t practice because she misses her boyfriend, and besides, she has an interview and a photo shoot. Politicians will still be lying rat bastards. Shit will still stink. Mothers will bake apple pies, and their kitchens will smell good, and some of us will wish we could go home again. Even if it’s just for a little while... * * * * * I humbly submit that our tools, as they exist now, are very good, and yet, they will still continue to evolve. Sadly, they are frequently better than what they are required to do. We need a balance.Now, back to our regularly scheduled program; currently in progress. |
The only time that I had a tool problem was when I got back from R and R in Saigon back in 67.
Post by: Eric Bridenbaker on September 28, 2005, 03:18:17 AM
| dcollins wrote on Wed, 28 September 2005 00:05 | ||
I've never tried it! There is a potential gotcha here since most generators match the peak-to-peak value as you switch waveforms, making the square wave a couple three dB louder unless otherwise matched. I cannot hear a 21k sine, that much I know for sure. But I will try this test, with an attempt at matching, at some point. I got a fin says you're hearing something else, though. DC |
Just tried it, DC is right, there is a slight level change between the square and sine. Could be my converters, but the square wave sounds brighter, fuller (at matched RMS level to the sine), and interestingly enough, slightly higher in pitch, though both are at 7KHz.
The difference was fairly discernable doing A/B comparisons, but not so confident that I would get it in a blind experiment hearing only one of the tones. All in all, they sound very similar.
Best Regards,
Eric
Post by: crm0922 on September 28, 2005, 03:37:15 AM
This I gotta try. I think my ribbon tweeters will get me to 21k. I'll have to use the computer to generate the signal though, so I suppose at least 42k sample rate will be required.
One thing I know is that I cannot hear a 21k sine wav, but I can hear a 19k one. At least I could before that loud-ass rock show I just got back from. Heavy-duty earplugs and hiding in the corner may not have spared me from ear-ruin. And now the whole band is asleep on the floor of my studio.
But I digress, yes Max, live music is good. I actually made a digital recording of the show this evening, and I expect it to sound as good or maybe better than it did live, not sterile, or weird, lifeless, etc. It won't sound as comforting and "fun" as an analog recording, but quite acceptable, I would say.
Chris
Post by: Jon Hodgson on September 28, 2005, 03:38:54 AM
| Eric Bridenbaker wrote on Wed, 28 September 2005 08:18 | ||||
Just tried it, DC is right, there is a slight level change between the square and sine. Could be my converters, but the square wave sounds brighter, fuller (at matched RMS level to the sine), and interestingly enough, slightly higher in pitch, though both are at 7KHz. The difference was fairly discernable doing A/B comparisons, but not so confident that I would get it in a blind experiment hearing only one of the tones. All in all, they sound very similar. Best Regards, Eric |
Your converters?
How did you generate this square wave, and how are you listening to it?
Post by: Johnny B on September 28, 2005, 04:43:28 AM
| dcollins wrote on Wed, 28 September 2005 05:10 | ||
And you're just the guy to bring a rigorous, scholarly, non-sectarian approach to the subject, so off to library with you! DC |
Dave, are you offering to fund a fully qualified team of researchers? If so, we could probably set up a 501c3 non-profit organisation and your generous contribution to advanced research will be tax deductible.
As for the "nonsectarian" aspect, I'd say "the religion" here goes in several directions. On the one hand, you have those evangelists who say that digital sound quality is just great the way it is, that in fact, no improvements are desireable or even possible...On the other hand, you have those worshipers who believe that digital sound quality does not match up to analogue's sound quality and some go so far as to say it never will.
In my "pud-like" view, I think digital has some serious challenges to overcome and that more research and applied science using a multidisciplinary approach can only help with those challenges. If people want to call me names or take pot shots at me for expressing that view, have at it.
My sense is that there are some Luddites in the digital arena who want to freeze the technology right where it is and prevent any advancement whatsoever.
Post by: maxdimario on September 28, 2005, 07:25:03 AM
| Quote: |
But I digress, yes Max, live music is good. I actually made a digital recording of the show this evening, and I expect it to sound as good or maybe better than it did live, not sterile, or weird, lifeless, etc. It won't sound as comforting and "fun" as an analog recording, but quite acceptable, I would say. |
fine..
but how do you know what the musicians actually sound like live... if you listened through a blaring pa, miked through a live board?
earplugs did you say?
the rolling stones in the early 60's with no mikes on drums or guitars would probably a more informative listening session for amplified rock music.
watch out for those ears...you can't get them back
Post by: compasspnt on September 28, 2005, 08:28:17 AM
My sense is that there are some Luddites in the digital arena who want to freeze the technology right where it is and prevent any advancement whatsoever.
[/quote]
Right. Such as Nika and Dan, whom you previously insulted?
By the way, you really don't have to put all that large amount of empty space at the end of every post.
Post by: tom eaton on September 28, 2005, 09:05:17 AM
-tom
Post by: Eric Bridenbaker on September 28, 2005, 11:24:31 AM
| Jon Hodgson wrote on Wed, 28 September 2005 03:38 |
Your converters? How did you generate this square wave, and how are you listening to it? |
Generated digitally using the test generator in Nuendo 3 @ 24Bit/48Khz.
Chain is Nuendo/Lynx/Bryston 3B, and a variety of monitors, one set has ribbon tweeters, but I tried this late at night (my home setup), so I used AKG 240 phones instead.
Would be nice if someone did this test all analog as well. Crack out your oscilliscopes!! A square wave sounds different from a sine at lower frequencies, and seems to retain it's flavor (ie harmonics, as the frequency rises, but that's as far as I'm going on this here thread.)
Best,
Eric
Post by: bobkatz on September 28, 2005, 11:29:32 AM
| electrical wrote on Tue, 27 September 2005 23:05 | ||
Dave, do you mean you can't hear the diference between a 7kHz sine wave and a 7kHz square wave? Its odd-order harmonics begin at 21kHz. For a professional listener, I'd be shocked if you couldn't. I know I can, and I don't believe I'm special. I'll put money on it gentlemen. I'll bet that Dave Collins, professional listener, can hear the difference between a 7kHz sine wave and a 7kHz square wave. |
Of course the first harmonic of 7 kHz is 14 kHz so we're likely to hear that difference...
But take a 20 kHz sine vs. triangle vs. square or a 12 kHz sine-triangle-square comparison, where, theoreticall, the first harmonic is above our hearing so we wouldn't hear a difference. Yet we do. Does this mean that we do have supersonic hearing?
(take a guess what I'm going to answer)
{scroll down for answer)
(scroll down for answer)
The sine-square-triangle test is a well known red herring. Most times it can be shown by simple measurement taht distortions IN THE AUDIBLE BAND have been produced by our non-linear loudspeakers. Whenever you perform a test like that, please make sure that your loudspeakers are NOT producing any audible products in the audible band! Nonlinear loudspeakers lead us to believe that we can hear supersonics when all it turns out to be is that they are adding distortion between 20 and 20 kHz....
But a positive result to the listening test doesn't disprove the assertion either, it only leaves room for doubt. Maybe we can hear above 20 khz, but sine versus square is not a good way to test for it.
A much better way is to try tests of low-pass filtering to see if you can hear a difference. Hey, wait a minute, I already did.... Ask George Massenburg about that one, too. It's controversial of course, but the gist of the evidence as far as I can see is that the filtering itself causes the anomalies, not the bandwidth reduction.
BK
Post by: archtop on September 28, 2005, 11:40:46 AM
| bobkatz wrote on Wed, 28 September 2005 08:29 |
A much better way is to try tests of low-pass filtering to see if you can hear a difference. Hey, wait a minute, I already did.... Ask George Massenburg about that one, too. It's controversial of course, but the gist of the evidence as far as I can see is that the filtering itself causes the anomalies, not the bandwidth reduction. BK |
But George doesn't like us any more.
Post by: timrob on September 28, 2005, 11:56:51 AM
| bobkatz wrote on Wed, 28 September 2005 10:29 |
Of course the first harmonic of 7 kHz is 14 kHz so we're likely to hear that difference... |
Hey Bob, Don't want to pick nits BUT... The way I learned harmonics is that the first harmonic is the Fundamental. So really 14k is the second harmonic of 7khz which would not be present in a Square wave which contains 1,3,5,7... It would be in a triangle wave though which contains 1,2,3,4... The relative amplitude and phase of the harmonics is also important. In a square wave the relative amplitudes fall out like this: 1, 1/3, 1/5, 1/7 and triangle: 1, 1/2, 1/3, 1/4. In either case the harmonics are in phase with the fundamental.
Sorry to throw this out there. I know you know what's happenin'.
I believe the only speakers I've heard of capable of reproducing square waves very well are the Meyer HD1s. Certainly, there may be others.
Post by: J.J. Blair on September 28, 2005, 12:01:08 PM
| Eric Bridenbaker wrote on Wed, 28 September 2005 00:18 |
Just tried it, DC is right, there is a slight level change between the square and sine. |
You need to L2 it to make sure you get equal gain.
| crm0922 wrote on Wed, 28 September 2005 00:37 |
But I digress, yes Max, live music is good. 2 |
Yes. Especially if the live music is made on a DX7, an M1 or a Triton, and you have a purer wave form that hasn't been corrupted by transistors, tubes, transformers, microphones, and all that other crap in the signal path that filters the emotion out of the music.
| Johnny B wrote on Wed, 28 September 2005 01:43 |
Dave, are you offering to fund a fully qualified team of researchers? If so, we could probably set up a 501c3 non-profit organisation and your generous contribution to advanced research will be tax deductible. |
Wow, this guy never ceases to amaze me. Where does he have the time to learn so much about audio AND study the tax code? Is it ironic to anybody other than me that he knows so much about 'non-profit' work? You know, it would be a shame if he actually started working as an engineer, because then he wouldn't have the time to educate us all. Not to mention, once you start actually putting things into practice, your experience catches up with your opinions.
Post by: PookyNMR on September 28, 2005, 12:14:53 PM
Seems like the perspective in many cases is still the same - ignorance.
Post by: Eric Bridenbaker on September 28, 2005, 12:25:12 PM
| J.J. Blair wrote on Wed, 28 September 2005 12:01 | ||
You need to L2 it to make sure you get equal gain. |
Yeah man, two birds with one stone, even out the volume and the waveshape at the same time. Turning a sine wave into a square.. it's like....magic. I'm amazed people were even able to make recordings before L type brickwall limiting.
I guess they didn't know what they were missing. Anyway it's not too late, there's still time to "remaster" all those classics. (Okay, now that's getting painful, sorry...)
Best,
Eric
Post by: lord on September 28, 2005, 12:27:23 PM
| electrical wrote on Tue, 27 September 2005 23:05 |
Dave, do you mean you can't hear the diference between a 7kHz sine wave and a 7kHz square wave? Its odd-order harmonics begin at 21kHz. For a professional listener, I'd be shocked if you couldn't. I know I can, and I don't believe I'm special. |
This question is a pretty pickle, but one that has come up before and (IMO) been logically explained.
If you cannot hear the 21k sine, then neither can you hear the true 3rd harmonic of a 7k square. However, the 7k sine and square tones present radically different signals to the amplifier and speakers, that can result in slightly different, audible non-linear distortions.
The higher the sampling rate, the more square your square wave will look, and most likely-- the more different it'll sound. But I am suggesting that the difference in sound is probably the result of the tweeter slapping around in there funny and not a true perception of the harmonic.
edit sorry, I see I'm late. thanks BK...
Post by: Norwood on September 28, 2005, 12:44:24 PM
| J.J. Blair wrote on Wed, 28 September 2005 09:01 | ||
Yes. Especially if the live music is made on a DX7, an M1 or a Triton, and you have a purer wave form that hasn't been corrupted by transistors, tubes, transformers, microphones, and all that other crap in the signal path that filters the emotion out of the music. |
Unless the Triton is a Triton EXTREME... it uses a gin-you-wine toob to make it feel all warm and cozy.
Post by: J.J. Blair on September 28, 2005, 02:32:22 PM
BTW, I'm surprised that the White House hasn't hired this Axis of Idiocy for their scientific advisory panel, for input on things like the 'theory' of evolution and global warming.
Post by: dcollins on September 28, 2005, 03:10:09 PM
| bobkatz wrote on Wed, 28 September 2005 08:29 |
Of course the first harmonic of 7 kHz is 14 kHz so we're likely to hear that difference... |
The square wave consist of only odd harmonics, so we multiply by 3 instead of two.
Although, I do agree that it's far more likely that what you're hearing is a distortion product than the actual 3rd harmonic.............
DC
Post by: Ronny on September 28, 2005, 03:33:06 PM
| J.J. Blair wrote on Wed, 28 September 2005 12:01 |
Not to mention, once you start actually putting things into practice, your experience catches up with your opinions. |
Usually it's your experience starts over riding your prior opinions and new logic forces them to change. Lest we not forget that as technology has reached such a rapid rate of improvement since the computer age went into full swing, that we need to constantly re-examine our prior views. What often wasn't possible yesterday, is often possible today and what isn't possible today will be possible tomorrow.
Post by: Johnny B on September 28, 2005, 03:36:26 PM
| J.J. Blair wrote on Wed, 28 September 2005 17:01 | ||
Wow, this guy never ceases to amaze me. Where does he have the time to learn so much about audio AND study the tax code? |
JJ, tax exempt 501(c)3's are fairly standard vehicles, these kinds of non-profit organisations are often used to benefit education or advance the arts and sciences. Run a google and you'll see what great things 501(c)3's can be.
Post by: Johnny B on September 28, 2005, 03:54:45 PM
| compasspnt wrote on Wed, 28 September 2005 13:28 |
Right. Such as Nika and Dan, whom you previously insulted? |
I did not insult them, I merely pointed out they were lacking in qualifications as to certain topics having to do with the human body's responses.
Although I disagree with some of their claims, I have nothing personal against either of them. For example, it's not as if I had a purchase from the Sweetwater Retail Music Store go sour.
Post by: J.J. Blair on September 28, 2005, 03:56:45 PM
| Johnny B wrote on Wed, 28 September 2005 12:36 |
JJ, tax exempt 501(c)3's are fairly standard vehicles, these kinds of non-profit organisations are often used to benefit education or advance the arts and sciences. Run a google and you'll see what great things 501(c)3's can be. |
He obviously understands comedy as well as he understands digital audio.
Post by: electrical on September 28, 2005, 04:04:49 PM
| dcollins wrote on Wed, 28 September 2005 00:05 |
I cannot hear a 21k sine, that much I know for sure. |
I can, or at least I could a year ago. I don't believe I'm special. I believe that if I filled your room with 21kHz you would hear it.
And wouldn't peak-to-peak matching for the test signals be appropriate? The 3dB comes from the harmonics you "can't hear" being added, so they shouldn't matter.
If your defense against this argument is, "you're hearing something, but it isn't the program," Then I don't know where to go from there. Maybe I only think I'm hearing that awesome bass drum on Jim White's drum kit too.
I don't believe this hyper-sonic stuff is the reason digital recordings sound different from analog ones, by the way. I just think that categorically denying that people percieve sound above 20kHz is contrary to my experience, and I believe it will be contrary to Dave Collins's experience once he tries this experiment.
Post by: Johnny B on September 28, 2005, 04:10:12 PM
| J.J. Blair wrote on Wed, 28 September 2005 19:32 |
BTW, I'm surprised that the White House hasn't hired this Axis of Idiocy for their scientific advisory panel, for input on things like the 'theory' of evolution and global warming. |
Emmm, "global warming" seems to be well accepted within the scientific community, despite the Bush Administration's attempts to thwart that knowledge from being widely accepted by the public-at-large. Some scientists assign "global warming" as one of the primary causes for the New Orleans disaster.
As for the Bush Administration's scientific advisory panel, the Bush White House would more likely than not be more interested in developing weapons of mass destruction based on digital sound systems.
Of course, some of us are more interested in devoting scarce research dollars to more peaceful and productive pursuits, like improving the sound of digital systems.
Post by: Fig on September 28, 2005, 04:44:31 PM
| electrical wrote on Wed, 28 September 2005 15:04 |
if I filled your room with 21kHz you would hear it. |
I agree here.
I've been messing around with an old HP tube test oscillator that goes up to 100 kHz. The ultrasonic stuff rolls off pretty fast after 24 kHz or so (to my ears) but when I add more gain to the system, I can perceive much higher frequencies.
Notice I did not say I can "hear" them. Its more like I can feel them through my sinuses and on my skin surface (not unlike the sensation of chest thumping bass, actually). I imagine I am pissing off a lot of neighborhood dogs and such.
| Quote: |
And wouldn't peak-to-peak matching for the test signals be appropriate? The 3dB comes from the harmonics you "can't hear" being added, so they shouldn't matter. |
Good point and very interesting. Are you saying if a listener hears ANY difference between the two, then they are hearing those harmonics?
| Quote: |
I don't believe this hyper-sonic stuff is the reason digital recordings sound different from analog ones, by the way. |
What then, Steve, is it?
Your views on archiving and keeping "permanent records" that will outlive the performers are well documented and I agree with them but, WHAT sonically do you believe is the difference?
Or is a workflow thing? No wait, you said "sound different".
I'm very curious.
| Quote: |
I just think that categorically denying that people percieve sound above 20kHz is contrary to my experience, and I believe it will be contrary to Dave Collins's experience once he tries this experiment. |
Yep.
$0.02,
Fig
Post by: RKrizman on September 28, 2005, 05:37:21 PM
| bobkatz wrote on Wed, 28 September 2005 11:29 |
The sine-square-triangle test is a well known red herring. Most times it can be shown by simple measurement taht distortions IN THE AUDIBLE BAND have been produced by our non-linear loudspeakers. Whenever you perform a test like that, please make sure that your loudspeakers are NOT producing any audible products in the audible band! Nonlinear loudspeakers lead us to believe that we can hear supersonics when all it turns out to be is that they are adding distortion between 20 and 20 kHz.... |
Hi Bob,
Question for you. In a high quality analog mix board is there typically enough nonlinearity for supersonic frequencies to create audible subtones? Particularly as a result of mixing tracks together? (sorry if this was answered a hundred pages and a thousand insults ago)
If the answer is yes, like I suspect it is, then that means that it may be useful to have information above 20 k when mixing analog in order to A) preserve the type of nonlinear effects that sounds may have when blendind in the air, and B) because that extra stuff just sounds good.
-R
Post by: dcollins on September 28, 2005, 06:29:17 PM
| RKrizman wrote on Wed, 28 September 2005 14:37 |
A) preserve the type of nonlinear effects that sounds may have when blendind in the air, and B) because that extra stuff just sounds good. |
Under normal conditions there are no new frequencies created when signals mix in air.
Although the mixing console may have .1% IM distortion......
DC
Post by: JamSync on September 28, 2005, 07:13:40 PM
| dcollins wrote on Wed, 28 September 2005 23:29 | ||
Under normal conditions there are no new frequencies created with signals mix in air. Although the mixing console may have .1% IM distortion...... DC |
You should read the works by MIT's Joseph Pompeii concerning the nonlinear interaction of ultrasonic frequencies. This is the basis for the Audio Spotlight. I was privileged to be at the AES where he first demonstrated the technology. Distortion by air causes audible by-products. It is this process that enables the Audio Spotlight to work. It's quite impressive to "hear" sound travelling across your face.
I can assure you that conditions were as normal as most AES gatherings allow.
>>>>>Pompeii, F. J., Proc.105th AES Conv, Preprint 4853 (1998)
Post by: Bob Olhsson on September 28, 2005, 08:26:20 PM
| Johnny B wrote on Tue, 27 September 2005 18:12 |
The biology and medical ref above comes from their claims about what amounts to their idle speculation about the frequency spectrum as far as the ear/brain/body interaction is concerned. They cannot prove their claims with sound medical science that only 20 to 20 is important, other contrary evidence suggests that the "important" frequency spectrum is far greater. |
EVIDENCE? such as...
I'm aware of many years of medical and psychoacoustic studies suggesting the opposite. I actually took a college course in psychoacoustics. We learned that there's a lot of flat-Earth science that is commonly cited in this area but the credible sources all seem to agree on the issue of bandwidth.
I'm sure that all of us are more than willing to learn something new.
Post by: Duardo on September 28, 2005, 08:48:33 PM
| Quote: |
I can't help it if 44k cannot be made to work, even though the theory says it should. |
Are you saying that you don't think that 44.1 kHz has improved over the past twenty years, and/or that it can't get any better?
| Quote: |
Or are they claiming to be experts in matters based purely on a layman's attempt at understanding, and giving opinions as to matters for which they are not really qualified. |
They are certainly more qualified than you are, and they obviously understand these things better than you do.
| Quote: |
The biology and medical ref above comes from their claims about what amounts to their idle speculation about the frequency spectrum as far as the ear/brain/body interaction is concerned. They cannot prove their claims with sound medical science that only 20 to 20 is important, other contrary evidence suggests that the "important" frequency spectrum is far greater. |
There is far more evidence...medical, biological, pick your discipline...that 20 to 20 is enough than not. If you were to poll a group of people from all of the disciplines you keep insisting need to weigh in on this you'd find that the majority of them would say that 20 to 20 is enough as well. The people who disagree with that are people who hang out in places like this...people who can hear that things aren't necessarily "right" but don't know exactly why, but have very strong convictions as to why without anything but speculation to back it up. There is much more evidence to show that the issues that you have with digital audio are not directly related to reduced bandwidth than there is to the contrary.
| Quote: |
On the one hand, you have those evangelists who say that digital sound quality is just great the way it is, that in fact, no improvements are desireable or even possible... |
Who has said that? Please, point me to one post on this thread where someone has said that.
-Duardo
Post by: i dig music on September 28, 2005, 10:58:27 PM
Anyhoo, it's getting close to 30 pages now, but it's no where near a resolution that anyone would care about, or even write to Bill or GM about.
What a shame, all the modern technology hasn't brought us to audio nirvana yet, but it has brought us to the on going digital analog summing sample-rate daw tape hardware software fuckfest.
Like it really matters now?
The reality is none of us would be able to sit in front of our CPU and belittle each other on a daily basis if it weren't for the digital audio revolution. It's cheap enough for everybody to get in on the fun.
Forums like this would not have existed in the days where it cost 1 to 2 million to build some walls and role tape for a $xxx an hr. Because in the old days, there were very few gigs to go around in this biz, and there was certainly no time to chat about bullshit that is irrelevant to the work and music. Actually, I would say were all pretty lucky to be able to talk about this stuff here, because many years ago the way things were done from studio to studio, it really was like living on different planets. Your experience was limited to a facility, big or small, and a place usually had a way of doing things their way. Plus you had to deal with actual faces on a day to day basis.
Now with the current technology and the internet it's all wide open to share and get connected, which does happen sometimes in a good way, but then there is always the guy that just can't seem to let something go, and his ego shows up and beats off to his own sorry ass drum sounds. And it wouldn't have mattered what format he tracks on.
Oh well, carry on. I'm not completely numb yet.
Post by: minister on September 28, 2005, 11:36:13 PM
Post by: dcollins on September 28, 2005, 11:40:04 PM
| electrical wrote on Wed, 28 September 2005 13:04 |
I can, or at least I could a year ago. I don't believe I'm special. I believe that if I filled your room with 21kHz you would hear it. |
Can you sweep it right up from 10k to 21k and hear the whole thing?
I'm still good for 16-17k so I guess I have a couple years left -- few complaints about my masters setting people's drapes on fire -- not yet anyway.
The last time I tried this, with a paltry 20k tone, I did hear something, but it went "pit........pit" and was the sound of $350 worth of tweeters. It's hard to keep your hand off the volume control when you just know it's there.
Ok, so I just connected two signal generators to the RMS meter, and guess what? I'm half wrong!
The HP 3325A barely changes from sine to square (about -0.1dB) however a cheapo generator is +3.3dB relative to the sine.
| Quote: |
I just think that categorically denying that people percieve sound above 20kHz is contrary to my experience, and I believe it will be contrary to Dave Collins's experience once he tries this experiment. |
For you, Steve, I will lug the HP into the studio and try it. I wonder if my Radio Shack SPL meter reads at 21k?
What speakers are you using?
I have also "sensed" these frequencies (at high yet non tweeter-blowing levels), but I want to know if the ear was involved.
DC
Post by: dcollins on September 29, 2005, 01:54:10 AM
| JamSync wrote on Wed, 28 September 2005 16:13 |
You should read the works by MIT's Joseph Pompeii concerning the nonlinear interaction of ultrasonic frequencies. This is the basis for the Audio Spotlight. |
Slapping the mat three times here, but I still wonder how this applies to "real-world" levels of audio.
| Quote: |
Distortion by air causes audible by-products. It is this process that enables the Audio Spotlight to work. It's quite impressive to "hear" sound travelling across your face. |
Is this desirable?
How loud is "nonlinear," anyway?
| Quote: |
I can assure you that conditions were as normal as most AES gatherings allow. |
Translation:
Don't order the Salisbury steak.
DC
Post by: JamSync on September 29, 2005, 02:17:44 AM
| dcollins wrote on Thu, 29 September 2005 06:54 | ||
Is this desirable? DC |
Not only is it desirable, it's being used in museums and other public places where a narrow cone of audio dispersion is desirable. It has *everything* to do with innovative surround technologies and *everything* to do with sound placement. It is not the transducer that is responsible for the narrow dispersion, but the character of the ultrasonic frequencies. It relates back to the '60's and the development of Sonar. Why don't you actually read the stuff before you simply dismiss it? (And BTW, the crazies who have glommed on it as a method of "mind control" due to its 4-degree angle at 400 Hz or whatever...yes, they really are completely off the wall and out of the ballpark, even though they cite the research. It's an incredible misuse of perfectly good science.)
Here's another reference, if you can handle the math: http://www.ia.csic.es/sea/sevilla02/ult04021.pdf. Westervelt was heading in the right direction; he just didn't nail the way to make the product. That was left to Pompeii.
My last post on this thread. If you want to be caustic, go ahead. Your statement "Under normal conditions there are no new frequencies created with signals mix in air" is inaccurate, misleading and untrue. The fact that you are not aware of the application of Pompeii's research does not make it irrelevant to this discussion. Have respect where respect is due. I really don't like the condescending tone in a lot of the posts around here, but if you don't know the research, you should admit it and learn it before you make smart-ass comments. This should also go for those who declare that "speakers can't repro ultrasonic frequencies". Bullshit.
Now, what to do with transducers that can and do repro those frequencies and how to use them in practical applications that relate to audible program content and the distribution of such content is the subject for a much different thread than this one has become.
Post by: Eric Bridenbaker on September 29, 2005, 03:49:49 AM
| JamSync wrote on Thu, 29 September 2005 02:17 | ||||
Not only is it desirable, it's being used in museums and other public places where a narrow cone of audio dispersion is desirable. It has *everything* to do with innovative surround technologies and *everything* to do with sound placement. It is not the transducer that is responsible for the narrow dispersion, but the character of the ultrasonic frequencies. It relates back to the '60's and the development of Sonar. Why don't you actually read the stuff before you simply dismiss it? (And BTW, the crazies who have glommed on it as a method of "mind control" due to its 4-degree angle at 400 Hz or whatever...yes, they really are completely off the wall and out of the ballpark, even though they cite the research. It's an incredible misuse of perfectly good science.) Here's another reference, if you can handle the math: http://www.ia.csic.es/sea/sevilla02/ult04021.pdf. Westervelt was heading in the right direction; he just didn't nail the way to make the product. That was left to Pompeii. My last post on this thread. If you want to be caustic, go ahead. Your statement "Under normal conditions there are no new frequencies created with signals mix in air" is inaccurate, misleading and untrue. The fact that you are not aware of the application of Pompeii's research does not make it irrelevant to this discussion. Have respect where respect is due. I really don't like the condescending tone in a lot of the posts around here, but if you don't know the research, you should admit it and learn it before you make smart-ass comments. This should also go for those who declare that "speakers can't repro ultrasonic frequencies". Bullshit. Now, what to do with transducers that can and do repro those frequencies and how to use them in practical applications that relate to audible program content and the distribution of such content is the subject for a much different thread than this one has become. |
Thanks a Mil, KK!!! Will definitely be checking out the Pompeii!!
I remember hearing about these things, seeing some of the prototypes on TV... Contained, highly directional sound, sort of like the benefits of headphones, without having to wear any. Last I heard about this was that they were still working on getting the bass frequencies to sound good, but for the purposes of this thread, who cares.
Having had a lifelong interest in ultrasonic phenomenon and applications, and because of it having to continually butt heads with the "20K" barrier, it's good to see that someone is actually doing something new, and useful with this technology. One that so many, even those committed to progress in the field of audio have already written off.
She blinded me.......
With science!!
Best,
Eric
Post by: crm0922 on September 29, 2005, 04:14:47 AM
I notice the difference in the "spatiality" of recordings on my 2" machine. Things sound good, rock bands rock, jazz stuff oozes sleaze. Etc. I love it. Here's the rub, though, if I transfer it to digital, the effect translates as well. At least as well as everthing else translates, given converter errors, jitter, cabling, yadda.
So analog sounds different, but I'll bet someone could A/B a lot of folks with audio sourced from tape and a digital copy and most couldn't tell the difference, given quality conversion.
I played some sine waves tonight. 19k I hear a clear tone. 20k I think I hear something, and 21k maybe. I am skeptical that my hearing still extends past 19k, let alone 20 or 21k. So must something else be happening, such as audible high frequency errors from the solid state electronics and such in the signal path? (PC->SPDIF->HEDD->passive attenuator->SS amplifier->spkr).
Chris
Post by: Johnny B on September 29, 2005, 04:19:27 AM
This is also my very last post until George gets back.
So in order to use this technology and control those extra high frequencies what will we need? Could it be?....emmmm....higher sample rates...
And for those who will be moving foward and doing more surround work....they might need those....what?....emmmm....higher sample rates
I'm just guessing, 'cuz...I've been slapped down here and have been called all kinds of names like "idiot" and so forth...
And Thanks Steve,
You got people to begin testing again, always a good thing.
I will also thank George in his absence.
To me, this has been a great thread because it has spurred some action and maybe some new thinking. Its only black marks are all the personal attacks which are uncalled for and perhaps could be considered unprofessional or rude...I hope that people can control their emotional outbursts in the future...I know I've been guilty of it myself a time or two, if I offended anyone....I'm truly sorry.
And thanks Brad,
You're a good guy too.
Someone has to mind the farm when the farmer goes to market.
I think you're doing a damn good job.
I'll finish with one last question:
20kHz---has the myth been shattered?
Post by: Jon Hodgson on September 29, 2005, 05:15:40 AM
| JamSync wrote on Thu, 29 September 2005 07:17 | ||||
Not only is it desirable, it's being used in museums and other public places where a narrow cone of audio dispersion is desirable. It has *everything* to do with innovative surround technologies and *everything* to do with sound placement. It is not the transducer that is responsible for the narrow dispersion, but the character of the ultrasonic frequencies. It relates back to the '60's and the development of Sonar. Why don't you actually read the stuff before you simply dismiss it? (And BTW, the crazies who have glommed on it as a method of "mind control" due to its 4-degree angle at 400 Hz or whatever...yes, they really are completely off the wall and out of the ballpark, even though they cite the research. It's an incredible misuse of perfectly good science.) Here's another reference, if you can handle the math: http://www.ia.csic.es/sea/sevilla02/ult04021.pdf. Westervelt was heading in the right direction; he just didn't nail the way to make the product. That was left to Pompeii. My last post on this thread. If you want to be caustic, go ahead. Your statement "Under normal conditions there are no new frequencies created with signals mix in air" is inaccurate, misleading and untrue. The fact that you are not aware of the application of Pompeii's research does not make it irrelevant to this discussion. Have respect where respect is due. I really don't like the condescending tone in a lot of the posts around here, but if you don't know the research, you should admit it and learn it before you make smart-ass comments. This should also go for those who declare that "speakers can't repro ultrasonic frequencies". Bullshit. Now, what to do with transducers that can and do repro those frequencies and how to use them in practical applications that relate to audible program content and the distribution of such content is the subject for a much different thread than this one has become. |
Firstly KK, thanks for posting that link. I've heard about this technology but this is the first time I've seen anything technical about it, so I'm very pleased to finally be able to learn about it properly. I'll have to try to get hold of Pompeii's paper too.
However if anything it would appear that rather than supporting a need to sample frequencies above the hearing threshold (be that threshold 20k or 200k), it may be a reason not to in most cases - certainly for the kind of thing that goes on in most studios.
At least here is my reasoning, I may be missing something....
So far my understanding is that the interaction of waves in air above the hearing threshold combined with the non-linearity of air generates frequency components that can be heard (sum and difference frequencies, a well known result of non linearities).
The original signals above the hearing threshold you don't hear, and the sum frequencies you don't hear, which leaves only the difference frequencies that you do hear.
Now the thing is that when you stand there in the room, those difference frequencies exist, they have already been created by the non linearity of the air, that's why you can hear them, so any transducer that responds to frequencies across your whole audible range will pick them up, so if you record a signal from that transducer onto an equally good recording medium and play it back on an equally good transducer, you'll get back what you heard... so far so good.
Now if you record that one signal at above your hearing threshold and play it back, then still you hear what you would have in the room.
But what happens if you record multiple signals from multiple locations, possibly at multiple times? As you often do in modern recording. You would then have a whole mishmash of ultrasonic components which when played at the same time might produce all kinds of unpleasant effects in the audible range... or they might be pleasant effects, I don't know... but I would suspect they would be unlikely to combine to generate exactly the same audible components they would have if they had all been playing in the room at the same time.
Shame about the lack of mind control though, I could really make good use of that!
Post by: Andy Simpson on September 29, 2005, 07:25:04 AM
| crm0922 wrote on Thu, 29 September 2005 09:14 |
Ok, back on target everyone. I notice the difference in the "spatiality" of recordings on my 2" machine. Things sound good, rock bands rock, jazz stuff oozes sleaze. Etc. I love it. Here's the rub, though, if I transfer it to digital, the effect translates as well. At least as well as everthing else translates, given converter errors, jitter, cabling, yadda. So analog sounds different, but I'll bet someone could A/B a lot of folks with audio sourced from tape and a digital copy and most couldn't tell the difference, given quality conversion. I played some sine waves tonight. 19k I hear a clear tone. 20k I think I hear something, and 21k maybe. I am skeptical that my hearing still extends past 19k, let alone 20 or 21k. So must something else be happening, such as audible high frequency errors from the solid state electronics and such in the signal path? (PC->SPDIF->HEDD->passive attenuator->SS amplifier->spkr). Chris |
Are you saying that if you record to tape you get good spatiality (which then is preserved when capturing digitally) but if you record to digital you don't get the good spatiality?
How about if you track digital and then bounce the whole lot to multitrack tape before mixing?
If we are talking about distortion, this should (in theory) give you the same spatiality 'distortion'.
...
Andy
Post by: maxdimario on September 29, 2005, 08:02:48 AM
if the tape gets fed the original waveform, it will highlight the original waveform in a way that the ear intuitively 'fills in the blanks'
if the distortion in the first stage of the signal path is related harmonically and 'musically' to the original waveform this helps the waveform translate better to different lower-resolution mediums and playback systems, much like the top of an acoustic guitar adds intelligibility to the string vibrations.
if ON THE OTHER HAND the distortion artifacs of the first stages are NOT RELATED harmonically etc. to the original waveform, it does NOT help intelligibility, but creates 'distance' or un-intelligibility.
the original waveform, when 100% undistorted, if played back on an excellent system will sound three-dimensional and open.
unfortunately most audio equipment, especially consumer reproducers reproduce very clean waveforms in a way that can sound sterile, and the fine detail, which clean audio relies on to 'make sense' to the ear, is lost.
so tape adds harmonics and artifacts which are related to the original signal, therefore 'spreading out' the waveform similar to the way that we see a thick line better than a fine line.
this way the ear can make sense of the original waveform even on inferior systems.
once the distortion is added, every piece of the audio chain afterwards will be elaborating on the sum of the original waveform and the added distortion, so it is important to use equipment which adds natural and 'sonically correct' distortion artifacts in the critical first parts of the chain.
I think, that math aside, everyone will agree that digital distortion, speaking in terms of quality and not quantity, is the worst kind known to man.
Post by: jimmyjazz on September 29, 2005, 10:45:12 AM
Post by: dcollins on September 29, 2005, 03:18:09 PM
| JamSync wrote on Wed, 28 September 2005 23:17 |
Your statement "Under normal conditions there are no new frequencies created with signals mix in air" is inaccurate, misleading and untrue. The fact that you are not aware of the application of Pompeii's research does not make it irrelevant to this discussion. Have respect where respect is due. |
Hey, if I'm wrong I'll be the first to admit it!
At what SPL do these things operate?
DC
Post by: Ronny on September 29, 2005, 03:36:35 PM
| dcollins wrote on Thu, 29 September 2005 15:18 | ||
Hey, if I'm wrong I'll be the first to admit it! At what SPL do these things operate? DC |
I have mixed two inaudible freq's in a DAW and heard a tone below both of them, but measuring the lower tone inside the DAW just shows a blank graph, where is it really coming from? I'm wondering what causes this, I'm pretty sure that it's not the summed frequencies and can only speculate that it's IM distortion or some type of effect of the DAC components. Doesn't happen in air, only inside the digital realm. Anyone have a clue as to why this phenomenon occurs?
Post by: Eric Bridenbaker on September 29, 2005, 06:20:35 PM
| Ronny wrote on Thu, 29 September 2005 15:36 | ||||
I have mixed two inaudible freq's in a DAW and heard a tone below both of them, but measuring the lower tone inside the DAW just shows a blank graph, where is it really coming from? I'm wondering what causes this, I'm pretty sure that it's not the summed frequencies and can only speculate that it's IM distortion or some type of effect of the DAC components. Doesn't happen in air, only inside the digital realm. Anyone have a clue as to why this phenomenon occurs? |
Hey Ronny,
A bunch of us here tried this, with mixed results. Something can definitely be heard, but it doesn't show up on FFT. Strangely enough, the tone survives conversion to mp3, though the waveshape is competely changed.
BTW: Beware with these types of tests, I'm pretty sure I blew up a pair of tweeters trying it, cause they didn't work after I was done (the volume was way too high), and if Tim de Paravicini is right, you can end up giving yourself temporary tinnitus, without even hearing a note.
Here's the thread, if you feel like checking it out...
http://recforums.prosoundweb.com/index.php/m/78014/4645/?SQ= 072f3101e405a6e4e3b5e27268c572a0
It got a little nasty, but not as crazy as this thread. I started it with a theory that's quite outside the box, and not everyone was down with it. I've refined my ideas since then as to what this thing is..... still don't know exactly, but I'm betting that it is a non-linear phenomenon.
Best,
Eric
Post by: dcollins on September 29, 2005, 07:18:57 PM
Post by: crm0922 on September 30, 2005, 01:44:13 AM
I haven't dumped from digital back to tape. That always seems to degrade things more than it helps. So I haven't listened for any better feeling of space, etc.
However, tracking to 2" and dumping the multitrack back to digital sounds identical to me, spatiality and all. Plus or minus a little bit that I would obviously expect from any analog-cabled transfer.
It seems easier to define the locations of the instruments and you get a little "excited" knowing when a particular track is going to kick in, or a drum fill or something because it leaps to life.
Want to know my guess as to why?
I didn't think so, but here it is:
I think the high noise floor of tape actually slighly reduces headphone bleed, instrument bleed, console crosstalk (maybe), and thus provides each track a more isolated "space" with less low-level interactions among virtually unrelated tracks. This combined with the fact that analog tape hiss is not offensive (to me) allows me to ignore said noise and focus on the music.
Thus, I reduce it to the argument that it is actually analog's "deficiencies in accuracy" that may provide this "improved spatiality" (as you would phrase it).
I also think this effect could be simulated on analog pretty easily. I was thinking of recording some tape hiss of my studer and individually mixing it with a pre-recorded digital set of tracks. I'm sure this has been done before, and is probably foolish, but I am curious.
Maybe someone will open a "vintage tape hiss" sample shop online with "the hiss used by the pros".
Now, Andy, please admit that Nyquist was right, and digital sampling does work as described!
Chris
Post by: Norwood on September 30, 2005, 02:06:43 AM
Post by: maxdimario on September 30, 2005, 08:32:59 AM
| Quote: |
Andy doesn't seem to deny that digital works as described, he just has slightly odd theories as to why it sounds the way it does. Max and Johnny... etc. |
Nyquist works as a mathematical model perfectly, although quantization was not part of the nyquist theory.
filters are far from perfect, as are complex digital machines.
nobody really knows what the ear looks for in sound in order to interpret it, we can only observe our reactions to be sure.
in analog, there are similar problems, and the way to get around them is to not rely on what theoretically should be, but on what actually works.
more harm has been done to music and society in general by people who preach an ideal but disjointed reality, ignoring the evident everyday truth of the matter.
digital is not perfect, only the nyquist theorem (as dc pointed out) is perfect.
no analog engineer would put all of the weight of a system's high end resolution on the 'shoulders' of a filter.... especially a boxcar filter, it is impossible to create such a filter with perfect characteristics.
perhaps this can be done in the digital domain, perhaps not.
I wonder if the added computing power needed to execute these filters etc, will not cause further distortions ...that will need to be weeded out further, and more time will pass.
the system is not perfect, and the imperfections although less in quantity are more damaging to sound.
Post by: Andy Simpson on September 30, 2005, 08:41:00 AM
| Norwood wrote on Fri, 30 September 2005 07:06 |
Andy doesn't seem to deny that digital works as described, he just has slightly odd theories as to why it sounds the way it does. Max and Johnny(no offense) need the convincing. |
Indeed, I agree that it works to the extent that we currently understand.
However, as the name of this thread suggests, I intended to get at digital from the perspective of spatial timing and we really don't know how much resolution we require in this area to be convinced.
I set out to consider why we might prefer 192 over 44 despite our hearing bandwidth.
The increased perceived bitdepth of higher sampling rates is a valid possible reason for hearing an improvement, since we all prefer 24 over 16 as a capture depth.
Infact, if (as Ronny stated in a different thread) 24bit dithered to 16bit sounds as good as the original 24bit, we might consider tape to be a high resolution format complete with dither that translates to digital well (because of the 'dither').
Andy
Post by: timrob on September 30, 2005, 11:01:23 AM
| maxdimario wrote on Fri, 30 September 2005 07:32 |
no analog engineer would put all of the weight of a system's high end resolution on the 'shoulders' of a filter.... especially a boxcar filter, it is impossible to create such a filter with perfect characteristics. |
Hate to break it to you. Magnetic heads are filters. As in Inductor.
Post by: maxdimario on September 30, 2005, 12:03:19 PM
yeah and tape machines run on AC current, does that mean that they have an inner 60 Hz clock??
the Nyquist theory is based on a perfect reconstruction filter, which doesn't exist.
a tape head is NOT a reconstruction filter, nor are tape eq filters.
Post by: timrob on September 30, 2005, 12:32:44 PM
| maxdimario wrote on Fri, 30 September 2005 11:03 |
what the heck does that have to do with it?? yeah and tape machines run on AC current, does that mean that they have an inner 60 Hz clock?? the Nyquist theory is based on a perfect reconstruction filter, which doesn't exist. a tape head is NOT a reconstruction filter, nor are tape eq filters. |
Well, you didn't say reconstruction filter did you.
Sorry, my friend, but analog is limited by its physical and electrical characteristics. It also has limitations based on the quality of AC, so yes it does in effect rely on that 60hz to run at the proper speed. We may prefer the sound of analog, but it ain't any more perfect than digital. In fact it is less perfect.
I think of analog in the sense of camera obscura. The picture has a nice dimension and impressionistic feel, but the image is obscured by the lens and the media used. Same with analog tape.
Post by: crm0922 on September 30, 2005, 03:39:01 PM
I think the argument is obscured here. My opinion is that, while the sound of analog is pleasing, it is less accurate and more prone to phase error, time smear, and other effects you attribute to digital.
Decent digital recording is far more accurate in every measureable way. The rest of the perception of apparent "sound quality" is more likely based on analog non-linearities and distortions. Ultimately, I believe these could be faked in a digital system with acceptable believability.
However, many engineers would not like this, and may prefer the "cold" sound of digital because it is more accurate to the original waveform produced by the source.
And yes, the reduced dynamic range of analog when recorded onto a digital device will place an analog noise-floor well above the digital noise, perhaps masking them. Whether this is perceived as "better sounding" or not is a matter of opinion.
Chris
Post by: J.J. Blair on September 30, 2005, 10:44:03 PM
| maxdimario wrote on Fri, 30 September 2005 09:03 |
tape machines run on AC current, does that mean that they have an inner 60 Hz clock?? |
Actually, most tape machines have a DC motor. I 'm surprised you don't know that, somebody as knowledgable as yourself. Only a handful of old machines have AC syncronous motors, and then the audio circuitry is DC. I thought you knew everything? Even I know this shit, and I'm the least technical guy in this conversation.
Post by: bobkatz on October 01, 2005, 01:19:53 AM
| timrob wrote on Wed, 28 September 2005 11:56 | ||
Hey Bob, Don't want to pick nits BUT... The way I learned harmonics is that the first harmonic is the Fundamental. |
You deserved to pick nits, I was technically incorrect! Sorry, I meant the "first overtone". You are correct, it's the second harmonic. I just wasn't thinking. Every time I say "harmonic" I have to count on my fingers
| Quote: |
So really 14k is the second harmonic of 7khz which would not be present in a Square wave which contains 1,3,5,7... It would be in a triangle wave though which contains 1,2,3,4... The relative amplitude and phase of the harmonics is also |
Ahhh, so, so I'm falling flat here on the example! My bad. I guess the only one of my points that remains is that the non-linear transducers are the probable cause of our hearing differences between high frequency sine, square, and triangle waves still remains...
Post by: timrob on October 01, 2005, 01:38:30 AM
| J.J. Blair wrote on Fri, 30 September 2005 21:44 | ||
Actually, most tape machines have a DC motor. I 'm surprised you don't know that, somebody as knowledgable as yourself. Only a handful of old machines have AC syncronous motors, and then the audio circuitry is DC. I thought you knew everything? Even I know this shit, and I'm the least technical guy in this conversation. |
Ouch J.J. I guess I stepped into that one myself. I was thinking about this old mono tape machine that my buddies dad had in his basement. It ran on bohemoth ac synchronous motors. IIRC, he had it rigged to run off a 220 circuit. Anyway, It was the basis for a lot of my earliest learning in electronics. The same guy also got me into Heathkit Projects. Anyway, you're right that modern machines with properly rectified Power supplies don't exhibit speed fluctuations due to line frequency the way that old thing would.
Post by: timrob on October 01, 2005, 01:46:19 AM
| bobkatz wrote on Sat, 01 October 2005 00:19 | ||||||
You deserved to pick nits, I was technically incorrect! Sorry, I meant the "first overtone". You are correct, it's the second harmonic. I just wasn't thinking. Every time I say "harmonic" I have to count on my fingers
Ahhh, so, so I'm falling flat here on the example! My bad. I guess the only one of my points that remains is that the non-linear transducers are the probable cause of our hearing differences between high frequency sine, square, and triangle waves still remains... |
Sorry, If I diverted attention from that point. I think it is an important one.
Post by: J.J. Blair on October 01, 2005, 04:15:39 AM
Post by: 12345 on October 01, 2005, 05:14:06 AM
DC, if I'm not mistaken a periodic wave must by nature be steady-state.
Post by: Eric Bridenbaker on October 01, 2005, 09:30:41 AM
http://www.ee.ualberta.ca/~beaulieu/online_pubs/ncb_cct_mod_ clett.pdf
Post by: maxdimario on October 01, 2005, 09:34:55 AM
| J.J. Blair wrote on Sat, 01 October 2005 04:44 | ||
Actually, most tape machines have a DC motor. I 'm surprised you don't know that, somebody as knowledgable as yourself. Only a handful of old machines have AC syncronous motors, and then the audio circuitry is DC. I thought you knew everything? Even I know this shit, and I'm the least technical guy in this conversation. |
I don't know who this was directed at..
anyway old machines ran on Ac and you could use sine wave oscillators and power amps to vary the speed.
I had a big custom LFO unit with a big tube power amp built-in, from a telefunken setup, once, designed for just this purpose.
the fact that those tape machines run on ac current and they are frequency dependent, doesn't mean they have an inner clock, or that 60 Hz ac current behaves like a clock in a digital system.
it only has to do with the way the motors are run.
you could have a car battery running the motors with dc motors and a governor-style speed control like on cheap cassette players, if you wanted to.
Post by: J.J. Blair on October 01, 2005, 12:05:53 PM
Post by: Level on October 01, 2005, 07:29:53 PM
| Quote: |
http://bigwww.epfl.ch/publications/unser0001.pdf |
Post by: Tim Gilles on October 03, 2005, 03:20:13 AM
| dcollins wrote on Mon, 26 September 2005 21:56 |
I'm reminded of the beginning of "Animal House," long tracking shot up the lawn of university -- close-up on the founders statue -- tilt down to the plaque that reads: "Knowledge is Good" But who needs books when we have the Internet? DC |
Dave. I just saw this!
Rolling on the floor laughing thinking about how I went to see this movie when it was new in theatres and my older brother Steve and I fell down laughing at the sheer absurdity of the opening shot....
...Only to find...
That...
No one....
Else....
in the completely jammed theatre...
...was....
Laughing.
.....
Ahh me.
Ya know yer alone when even the penguins are heading in the other direction.
Best regards to all,
Tim "Rumblefish" Gilles
Post by: PP on October 03, 2005, 11:05:59 AM
Post by: Invisible Member on October 03, 2005, 04:32:42 PM
I didn't have time to sort through every post but there seems to me a missing part(s) of the discussion.
Simply put, not everything sounds the same to everyone. You could put 2 people in a experiment that reproduces a sound with minimal coloration and the two people will percieve the samething in two different ways.
The problem that science has is when data must be evaluated via the 5 senses. While mathmatically it may work, percieving is completely unscientific. When trying to describe the taste of a food which has been engineered and reproduced with stunning accuracy the problem of perception and description falls outside of "Science".
Your discussing 2 parts of an nearly infinitely long equation.
I'm doing some studying on vibration and the class shows how complex just particle propagation is on Carbon reinforced structure. . Using B&K arrays and Near Field Acoustic Halography take a solid panel and introduce a sinewave vibration and watch not only the fequency morph but the centroid of the load move around.
Every listening environment is going to do the same thing a little different everytime.
No matter how good you think your ears are and how balanced your room is, there is alot of things going on outside of traditional analytical methods.
Sevareid's Law: "the chief source of problems is solutions"
"For every noise we control we uncover the next noise that we need to control"
You might start thinking about what size and orientation your coffee cup is because it probably changes what you hear more than you think.
Peace,
Dennis
Post by: dcollins on October 03, 2005, 05:28:02 PM
| notyournamehere wrote on Mon, 03 October 2005 13:32 |
Sevareid's Law: "the chief source of problems is solutions" |
"Problems are the price of progress. Don't bring me anything but trouble. Good news weakens me."
-- Charles Kettering
DC
Post by: compasspnt on October 03, 2005, 05:43:07 PM
| dcollins wrote on Mon, 03 October 2005 17:28 | ||
"Problems are the price of progress. Don't bring me anything but trouble. Good news weakens me." -- Charles Kettering DC |
Wow, Dave, that's good news...I mean, bad news...I mean...I must really be progressing then!
Seriously, a great quotation!
Post by: Invisible Member on October 04, 2005, 09:50:11 AM
| dcollins wrote on Mon, 03 October 2005 14:28 | ||
"Problems are the price of progress. Don't bring me anything but trouble. Good news weakens me." -- Charles Kettering DC |
Im sure this is on your front door to warn clients. I see problems somewhat similar if not just for engaging challenge which often leads to unsatisfactory results. But on the other side problems are also the warning for disease and electrical fires.
Maybe you should host a tv series called Survivor: Mastering Audio Edition.
When something does what you want it's called science. When something doesn't do what we want it's called art. Reconciliation of the two may be difficult.
What's the goal again?
Accurate reproduction or Artistic interpretation?
Peace,
Dennis
Post by: maxdimario on October 12, 2005, 07:39:08 PM
progress can better man's life, make him happy, or it can make some men more happy than others, or it can make happy men out of men that were less happy..etc.
who is happy with digital?
what kind of clients, what kind of operators, what kind of manufacturers?
Post by: compasspnt on October 12, 2005, 09:07:52 PM
| maxdimario wrote on Wed, 12 October 2005 19:39 |
who is happy with digital? what kind of clients, what kind of operators, what kind of manufacturers? |
Max,
Back "in the day" of only-analogue, speaking here about the 60's and 70's especially, but also extending into the 80's a bit, almost no one was truly satisfied with analogue. We all wanted much less noise. So tape manufacturers were pushed to come up with new formulations which allowed higher recording levels and less noise. We all wanted less wow and flutter. So tape machine manufacturers were pushed to come up with better transport systems. We all wanted less maintainence and easier alignment procedures. So tape machine manufacturers came up with one button alignment, modular "works in a drawer" cards, etc. We all wanted higher track count capability. So manufacturers came up with wider tape widths, smaller track width, and sync devices. We all wanted easier punching ability. Manufacturers did the best they could. We all wanted better editing possibilities. So manufacturers provided easier access to heads, "rocking" capability, magic scissors, etc. We all wanted lower costs. That never happened.
Now all of these things, and much more, have come to pass. And it is available to almost everyone with a few dollars or pounds or euros or yen or dinars in their pocket.
And what do so many want?
More analogue sounding plug-ins.
Humans are never satisfied.
In general, that is a good thing. Improvment is possible in all platforms, and we should strive for it. But we should learn to use what we have at any one time to the best of our abilities, and be "humanly unsatisfied" with our performances.
Analogue has not disappeared off the face of the earth (yet, at least). So you use it. I'll use it.
Digital has opened up a wonderful world of convenience and capability. I will use that also. But digital can be abused by pretenders, just as analogue could be. If you don't learn how to use digital properly, it will sound bad, unmasked by analogue properties.
I wonder if the users of the next big format will be looking back to the golden days of 16 bit, and trying to achieve it?
Post by: dcollins on October 12, 2005, 09:54:45 PM
| compasspnt wrote on Wed, 12 October 2005 18:07 |
I wonder if the users of the next big format will be looking back to the golden days of 16 bit, and trying to achieve it? |
There's a guy looking for a 1630 system over on Brads board. He hopes to eliminate that evil computer, I suppose.
When I hear people talk of "Vintage Black-face" and they mean ADAT's................
DC
Post by: Duardo on October 12, 2005, 10:17:20 PM
| Quote: |
who is happy with digital? |
I'm pretty happy with digital. That doesn't mean that I think it's perfect, but analog wasn't either. I worked in a 2" facility for years and I like tape, but for my home studio I'm much happier with my digital setup for many reasons...cost, maintenance, space, noise, size...and it sounds really good. I'd say that the system overall sounds at least as good as the analog studio did. Not the same, and there are times I'd certainly like to have all the analog stuff at my disposal, but I'd say I'm happy with it.
I'm also excited to see what improvements can be made in the coming years.
-Duardo
Post by: danickstr on October 12, 2005, 10:34:06 PM
Post by: vernier on October 12, 2005, 11:30:12 PM
Post by: Jon Hodgson on October 13, 2005, 05:16:49 AM
| vernier wrote on Thu, 13 October 2005 04:30 |
Analog hasn't been more costly for me ..maintenance consists of cleaning heads and occasional degauss. I've upgraded computers and disk drives more times than I can count, but my all-analog settup has required two things in the past ten years ..pot for a limiter and playback head for the two-track. I won't go in to detail about the hours messing with computers, the costs, disk-space, backup, upgrades ..it's just not a better mouse-trap. |
But what if you'd simply not upgraded your DAW in that time... or what if you'd upgraded your analogue system as often?
I think that in either domain if you have a system that works to your satisfaction and just stick with it, the maintainance hassles are probably similar. The problem is often that since it is usually so easy to update computer software and even hardware (95% of the time you just download, install and go) that people get in the habit of doing it any time, and if it goes wrong then Murphy's law says it'll probably be when you've got a really important session looming, then the stress helps blow the whole problem out of proportion.
If you decided to upgrade your desk, or it had a faulty channel that needed work, you'd probably wait until you had a gap between projects (or make one) and give yourself a nice healthy chunk of free time to do it, enough for the job itself and enough for any problems should they arise.
If more people did the same thing with software, there'd be fewer grey hairs in the industry
Post by: maxdimario on October 13, 2005, 07:41:03 AM
What do you set out to do?
What do you want to achieve.
I have done stuff on a DAW that would have taken a huge studio with 2 assistants working full time as well as technical staff etc. all by myself in a room.
When there are more people involved in a DAW environment they are there usually to help make production decisions of the microscopic type, such as: should we shift this snare hit 1 mS back so that it plays along with the vocals etc.
I don't like that string sound..change it.. I've got one on my hard disk..
this makes for some perfectly constructed productions.
I guess I am just not a big fan of those productions, and would rather see less tracks recorded in a way that it sounds live (and I still stand by the fact that analog recording sounds more live than digital...in the end-product) with better musicians with something to say, as opposed to 'let's see where it takes us'.
I don't think noise is really such an issue, especially with big tape and automation.
if anything digital would be great to implement for gain-riding, which is my favourite aspect of the daw.... even though sometimes I feel that having to rehearse ANY physical movement in order to get it right communicates something extra to the listener.
Post by: Tomas Danko on October 16, 2005, 08:27:28 AM
And it's always been like that.
Anyway, regarding the next generation reaching for old 16 bit recorders to get "that sound", it's already going on within the hip hop- and rap community. I've heard about people choosing early ADAT machines on some songs, when they are usually running PTHD rigs and whatnot, just to get that "old school" sound.
Just in the same way some people are running their old Akai S900-series samplers to get that certain character to their drum samples.
Crunchy.
Of course, in the future people who can not afford pricey vintage ADAT units will just have to pay $99 to get the latest Behringer Crunchalizer Platinum Pro 32/384.
Post by: compasspnt on October 16, 2005, 10:46:47 AM
| Jon Hodgson wrote on Thu, 13 October 2005 05:16 |
But what if you'd simply not upgraded your DAW in that time... or what if you'd upgraded your analogue system as often? |
It seems to me that the manufacturers of the older analogue gear (Ampex, Studer, Neve, API, etc., etc.) were usually trying to sell one big, expensive item at a time, which was designed to have quite a lifespan. Today, the "scam" (dare I call it that...oh well, I've done it) has been figured out. Sell it for less (in most cases), but obsolete it every year or six months. Keep everybody hooked on updating software. "Improve" everything every ____ weeks. Find catchy new marketing angles and product names and artistic designs which will make their "old stuff" (10 months old) appear and look totally uncool.
I know there ARE improvements being made, and I'm not saying that one should keep recording on a PT3 16 bit system for their entire career. I've just found that, in many cases, you can indeed successfully continue using a particular system (with occasional third party upgrades, especially converters) much longer than the manufacturers want you to believe.
Post by: CCC on October 16, 2005, 11:10:03 AM
| compasspnt wrote on Sun, 16 October 2005 15:46 |
I know there ARE improvements being made, and I'm not saying that one should keep recording on a PT3 16 bit system for their entire career. I've just found that, in many cases, you can indeed successfully continue using a particular system (with occasional third party upgrades, especially converters) much longer than the manufacturers want you to believe. |
Just as long as the system doesn't quantize the spatial timing aspects of a recording into chunks of 1.7 cm I'm happy. I understand that this danger can be minimized by using an all tube front end on ProTools HD, mastering to vinyl and then bouncing the vinyl master to red book standard CD.
I'm kidding, of course.
.
.
.
.
.
.
...
Or am I?
Post by: mandel on October 16, 2005, 11:49:15 AM
Concerning this debate, may I ask: are there really no gains in higher sampling rates in that all extraneous information beyond 22KHz are really superfluous to recreating sound wave? Does, for example, sampling at 192KHz reproduces transients more effectively? I've been reading lots of literature but have difficulties understanding the technical explanations.
Post by: Bill Mueller on October 16, 2005, 01:15:23 PM
| compasspnt wrote on Sun, 16 October 2005 10:46 | ||
It seems to me that the manufacturers of the older analogue gear (Ampex, Studer, Neve, API, etc., etc.) were usually trying to sell one big, expensive item at a time, which was designed to have quite a lifespan. Today, the "scam" (dare I call it that...oh well, I've done it) has been figured out. Sell it for less (in most cases), but obsolete it every year or six months. Keep everybody hooked on updating software. "Improve" everything every ____ weeks. Find catchy new marketing angles and product names and artistic designs which will make their "old stuff" (10 months old) appear and look totally uncool. I know there ARE improvements being made, and I'm not saying that one should keep recording on a PT3 16 bit system for their entire career. I've just found that, in many cases, you can indeed successfully continue using a particular system (with occasional third party upgrades, especially converters) much longer than the manufacturers want you to believe. |
Terry,
You are being too kind. The real scam is to build software that does not work and then "upgrade" it for years to get it to do what it was supposed to do in the first place. I was a first adopter of Sound Designer and then Pro Tools. It took YEARS for that software to become stable enough to base your livelihood on. My Tascam MX2424 just got stable on the last software version and then they stopped making it. Operating systems have also just become stable with the latest versions of XP and System X. To try to be fair, software development is an entirely new business, only a blink of time in the continuum. However, we the users, have been paying for dysfunctional crap for too many years.
Best Regards,
Bill
Post by: Jon Hodgson on October 16, 2005, 01:15:28 PM
| compasspnt wrote on Sun, 16 October 2005 15:46 | ||
It seems to me that the manufacturers of the older analogue gear (Ampex, Studer, Neve, API, etc., etc.) were usually trying to sell one big, expensive item at a time, which was designed to have quite a lifespan. Today, the "scam" (dare I call it that...oh well, I've done it) has been figured out. Sell it for less (in most cases), but obsolete it every year or six months. Keep everybody hooked on updating software. "Improve" everything every ____ weeks. Find catchy new marketing angles and product names and artistic designs which will make their "old stuff" (10 months old) appear and look totally uncool. I know there ARE improvements being made, and I'm not saying that one should keep recording on a PT3 16 bit system for their entire career. I've just found that, in many cases, you can indeed successfully continue using a particular system (with occasional third party upgrades, especially converters) much longer than the manufacturers want you to believe. |
I agree, my point was mostly that a lot of the complaints people have about the amount of time they spend dealing with computer problems would be far less if they approached things with more of the attitude they would when buying hardware
1) Unless it's a show stopper, don't upgrade at a time when having any problems would be a big deal, like in the middle of a busy week of sessions.
2) When you do upgrade, timetable in some buffer time in case things go wrong.
Post by: Jon Hodgson on October 16, 2005, 01:31:44 PM
| mandel wrote on Sun, 16 October 2005 16:49 |
I'm a layperson who has been following this thread for some time. Pls note I'm no audio expert or engineer, just someone interested in hi-fi audio. Concerning this debate, may I ask: are there really no gains in higher sampling rates in that all extraneous information beyond 22KHz are really superfluous to recreating sound wave? Does, for example, sampling at 192KHz reproduces transients more effectively? I've been reading lots of literature but have difficulties understanding the technical explanations. |
ok in simple terms...
If you can't hear anything above 20kHz, then sampling at 192kHz is unneccessary, and in fact probably detrimental in the real workd (since you can make a more accurate 96kHz converter than 192kHz).
You can't seperate transients and bandwidth, the two are different views of the same thing. Sharper transients mean higher frequencies and therefire require higher bandwidth. However if you can hear the difference in the transient then this also means that you can hear the higher frequencies.
So if capturing sharper transients makes a difference, that means you can hear over 20kHz, if you can't hear over 20kHz, then that means you don't need to capture sharper transients.
However if you're talking about the accuracy of the transient in time, then that isn't affected by sample rate.
Post by: compasspnt on October 16, 2005, 02:15:39 PM
What Jon said, plus please check this out:
http://recforums.prosoundweb.com/index.php/t/2997/6490/?SQ=5 c62170fa2901f2a00976812f4bd4581
Personally, for tracking I think 48/24 is totally fine for now (as good as any SR/24). I do mix to 96/24, though (although this is as much for having higher SR available for SACD or DVD-A or future formats as it is for better CD quality). Others differ, of course, in their assessments. But Dan Lavry gives good, sound, technical arguments against 192, as we know it today at least.
Post by: maxim on October 16, 2005, 10:56:42 PM
"But Dan Lavry gives good, sound, technical arguments against 192, as we know it today at least. "
as i (?mis)understood it, it all gets too confusing for the computer, and more errors are made
fwiw, i haven't upgraded my hard/soft system for 5 years now, partly out of stinginess, partly out of superstitionness/experience (digital rule #2: don't change horses in midstream), but mainly coz it's working as fine as it did 5 years ago
Post by: Jon Hodgson on October 17, 2005, 02:54:09 AM
| maxim wrote on Mon, 17 October 2005 03:56 |
terry wrote: "But Dan Lavry gives good, sound, technical arguments against 192, as we know it today at least. " as i (?mis)understood it, it all gets too confusing for the computer, and more errors are made fwiw, i haven't upgraded my hard/soft system for 5 years now, partly out of stinginess, partly out of superstitionness/experience (digital rule #2: don't change horses in midstream), but mainly coz it's working as fine as it did 5 years ago |
Not too confusing for the computer, but for the converter.
Well confusing is not exactly the right word.
Basically you have to remember that electrons don't move instantaneously, and don't sit exactly still when they get to their destination, if they did then every electronic engineer would be a Rupert Neve, because achieving high quality low noise circuitry would be simple.
In layman's terms, it takes a while for things to settle down between samples so that you can get an accurate sample, the faster you sample the harder it is to get all those electrons in place and in their seats in time for the next sample, and the more errors that occur.
Add to that the fact that it gets harder and harder to keep the clock accurate and you can get more problems.
Basically you can always make a more accurate lower rate converter.
Post by: mark fassett on October 17, 2005, 02:58:08 AM
| My World wrote on Sat, 01 October 2005 02:14 |
Mr. Fassett, if you would like to write a check to me I would be happy to cash it. |
I suppose the rules about real names are no longer being enforced? Oh well, it was good while it lasted.
Post by: mandel on October 17, 2005, 03:23:44 AM
| compasspnt wrote on Sun, 16 October 2005 19:15 |
KY, What Jon said, plus please check this out: http://recforums.prosoundweb.com/index.php/t/2997/6490/?SQ=5 c62170fa2901f2a00976812f4bd4581 Personally, for tracking I think 48/24 is totally fine for now (as good as any SR/24). I do mix to 96/24, though (although this is as much for having higher SR available for SACD or DVD-A or future formats as it is for better CD quality). Others differ, of course, in their assessments. But Dan Lavry gives good, sound, technical arguments against 192, as we know it today at least. |
I've read Dan's paper, which is very good and clear. In other words there's no technical gain in having higher sampling rates > than human audible range of 20KHz, other than slower roll-off filter argument? I've asked this because there are people who argue that upsampled 24/192 CD does not sound as good as 24/192 DVD-audio or SACD. Also, I've read elsewhere that at 192kHz the shape of the wave is reproduced more accurately than at 44/8 (fallacy? if true why does it matter?). I can see where 24-bits come in, but for all I know 192Khz wouldn't thereotically make a difference from 44 or 48 KHz (unless ultrasonics come in of course).
So basically it all boils down to filter and bits, and not sampling rate?
Out of interest, what does affect the accuracy of the transient in time?
Pls bear with me as I'm not as technically equipped as most on this board.
Post by: Jon Hodgson on October 17, 2005, 05:35:57 AM
| mandel wrote on Mon, 17 October 2005 08:23 |
I've read Dan's paper, which is very good and clear. In other words there's no technical gain in having higher sampling rates > than human audible range of 20KHz, other than slower roll-off filter argument? I've asked this because there are people who argue that upsampled 24/192 CD does not sound as good as 24/192 DVD-audio or SACD. |
There are three possibilities that spring to mind here.
1) They don't hear a difference but convince themselves they do, this is human nature.
2) The downsampling or upsampling was less than perfect, and affected the sound quality.
3) They're vampires and actually can hear something over 20k
| mandel wrote on Mon, 17 October 2005 08:23 |
Also, I've read elsewhere that at 192kHz the shape of the wave is reproduced more accurately than at 44/8 (fallacy? if true why does it matter?). |
Fallacy..ish
It is true that it captures the shape of the wave more accurately, if you look at the samples on your screen... but it doesn't actually capture any more information that the reconstruction filter can use (sampling at more than twice your highest desired frequency already captures all the information neccessary), so it doesn't change anything at playback.
Think of it like this.
1. Draw a circle on a piece of paper
2. How many points do you have to sample on the circumference of that circle to reconstruct it exactly? (so that if you gave someone a sheet of paper with that many dots on it, and told them to draw a circle going through those points, that you could guarantee they drew the same circle)
The answer is 3
3 dots don't LOOK like a circle, but they contain all the information neccessary to recreate that circle. Now if you plot 20 dots on the circumference, then you'll have something that LOOKS more like a circle, but doesn't actually convey anything more to the circle re-drawer.
Well the reconstruction filter is your circle re-drawer, except he's not drawing circles, he's drawing lines made up of sine waves added together, and it turns out that the minimum number of points on a sine wave you need is 2.x where x is any number greater than 0. So you have to sample at just over twice the highest frequency you need.
| mandel wrote on Mon, 17 October 2005 08:23 |
I can see where 24-bits come in, but for all I know 192Khz wouldn't thereotically make a difference from 44 or 48 KHz (unless ultrasonics come in of course). |
You've got it
| mandel wrote on Mon, 17 October 2005 08:23 |
So basically it all boils down to filter and bits, and not sampling rate? |
Yes.
And when it comes to capture and playback, you are unlikely to gain from more than 24 bits, but processing is a different matter.
| mandel wrote on Mon, 17 October 2005 08:23 |
Out of interest, what does affect the accuracy of the transient in time? |
The phase response of the various analogue stages before sampling occurs..
your mic
your preamp
the analogue antialiasing filter.
Of course processes inside your DAW can also affect phase, though that is entirely dependent on the algorithm.
But there is nothing in the process of sampling and playback that does - assuming quality anti-aliasing and reconstruction filters.
| mandel wrote on Mon, 17 October 2005 08:23 |
Pls bear with me as I'm not as technically equipped as most on this board. |
Nor were any of us until we learned about it
Actually I've been involved in this stuff for years, and I still learned some things when I read Dan's papers a week or so back, sometimes seeing the same information presented in a slightly different fashion can make a light come on.
Post by: mandel on October 17, 2005, 09:53:08 AM
| Jon Hodgson wrote on Mon, 17 October 2005 10:35 |
Well the reconstruction filter is your circle re-drawer, except he's not drawing circles, he's drawing lines made up of sine waves added together, and it turns out that the minimum number of points on a sine wave you need is 2.x where x is any number greater than 0. So you have to sample at just over twice the highest frequency you need. |
You've been a big help. However, another question. This only refers to sine waves, right? How about more complex sounds like timpani, drums and cymbals which doesn't conform to sine curves?
Post by: Jon Hodgson on October 17, 2005, 10:01:08 AM
| mandel wrote on Mon, 17 October 2005 14:53 | ||
You've been a big help. However, another question. This only refers to sine waves, right? How about more complex sounds like timpani, drums and cymbals which doesn't conform to sine curves? |
They're all sine waves added up.
Re. Fourier
Post by: jimmyjazz on October 17, 2005, 11:14:18 AM
| Jon Hodgson wrote on Mon, 17 October 2005 05:35 |
Think of it like this. 1. Draw a circle on a piece of paper 2. How many points do you have to sample on the circumference of that circle to reconstruct it exactly? (so that if you gave someone a sheet of paper with that many dots on it, and told them to draw a circle going through those points, that you could guarantee they drew the same circle) The answer is 3 3 dots don't LOOK like a circle, but they contain all the information neccessary to recreate that circle. Now if you plot 20 dots on the circumference, then you'll have something that LOOKS more like a circle, but doesn't actually convey anything more to the circle re-drawer. |
Jon, this is an outstanding analogy. Great job.
Post by: mandel on October 17, 2005, 11:44:14 AM
There's an interesting article which argues that wavelets be used as a supplement to current Fourier-based frequency analysis to yield a better representation of transients.
http://www.regonaudio.com/Audio%20Measurement%20via%20Wavele ts.html
(This article is a pretty old one dating back to 1991; however, much of it still seems applicable)
Quote from DCs Mike Story on wavelet:
'This conventional analysis starts from the viewpoint that the behavior of the ear can be described in mathematical terms using Fourier analysis. This assumption is probably pretty good — it means we are interested in frequency responses, for example, and these do provide good guides to the performance of equipment and to descriptions of what we hear. The analysis was right at the heart of the definition of the audio coding used on CDs.'
‘For those working with audio, it is also apparent that theories based on these descriptions are not completely adequate, and that there can be significant differences in the performances of pieces of equipment with similar "conventional" specifications. It seems that two things are going on here: the ear may have more than one mechanism at work; and sine waves may not be the best function to use as the basis for analysis. On the mechanism front, it seems highly likely that the ear has a sound localization mechanism ("where is it?") that is fast, and independent of the mechanism that says "it’s a violin," and that is related to transient response. There may also be a third mechanism at work.
On the analysis front, it may be that some form of wavelet is the best basis for mathematical modeling. The problem here is that sine-wave theory is relatively simple, and has been fully worked out by generations of mathematicians, following on from Fourier. Wavelet maths is just plain hard work, and does not yet have anything like such a solid core of mathematical results to call upon. Our ears, however, are not waiting.’
While digital audio nowadays are based on Fourier Theorem which provides an adequate mathematical representation of audio signals, it'll be good if we can implement both Fourier and wavelet analysis in digital audio in a way that yield sounds as close to the analog as possible. So if extra sampling rates isn't the way to go, maybe we should focus on wavelet research, then using the extra bytes there.
Just a thought.
Post by: Jon Hodgson on October 17, 2005, 12:22:37 PM
| mandel wrote on Mon, 17 October 2005 16:44 |
I'm becoming a real pain in the arse. |
Pah! You're an amateur in those stakes... compared to what I've dealt with lately you're hardly even a mild irritant!
Thanks for the link, I'll read it more carefully later...
However basically the article does not disagree that all waveforms can be recreated by a combination of sine waves, rather it suggests that this model is not the best matched to the ear, and that a model which included wavelets would be a better match.
Now this model would not allow you to capture any more information when recording, because it can already be shown that given a high enough sample rate (more than twice the highest frequency you can perceive, whether that be steady state or in transients) you can capture the signal in its entirety.
What such a model might allow however includes...
1) The ability to capture the important information with fewer samples, so for example if it was shown that people can somehow detect 60kHz in transients and only 16 kHz steady state, then rather than sampling the whole signal at over 120kHz, it might be better to have some kind of hybrid sampling system which gathered your steady state and transient information seperately.
2) Better lossy audio compression (basically and extension of the above, just done after the initial sampling).
3) New ways to analyze and process sounds.
It might also be interesting for synthesizing new sounds.
Whether such a model would be a better representation of hearing I don't know yet... need to do some more reading... and even if it is, then it doesn't mean you're going to need new converters. As a general storage and processing medium, linear PCM is unbeatable, even if sometimes there might be advantages in using wavelet analysis and synthesis on that stream.
Post by: bblackwood on October 17, 2005, 12:47:48 PM
| mark fassett wrote on Mon, 17 October 2005 01:58 | ||
I suppose the rules about real names are no longer being enforced? Oh well, it was good while it lasted. |
No, they are being enforced, but My World has apparently legally changed his name to My World and has provided proof to GM.
It takes all kinds...
Post by: wavdoctor on October 17, 2005, 12:51:47 PM
my 2 cents..
Post by: mark fassett on October 17, 2005, 01:00:38 PM
| bblackwood wrote on Mon, 17 October 2005 09:47 |
but My World has apparently legally changed his name to My World and has provided proof to GM. It takes all kinds... |
I stand corrected... thanks for the clarification.
Post by: Duardo on October 17, 2005, 01:39:30 PM
| Quote: |
Analog hasn't been more costly for me ..maintenance consists of cleaning heads and occasional degauss. |
How about tape costs?
Post by: SteveBoker on October 17, 2005, 02:25:10 PM
| Jon Hodgson wrote on Mon, 17 October 2005 04:35 |
Think of it like this. 1. Draw a circle on a piece of paper 2. How many points do you have to sample on the circumference of that circle to reconstruct it exactly? (so that if you gave someone a sheet of paper with that many dots on it, and told them to draw a circle going through those points, that you could guarantee they drew the same circle) The answer is 3 3 dots don't LOOK like a circle, but they contain all the information neccessary to recreate that circle. Now if you plot 20 dots on the circumference, then you'll have something that LOOKS more like a circle, but doesn't actually convey anything more to the circle re-drawer. Well the reconstruction filter is your circle re-drawer, except he's not drawing circles, he's drawing lines made up of sine waves added together, and it turns out that the minimum number of points on a sine wave you need is 2.x where x is any number greater than 0. So you have to sample at just over twice the highest frequency you need. |
Great post! Gets at the fundamental points accurately and understandably.
Now, suppose you have a clock driving your circle redrawer. When a circle is redrawn is transient response for that circle (frequency). As the frequency approaches the Nyquist limit (2.x dots per circle) the circle redrawer has less control over when the circle begins to be drawn. In order to know both what circle to draw and when to draw it you need 4.x dots per circle.
Does this matter? Not in a mono signal. Possibly in a stereo signal. Why? Because we are sensitive to when a signal occurs as a cue to position in space. Interaural arrival times give us this cue.
However, it is a long shot as to whether this is detectable. We are sensitive to interaural phase differences in steady state frequencies. In the 80s Kubovy demonstrated that you can create the perception of a sine wave emerging from white noise simply by creating smooth changes in interaural differences in phase for a selected frequency in white noise.
Again, this is a long shot. For 48k sampling, this possibly detectable effect would only apply to spatialization of frequencies above 12kHz. At 12kHz, the wavelength of sound in air is significantly less than the distance between the ears. So, it is really a long shot as to whether this would be detectable. However, I'm willing to entertain the possibility that at 96kHz, high harmonics might sound more "focused" in the sound field than at 48kHz. I can't think of a reason why 192kHz makes any sense.
Against all of that, consider this. The placebo effect is real. Tell subjects that they are getting "advanced" 192kHz sampling and a significant proportion of subjects will report that they like it better whether or not they could actually detect the difference in a psychophysical experiment.
Post by: 12345 on October 17, 2005, 03:07:43 PM
MW
Post by: malice on October 17, 2005, 03:35:51 PM
You don't seem to advertise for a service
You have no link to any of your work
You seem to have an interesting business relationship with clients although you don't have any means of communication apart from a comment window
malice
MARSH moderator
France
Post by: Jon Hodgson on October 17, 2005, 03:52:37 PM
| SteveBoker wrote on Mon, 17 October 2005 19:25 | ||
Great post! Gets at the fundamental points accurately and understandably. Now, suppose you have a clock driving your circle redrawer. When a circle is redrawn is transient response for that circle (frequency). As the frequency approaches the Nyquist limit (2.x dots per circle) the circle redrawer has less control over when the circle begins to be drawn. In order to know both what circle to draw and when to draw it you need 4.x dots per circle. Does this matter? Not in a mono signal. Possibly in a stereo signal. Why? Because we are sensitive to when a signal occurs as a cue to position in space. Interaural arrival times give us this cue. However, it is a long shot as to whether this is detectable. We are sensitive to interaural phase differences in steady state frequencies. In the 80s Kubovy demonstrated that you can create the perception of a sine wave emerging from white noise simply by creating smooth changes in interaural differences in phase for a selected frequency in white noise. Again, this is a long shot. For 48k sampling, this possibly detectable effect would only apply to spatialization of frequencies above 12kHz. At 12kHz, the wavelength of sound in air is significantly less than the distance between the ears. So, it is really a long shot as to whether this would be detectable. However, I'm willing to entertain the possibility that at 96kHz, high harmonics might sound more "focused" in the sound field than at 48kHz. I can't think of a reason why 192kHz makes any sense. Against all of that, consider this. The placebo effect is real. Tell subjects that they are getting "advanced" 192kHz sampling and a significant proportion of subjects will report that they like it better whether or not they could actually detect the difference in a psychophysical experiment. |
Hmmm, I think you may be taking the circle analogy a little far, since phase is correctly reproduced in a sampling system right up to the Nyquist frequency... try it.
Post by: 12345 on October 17, 2005, 04:23:20 PM
Much research has been put into the "minimum number of points to accurately measure a circle," and there are several standards that dictate this. Some statistics, for example show diminishing returns at 2n+1 where n is the theoretical minimum, so the accepted number would be 7 points. Also, some standards provide for the circle's intended purpose, and therefore maximum material condition, bonus tolerance, minimum material condition, roundness, position, or ANY NUMBER of methods to try to express a diameter of a real-world circle within its intended context. Which is exactly why trying to apply a UNIVERSAL rule to the sampling theory debate is ridiculous. There will always be people who want more information, and there will always be those where the supplied information is "good enough." A person designing a space shuttle whose trajectory has to be triangulated out to 17 decimal places will surely be more interested in accuracy of the 17th decimal place than someone bolting together pieces of wood for an art sculpture.
All this without even going into cylinders (extruded circles), projection planes, cosine errors, calibrations, certifications, traceability, yada yada.
Sincerely,
MW
Post by: Jon Hodgson on October 17, 2005, 04:29:21 PM
| My World wrote on Mon, 17 October 2005 21:23 | ||
Correct...I don't really advertise. I try to do everything word of mouth... Regarding the "circle" analogy, yes, that's true that you only need 3 points to "DRAW" a circle in the "design" world, but what the converter is actually trying to do is "MEASURE" a circle...that is, each point in the circle has some uncertainty... http://scholar.google.com/scholar?hl=en&hs=4oF&lr=&a mp;a mp;client=firefox-a&rls=org.mozilla:en-US:official&s a=X&oi=scholart&q=ANSI+b89+minimum+number+of+points+ for+circle Much research has been put into the "minimum number of points to accurately measure a circle," and there are several manufacturing standards that dictate this, including ANSI, ISO, and VDI/VDE (depending on what country you live in, and what kind of circles you have to measure). Some statistics, for example, (ANSI B89 standard I think), for example show diminishing returns at 2n+1 where n is the theoretical minimum, so the accepted number would be 7 points. Newer manufacturing standards provide for a total runout, maximum material condition, bonus tolerance, minimum material conditions, roundness, true position, or ANY NUMBER of methods to try to determine a circle diameter of a real-world circle. Which is exactly why trying to apply a UNIVERSAL rule to the sampling theory debate is ridiculous. There will always be people who want more information, and there will always be those where the supplied information is "good enough." A person designing a space shuttle whose trajectory has to be triangulated out to 17 decimal places will surely be more interested in accuracy than someone bolting together pieces of wood for an art sculpture. All this without even going into cylinders (extruded circles), projection planes, cosine errors, calibrations, certifications, yada yada. And all of this without each manufacturer of "circle measurement devices" trying to interpret and market to show their products in a positive light. It is a "lifetime" effort to understand all of this. Sincerely, MW |
A circle is a mathematically defined shape, anything that requires more than three points to define it is by definition not a circle.
Post by: Jon Hodgson on October 17, 2005, 04:45:40 PM
Post by: 12345 on October 17, 2005, 04:47:03 PM
Jon Hodgson wrote:
"A circle is a mathematically defined shape, anything that requires more than three points to define it is by definition not a circle."
To which I say:
So is a sine wave. So why bother sampling audio? Why not just ask our computers to create a bunch of sine waves, and call it music?
MW
Post by: Jon Hodgson on October 17, 2005, 04:48:08 PM
| My World wrote on Mon, 17 October 2005 21:47 |
So is a sine wave. So why bother sampling audio? Why not just ask our computers to create a bunch of sine waves, and call it music? |
We do, that's what sample playback is.
In fact that's what any audio generation is
Post by: 12345 on October 17, 2005, 04:49:30 PM
Post by: 12345 on October 17, 2005, 04:50:56 PM
Post by: Jon Hodgson on October 17, 2005, 04:53:01 PM
| My World wrote on Mon, 17 October 2005 21:50 |
But on the "input" side, we have to measure it. |
Yes, you are right.
And the way to do that, and the effect of inaccuracies in that measurement, are well researched and documented.
Post by: 12345 on October 17, 2005, 05:58:38 PM
And what remains most important in all of this is that those who make their decisions do so with an understanding of the trade-offs involved.
MW
Post by: Jon Hodgson on October 17, 2005, 06:03:47 PM
| My World wrote on Mon, 17 October 2005 22:58 |
Yet somehow the debate rages on... And what remains most important in all of this is that those who make their decisions do so with an understanding of the trade-offs involved. MW |
Yes, because some people have trouble accepting the mathematics, and also some still base their opinions on the state of converter technology a decade ago.
Post by: 12345 on October 17, 2005, 06:27:16 PM
This debate is not about math. It is about design constraints, and how one designer chooses to implement converter design.
MW
Post by: Jon Hodgson on October 17, 2005, 06:29:22 PM
| My World wrote on Mon, 17 October 2005 23:27 |
And there is yet another category of people who grasp the mathematics, and also the set of design criteria...yet continue to challenge the design constraints themselves. This debate is not about math. It is about design constraints, and how one designer chooses to implement converter design. MW |
Oh yes?
Enlighten me
Post by: 12345 on October 17, 2005, 08:14:11 PM
| Jon Hodgson wrote on Mon, 17 October 2005 15:29 | ||
Oh yes? Enlighten me |
I'll start with one example. Let's say the microphone of the future will not consist of simply a pressure transducer like we are used to. Let's say there is an input reference frequency that the microphone must also transmit through the converter. This reference frequency might be in the Hz range, the kHz range, the 100kHz range, the MHz range (absolutely necessitating the redesign at this point)...whatever...but let's also say that its relationship to the microphone is completely known (to within some specification). Let's say it is so well-specified that as the microphone transmits the sound within the room, there is enough information for the computer to decode not only the sound, but also where it came from, how tall the singer was, and perhaps even the brand of drum set. Not only that, but it can also send a mapping of its own errors and check its performance against the design criteria. So now the microphone functions as a device capable of transmitting perhaps much more than just a 20-some-odd kHz frequency into a band-limited converter, and now has to pass whatever information the system designer intended. This scenario would certainly require a re-evaluation, and perhaps a redesign on the converter end. To make a long story short, the sheer existence of a high-kHZ or MHz reference tone, should someone choose to implement it, necessitates the converter design, and would also necessitate a microphone redesign...and the high-frequency energies would have to be matched to the detection device, etc.
So that is one example where because the design of the input is changing, the design of the digitizer (converter) must operate under a newly-defined set of parameters. This really isn't so far-fetched considering FM radio operates at 88MHz to 108MHz.
This is even less far-fetched when we consider that the world of optics often mixes low-frequency (visible patterns) and high-frequency (visible light) to extract information that neither bandwidth is capable of supplying alone.
Sincerely,
MW
Post by: Ronny on October 17, 2005, 08:35:13 PM
| Duardo wrote on Mon, 17 October 2005 13:39 | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
How about tape costs? Post by: Ronny on October 17, 2005, 08:36:28 PM You only need one point to determine a circle. Just zoom in on it and you'll see a round dot.
Post by: Duardo on October 17, 2005, 08:44:54 PM
It's been a couple years. I haven't bought any recently as I just reuse the ones I have and archive. My point was that even a couple years ago, the costs for digital media are much smaller than analog. I have had problems with hard drives crashing and losing data on them, though, so I'm more comfortable archiving to CD-R and DVD-R. I know they may be problematic in the future but haven't had any problems yet. I supposed I don't do enough work for the time it takes to burn and label them to really be much of an issue... -Duardo Post by: Ronny on October 17, 2005, 09:06:52 PM
Hard drives that are in PCs and Macs will crash because there are other things going on. The dedicated to audio only HD-R's don't have that problem as the proprietary linear recording formats are very stable. At least the D2424, MX2424 and HDR2496, I don't hear of data loss problems, the Alesis HD24's have had a rare occurence, but nothing like PC or Mac based hd recording that has Windows or Apple operating systems and fragmentation. I wouldn't attempt to archive in Windows or Apple systems and am thankful that I don't have to rely on them for multi-tracking or mastering, recording, transferring, archiving, burning to archive and especially can't stand mixing with a damn mouse. To each his own though. Post by: lord on October 17, 2005, 10:07:24 PM My World... Nice idea. The optical techniques that you speak of work by modulating known light sources. The audio analog would be singing into a blasting test tone. That would be interesting... Sennheiser RF microphones sort of work on this concept, using radio frequencies to measure the deflection of a microphone diaphragm. At least, that's how I understand it. This is kind of an important simplification of your concept because ultimately existing audio equipment expects to deal with single channel continuous pressure gradients. And eventually, that is what will get reproduced via speakers, right? or are we redesigning all speaker technology too while we're at it? But there is no reason that you cannot sample lots more points in space. There was someone who had an array of thousands of closely spaced microphones. The post processing that could be done on the signal was quite sophisticated. I wish I could remember who was doing this research. Ultimately, I expect that we'll be able to know exactly where every air molecule in the room is at any given time, its direction of travel, and how fast it's spinning. But how do you intend to make use of this information? Post by: Ronny on October 17, 2005, 10:33:18 PM
Wow, you've just cracked the speed versus distance problem in space travel. Post by: 12345 on October 18, 2005, 12:44:24 AM And yes, why not?...control the particles in the room...Let's break down sound into molecules which make up air and dust in the room. Fun stuff...some universities are working on "self-networking nanoparticles" which can be spread like dust into a forest, and the particles will self-network, might be configurable on-the fly to a wide range of input/output parameters, and might develop "intelligence" over time (whatever that means). So why not apply this to sound recording? MW Post by: dcollins on October 18, 2005, 12:46:57 AM
Are you really Bob Lazar? DC Post by: 12345 on October 18, 2005, 01:01:01 AM http://www.abovetopsecret.com/pages/airspike.html Cool! MW Post by: dcollins on October 18, 2005, 01:12:59 AM
C'mon, admit it! DC Post by: 12345 on October 18, 2005, 01:19:27 AM http://www-bsac.eecs.berkeley.edu/archive/users/warneke-bret t/SmartDust/ Post by: 12345 on October 18, 2005, 01:27:25 AM
That is some funny stuff, DC MW Post by: dcollins on October 18, 2005, 01:27:55 AM
http://www.technologynewsdaily.com/node/1538 Another form of smart dust............. DC Post by: Ronny on October 18, 2005, 03:11:21 AM
More like parasites than dust. Post by: mandel on October 18, 2005, 04:38:39 AM if we record and store at 24/48, the data playback should theoretically be indistinguishable from 24/192 or 24/384, using Nyquist and Fourier laws...more info does not mean more accuracy or resolution here, since Fourier analysis says all one needs is frequencies up to the Nyquist frequency to break down all audible signals into integral of sine curves...The only difficulty, and the reason why digital can sound bad, is in the implementation of filters and other errors...if "perfect" filters are used and quantization errors are neglible, then sound reproduction should be "perfect" - no different from that in a "live" stage... am I right here? so, the question...I'm not challenging Nyquist and Fourier, but Fourier analysis is after all mathematical theorem, not a physics one...so could it be possible that using wavelet technology as complement, sound reproduction could be improved...better filters or easier filters to implement, and/or since wavelet does not deal with sampling rate, could there be something outside sampling rate that one could measure and could better represent the ear-brain thingy? or would you argue that perfect is already perfect? Post by: Jon Hodgson on October 18, 2005, 07:29:47 AM
Well the system is only going to be as good as the mics and speakers, but you get the general idea.
Filters and Fourier work very well together, so I doubt wavelets would make life any easier there. It is possible that a model which more closely matched the ear/brain response, in both its strengths and weaknesses, would allow a better capture system, however I don't find it very likely because 1) The system would inevitably be more complex than a constant rate sampling system, complexity tends to create more problems. 2) We can already sample what most research tells us is more than the neccessary bandwidth with noise and distortion levels lower than the other links in the chain presently achieve, and perhaps will ever achieve. So barring some radical shakeup of our knowledge of audio perception, I would say that constant rate sampling will probably remain the best option as an interface between analogue and digital. IMHO A wavelet based model is more likely to be of use in lossy compression, signal analysiz and processing, and sound synthesis. So basically things you do to the PCM stream in between sampling it and playing it back. Post by: mandel on October 18, 2005, 08:15:37 AM
Signal analysis and processing...would this make a difference to sound reproduction ie. more transparent sound produced? (since we've established no higher sampling rate or data is required, at least not with our present knowledge of audio physics...) Im asking lots of questions... Post by: maxdimario on October 18, 2005, 08:20:40 AM
they use RF to lower the impedance of the output coming out of the capsule and get a good output level in relation to that impedance. in standard condensers a DC voltage is applied, and as the capsule membrane gets closer or more distant from the backplate the change in capacitance generates a voltage which is extremely low in current, hence the need for tubes and fets and 400 Meg resistors etc. the sennheiser mic uses a high frequency that passes through the capsule. as you might know the higher the frequency the lower the resistance of a capacitor. so they use the rf to go through the capsule (which modulates the RF signal), then rectify it to get the analog waveform of the capsule movement. sort of like decoding an AM radio signal, but lower frequency. nothing like he was talking about I'm sure. Post by: Jon Hodgson on October 18, 2005, 09:10:04 AM
Well combining knowledge of the sound with knowledge of the speakers and possibly of the microphones, might possibly allow you to process the signal in such a way as to reduce the perceived distortions created by them, and thus make the sound more natural. As a far simpler example of what I mean, BBE claim that part of their processing, by changing the phase relationship of high and low frequencies, cancels out the deficiencies of speakers in this respect and the sound that reaches your ears is more natural as a result. Post by: lord on October 18, 2005, 09:50:05 AM Fascinating. I want everyone to keep taking their meds, mmm-kay? Post by: dcollins on October 20, 2005, 04:17:23 PM
How do they know what kind of speakers I use? Or is there a DIP switch? DC Post by: Jon Hodgson on October 20, 2005, 04:28:36 PM
They've been peeking... What's more disturbing is they also told me what colour pyjamas you wear. But you're quite right, even if they're right about the reason for what you perceive, they can only do it according to some arbitrary reference... what they judge to be the average case perhaps. Personally I think that compensation for speakers, if it's going to happen, should really be happening in the playback system. Post by: Ronny on October 20, 2005, 04:51:49 PM
The only real way is to tri-amp and delay the speaker lines relative to the distance between the horns, mids and bottom end. I evaluated the BBE for Thorobred when it first came out, must have been 12 years or so ago. I mirrored the sound by eq'ing 6 to 16k with a shelf boosted by +5dB and used a parametric on the low end, boosting, IIRC +3dB with a bell curve around 100Hz. Had several people listen in the blind while I toggled between the BBE and the eq'd no BBE path and they couldn't pick the two out. My conclusion is that the BBE mainly raises gain and relies on the louder is better phenomenon to sell the device. Post by: compasspnt on October 20, 2005, 06:54:08 PM
Dave and Jon, Perhaps it works like the FAA's "average person's weight" rule on the airlines. Since they don't have time (and don't want to offend the [heavier] passengers), instead of doing proper weight and balance for every commercial flight, they just multiply by the "average FAA person" weight of 180 pounds (used to be 156 pounds!). Although, maybe that's about to change... http://www.bigfatblog.com/archives/000390.php Should work for you the same way in monitors. Oh wait, it didn't work for these poor passengers... http://www.ntsb.gov/Pressrel/2004/040226.htm Post by: Level on October 22, 2005, 01:27:59 PM Until the paradigm shifts.. in favor of new ways of thinking and implementation, we as audio engineers must amuse ourselves with baby steps such as converter technology and the "whys" of how well pure analogue recording..to this day meshes with our sensory system in very adequate ways. Newer and bolder models of the way a reproduction system (from the capture to the audience) is implemented have proven to yield incredibly fascinating and rewarding results. Those who are thinking, and designing (and implementing) such novel approaches are certain to go forward while everyone else stands in amazement that such absurdly simple ideas can produce such vastly Superior results. The old adage of "thinking outside of the box" is long overdue of acceptance by the status quo..they just don't want to see their pie sliced up and eaten..right before their very eyes! The very security of copywritten artistic material also hangs in the balance,(!) and for every problem...there is ONE solution and many an attempt. Post by: dcollins on October 23, 2005, 09:06:30 PM
I thought "modern" jets weighed themselves?
Or poor Aaliyah. PIC on blow and alcohol, 700lbs overweight. Aft CG. Everybody dead. DC Post by: compasspnt on October 23, 2005, 09:43:56 PM
And the worst part is Virgin had a chartered Lear 35 on the way to get her, but she had places to go, things to do, party to attend, and wouldn't wait. That whole thing was about as dumb as it gets. Post by: 12345 on October 27, 2005, 06:12:33 AM Shrek 4D is cool. [Edited yet again so as not to reveal the "surprise"] MW |