Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Andy Simpson]

[quote title=Gerald Leitch wrote on Thu, 15 September 2005 23:16]

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

[Eric Bridenbaker]

This is a very interesting way of looking at the issue, never heard that one before. 1.7cm is definitely crude.

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

[Loco]

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.

[tom eaton]

Okay, here's the spec for the Benchmark DAC1:

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

[timrob]

andy_simpson wrote on Thu, 15 September 2005 17:10

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

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

[dcollins]

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

[jimmyjazz]

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.

[Eric Bridenbaker]

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

[J.J. Blair]

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.  

[Eric Bridenbaker]

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!

[maxdimario]

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 .

[dcollins]

Eric Bridenbaker wrote on Thu, 15 September 2005 22:37

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

That's two nanoseconds.  You can forgive them, if combined.....

DC

[Andy Simpson]

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

[Bill Mueller]

andy_simpson wrote on Fri, 16 September 2005 05:12

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

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

[Loco]

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.