R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: [1] 2 3 ... 53   Go Down

Author Topic: 192KHz sample rate for audio  (Read 179577 times)

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
192KHz sample rate for audio
« on: April 29, 2004, 03:09:38 pm »

My  white paper "Sampling Theory" is at www.lavryengineering.com
under "suport". I posted it a few weeks ago, and have since been involved in conversations on the proaudio NG, comp.dsp, mastering web board. I am pleased with the response. A number of  highly respected engineers know the push to 192KH is about manufacturers making money, not about scientific and engineering issues.

The few that argue on behalf of 192KHz talk about "things we do not know" and about listening tests. Such listening tests, taking a device designed to do 192 and switching in a final X2 decimation to 96KHz is NOT VALID, because it tests the last stage decimator and nothing else. The real test should be based on the comparing a 192KHz device against a 96KHz device. By the time you do 192KHz, way too far from what is optimal for audio, the damage is already done.

I am not saying that all distortions are bad. There are those that like tube distortions and more. I am saying that whatever you hear is NOT outside the audio hearing range, thus can be contained within 20KHz or so, certainly within 44-48KHz (88-96KHz sampling). Moreover, in audio, the lowest bandwidth device is the bottleneck (week link in the chain). Lets see:
1. Mics? Who is using 96KHz mics? Most Mics drop at about 20KHz...
2. Speakers? Who is using 96KHz mics? Most Mics drop at about 20KHz...
3. How much energy do musical instrument put out above 40KHz?
4. Ear - my dog does not hear 48KHz

So all of a sudden we are presented with that gizmo- 192KHz?

It is true that there is a lot we can not explain. I have been catering to the ear for many years, and do not know all the answers. But that does not mean that everything is fuzzy. There are some things we know, and know well. We know that sampling must exceed twice the bandwidth of interest. Not by much, just a tiny bit.

Only in audio there is such a disconnect between content bandwidth and sampling. No one else (medical, instrumentation, telecom, video) goes nuts with faster sampling because it yields no positives, only negatives:
1. Less accuracy (there is always a tradeoff between speed and accuracy)
2. More data (storage, data transfer...)
3. More processing required (often traded with lower quality processing)

Again, for those that like a certain distortion associated with 192KH sampling, it is all contained under 20KHz or so. One should not take a whole industry into having to lower transparency, twice the storage requirement and so on. Why not instead, manufacture that distortion (if you like it) with a 96KH device? (96KHz is already an overkill, 60KHz would have been an optimal rate, taking care of ALL the issues including filters, pre ringing and there is nothing else to worry about!).

My paper may be of interest to some, while other "ear types" may find it "too much". I kept the math to a minimum, so it is about graphs, plots and text. I put some energy to explaining what Nyquist finding is all about (he was a major contributor to modern  technology).
Common sense may be misleading folks to think that "the more the better". What is true for say pixels and video or computer screen, is not true for sampling limited bandwidth signal. Making an analogy here is wrong! You need 2 points to draw a straight line. No need for more. You need 3 point for a circle. Well, the bandwidth restriction ends up with: you need only to exceed twice the highest frequency you deal with, thus 88.2KHz accommodates 44.1KH of audio.

I have heard numerous folks that understand math, engineering and science talk about the 192KHz being all about selling new gear to make money. In fact, there are some well respected engineers making such statements on the PRO Audio NG right now. I talked to a lot of folks that admitted privately that they are afraid, uncomfortable, consider it unwise... to raise objections to their employer. I see a strong correlation between the who promotes 192 and who sells such gear. I get ZERO scientific arguments suggesting that my comments are not on solid grounds. I get some folks that want to promote 192 to IGNORE all the science and engineering and math. All I hear is "maybe it is something we do not understand", and “golden folks hear this” (a 192KHz design, against a 192KHz design plus a less than perfect decimator to 96KHz, which is how you listen to a final stage decimator). Folks, if that is what you like, lets do a real good X2 up-sampler than use the imperfect X2 down sampler (decimator), and there is your 192KHz sound. This is the ONLY VARIABLE in most of the listening test. Many folks reported that they like the sound of that X2 last stage decimator. They were confused into thinking it is a test comparing 192KHz to 96KHz.  

I am sad to see audio going in the wrong direction. I am not going to be a part of it, though it does impact my economics. The upside: I can look at myself in the morning. I did not sell out.

I few preemptive comments:
1. Nyquist is immuned to real life imperfections. All the noise, be it analog or quantization or any linearity causing distortions follows Nyquist. It is solid and does not need verification for every (or any) piece of gear.  
2. Again, I do not tell folks what they hear and what they do not. The argument is not about what is good or bad. It is about signals distortions and what not. The standard for “people with ears” is to go for a listening test. My points are valid independently of listening test, because I stay away from tests and good vs bad.
It is possible to figure some things with math. I stop where math and science and real world engineering do, and it still lets me tell a story full of facts.    

BR
Dan Lavry
Lavry Engineering
Logged

bblackwood

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 7036
Re: 192KHz sample rate for audio
« Reply #1 on: April 29, 2004, 03:21:49 pm »

Dan, great to have you here. Also great to see a manufacturer standing up for their beliefs and not trying to cram un-needed product down our throats.

Thank you.
Logged
Brad Blackwood
euphonic masters

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 192KHz sample rate for audio
« Reply #2 on: April 29, 2004, 04:27:13 pm »

Dan,

We meet yet again.  Smile

Nika.
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.

Johnny B

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1134
Re: 192KHz sample rate for audio
« Reply #3 on: April 29, 2004, 09:07:04 pm »

Correct if I'm wrong, but isn't a sound wave not some 2 dimensional wave, but rather a very complex wave that has
very different rise and fall times for many frequencies?
The pictures I've seen make a sound wave look 3-D, sort of like looking at a bunch of mountains and valleys. And if that's true,
don't you really need something very fast to go out
a digitize such complex wave forms.

It would be different if it were just simple sine waves on a X-Y plot.

Perhaps some of the answers lie with the geo-physicists who are trying to do oil exploration and earthquake research.
They are trying to use FFT to look at complex waves too, but they may have more money for the research. I dunno, could be wrong. Often am.



 
Logged
"As far as the laws of mathematics refer to reality,
they are not certain; as far as they are certain,
they do not refer to reality."
---Albert Einstein---

I'm also uncertain about everything.

natpub

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 394
Re: 192KHz sample rate for audio
« Reply #4 on: April 29, 2004, 10:00:04 pm »

Can someone give me a very simple explination about something I heard that high sample rates were displacing a massive amount of distortion and cramming into supposedly inaudible ranges? Does any of that bleed over?

Also, though this may be irrelevant, in a Thesis I did on sound and the central nervous system, there were effects found for extreme hypersonic frequencies. However, none in the range you are suggesting (192 cycles/s). My only point is that the body may recognize inaudible tones in ways other than the ear. I can post references if anyone is interested. The data was mostly compiled from studies connected with the Apollo programs in the 60's, and was in the field of psychology, not acoustics.


Regards,

KT
Logged
Kurt Thompson
Vibrational Arts, Inc.
Blue Skyway Music
Sonic Sorcery Studios
Austin,TX/Columbus,OH

Mark_W

  • Newbie
  • *
  • Offline Offline
  • Posts: 8
Re: 192KHz sample rate for audio
« Reply #5 on: April 30, 2004, 12:27:37 am »

Dan, I understand what you are saying and it makes sense, but I have a question. I understand that some effect plugins convert to a higher sample rate for signal processing. I think I recall reading somewhere that George's GML EQ plugin does that. My question is if it makes no sense to use higher sample rates for A/D and D/A conversion, then what might be the rational for up converting sample rates for plugin effects signal processing? Or do you think that also offers no real benefit?

-Mark Wanlass
Logged

Loco

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 508
Re: 192KHz sample rate for audio
« Reply #6 on: April 30, 2004, 01:07:10 am »

natpub wrote on Thu, 29 April 2004 22:00

Can someone give me a very simple explination about something I heard that high sample rates were displacing a massive amount of distortion and cramming into supposedly inaudible ranges? Does any of that bleed over?


That's what dither and noiseshaping are about. You need to crank your speakers really loud in order to hear something in there. However, at normal listening levels, the results are a lot better than without it. (UV22 anyone?)

Just like what you do when mixing. You push unwanted stuff into almost inaudible regions or mask it with more sensitive information. However, when a converter goes higher than what you can hear, the designer can place a lower quality combination of dither and noiseshaping above your hearing range thus making the converter cheaper. However, in order to work, you have to run at the intended frequency.

As for the megasampling frenzy, it's a matter of convenience more than if you can hear it or not. Using a Genelec S30c and Apogee DA16 I could still hear things going on at around 24K. However, they may not be musically helpful and eventually could become a problem in your mix. Also, I've found that the same EQ plugin working both at 44.1 and 88.2 behaves different when tweaking stuff above 12K. You can abuse it at 88.2 without artifacts or aliasing, but I read some plugin manufacturers noticed this and made their process to happen upsampling data.

But the real problem with working at higher sampling rates is the amount of data flowing. Not only you get your DSP capabilities and track count cut in half, eventually you start generating data faster than you can actually perform reliable Backups.
Logged
Carlos "El Loco" Bedoya

"There's no right, there's no wrong. There's only popular opinion"   Jeffrey Goines
http://www.tukanart.com

natpub

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 394
Re: 192KHz sample rate for audio
« Reply #7 on: April 30, 2004, 03:15:49 am »

Thanks Loco, I think I get it now Smile
Logged
Kurt Thompson
Vibrational Arts, Inc.
Blue Skyway Music
Sonic Sorcery Studios
Austin,TX/Columbus,OH

Peter Simonsen

  • Full Member
  • ***
  • Offline Offline
  • Posts: 225
Re: 192KHz sample rate for audio
« Reply #8 on: April 30, 2004, 07:56:10 am »

Dan,

Thanks for the link, and your honest interest regarding these matters.

I´ll DL the paper and read it as soon as time/work let me do so..

Thanks again

Kind regards

Peter
Logged
http://www.psl-design.dk
 
Life can only be understood backwards; but it must be lived forwards.>>Søren kirkegaard>
>>Its not a dream Peter is real<<

sdevino

  • Full Member
  • ***
  • Offline Offline
  • Posts: 153
Re: 192KHz sample rate for audio
« Reply #9 on: April 30, 2004, 08:55:12 am »

Johnny B wrote on Thu, 29 April 2004 21:07

Correct if I'm wrong, but isn't a sound wave not some 2 dimensional wave, but rather a very complex wave that has
very different rise and fall times for many frequencies?
The pictures I've seen make a sound wave look 3-D, sort of like looking at a bunch of mountains and valleys. And if that's true,
don't you really need something very fast to go out
a digitize such complex wave forms.

It would be different if it were just simple sine waves on a X-Y plot.

Perhaps some of the answers lie with the geo-physicists who are trying to do oil exploration and earthquake research.
They are trying to use FFT to look at complex waves too, but they may have more money for the research. I dunno, could be wrong. Often am.



 




and all the little peaks and valleys need to be 20kHz or less in bandwidth to be heard by humans (at least to the best of our current knowledge).  In other words all the peaks and valleys are the sum of lots of little sinewaves at various phases and magnitudes. All of those sinewaves need to have a freequency of 20kHz or less or they are not audio.

Logged
Steve Devino

Granite Rocks Recording Studios
Studio gear design and setup

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 192KHz sample rate for audio
« Reply #10 on: April 30, 2004, 11:08:50 am »

Johnny B wrote on Fri, 30 April 2004 02:07

Correct if I'm wrong, but isn't a sound wave not some 2 dimensional wave, but rather a very complex wave that has
very different rise and fall times for many frequencies?
The pictures I've seen make a sound wave look 3-D, sort of like looking at a bunch of mountains and valleys. And if that's true,
don't you really need something very fast to go out
a digitize such complex wave forms.


Sound is simply changes in air pressure that registers as discernable by the ear.  Such changes in air pressure do emanate in a 3D pattern from a given place, but at any given point in space, the changes in air pressure are 2 dimensional entirely.  Either the air pressure increases or it decreases.

Quote:

It would be different if it were just simple sine waves on a X-Y plot.


Oh, but it is.  The X axis represents time and the Y axis represents air pressure.  It's that simple.  The only reason we sometimes look at 3D analysis of sound is either in showing how it propagates from a given source, or in waveform analysis where we dissect what amounts to the changes in the changes in the air pressure over time (one step removed and more analytical than looking at a simple representation of what sound actually is).  For example, a 3D chart can show us how individual frequencies change over time.  Sound, however, is simple changes in air pressure, and those are two dimensional: time and amount of change.

Quote:

Perhaps some of the answers lie with the geo-physicists who are trying to do oil exploration and earthquake research.
They are trying to use FFT to look at complex waves too, but they may have more money for the research. I dunno, could be wrong. Often am.


Which "answers" are you looking for?  Waveform analysis is used in many fields of research, many of which are interrelated on levels that don't hit the end user.  Waveform analysis is used in telecom, audio, video, physics, astronomy, radar, oceanography, weather, electrical engineering, structural engineering, and many more fields.  Waveform theory is simply a tool, however, for analyzing what happens.  The field of study, unto itself, is not explored much anymore because it is mathematically as finite and determinable as, say, addition, or fractions, or matrices, or other maths.  There is no reason continuing to explore waveform analysis just like there is no reason to continue to explore addition and subtraction.  Those are all just tools to do real work.

Nika.
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 192KHz sample rate for audio
« Reply #11 on: April 30, 2004, 11:15:52 am »

natpub wrote on Fri, 30 April 2004 03:00

Can someone give me a very simple explination about something I heard that high sample rates were displacing a massive amount of distortion and cramming into supposedly inaudible ranges? Does any of that bleed over?


This is one of the most posited theories right now on why people can hear a difference between sample rates - because the excess noise and distortions above the audible range, when shoved into systems that weren't designed to handle them, end up causing distortion that bleeds into the audible range.  Ergo, some studies and whitepapers have been written on the subject (David Griesinger wrote one, for example) and apparently this theory has strong substantiation.  In other words, yes.

Quote:

Also, though this may be irrelevant, in a Thesis I did on sound and the central nervous system, there were effects found for extreme hypersonic frequencies. However, none in the range you are suggesting (192 cycles/s). My only point is that the body may recognize inaudible tones in ways other than the ear. I can post references if anyone is interested. The data was mostly compiled from studies connected with the Apollo programs in the 60's, and was in the field of psychology, not acoustics.


The issue of affecting the body in ways other than audible is a favorite amongst enthusiasts of high sample frequencies, though no substantiation has been found yet that frequency content (at the level at which music produces it) above 20KHz is discernable by the body (Oohashi excluded, and for substantiatable reasons).

Nika.
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 192KHz sample rate for audio
« Reply #12 on: April 30, 2004, 11:20:28 am »

Mark_W wrote on Fri, 30 April 2004 05:27

Dan, I understand what you are saying and it makes sense, but I have a question. I understand that some effect plugins convert to a higher sample rate for signal processing. I think I recall reading somewhere that George's GML EQ plugin does that. My question is if it makes no sense to use higher sample rates for A/D and D/A conversion, then what might be the rational for up converting sample rates for plugin effects signal processing? Or do you think that also offers no real benefit?

-Mark Wanlass


It is unequivocal that certain processes are abetted by upsampling for the sake of processing.  However, more processes are typically thrown into that theoretical pool than actually require it.

However, there are disadvantages in a few regards to recording at the higher rate to start with in order to accommodate for the rare case where processing can be improved.

1.  It is not required, so this chews up resources.
2.  Keeping the material at higher frequencies throughout allows high frequency noise and distortion into the rest of the signal chain that, through processing, can end up bleeding into the audible range.
3.  Some processes are actually disadvantaged by operating at higher rates.

Nika.
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.

Mark_W

  • Newbie
  • *
  • Offline Offline
  • Posts: 8
Re: 192KHz sample rate for audio
« Reply #13 on: April 30, 2004, 11:50:40 am »

Nika, could you be more specific about what types of processing require upsampling and the technical reasons for that?

-Mark
Logged

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 192KHz sample rate for audio
« Reply #14 on: April 30, 2004, 12:11:17 pm »

That's a little bit involved, but let's just try it this way:

Nonlinear processes (compression, limiting, gating, expansion, etc) all add distortion as an inherent aspect of changing the shape of the waveform the way they do.  Adding distortion means adding high frequency content.  If that content is above the Nyquist frequency then it will inherently alias back into the audible range.   What we need to do is add an anti-aliasing filter.  That type of filter can only be added if we upsample first to a higher rate, allowing the anti-aliasing filter to be added prior to downsampling to the lower rate again.  

Think of it this way - you take a sine wave at 5KHz and compress it.  This will inherently add some 15KHz to the waveform.  You compress it some more and you add some 25KHz.  You keep compressing and you add some 35KHz, 45KHz, 55KHz, etc.  But at a 44.1KHz sample rate the 25KHz content and up will all alias back into the audible range, so we end up with:

5KHz
15KHz
19.1KHz (alias of 25KHz)
9.1KHz (alias of 35KHz)
900Hz (alias of 45KHz)
10.9KHz (alias of 55KHz)

What we need to do is upsample to a rate of, let's say, 176.4KS/s, so that the aliasing is not created but instead the actual frequencies are allowed to be created.  Then we can add a filter, removing everything above 22.05KHz, leaving only the 5KHz and 15KHz that were "legal" anyway, and downsample back to 44.1KS/s.

Upsampling during processing needs to happen, therefore, with non-linear processes, and these include the ones I listed above and their derivations, such as tube distortion, ring modulators, and more.  They do not need to happen with linear processes such as EQ, reverb, pitch correction, simple gain changes, etc.*

Also, the simple use of compression does not inherently require upsampling.  It only requires it if the amount of compression is enough to create the distortion I have discussed at levels that are audible - a heavily debated issue in some circles.  

For another issue regarding upsampling of processes you can look here:

http://www.tllabs.com/index.php?option=content&task=view &id=37

Make sense?

Nika.


* With due respect to George and his upsampled EQ plugin, there are arguments for why this should not be necessary if certain implementations are used.
Logged
"Digital Audio Explained" now available on sale.

Click above for sample chapter, table of contents, and more.
Pages: [1] 2 3 ... 53   Go Up