R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 [3] 4 5 ... 53   Go Down

Author Topic: 192KHz sample rate for audio  (Read 179811 times)

Zoesch

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 269
Re: 192KHz sample rate for audio
« Reply #30 on: May 03, 2004, 02:23:05 am »

Some corrections here, just terminology...

Shannon's theorem is not interchangeable with Nyquist theorem.

Nyquist's theorem dealt with the minimum sample rate necessary to accurately digitize a signal. Shannon provided the proof that Nyquist theorem was correct and earned having his name on the theorem (Which some people do refer to as the Shannon-Nyquist theorem)

Shannon's theorem dealt with the efficiency of the data channel (in bits per second) as a function of bandwidth and the signal to noise ratio... Shannon's theorem is what allows us to calculate the actual capacity use for different coding methods (Since each coding method has its own S/N-ratio characteristics) he then defined (With Hartley) the maximum amount of information that can be transmitted over a channel with multilevel encoding (Like QAM for example) and proceeded to delve into Information Theory and signal entropy.

Variable upsampling (As used on MP3 VBR encoding) is a solution that is unfortunately too expensive in computational terms, more elegant solutions (Like apodization as proposed by Graven -see my thread on Klett's forum-) are preferable IMO.

Logged
It has always been Ringo's fault

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 192KHz sample rate for audio
« Reply #31 on: May 03, 2004, 09:00:18 am »

Zoesch wrote on Mon, 03 May 2004 07:23

Shannon's theorem is not interchangeable with Nyquist theorem.


Nyquist didn't have a "theorem."  He only had a theory.  Shannon turned Nyquist's "theory" into a "theorem."  Shannon proved many things, included the validization of Nyquists posit and the on that Zoesch referred to, amongst dozens or hundreds of others.

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

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

ted nightshade

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1272
Re: 192KHz sample rate for audio
« Reply #32 on: May 03, 2004, 11:30:32 am »

It is awfully counterintuitive, this whole Nyquist-Shannon business- basically, if you haven't been through a lot of experiments to prove it yourself, you have to take it on faith. Yes, faith.

I can do that, almost, as a lot of knowledgable and careful people buy it wholesale...

Nonetheless, regardless of Fourier etc., actual sound waves are extremely complex- if a percussion transient can be considered the sum of a number of pure sine waves, that's a whole heck of a lot of sine waves! It's very difficult to accept that these extremely useful theories offer anything other than very good approximations.

It's so easy to imagine several different lines that pass through the same 2 points... That's why it's so difficult to accept these ideas, however well they may work in practice.


Logged
Ted Nightshade aka Cowan

There's a sex industry too.
Or maybe you prefer home cookin'?

George Massenburg

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 349
Re: 192KHz sample rate for audio
« Reply #33 on: May 03, 2004, 12:29:16 pm »

I'm loathe to throw a curveball in here, but I must.

Dan, in these years you and I have discussed the possible impact of wide-band signals, one idea that we talked about sticks in my mind.

You, and presumably others, interpret Shannon-Nyquist which is stated in the frequency domain to have a correlary in the time domain.

This may be correct, but...

I am here to tell you two things:

1. I am first and foremost an engineer who deeply respects the scientific method, and I am here to tell you that there is no science so far that supports the physiology of perceptive differentiation of very high sample rates.  We have some ideas as to what a credible testing protocol might look like.

2. I have heard some stunning, compelling presentations in the high-sample rate, high-resolution contexts, particularly in 384kHz & DSD (SACD technology).  Perhaps the impressions of these technologies wasn't attributable to high-frequency response; perhaps it is.  Lately our thinking is that the really impressive localization in the hi-rez formats might be attributable to much better cross-channel resolution in the time domain.

I don't know.

And you don't know.

Not until we get something proven.  Hard proven.  And although I love your work (and, as you know, use your converters regularly), you haven't really proven the case to me.

George

Logged

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 192KHz sample rate for audio
« Reply #34 on: May 03, 2004, 12:48:20 pm »

George Massenburg wrote on Mon, 03 May 2004 17:29

Lately our thinking is that the really impressive localization in the hi-rez formats might be attributable to much better cross-channel resolution in the time domain.


George,

How is it possible that there is ANY "cross channel resolution" problem in the time domain with base rate sample frequencies.  We know that there are possibilities for time domain problems pending the filters, but how would these problems manifest themselves "cross channel?"

Am I missing something dumb, here?

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

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

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: 192KHz sample rate for audio
« Reply #35 on: May 03, 2004, 01:49:56 pm »

ted nightshade wrote on Mon, 03 May 2004 16:30

It is awfully counterintuitive, this whole Nyquist-Shannon business- basically, if you haven't been through a lot of experiments to prove it yourself, you have to take it on faith. Yes, faith.

I can do that, almost, as a lot of knowledgable and careful people buy it wholesale...

Nonetheless, regardless of Fourier etc., actual sound waves are extremely complex- if a percussion transient can be considered the sum of a number of pure sine waves, that's a whole heck of a lot of sine waves! It's very difficult to accept that these extremely useful theories offer anything other than very good approximations.

It's so easy to imagine several different lines that pass through the same 2 points... That's why it's so difficult to accept these ideas, however well they may work in practice.




Logged

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: 192KHz sample rate for audio
« Reply #36 on: May 03, 2004, 02:28:08 pm »

George Massenburg wrote on Mon, 03 May 2004 17:29

I'm loathe to throw a curveball in here, but I must.

Dan, in these years you and I have discussed the possible impact of wide-band signals, one idea that we talked about sticks in my mind.

You, and presumably others, interpret Shannon-Nyquist which is stated in the frequency domain to have a correlary in the time domain.

This may be correct, but...

I am here to tell you two things:

1. I am first and foremost an engineer who deeply respects the scientific method, and I am here to tell you that there is no science so far that supports the physiology of perceptive differentiation of very high sample rates.  We have some ideas as to what a credible testing protocol might look like.

2. I have heard some stunning, compelling presentations in the high-sample rate, high-resolution contexts, particularly in 384kHz & DSD (SACD technology).  Perhaps the impressions of these technologies wasn't attributable to high-frequency response; perhaps it is.  Lately our thinking is that the really impressive localization in the hi-rez formats might be attributable to much better cross-channel resolution in the time domain.

I don't know.

And you don't know.

Not until we get something proven.  Hard proven.  And although I love your work (and, as you know, use your converters regularly), you haven't really proven the case to me.

George



Logged

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: 192KHz sample rate for audio
« Reply #37 on: May 03, 2004, 02:30:17 pm »

[quote title=George Massenburg wrote on Mon, 03 May 2004 17:29]I'm loathe to throw a curveball in here, but I must.

Dan, in these years you and I have discussed the possible impact of wide-band signals, one idea that we talked about sticks in my mind.

"1. I am first and foremost an engineer who deeply respects the scientific method, and I am here to tell you that there is no science so far that supports the physiology of perceptive differentiation of very high sample rates.  We have some ideas as to what a credible testing protocol might look like."

Did you read my paper? Everything I said is backed by math. The conclusions are rock solid: The difference between an analog wave and the sampled signals is high frequency content. By removing such content, you remove the difference, thus get back to the original. The high frequency difference is just too often confused with the subject of aliasing. It is the sampling process that makes for the high frequency difference, and removing the high frequency ends up as is filtering at the DA side.

"2. I have heard some stunning, compelling presentations in the high-sample rate, high-resolution contexts, particularly in 384kHz & DSD (SACD technology).  Perhaps the impressions of these technologies wasn't attributable to high-frequency response; perhaps it is.  Lately our thinking is that the really impressive localization in the hi-rez formats might be attributable to much better cross-channel resolution in the time domain."

I have taken great care to state that I do not say that people do not hear what they hear. I also do not claim what is good sound vs. bad sound. All I am saying is that whatever you hear, be it good, bad, nice or not, is not energy that requires 192KHz sampling. Please read the article, and point out what is it specifically you disagree with.

I am finding myself arguing such simple things such as:
material that was recorded at 192 and than decimated to 44.1KHz can not possibly have energy above 22.05KHz. You say  

I don't know.
And you don't know.

I think I know what I am saying, and I am being careful not to say what I do not understand.

"Not until we get something proven.  Hard proven.  And although I love your work (and, as you know, use your converters regularly), you haven't really proven the case to me."

I do not need to prove fundamentals again and again. I have to conclude that you did not take the time to read my article.

I recall our talk about "aliasing" from a time domain perspective. But it is just a more convenient way for some to view the same thing. The time/frequency domain relationships are just math, but the same thing. Nyquist holds for non linearity, noise and anything you will throw at it.

I too love your work. I am not crazy to think that you, of all people can not hear. What I said does not challenge it at all. It is about where the energy resides, is 192KHz needed for audio, where the signal itself is not even there (mics, ear, speakers... all in series....)  
 
BR
Dan Lavry
Logged

Level

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1811
Re: 192KHz sample rate for audio
« Reply #38 on: May 03, 2004, 02:42:43 pm »

It is all agreed we simply cannot hear fundamentals above 20K. Lets look at things from this perspective proceeding:

Interferance patterns from upper harmonics do interfere with the audible frequency range. This interference BECAUSE the upper harmonics found in acoustic instruments is present, is my argument for high sampling frequencies. This is simply one therory. Another is the high pass filters can be made to roll off much more gradually to accomidate the nyquist therory so the absolute brick wall frquency transponded is 1/2 of the sample frequency without any filter degradation. We all know that steep filters cause ringing in the audible range and the reconstruction of the data leads to errors in absolute sinusodul shapes of the waves AFTER the interference pattern that is supposed to be there are filtered out.

It is very common for microphones, microphone preamplifiers and line amplifiers to have notible dynamics and frequency display upwards of 40Khz. The lack of a steep filter will result in the intended and natural display of the acoustical event.

Now for the sake of getting down to basics, most people can sense the absence of the upper harmonics due to the interference pattern not being there, that is natural in the acoustic event. This said, it makes perfect sense to have unimpeded flat frequency resolution and dynamics up to and above 50Khz. Above that, the jury is still out. If you have this ideal of proper capturing, recording and reproducing to 50Khz (remember, acoustic events, not electronically produced) it makes perfect sense to have the sampling rate at up to 6 times the audible frequency X 2 divided by 2 to get to 50Khz without filtering effects, dynamic coherency problems or interference patterns not being assimalated or stored.

Doing the math:

20K X 6 = 120K
120K X 2.0 = 240,000

Divide this by half = 120,000 and half again for nyquist, 60,000kHZ

So if we have our sample rate at 240KHZ and make the filter half as steep and compensate for nyquist, then you can have your 60Khz without problems.

Remember again, it is not that you can hear the fundamental frequency. It is that...the fundamental frequencies contain harmonics which cause interference patterns in the frequencies we do hear. These interference patterns need to remain intact for us to enjoy a reproduction as it originally produced acoustically.

I am certain that the level in which these upper harmonics are properly displayed are -30 to -50 dB down but they are a natural event and to capture and record them digitally for proper playback of acoustic instruments, it is vital to run the sampling frequency as high as practically possible with PCM. Now, DSD, that is another can of worms entirely with the 100khz noise and the poor mastering I have heard as of late.
Logged
http://balancedmastering.com

"Listen and Learn"
---Since 1975---

Innominandum

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 52
Re: 192KHz sample rate for audio
« Reply #39 on: May 03, 2004, 03:44:08 pm »

Level wrote on Mon, 03 May 2004 12:42

We all know that steep filters cause ringing in the audible range


I'm wondering about this - can phase distortion be corrected?
Logged

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 192KHz sample rate for audio
« Reply #40 on: May 03, 2004, 03:52:03 pm »

Dylan Bowker wrote on Mon, 03 May 2004 20:44

Level wrote on Mon, 03 May 2004 12:42

We all know that steep filters cause ringing in the audible range


I'm wondering about this - can phase distortion be corrected?



Of course.  But it is also important to note that the ringing in the audible range is not, per se, phase distortion.  

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

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

George Massenburg

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 349
Re: 192KHz sample rate for audio
« Reply #41 on: May 03, 2004, 04:54:19 pm »

Nika Aldrich wrote on Mon, 03 May 2004 11:48

George Massenburg wrote on Mon, 03 May 2004 17:29

Lately our thinking is that the really impressive localization in the hi-rez formats might be attributable to much better cross-channel resolution in the time domain.


George,

How is it possible that there is ANY "cross channel resolution" problem in the time domain with base rate sample frequencies.  We know that there are possibilities for time domain problems pending the filters, but how would these problems manifest themselves "cross channel?"

Am I missing something dumb, here?

Nika.


Geez, I don't know the numbers (in particular I can't quote the exact time resolutions of the different SR formats), but they obviously get progressively finer.  And any two (or more) channels will have more and more ability to resolve time differences.

George
Logged

Nika Aldrich

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 832
Re: 192KHz sample rate for audio
« Reply #42 on: May 03, 2004, 05:11:20 pm »

Level,

First, George has requested that all posters to this forum use their full name in their posts.  FYI.  On to your post:

Level wrote on Mon, 03 May 2004 19:42


Interferance patterns from upper harmonics do interfere with the audible frequency range.


Only in non-linear devices.  And the upper harmonics can only create subtones within the non-linearity of the ear if both of the notes being combined are audible to the ear.  If we are expecting upper harmonics to mix in such a way as to cause a new, audible frequency then both of the upper harmonics have to, by themselves, be audible.  If either is not then we won't hear the "created" subtone as it can't be created by the ear.

Quote:

This interference BECAUSE the upper harmonics found in acoustic instruments is present, is my argument for high sampling frequencies. This is simply one therory.


And this theory is not valid.

Quote:

Another is the high pass filters can be made to roll off much more gradually to accomidate the nyquist therory so the absolute brick wall frquency transponded is 1/2 of the sample frequency without any filter degradation. We all know that steep filters cause ringing in the audible range


Not really.  They cause ringing in the transition band.  Any frequency affected by the rolloff of the filter will be manifested in like amplitude as a ripple extending prior to the transient due to the use of linear phase filters.  If, for instance, no frequency below 20KHz is affected by a filter more than x amount then no frequency lower than 20KHz will ring with any more amplitude than the inverse of x.  

Quote:

and the reconstruction of the data leads to errors in absolute sinusodul shapes of the waves AFTER the interference pattern that is supposed to be there are filtered out.


I'm lost on this sentence?

Quote:

Now for the sake of getting down to basics, most people can sense the absence of the upper harmonics due to the interference pattern not being there, that is natural in the acoustic event.


That is absolutely erroneous.  Do you have any form of substantiation to this that would therefore counter the studies that report the contrary?  

Quote:

This said, it makes perfect sense to have unimpeded flat frequency resolution and dynamics up to and above 50Khz. Above that, the jury is still out. If you have this ideal of proper capturing, recording and reproducing to 50Khz (remember, acoustic events, not electronically produced) it makes perfect sense to have the sampling rate at up to 6 times the audible frequency X 2 divided by 2 to get to 50Khz without filtering effects, dynamic coherency problems or interference patterns not being assimalated or stored.


This sounds like conjecture?

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 #43 on: May 03, 2004, 05:34:48 pm »

George Massenburg wrote on Mon, 03 May 2004 21:54


Geez, I don't know the numbers (in particular I can't quote the exact time resolutions of the different SR formats), but they obviously get progressively finer.  And any two (or more) channels will have more and more ability to resolve time differences.

George



But the timing "resolution" of a waveform issue is in violation of Nyquist - unless we aren't communicating through a semantical barrier?  Can you explain more about what it is you are positing may be the issue at hand?

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

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

gasman

  • Newbie
  • *
  • Offline Offline
  • Posts: 17
Re: 192KHz sample rate for audio
« Reply #44 on: May 03, 2004, 05:53:56 pm »

Howdy folks...first, a disclaimer:

I have little experience or solid understanding of anykind about these matters of anti-aliasing/nyquist/bit depth issues and what-not. I'm more of an acoustic/speaker guy with musician's disease(I keep thinking I'm a player!).

But I do have a related experiment I'm invoved with that I think is at least a practical application of beyond-human range sampling values...

I'm currently (okay, not right now) gathering field audio via 24/96 converters (okay, it's a firestation, sue me.) and a pair of earthworks, the sources being very large and very small animals and insects that communicate beyord human ranges(slow down some samples of this year's cicada swarms for spooky orchestralish melodies, for example)...I then freak em in the DAW and my associate scientists are going to town on it...

Now, my thinking is the 2 octaves above and below human range (5-80khz) would require a sample rate of 160khz in order to get the full 80khz, but that's just retarded-reverse-assumption based on 44k/20-20k relationship, as I was possibly erroneously informed as I got my earthworks(4-40Khz) that in order to capture the 40k I would need the 96k converters.

So please forgive as my thing may not have anything to do with anything, but rather, riddle me this...in order to properly capture 80khz(I know, first I have to find a mic that'll do it)would I have take a step up to the 192khz converters?

Also, what would be the capturable frequency range of the newfangled DSD bit for bit or whatever?

Thanks in advance!

GM
Logged
I'm gonna punch you right in the Ovary.
Pages: 1 2 [3] 4 5 ... 53   Go Up