Head-Fi.org › Forums › Equipment Forums › Sound Science › Hi-Rez - Another Myth Exploded!
New Posts  All Forums:Forum Nav:

Hi-Rez - Another Myth Exploded! - Page 4

post #46 of 156
Quote:
Originally Posted by gregorio View Post

From my understanding, there will not be any break throughs in signal processing when it comes to maintaining accuracy at very high sample rates. To do so would require a re-write of the laws of physics. Dan Lavry explained it very well in his white paper (linked to on the first post) and it's still just as true today as when it was written 7 years ago.

G


I'm talking more like 70 years from now wink.gif

 

Take the problem of lithography for silicon-based CMOS. 7 Years ago many people claimed that an 11nm cut would be impossible due to limitations in optics and that the gates wouldn't work because of interference from quantum tunneling. But then came along multiple patterning and nano-level constructs (monolayer dielectric materials and other things that were not in the EE lexicon a decade ago). Although nanotech is taking its sweet time and will deliver much less than promised, advances in molecular engineering are already being tested in labs around the world due to the multi-billion dollar R&D into the chip market. You don't have to rewrite the laws of physics if you're using radically different devices, like the proposed designs for single-electron transistors.


Edited by anetode - 9/13/11 at 8:26am
post #47 of 156

I always say "Capture in 48 because of ADC limitations, process in 192 because of program limitations."

post #48 of 156
Thread Starter 
Quote:
Originally Posted by PelPix View Post

I always say "Capture in 48 because of ADC limitations, process in 192 because of program limitations."


I wouldn't say that is good advice. Capture at 48kS/s, fine but why would you want to then process at 192kS/s? There is nothing to be gained by that but a fair amount of accuracy to loose.

G
post #49 of 156

You need to apply effects at a higher sample rate because a lot of software effects have inaccuracies and artifacts from being optimized to run in real-time.  If you run them at the playback sampling rate, they'll introduce these artifacts into the audible band, but if you run them at a ridiculously high sampling rate, they'll be above the nyquist frequency and you can LPF them out.

post #50 of 156
Thread Starter 
Quote:
Originally Posted by PelPix View Post

You need to apply effects at a higher sample rate because a lot of software effects have inaccuracies and artifacts from being optimized to run in real-time.  If you run them at the playback sampling rate, they'll introduce these artifacts into the audible band, but if you run them at a ridiculously high sampling rate, they'll be above the nyquist frequency and you can LPF them out.


You are joking?
post #51 of 156
Quote:
Originally Posted by gregorio View Post



You are joking?


Sadly not.  You'd be horrified at what software developers will do for performance at the expense of quality.

Check this out:

http://www.pelpix.info/verb441.flac

http://www.pelpix.info/verb96.flac

 

The oversampled one has high frequency detail, while the natively sampled one has them rolled off.  Variances like this do happen in software effects.  I'm still trying to find an example of the artifacting.  It's usually much less audible than an obvious difference like this.


Edited by PelPix - 9/13/11 at 12:42pm
post #52 of 156
Thread Starter 
Quote:
Originally Posted by PelPix View Post


Sadly not.  You'd be horrified at what software developers will do for performance at the expense of quality.


Exactly and at 192kS/s you need anywhere from 4 to 16 times the performance and still not attain the quality of 48kS/s. I am aware that some programmers put more effort and attention into the signal processing at 96kS/s but I have never experienced any plugin that didn't produce more non-linearity at 192kS/s than at 96kS/s. But then I'm generally not too keen on many of the VST plugins anyway. It could be that you prefer the additional distortion of processing at 192kS/s but personally I try to avoid it.

I can't check out your files at the moment as I'm not in my studio.

G
post #53 of 156
Quote:
Originally Posted by PelPix View Post


Sadly not.  You'd be horrified at what software developers will do for performance at the expense of quality.

Check this out:

http://www.pelpix.info/verb441.flac

http://www.pelpix.info/verb96.flac

 

The oversampled one has high frequency detail, while the natively sampled one has them rolled off.  Variances like this do happen in software effects.  I'm still trying to find an example of the artifacting.  It's usually much less audible than an obvious difference like this.

Almost off topic, but good SOP.... When using real time reverb, which does suck, route the dry signal to another track, set the reverb effect to 100% wet, and then blend the reverb in with dry, just like you would on an analog console.  That way the reverb can't damage the original track.

 

Doesn't completely fix the problem with reverb adding artifacts, but at least they are only on the reverb track.

 


Edited by barleyguy - 9/13/11 at 1:53pm
post #54 of 156
Quote:
Originally Posted by gregorio View Post



Exactly and at 192kS/s you need anywhere from 4 to 16 times the performance and still not attain the quality of 48kS/s. I am aware that some programmers put more effort and attention into the signal processing at 96kS/s but I have never experienced any plugin that didn't produce more non-linearity at 192kS/s than at 96kS/s. But then I'm generally not too keen on many of the VST plugins anyway. It could be that you prefer the additional distortion of processing at 192kS/s but personally I try to avoid it.

I can't check out your files at the moment as I'm not in my studio.

G

Here, have some spectrograms!

Sorry for the dithering!

Native:

 

verb441.png

 

Oversampled:

verb96.png

post #55 of 156

The study you mention, and I know which one because it inevitably gets brought up in these discussions, has about as many holes in it as the study that purported to link vaccines to autism. A more-or-less neutral take on it is available @ [Wikipedia link].""  " """

 

Thanks, I actually am grateful when anything I say that's inaccurate is corrected.  I'll stop "bringing it up in these discussions."  :-)

 

Gregorio, it does no good simply to say "I stand by what I said."  Read about Keith O. Johnson and/or Spectral and you'll see: (1) They're not on any "bandwagon."  They've been making state-of-the-art audio equipment for three decades or more, they bring out new products very slowly, and their dealers tend to despair of the fact that they do so little marketing.  Ask any audio dealer you like or audio industry person about them.  (2) "Ill informed"?  Keith O. Johnson has several important patents concerning the A/D and D/A processes used in modern recording studios.

 

Really - even aside from any discussion about digital audio, I feel it would be a good thing if you knew more about Mr. Johnson before throwing out general statements about what sort of person he must be.

 

You asked about myths.  To me it's a myth to take a single idea (Shannon/Nyquist), even given the fact that the idea is provably correct as a mathematical theorem (though with certain ideal assumptions that do not in fact obtain in reality) and say this is absolutely everything one needs to know.  I don't know if you're old enough, but for me this is the second time around with this sort of stuff.  People were waving the Shannon/Nyquist "flag" when CDs first came out.  No one knew anything about jitter.  And they told everyone who was unsatisfied with CD sound that they weren't hearing what they said they were, because Shannon/Nyquist proved them wrong.  Then people began to be able to measure jitter and relate it to audible distortion and lo! Shannon/Nyquist no longer proved CDs were perfect.

 

I realize you are specifically referring to sampling rate, and not saying we should ignore other issues like jitter.  But even specifically referring to sampling rate, here is what I see as a non-expert, and therefore someone who needs to look to others who know more:

 

(1) The Lavry paper says 88.2 or 96 is all we'll ever need, on the basis of Shannon/Nyquist (though according to Shannon/Nyquist it is really 44.1, so 88.2 and 96 are straying from the "true religion" somewhat already).

 

(2) Numerous other people in the industry who've been deeply respected for a long time (several of whom have designed products I own, so I know from personal experience they're very good - Keith O. Johnson of Spectral, Mike Moffat who was with Theta, now with Schiit) think 176.2 and 192 resolutions sound better.

 

(3) To explain the difference, you can go with conspiracy theories - only Lavry is brave enough to stand against the tide!  But I no more credit this than I would say Lavry takes his position because his DAC won't accept higher-res inputs.  I believe he designed his DAC that way because he sincerely thinks higher resolutions don't add anything.  And I think by the same token if you do some research, you will find that not all the people supporting higher res are ignorant or scam artists.

 

(4) To really make up my mind on this, I will have to listen for myself.  I haven't had an opportunity to do that yet, because my old Theta DAC only accepts up to 16/48, and the one time I did have an extended listening session with high-res at a friend's store, it was to listen for other things (which audiophile player I preferred on the Mac), so I paid no attention to what the resolution was on the pieces we listened to.  But the Bifrost is coming, so I'll at last have the opportunity to see if I personally feel there's any audible difference.

 

(An explanation for this last sentence: I certainly feel scientific inquiry is the best tool to find out the answers to questions like this.  Now for all I know there are academic papers talking about why 176.4 and 192 are better than 88.2 and 96.  I haven't done the research to be able to say yes or no, and if Lavry is what you're citing I'm going to assume you haven't done the academic research either.  Also, I have experienced science and engineering having to catch up to what people's ears are telling them before (see reference to when CDs first came out above).  So if good careful listening on my own, and putting my wife through some blind testing, indicate higher or lower res is better, I'll treat it as a data point in search of an explanation rather than dismissing it out of hand on the basis of Lavry's take on Shannon/Nyquist.)

 

 

post #56 of 156
Thread Starter 
Quote:
Originally Posted by barleyguy View Post

Almost off topic, but good SOP.... When using real time reverb, which does suck, route the dry signal to another track, set the reverb effect to 100% wet, and then blend the reverb in with dry, just like you would on an analog console.  That way the reverb can't damage the original track.

 

Doesn't completely fix the problem with reverb adding artifacts, but at least they are only on the reverb track. 


Not sure I understand, are you saying real time reverbs suck? Even a very mediocre reverb should not affect the dry signal when used as an insert. Aux sending/returning is only useful when sending more than one channel to the reverb.

Quote:
Originally Posted by PelPix View Post

Check this out:

http://www.pelpix.info/verb441.flac

http://www.pelpix.info/verb96.flac

 

The oversampled one has high frequency detail, while the natively sampled one has them rolled off.


OK, I've just had a listen through my laptop, not very good but at least it gives me a vague idea. There is a huge difference between the two files, to the point that they actually sound like similar but different reverb presets. The size and RT sound the same but the PreDelay is different, the ERs are more defined in Verb441, whereas the diffusion appears far higher in level and earlier in the Verb96 but obviously the biggest difference is the mid and high frequency content. None of this is helped by the fact that the dry signal sounds badly clipped. It's not a decent quality reverb (at either sample rate) but if it really is the same settings on both files then it easily qualifies as the worst programmed reverb I've come across in my 20 years of digital recording. If that was my reverb I'd be demanding a refund. Don't mess around with the sample rate, leave it at 44.1kS/s but do yourself a favour and get a decent reverb plugin.

G
post #57 of 156


 

Quote:
Originally Posted by judmarc View Post""  " """

 

Thanks, I actually am grateful when anything I say that's inaccurate is corrected.  I'll stop "bringing it up in these discussions."  :-)


The same thing happens whenever anyone tries to promote an outdated or fringe bit of science: a few pundits seize to a study that supports their preconceptions without reading up on the field or critically considering the methodology. People who read these pundits take it as some sort of affirmation from authority and disseminate it on public forums and the next thing you know it's another nebulous thruthiness that others rely on. I don't mean that the study was badly intentioned, few studies claiming paradigmatic shifts eventually pass muster. Extraordinary claims and all that.

 

 

Quote:
Originally Posted by judmarc View Post

 

You asked about myths.  To me it's a myth to take a single idea (Shannon/Nyquist), even given the fact that the idea is provably correct as a mathematical theorem (though with certain ideal assumptions that do not in fact obtain in reality)

 

You have it a little bit backward: Shannon/Nyquist is a mathematical theorem that is provably correct as an idea. Whatever ideal assumptions contained therein are nothing but mere math, which is ubiquitous in reality.

 

More to the point, Shannon/Nyquist isn't meant to be an ideological force, it is more of a naturally observed limit, much like we can calculate the maximum speed of a photon through a vacuum. The practical limitations of the theorem are why we need more than the 10% headroom afforded by 44.1khz.

 

I see a fair bit of cynicism in your comments which appears to hide a lack of familiarity with the science at hand. If you take some time to read an introductory text to sampling theory and Lavry's paper I think you would find most of your concerns accounted for.

 

As for the personality bit, this isn't a reputation contest of authorities, I'm guessing the Lavry paper is linked to because it is well written and approachable, not necessarily because of who wrote it. This isn't esoteric knowledge and anyone with a high tolerance for dry reading can go to the library and check out a textbook on this stuff that says the same thing. Personally, I'd rather do that than read the hagiographies of selected engineers in audiophile magazines.

 

---

Looking back over what I wrote it seems a touch confrontational, that's an artifact of my personality rather than any bad faith redface.gif

 

post #58 of 156
Quote:
Originally Posted by gregorio View Post



Not sure I understand, are you saying real time reverbs suck? Even a very mediocre reverb should not affect the dry signal when used as an insert. Aux sending/returning is only useful when sending more than one channel to the reverb.




OK, I've just had a listen through my laptop, not very good but at least it gives me a vague idea. There is a huge difference between the two files, to the point that they actually sound like similar but different reverb presets. The size and RT sound the same but the PreDelay is different, the ERs are more defined in Verb441, whereas the diffusion appears far higher in level and earlier in the Verb96 but obviously the biggest difference is the mid and high frequency content. None of this is helped by the fact that the dry signal sounds badly clipped. It's not a decent quality reverb (at either sample rate) but if it really is the same settings on both files then it easily qualifies as the worst programmed reverb I've come across in my 20 years of digital recording. If that was my reverb I'd be demanding a refund. Don't mess around with the sample rate, leave it at 44.1kS/s but do yourself a favour and get a decent reverb plugin.

G


The dry signal isn't clipped.  It's a dirac pulse.

Also, this VST is from Lexicon lol.  I paid about 1.5k for it.

 

Oh!  I see what happened!  The problem lies in what the oversampling algorithm I used did to the pulse!

 

I fixed it.  They now have the same pulse!

 

http://www.pelpix.info/native.flac

http://www.pelpix.info/oversampled.flac

 

They now sound exactly the same except for the high-end roll-off!


Edited by PelPix - 9/13/11 at 3:39pm
post #59 of 156
Quote:
Originally Posted by PelPix View Post

Also, this VST is from Lexicon lol.  I paid about 1.5k for it.



blink.gif

 

Good god, it would be cheaper for you to hire a CS student to write you a custom one

post #60 of 156
Thread Starter 
Quote:
Originally Posted by judmarc View Post

(1) The Lavry paper says 88.2 or 96 is all we'll ever need, on the basis of Shannon/Nyquist (though according to Shannon/Nyquist it is really 44.1, so 88.2 and 96 are straying from the "true religion" somewhat already).

 

(2) Numerous other people in the industry who've been deeply respected for a long time (several of whom have designed products I own, so I know from personal experience they're very good - Keith O. Johnson of Spectral, Mike Moffat who was with Theta, now with Schiit) think 176.2 and 192 resolutions sound better.


Judmarc I've been doing "hi-rez" digital recording professionally for nearly 20 years, in fact many years before it started to be called hi-rez. I have done work at a number of the world's top recording studios, with some of the very finest equipment in the world. I am a recording engineer and producer, I do not have PhDs in maths, Electronic Engineering and Physics so there's a limit to my understanding of digital audio, even if I have picked up a fair bit along the way.

But, you seem to think that everything I'm saying is just quoting Dan Lavry. I'm quoting Lavry because his paper was the first to tackle this issue and it is widely accepted by the scientific and recording communities. Even some of the other manufacturers (Prism, Benchmark and Apogee) have publicly endorsed Lavry's paper, and I believe Weiss has too, certainly their top of the line pro converter does not support sample rates >96kS/s. Lavry, Prism and Weiss represent the top professional converter manufacturers in the world.

"Benchmark has no evidence that 192kHz performs better than 96kHz, but we have a substantial body of evidence that shows that 192kHz has defects that are not present at 96kHz. These issues are also shared openly by one of our competitors: Lavry Engineering. We suspect many other manufacturers are aware of these issues, but choose not to talk about them." John Siau - V.P., Benchmark Media Systems, Inc.

I have yet to hear of anyone scientifically contesting Lavry's paper. I would certainly be interested in any rebuttal from your Mr. Johnson. But I don't see how knowing more about Mr. Johnson is going to change the scientific facts and issues.

I don't believe it was Shannon who stated 44.1kS/s was sufficient. Regardless of who proposed the theory, they were correct. The only reason for using sample rates of 88.2kS/s and 96kS/s is because we do not currently have the technology to implement (virtually) artefact free filters at 44.1kS/s as easily and cheaply as at 88.2k and 96k. The problems with >176.4kS/s are not caused by the limitations of current technology, they are caused by the limitations of the laws of physics and therefore not likely to change any time soon.

G
Edited by gregorio - 9/13/11 at 3:46pm
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Hi-Rez - Another Myth Exploded!