Hi-Rez - Another Myth Exploded!

Sep 30, 2011 at 10:13 AM Post #136 of 156


Quote:
No, I didn't say SACD was worse than 16/44, just no better. There are both technical advantages and disadvantages to the SACD format compared to 16/44. However in a year long study by the AES with over 500 DBTs, using professionals and non-professionals, the success rate in distinguishing SACD from CD was almost exactly 50%.
G


This is subtly but criticaly misleading on a number of points. Firstly the study was not done by the AES it was done by Brad Meyer and David Moran (both AES members) and was published by the AES ( it was not an official AES research initiative)  and was done in conjunction with numerous members of the BAS (Boston Audio Society) , i.e the AES as a body is not sayingSACD is = CD.
 
Secondly the tests did not directly compare SACD and CD. What the researchers did was use high res players i.e SACD and DVD-A players and split the analog outputs into 2 LR pairs sending one pair to an A/D/A chain------->Level matching preamps------->ABX box and the other pair directly to the ABX box.
 
So the players analog signal is compared against the same signal that has been digitized back to 16/44.1 and then converted back to analog again. In theory this should be even more damaging than a (competent) purely digital domain conversion due to the extra plumbing
 
While I broadly accept the M and M findings there are two methodological issues which they should have addressed better. Firstly their High res players were relatively poor in the context of the theoretical capabilities of high res recordings i.e in SNR terms they were equivalent or marginally better than a good CD player this may or may not be important but should have been sorted out. Secondly a substantial number of their high res recordings were not high res born being taken from older sources, some were native high born - both of these points have been picked up by critics of the paper
 
 
 
 
 
Sep 30, 2011 at 10:40 AM Post #137 of 156
This is subtly but criticaly misleading on a number of points.


I agree with virtually everything you said, except that I was critically misleading. I accept the paper was not commissioned by the AES but it was (as you stated), conducted by AES members, peer reviewed by the AES and published by the AES. Also, I consider they chose a wide range of source material, from original hi-rez SACD to hi-rez recordings created from old analogue masters.

The relatively average SNR of the equipment used, is not particularly relevant, provided it exceeded the 96dB SNR of CD. Also, as you stated, the tests if anything were weighted towards a more obvious difference between CD and SACD/hi-rez, on the basis of the additional AD/DA loop, which would do nothing other than deteriorate the 16/44 playback chain. Certainly if the results had proved a perceivable difference, the additional AD/DA loop would have been cause for considerable and justifiable criticism.

I was guilty of over simplification but I do not accept that I was critically misleading.

G
 
Sep 30, 2011 at 12:03 PM Post #138 of 156


Quote:
I agree with virtually everything you said, except that I was critically misleading. I accept the paper was not commissioned by the AES but it was (as you stated), conducted by AES members, peer reviewed by the AES and published by the AES. Also, I consider they chose a wide range of source material, from original hi-rez SACD to hi-rez recordings created from old analogue masters.
The relatively average SNR of the equipment used, is not particularly relevant, provided it exceeded the 96dB SNR of CD. Also, as you stated, the tests if anything were weighted towards a more obvious difference between CD and SACD/hi-rez, on the basis of the additional AD/DA loop, which would do nothing other than deteriorate the 16/44 playback chain. Certainly if the results had proved a perceivable difference, the additional AD/DA loop would have been cause for considerable and justifiable criticism.
I was guilty of over simplification but I do not accept that I was critically misleading.
G

 
Gregorio, I was not saying that you were critically misleading or that you meant to mislead. What I mean was that the way it was presented was subtly but importantly misleading. I did not mean to imply intent to mislead. I did not mean to give offence.
 
How we say things is important. For instance if I say "an AES study" and you say "a study by two researchers who are also AES members" they are similar but they definitely have different implications - one implies a stamp of authority.
 
The point about the SNR is that some of the the people who advocate SACD/DVD-A believe that the technical superiority of the medium is important, with 24bit PCM you have a theroretical SNR of about 144db - granted nothing manages this or close to it. Say we settle on a high res device that has an SNR of 124 db then the A/D/A process degrades this to no more than ~96db or a great deal. If the high res device has an SNR of 100db the A/D/A process does less damage in terms of extra noise so any difference would be hard/impossible to pick up.
 
Personally I think this argument is a load of fetid dingo's kidneys, studies show it is difficult to tell the difference between 16 bit and 15 bit, but that is the argument made.
 
Similarly if the source material has a poor dynamic range and frequency extension then mangling it to 16.44,1 is less damaging than if the material has a huge dynamic range that is materially nobbled by the A/D/A - again A sarcina foetidus dingos renes as I want to know where they find a DVD-A with that kind of dynamic range......
 
Sep 30, 2011 at 1:21 PM Post #139 of 156
No offence taken Nick, my wording was a little sloppy!

G
 
Sep 30, 2011 at 4:12 PM Post #140 of 156
In a further explanation of their methods on the Boston Audio Society web site, it appears that the experimenters may simply have failed to "turn it up."  :-)
 
One of the authors, using a short repeated section of room tone on the Hartke disc mentioned above, obtained a positive result (15/15) at a gain of only 10 dB above our standard level.
 
See http://www.bostonaudiosociety.org/explanation.htm
 
As I've mentioned before in this thread, I don't have a DAC at the moment capable of playing high-res files.  When I do get it (the Bifrost, should be here within 2-3 weeks I'd guess), I'll be interested to listen to well-recorded 96kHz and 192kHz versions of the same track (Linn has a good selection of such downloads) to see if I personally can hear any difference at all, and to see whether my wife can tell in blind testing.
 
Sep 30, 2011 at 4:55 PM Post #141 of 156
Quote:
In a further explanation of their methods on the Boston Audio Society web site, it appears that the experimenters may simply have failed to "turn it up."  :-)
 
One of the authors, using a short repeated section of room tone on the Hartke disc mentioned above, obtained a positive result (15/15) at a gain of only 10 dB above our standard level.


Careful with that selective quotation there! Just before that line was this:
 
[size=small]
Our standard system gain was calibrated using an octave of pink noise recorded at -16 dBFS, which produced a wideband SPL of 85 dB at the listening chair.[/size]

 
And just after was this:
 
[size=small]
This setting produced sound levels clearly higher than those at the site, as the peak levels for this small vocal/percussion ensemble would have been 111 dB SPL on the loudest part of the disc.[/size]

 
111dB! Who listens at that volume? 85dB is considerably more realistic as a wideband volume, as are the apparent 101dB peaks that would have resulted. I listen less than half that loud, myself. Their standard volume would be uncomfortable for me, let alone the increased volume. That could reveal the noise floor of the low bit depth tests, which could make a difference. This is like saying "We could hear a difference when there was a difference". Of course! But who listens loud enough to reveal the 96dB noise floor of 16 bit files? Who listens loud enough to reveal the reduced noise floor of dithered 16 bit files? If you do, then by all means go for 24 bit, but in a few years you might not hear much at all.
 
Failed to turn it up is right! If they forced listeners to use that volume, they could be sued!
 
Sep 30, 2011 at 5:55 PM Post #142 of 156
One of the authors, using a short repeated section of room tone on the Hartke disc mentioned above, obtained a positive result (15/15) at a gain of only 10 dB above our standard level.


That is entirely consistent with what I would expect. 10dB is over three times the output level than an 85dB setting, which is already very high for a relatively small room and particularly excessive in my opinion considering the noise floor of the room was only 19dBA! 75dB would have been a reasonable setting for that room! Peak levels at this +10dB setting would have been 111dB, in a small very quiet room that would have been painful! In another experiment they turned up the level by 20dB during a very quiet passage and again heard the digital noise floor on CD, unsurprising really as the noise floor of CD is 96dB (lower though with noise-shaped dither). But even at higher than normal listening levels (85dB in that room) no one could identify a difference with 16bit. Consistent with experiments I've done myself and with predictions for 16bit source material.

G
 
Oct 3, 2011 at 8:18 AM Post #145 of 156
Careful with that selective quotation there!
 
Guilty as charged.  But I did wish to point out that additional bit of background information about Meyer and Moran's testing.  It may just be that as you and Gregorio say, turning up the volume to near-painful levels allowed the noise floor to be heard on 16-bit material.
 
Meanwhile, two things:
 
- Thanks very much to Anetode and Gregorio, for pointing to or making available academic-level references on digital audio.  I appreciate the opportunity to further my education on a topic in which I'm tremendously interested.
 
- I continue to eagerly await the arrival of a DAC in the next couple of weeks with high-res-capable inputs, that will allow me to carry out my own (albeit anecdotal and unreliable) 'experiments' on myself and my wife, to determine whether we think we consistently hear any differences at all between well-recorded 24/96 material and the same material at 24/192.  Should be interesting!  I'll report back....
 
May 11, 2012 at 7:26 AM Post #146 of 156
None of that answered any part of my question.  I am not talking about the waveform represented by the samples at all.  The reverb, being an entirely digital process, does not see the final waveform between the samples.  It sees a series of dots and it processes those dots.  What I'm wondering is if processing at a higher sampling resolution produces a more perceptually coherent (Not necessarily mathematically correct) reverb when resampled back down in the same way that rendering a CGI image at a higher sampling resolution and resampling it produces a more coherent image.
Please open the full images.
 
This image has one sample per pixel and, from a mathematical standpoint, contains all the information needed to reconstruct the image within the bandwidth constraints of the sampling resolution of the image.
 
http://upload.wikimedia.org/wikipedia/en/8/84/Mandelbrot-spiral-original.png
 
This image has multiple samples per pixel that have been averaged.  It is resampled from an image that has a 400x higher sampling resolution.
 
http://upload.wikimedia.org/wikipedia/en/6/63/Mandelbrot-spiral-antialiased-400-samples.png
 
While the first image is the closest mathematical representation of the digital equation that can exist within the constraints of the image size, which looks better?

 
http://www.pelpix.info/silverplate1.flac
http://www.pelpix.info/silverplate2.flac
 
One of these has the original samples fed into the reverb process and one of these is oversampled.
Does one sound better?  If so, which?

 
Let me simplify it.  I misspoke:
If you feed a digital reverb three samples, it effectively sees a triangle wave, not a sine wave.  It connects the samples with straight lines.

 
We're not talking about the waveform here.  A digital reverb does not see a waveform nor a sound.  It does not see the samples as connected in any way, shape, or form.  It applies the process to each individual sample as its own entity.  All laws related to waveform capture and reproduction aren't really applicable in digital processing.
 
Here are three samples

 
They represent this sine wave portion when traced as such (My line work is bad, so the curve isn't right.  My apologies.)
 

 
 
However, the average digital reverb simply connects the dots into a triangle wave:
 

 
Digital processors that aren't very, very bad like the one demonstrated above use interpolation, but it is cheap and full of aliasing.  The only reverb I have ever seen that doesn't alias without resampling is 2C-Audio's Aether.  It's clean for the whole band and has zero aliasing, but processing time can be up to two hours for a 5 minute audio file.

 
I'm aware, it's just that everyone's implementation is terrible.  For some reason interpolating the points between is difficult for VST plugins because they have to be real-time.  I'm not sure of the actual mechanics.
 

 
Now, there's something of note anyone who studies waveforms and how they're sampled needs to understand:
"A higher sampling rate does not increase resolution, it increases bandwidth."
This is a contradictory statement.  We're talking about a line that is going up and down.  The more times it goes up and down within a certain period of time, the higher the frequency.
Resolution is bandwidth!
 
 
Common sense would say that higher sampling rate waveforms are better because you get the little tiny details in the waveform.  However, waveforms as a visual representation make no sense.  Even though everything about them tells your brainn that those details are shorter and quieter, the little tiny details you're seeing aren't shorter or quieter than the big ones, they're higher in pitch.

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

 
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.

 
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.

 
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!

 
Thanks for the interesting read.
 
 
 
Let's, for the sake of argument say that hypothetically there are no physics or processing limitation problems with sample rates greater than 96kS/s. That leaves us with a 192kS/s sample rate which performs just as well as 96kS/s with the added benefit of being able to record frequencies between 48kHz and 96kHz. My dog can't hear 48kHz and even some bats can't hear as high as 96kHz. Between 48kHz and 96kHz we are looking at virtually zero energy produced by musical instruments, no standard studio mics which can record that high and virtually no cans or speakers (and probably not many amps) which can accurately reproduce 96kHz, in addition of course to the fact that 96kHz is nearly 5 times beyond the limit of human hearing. So even if there weren't any problems with 192kS/s it would still be absolutely pointless. I would like you to ask your Mr. Johnson how he EQ's and mixes all these frequencies which he can't even hear?
 

 
There are still a number of problems, not just this one: After many decades and many tests there is still no proof that humans can perceive anything above 22kHz, let alone 48kHz or 96kHz.

Apart from bell percussion no instruments produce any energy above 48kHz, for example a violin produces only 0.04% of it's sound energy above 20kHz. No standard studio mics can record above 48kHz (very few go beyond 20kHz). There are virtually no head phones or speakers which reproduce 96kHz. And lastly, no producer can mix frequencies which they can't hear! Haven't I stated all this before?

 
There are (recent) studies which support that people can differentiate between 44.1kHz and higher sampling rates with the statistical outcome of chance at less than 1%.
 
May 11, 2012 at 10:39 AM Post #147 of 156
Quote:
...There are (recent) studies which support that people can differentiate between 44.1kHz and higher sampling rates with the statistical outcome of chance at less than 1%.

so cite them - lets see how extensively they have been tested by replication by other researchers - many problems prevent isolated individual studies, even those following the form of the scientific method, from being accepted as "proof" until replicated, validated by multiple groups
 
 
and please read up on digital/analog theory, Nyquist criteria, waveform reconstruction instead of repeating the same tired - wrong headed  "poor interpolation" comments/illustrations
 
May 11, 2012 at 3:36 PM Post #148 of 156
Originally Posted by jcx /img/forum/go_quote.gif
 
so cite them - lets see how extensively they have been tested by replication by other researchers - many problems prevent isolated individual studies, even those following the form of the scientific method, from being accepted as "proof" until replicated, validated by multiple groups
 
and please read up on digital/analog theory, Nyquist criteria, waveform reconstruction instead of repeating the same tired - wrong headed "poor interpolation" comments/illustrations
 

 
... it's interesting you're pre-dismissing studies as invalid until replicated and validated by several independant universities.
 
As far as I know, there's very little testing, replication, or validation on CD quality versus higher resolutions, all I see is the Meyer & Moran paper linked over and over.  They concluded via subjective listening that their SACD and DVD-A selections sounded almost universally better than CD quality, however... when the SACD / DVD-A was downsampled to 16/44.1 no listener could correctly ABX the new copy with a 10/10 success rate.
 
A lot of the SACD's were actually upsampled CD's, and the success rates hovered a little too strictly under 7/10, (so in other words, if their ABX setup was a roulette table at a casino they would have been arrested).
 
A more recent study says the difference between 24/44.1 and 24/88.2 is audible, but it's not clear if that one is statistically correct either.
 
May 11, 2012 at 3:55 PM Post #149 of 156
[size=12pt]"They concluded via subjective listening that their SACD and DVD-A selections sounded almost universally better than CD quality" is [size=12pt]a gross misrepresentation - [size=12pt]as has been discussed they did not come to any such conclusion while listening to music at safe levels - rather by turning up gain to problematic levels[/size][/size][/size]
 
[size=12pt][size=12pt][size=12pt]Meyer-Moran - like any such study can't "prove" "no one can hear X" - but the lack of positive result, supported by other studies of human hearing limits does seem to indicate that certainly "everyone" doesn't have this remarkable ultrasonic hearing ability[/size][/size][/size]
 
[size=12pt][size=12pt][size=12pt]I don't claim the question is closed - only that declaring flatly that there is "proof" that 44k sampling is inadequate is not respecting the poor quality of the "evidence" - it a speculative statement at this point[/size][/size][/size]
 
May 12, 2012 at 4:10 AM Post #150 of 156
[size=12pt]"They concluded via subjective listening that their SACD and DVD-A selections sounded almost universally better than CD quality" is [size=12pt]a gross misrepresentation - [size=12pt]as has been discussed they did not come to any such conclusion while listening to music at safe levels - rather by turning up gain to problematic levels[/size][/size][/size]

 
You're referring to this
 
"In one brief test with two subjects we added 14 dB of gain to the reference level quoted and tested the two sources with no input signal, to see whether the noise level of the CD audio channel would prove audible. Although one of the subjects was uncertain of his ability to hear the noise, both achieved results of 10/10 in detecting the CD loop. (We have not yet determined the threshold of this effect. With gain of more than 14 dB above reference, detection of the CD chain?s higher noise floor was easy, with no uncertainty. Tests with other subjects bore this out.)"
 
I'm referring to this
 
"4 A NOTE ON HIGH-RESOLUTION RECORDINGS
Though our tests failed to substantiate the claimed advantages of high-resolution encoding for two-channel audio, one trend became obvious very quickly and held up throughout our testing: virtually all of the SACD and DVD-A recordings sounded better than most CDs? sometimes much better. Had we not ?degraded? the sound to CD quality and blind-tested for audible differences, we would have been tempted to ascribe this sonic superiority to the recording processes used to make them.

Plausible reasons for the remarkable sound quality of these recordings emerged in discussions with some of the engineers currently working on such projects. This portion of the business is a niche market in which the end users are preselected, both for their aural acuity and for their willingness to buy expensive equipment, set it up correctly, and listen carefully in a low-noise environment. Partly because these recordings have not captured a large portion of the consumer market for music, engineers and producers are being given the freedom to produce recordings that sound as good as they can make them, without having to compress or equalize the signal to suit lesser systems and casual listening conditions. These recordings seem to have been made with great care and manifest affection, by engineers trying to please themselves and their peers. They sound like it, label after label. High-resolution audio discs do not have the overwhelming majority of the program material crammed into the top 20 (or even 10) dB of the available dynamic range, as so many CDs today do."
 
 
 
 
[size=12pt][size=12pt][size=12pt]Meyer-Moran - like any such study can't "prove" "no one can hear X" - but the lack of positive result, [/][/size][/size][/size]

 
There is no lack of a positive result - http://www.aes.org/e-lib/browse.cfm?elib=15398
 
If you'd like to replicate it ...
 
- Recording microphones (a pair of Sennheiser MKH 8020), FR of 10Hz-60kHz.
- Two stereo feeds from the mic preamp (Millennia HV-3D) to two Micstasy ADCs, one set to 44.1/24 the other to 88.2/24
- Test signals: five musical/instrumental (orchestra, classical guitar, cymbals , voice, violin) recordings from live performances, taking place in several halls/rooms with varying dimensions & acoustics.

- For use in tests, 5-8 sec excerpts were used, with no processing except fade in.out via Pyramix 6 software.
- The 88.2 excerpts were downsampled to 44.1 via Pyramix software, so there were three sets of signals, native 44.1, native 88.2, and downsampled 88.2-->44.1.
 
- Playback: 5 blocks corresponding to the 5 excerpts, 12 trials per block ( i.e. all pairwise combinations of the three versions, each presented 4 times, twice in each of the two presentation orders), randomized ABX protocol.
- Playback hardware was an RME Fireface 800 DAC, a Grace m906 monitor controller, and a Classe CA-5200 stereo amp, feeding a pair of B&W 802D loudspeakers (FR 70Hz-33kHz).
- To avoid clipping a 750ms switching interval was employed, set in the user interface which was Cycle '74's Max/MSP/Jitter software package, all playback was at 24 bits.
 
 
Note:  In my personal opinion, I think the 750ms interval is good.  I think rapid or immediate time aligned switching in ABX could cause an illusion that two sounds are the same.
 
This test used very well controlled recordings completely under their own jurisdiction (unlike other tests selecting titles randomly, without knowing the recording technique involved).
 
 
 
[size=12pt][size=12pt][size=12pt]supported by other studies of human hearing limits does seem to indicate that certainly "everyone" doesn't have this remarkable ultrasonic hearing ability[/size][/size][/size]

 
High-rez isn't necessarily about hearing above 22.05kHz.  If that were the case, I'd suspect an extremely limited number of people could perhaps hear information in the 22.05~26.00kHz range, with a very subtle difference in overtones and such, however it that were the case I'd declare it as a useless format.
 
Clearly, recording at 1Hz~100kHz is truer to reality, especially if it removes the need for all kinds of filters.
 
When you start discussing the various side-effects and designs of these filters, whether the extended frequency range is actually perceivable or not, and discuss the hardware limitations or performance of various ADC / DAC chips or if your speakers have IMD, it starts to become very academic.  You could ill-quote different spec sheets all night, it just depends on the scenario of the situation, and the same rule applies to all listening tests looking for transparency, they are limited to the equipment in the scenario and human perceptivity.
 
 
Just to OT for a moment, Lavry says about 192kHz " the negative consequences of the reduced oversampling ratio far outweigh any benefits derived from the higher sample rates."
 
Really?  My DAC states "...[size=x-small]this DAC has up to 10 times less out-of-band noise compared to anything similar on the market today.  It also has the capability to accept a 216KHz sample rate while keeping the same digital filter oversampling rate and the same speed of the modulator."[/size]
 
It all seems situational to me.
 
My current DAC costs less than $200 and SACD's are pretty cheap, with lamented higher sound quality, so I don't see how the marketing side is an issue either, DSD/.DSF playback will come down in price eventually.
 
The marketing to watch out for is fake high-rez and fake DSD.  You can do this with a program like http://spectro.enpts.com/, but as high-rez content becomes more popular the content above 22kHz could be fabricated as well.
 
All in all microphones, speakers and the music itself is 6000 times more important than this sector, but I really fail to see the issue in trying to improve every component in audio playback to it's highest potential or slightly beyond.
 
For musical enjoyment I'll keep listening to 92 kbps MP3 with $10 Sony earbuds, but for the curiosity and intimacy of audio playback I'm very happy to try DSD and headphones with an FR like this... whyever not?
 

 

Users who are viewing this thread

Back
Top