Conclusive "Proof" that higher resolution audio sounds different

[Tony Lauck]

Well was wondering if I was missing something there. Even if you compare bits in to bits out via a loop through analog with the clocks locked together you can get enough timing error to start corrupting results. A meter of cable connecting the two causes something over 3 nanoseconds of time difference which will effect higher frequencies enough to be out of the noise floor in the residuals.

You can time compensate after the fact in the digital domain. Software to do this exists. Programs such as Diffmaker may not have the resolution needed, but one can get more resolution by burning more computer cycles (time and electricity). But as I suggested, if one is concerned with the digital portions of a DAC, such as the modulator. one can do this testing entirely in the digital domain. This has the further advantage of not having any analog gear with the opportunity to confuse results if one is looking at the performance of purely digital algorithms.

When I tested Reefman's modulator I took his code and translated it into Python and then ran a simple program that did various low pass filters (including brick wall at 22.05 kHz) to generate the output of the DSD "DAC" so it could be compared with the input to the modulator. Given that the modulator failed badly on DC it wasn't even necessary to do any time compensation. If one tests with sine waves that are computer synthesized one can generate these with any desired phase shift using math library routines. Here one searches for the best null for a given frequency and if this is bad it proves the modulator is no good. With music signals more complex processing is required if one needs to take into account sub-sample delays, but my software was so far from real time that I didn't get involved with this back when I was playing around with this modulator.

In the analog case one must compensate for cable delays. One can deal with sub-sample delays by making the cable delay be an integral multiple of the clock rate plus a desired offset for a fractional part and then use a shift register to make up for the integral difference. I'm not serious about this approach, but it can be made to work by careful adjustment of cable properties. The following story illustrates one aspect of this approach. We were trying to measure the performance of a fiber optic receiver before we got our phase lock loop circuit working, so we had to make sure that the delay at the receiver was exactly a bit multiple. We could watch the transmitter and receiver signals on a scope. When things drifted out too much we started getting bit errors. We couldn't understand the source of the drift. We thought our oscillator was drifting and so we slaved it to an atomic clock that we had used in an earlier project and was lying around in the lab. In the end, we realized that we had built a thermometer using more than $50K of test equipment. The fiber delay was a function of temperature. If we had wanted, we could have fine tuned the delay by using a light bulb and a dimmer to control the temperature of the fiber.

I think this discussion is off topic for this thread, but I'll be glad to discuss it further with you off-line or in another thread.

 

[Orb]

I'm not sure exactly what you mean by "noise stability" so I would appreciate clarification.

I'm 100% with you on the question of noise modulation. I consider uncorrelated noise to be benign, provided it is at a sufficiently low level. However, correlated noise is distortion and the type of correlated noise created by DACs is potentially ugly. If you look at DAC chip data sheets and published DAC measurements it is often possible to see certain types of noise modulation by comparing graphs. There are DAC architecture techniques that convert correlated noise into uncorrelated noise, at the cost of extra hardware and possibly a higher total noise floor. I suspect chip designers often make tradeoffs to make their product "measure" better at the expense of potentially sounding worse. (DAC designers of high end products marketed to subjective listeners don't commit this particular design "sin" since they eschew published measurements.)

This is a good point although is all uncorrelated noise benign?
Case in point that of the ultrasonic hash generated from DSD modulation and implications this may have at the output; in terms of noise floor from RMS perspective when considering the generated swell from 20khz to 40khz, and possible implications for wideband audio equipment, as a sidenote this is technically not white noise like a PCM solution so no idea if this has some consideration as well.

Coming back to 16bit vs 24bit discussion; would be interesting to see how this also compares when transcoding native DSD to 16bit/96khz and 24bit/96khz with scope-focus on the whole range including ultrasonics that will have the DSD hash-swell.
Any discussion regarding transparency or lack of at 16bit that includes processing such as downsampling/decimation, this should also be applicable to transcoding; although in reality as mentioned earlier the general debate of 16bit vs 24bit should start with the recording/studio stage without relying upon tests that only show the transparency of downsampling/decimation.
I guess historically such discussions should had started with "what for all is the acceptable definition of 16bit in real world audio environment when comparing performance-transparency to 24bit" but that would be for another thread IMO and the horse has already bolted anyway.

Tony, you worked with studios when they 1st started with 16bit/44.1khz native recordings and if so what was your experience in difference between that and when moved to 24bit/88.2khz+ ?
I assume some benefits were possibly noticeable when it came to editing/mixing/mastering.
Native being the original sampling rate of recording-to-mastering file.

Thanks
Orb

 

[jkeny]

There are lots of good recent posts in here that are orthogonal to the topic of the thread & it seems a pity to curtail discussion on them because they are somewhat OT, so why not open another thread where the points raised here can be teased out some more?
I see Tony is involved in a thread on AA which has become contentious, called "Anyone can hear a difference" & that borders on these areas. Not wishing to start a contentious thread but lacking the skills of being a good headline writer, I'm going to use a variation of this title for the new thread "Can anyone hear a difference?"

Edit: The thread I started I called "Audio Theory Vs Implementation Reality" for want of a better name

 

[Tony Lauck]

This is a good point although is all uncorrelated noise benign?
Case in point that of the ultrasonic hash generated from DSD modulation and implications this may have at the output; in terms of noise floor from RMS perspective when considering the generated swell from 20khz to 40khz, and possible implications for wideband audio equipment, as a sidenote this is technically not white noise like a PCM solution so no idea if this has some consideration as well.

Coming back to 16bit vs 24bit discussion; would be interesting to see how this also compares when transcoding native DSD to 16bit/96khz and 24bit/96khz with scope-focus on the whole range including ultrasonics that will have the DSD hash-swell.
Any discussion regarding transparency or lack of at 16bit that includes processing such as downsampling/decimation, this should also be applicable to transcoding; although in reality as mentioned earlier the general debate of 16bit vs 24bit should start with the recording/studio stage without relying upon tests that only show the transparency of downsampling/decimation.
I guess historically such discussions should had started with "what for all is the acceptable definition of 16bit in real world audio environment when comparing performance-transparency to 24bit" but that would be for another thread IMO and the horse has already bolted anyway.

Tony, you worked with studios when they 1st started with 16bit/44.1khz native recordings and if so what was your experience in difference between that and when moved to 24bit/88.2khz+ ?
I assume some benefits were possibly noticeable when it came to editing/mixing/mastering.
Native being the original sampling rate of recording-to-mastering file.

Thanks
Orb

To reply to your questions I will say that with digital audio there is no uncorrelated noise. It's always correlated, the questions are how strong is this correlated noise, what is its frequency spectrum and what is the nature of the correlation? With TPDF dither and multi-bit PCM one can eliminate first order correlation (distortion) and second order correlation (noise power) but higher order correlation remains in the quantization noise. One can use an audio editor and word length truncation to create 1st order correlated noise, rectangular dither to create 2nd order correlated noise and TPDF dither to create 3rd order correlated noise. One can mix in white noise from a separate file to create uncorrelated noise. Then one can compare how these different files sound with the original. I have found this to be useful training. The differences are clearly audible when operating at the 8 bit level, so all of the noise is clearly audible without ruining one's hearing.

The ultra-sonic hash created by DSD is always correlated to the signal. It is mathematically impossible to remove the second or higher order moments of the correlated noise from a 1 bit format. The noise can be benign only if it is at a sufficiently low level and this will depend on the playback equipment and the listener's taste and experience. (I don't mean to knock DSD. I am a big fan of DSD, but there are a lot of false technical claims floating about that deserve to be debunked.)

When I first started working with digital audio (as opposed to playing recordings) I started with 88/24 and higher resolution. I had no interest in the 44/16 format as I had already concluded that it was not and could not be transparent. By that time computers, network bandwidth and storage were sufficiently cheap that the extra cost of high resolution was irrelevant.

For more on the theory of dither and noise correlation, read Wanamaker's PhD thesis, if you can follow graduate level math.

 

[microstrip]

(...) There are DAC architecture techniques that convert correlated noise into uncorrelated noise, at the cost of extra hardware and possibly a higher total noise floor. I suspect chip designers often make tradeoffs to make their product "measure" better at the expense of potentially sounding worse. (DAC designers of high end products marketed to subjective listeners don't commit this particular design "sin" since they eschew published measurements.) (...)

Could you give us a few examples of these architecture techniques? You are referring to a really interesting subject.

 

[Tony Lauck]

Could you give us a few examples of these architecture techniques? You are referring to a really interesting subject.

Post your question in jkeny's new thread and I'll be glad to answer.

 

[Orb]

To reply to your questions I will say that with digital audio there is no uncorrelated noise. It's always correlated, the questions are how strong is this correlated noise, what is its frequency spectrum and what is the nature of the correlation? With TPDF dither and multi-bit PCM one can eliminate first order correlation (distortion) and second order correlation (noise power) but higher order correlation remains in the quantization noise. One can use an audio editor and word length truncation to create 1st order correlated noise, rectangular dither to create 2nd order correlated noise and TPDF dither to create 3rd order correlated noise. One can mix in white noise from a separate file to create uncorrelated noise. Then one can compare how these different files sound with the original. I have found this to be useful training. The differences are clearly audible when operating at the 8 bit level, so all of the noise is clearly audible without ruining one's hearing.

The ultra-sonic hash created by DSD is always correlated to the signal. It is mathematically impossible to remove the second or higher order moments of the correlated noise from a 1 bit format. The noise can be benign only if it is at a sufficiently low level and this will depend on the playback equipment and the listener's taste and experience. (I don't mean to knock DSD. I am a big fan of DSD, but there are a lot of false technical claims floating about that deserve to be debunked.)

When I first started working with digital audio (as opposed to playing recordings) I started with 88/24 and higher resolution. I had no interest in the 44/16 format as I had already concluded that it was not and could not be transparent. By that time computers, network bandwidth and storage were sufficiently cheap that the extra cost of high resolution was irrelevant.

For more on the theory of dither and noise correlation, read Wanamaker's PhD thesis, if you can follow graduate level math.

Maybe we are debating semantics or looking from different POV-focus but to me white noise is technically uncorrelated from the signal due to possibly nature of the wideband input signal, the possibility of small amount of system noise (especially when involving tapes), and of course additive dither such as when truncating bit depth (such as flat TPDF).
That said this topic regarding dither has been done at length in the past including other types used-how studios may misuse it-etc, however never seen that paper before so thank you (I usually refer to work by Vanderkooy/Lipshitz and a few others).
Where it becomes more complicated is with DSD, but I sort of put that into the context the noise shaping pushes this outside of the audioband signal and so to my perspective not fully correlated but then has a characteristic that can be seen on the DAC/modulation output; not really uncorrelated I agree
Also there is the nature of the noise as I mentioned with PCM-Quantization-decimation usually white noise but this is not so with DSD.

With all that said, my interest regarding DSD is the potential for "side effects" further down the chain due to the modulation noise/output characteristics of DSD and that from what I understand this raises the noise floor a fair bit when taking the ultrasonic hash into consideration.
And curious if anyone ever bothered to test 16bit as transparent as 24bit by transcoding native DSD master file to PCM for both bit depths at 96khz (Philips theory/recommendation in the past is that one requires 32bit/384khz).

Thanks again for that link.
Orb

 

[Blazar]

damn this is a LONG thread...

I hope the studios are taking actual recordings so seriously as folks are worrying about CD bit rates. The "audiophile" recorded CD's are rarely the actual music I listen to. Standard redbook CD can sound dramatically different based on the recorded content as opposed to the actual bitrate, sampling frequency.

Adele 21 is an example of an album that sounds TERRIBLE considering how popular the CD became. I honestly think she should re-record that one to satisfy fans.

 

[amirm]

Adele 21 is an example of an album that sounds TERRIBLE considering how popular the CD became. I honestly think she should re-record that one to satisfy fans.

Is that the truth. I have her CD and Blu-ray version of her Live at Albert Hall performance and the CD is just horrid. I can't listen to it. The Blu-ray version is in a different class. Clearly someone butcherd the CD master. This is why I say regardless of spec differences, getting the upstream bits before production of CD has a better chance of having higher fideltiy.

 

[arnyk]

Is that the truth. I have her CD and Blu-ray version of her Live at Albert Hall performance and the CD is just horrid. I can't listen to it. The Blu-ray version is in a different class. Clearly someone butcherd the CD master. This is why I say regardless of spec differences, getting the upstream bits before production of CD has a better chance of having higher fideltiy.

QAmir, please do tell about the DBTs you did to support the conclusion that the CD sounds horrid, including the listener training and conformance with all of the important points of BS 1116. IOW, isn't your sauce for Meyer & Moran also sauce for you? ;-)

 

[amirm]

QAmir, please do tell about the DBTs you did to support the conclusion that the CD sounds horrid, including the listener training and conformance with all of the important points of BS 1116. IOW, isn't your sauce for Meyer & Moran also sauce for you? ;-)

Hi Arny.

Here is a bit about the comparison: http://www.whatsbestforum.com/showt...oyal-Albert-Hall&p=91019&viewfull=1#post91019

FYI, I finished an article on this topic for the Widescreen Review magazine this week. Hopefully I met the deadline for it to get published in the January issue. Once there, I will put a copy of it online. It reviews both Meyer and Moran tests, and Stuart et al. listening tests countering theirs. Plus some of the business components of high resolution audio. Should free me up from constantly explaining the same thing over and over again . And oh it also includes the above anecdotal reference to Adele concert .

 

[cooljazz]

Waveform of Adele song from 2012...



Waveform of randomly picked song of older music.....A Linda Ronstadt cut from 1974.



"Nuff said!

CJ

 

[arnyk]

Hi Arny.

Here is a bit about the comparison: http://www.whatsbestforum.com/showt...oyal-Albert-Hall&p=91019&viewfull=1#post91019

Amir, you could have just admitted that you have one very lofty standard for others and routinely set a far lower bar for your own work. Thanks for documenting the same, so clearly.

 

[Orb]

QAmir, please do tell about the DBTs you did to support the conclusion that the CD sounds horrid, including the listener training and conformance with all of the important points of BS 1116. IOW, isn't your sauce for Meyer & Moran also sauce for you? ;-)

I appreciate this is not your point, but listen to Adele 21; she is a great talent but yeah this CD is effing awful in terms of sound quality, really annoying as there are really good songs on the album.
And the trend on her CD is seen with some other recent talented female musicians/singers CD releases.

Cheers
Orb

 

[arnyk]

I appreciate this is not your point, but listen to Adele 21; she is a great talent but yeah this CD is effing awful in terms of sound quality, really annoying as there are really good songs on the album.
And the trend on her CD is seen with some other recent talented female musicians/singers CD releases.

The above appears to be a suggestion that I perform a sighted evaluation without any reliable reference. I might commit such a ridiculous act in private due to curiosity, but I would never report its outcome on a public forum along with any claim worth posting.

If there is no reliable reference for comparison, then any perceptions are about as speculative as they come.

There seems to be this audiophile mythological perception that compression can only harm sound quality, but that is not always true. The axiom that all generalizations are false stands despite its self-contradictory nature. ;-)

For example, the apparent artistic skill of a vocalist can be improved with manual gain riding, algorithmic compression and pitch correction when applied skillfully and in moderation.

These effects are clearly audible when applied to excess, but in moderation they can improve the apparent quality of the artistry and technique for both instrumentalists and vocalists.

 

[amirm]

Amir, you could have just admitted that you have one very lofty standard for others and routinely set a far lower bar for your own work. Thanks for documenting the same, so clearly.

Hi Arny. Not sure what you think I have documented here as you are not being specific. Would you please explain?

 

[Orb]

The above appears to be a suggestion that I perform a sighted evaluation without any reliable reference. I might commit such a ridiculous act in private due to curiosity, but I would never report its outcome on a public forum along with any claim worth posting.

If there is no reliable reference for comparison, then any perceptions are about as speculative as they come.

There seems to be this audiophile mythological perception that compression can only harm sound quality, but that is not always true. The axiom that all generalizations are false stands despite its self-contradictory nature. ;-)

For example, the apparent artistic skill of a vocalist can be improved with manual gain riding, algorithmic compression and pitch correction when applied skillfully and in moderation.

These effects are clearly audible when applied to excess, but in moderation they can improve the apparent quality of the artistry and technique for both instrumentalists and vocalists.

Err,
as I SAID " I appreciate this is not your point, but listen to Adele 21"...... I then went on to mention how the CD and other modern talented female artists do actually sound bad from a recording-mastering perspective on CD/[iTunes]/etc and so might find it of interest to have a listen.
So your response is not applicable; unless you have listened to Adele and other comparable female artists and done an analysis that you reported upon?
Cheers
Orb

 

[amirm]

The above appears to be a suggestion that I perform a sighted evaluation without any reliable reference. I might commit such a ridiculous act in private due to curiosity, but I would never report its outcome on a public forum along with any claim worth posting.

That is easy for you to say Arny because according to your own statements, you have never, ever documented the results of any listening tests online! I asked you repeatedly to give us a link to such tests and all you would do is point to this one amplifier test:

If you have documented other tests and they have all been blind, I would love to evaluate them. So please provide a link .

For now, my evaluation was blind. But before I get into more of that, this is from Meyer and Moran test published in the Journal of AES as I have quoted in my upcoming article:

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.

So with respect to difference in mastering, the AES Journal allowed inclusion of such sighted evaluation. You say that sighted evaluation of difference in mastering is not to be mentioned online yet it is part of the JAES. You think the review board made a grave mistake by not flagging this for deletion?

 

[arnyk]

That is easy for you to say Arny because according to your own statements, you have never, ever documented the results of any listening tests online!

It is true that I worked with others who have helped me document some of the DBTs that I was involved with:

Greenhill, Laurence , "Speaker Cables: Can you Hear the Difference?" Stereo Review, ( Aug 1983)
Greenhill, L. L. and Clark, D. L., "Equipment Profile", Audio, (April 1985)
Nousaine, Thomas, "Wired Wisdom: The Great Chicago Cable Caper", Sound and Vision, Vol. 11 No. 3 (1995)
Nousaine, Thomas, "Flying Blind: The Case Against Long Term Testing", Audio, pp. 26-30, Vol. 81 No. 3 (March 1997)
Nousaine, Thomas, "Can You Trust Your Ears?", Stereo Review, pp. 53-55, Vol. 62 No. 8 (August 1997)
Masters, I. G. and Clark, D. L., "Do All Amplifiers Sound the Same?", Stereo Review, pp. 78-84 (January 1987)
Masters, Ian G. and Clark, D. L., "Do All CD Players Sound the Same?", Stereo Review, pp.50-57 (January 1986)
Masters, Ian G. and Clark, D. L., "The Audibility of Distortion", Stereo Review, pp.72-78 (January 1989)
Carlstrom, David, Greenhill, Laurence, Krueger, Arnold, "Some Amplifiers Do Sound Different", The Audio Amateur, 3/82, p. 30, 31, and Hi-Fi News & Record Review, Link House Magazines, United Kingdom, Dec 1982, p. 37.

etc. etc. etc.

 

[Joe Whip]

I have the Adele 21 CD. It does sound dreadful.