Quantcast

Meyer & Moran Updated: Proof that CD audio isn't transparent!

Welwynnick

Distinguished Member
In September 2007, Brad Meyer and David Moran of the Boston Audio Society published their seminal scientific paper through the Audio Engineering Society entitled "Audibility of a CD-Standard A/DA/A Loop Inserted into High-Resolution Audio Playback". In summary this stated that:
Claims both published and anecdotal are regularly made for audibly superior sound quality for two-channel audio encoded with longer word lengths and/or at higher sampling rates than the 16-bit/44.1-kHz CD standard. The authors report on a series of double-blind tests comparing the analog output of high-resolution players playing high-resolution recordings with the same signal passed through a 16-bit/44.1-kHz “bottleneck.” The tests were conducted for over a year using different systems and a variety of subjects. The systems included expensive professional monitors and one high-end system with electrostatic loudspeakers and expensive components and cables. The subjects included professional recording engineers, students in a university recording program, and dedicated audiophiles. The test results show that the CD-quality A/D/A loop was undetectable at normal-to-loud listening levels, by any of the subjects, on any of the playback systems. The noise of the CD-quality loop was audible only at very elevated levels.
This has been quoted ever since, and is the de facto proof that audiophools are only able to hear the benefit of high definition audio in sighted tests. Grumblings aside, these findings have hardly been challenged ever since, but AVS launched an initiative this year for members to perform rigorous double-blind ABX tests on themselves, to see who could hear the limitations of 16/44.1 digital audio.

One of the people who took up the challenge was Amir Majidimehr (amirm), the guy who led Microsoft's audio and video codec devlopment, and who spearheaded the HD-DVD effort in the HD video format war. He was able to reliably distinguish when the original 24/96 tracks were down-sampled to 16/44. Not just reliably, he was able to get the DBT ABX test right every time and give an uncertainly score of 0.0%. And he did this on a laptop, not an esoteric audio system.

Conclusive "Proof" that higher resolution audio sounds different

Granted, some people cannot hear this difference, but Amir and a few others were able to demonstrate discrimination to a level that was beyond any statistical uncertainty or debate. This appears to prove that CD standard audio (16 bit/44.1kHz) is not transparent after all.
The above discussion was triggered by a formal effort on AVS Forum to compare CD sampling rate and bit depth to "High res music" (24-bit/96 Khz). You can read more about it here including a link to the files: http://www.avsforum.com/forum/91-aud...-take-2-a.html
Yet again, I did not think I could hear the differences. But the results proved otherwise . Here are my double blind ABX tests of the files:
foo_abx 1.3.4 report
foobar2000 v1.3.2
2014/07/10 21:01:16

File A: C:\Users\Amir\Music\AIX AVS Test files\Just_My_Imagination_A2.wav
File B: C:\Users\Amir\Music\AIX AVS Test files\Just_My_Imagination_B2.wav

21:01:16 : Test started.
21:02:11 : 01/01 50.0%
21:02:20 : 02/02 25.0%
21:02:28 : 03/03 12.5%
21:02:38 : 04/04 6.3%
21:02:47 : 05/05 3.1%
21:02:56 : 06/06 1.6%
21:03:06 : 07/07 0.8%
21:03:16 : 08/08 0.4%
21:03:26 : 09/09 0.2%
21:03:45 : 10/10 0.1%
21:03:54 : 11/11 0.0%
21:04:11 : 12/12 0.0%
21:04:24 : Test finished.

---------- Total: 12/12 (0.0%)
Summarizing, I managed to consistently tell all three files apart from their downsampled 44.1 Khz/16 bit versions.

Track 1 and 3 were relatively easy to tell apart but both took fair amount of effort to find the critical segments were the difference could be heard. The middle track #2 was essentially not distinguishable but I managed to finally find the difference.

Again, remember all of these tests are double blind ABX, exactly the kind demanded to be provided for "proof." Levels were matched. So there is really no wiggle room here. I was able to find all the differences.
I have to say that reading these results made my day.

Nick
 
Last edited:

Welwynnick

Distinguished Member
There's another AES paper here by Pras and Guastavino at McGill University, Montreal:

AES E-Library » Sampling Rate Discrimination: 44.1 kHz vs. 88.2 kHz

This also states that trained listeners can distinguish between digital audio files recorded at 88.2 and those recorded at (or downsampled to) 44.1kHz:

It is currently common practice for sound engineers to record digital music using high-resolution formats, and then down sample the files to 44.1kHz for commercial release. This study aims at investigating whether listeners can perceive differences between musical files recorded at 44.1kHz and 88.2kHz with the same analog chain and type of AD-converter. Sixteen expert listeners were asked to compare 3 versions (44.1kHz, 88.2kHz and the 88.2kHz version down-sampled to 44.1kHz) of 5 musical excerpts in a blind ABX task. Overall, participants were able to discriminate between files recorded at 88.2kHz and their 44.1kHz down-sampled version. Furthermore, for the orchestral excerpt, they were able to discriminate between files recorded at 88.2kHz and files recorded at 44.1kHz.
Nick
 

simon ess

Well-known Member
Interesting, but there's something I don't understand. I'm perhaps just missing it so am looking for clarification.

How do you do abx testing on yourself?

If those same individuals subject themselves to rigorous, independent scientific tests, would you expect the results to be the same?
 

andy1249

Distinguished Member
The Meyer/Moran paper is free and freely distributable, this is the main reason it's so widely read , taken as fact, and used so much as a stick to beat people with.

It's been discredited and contradicted lots of times, as in the paper linked to above, but those papers are not free and not freely distributable , costing approx 20 dollars a pop and your not allowed to share or quote other than the abstract, as above.

In short, the free material gets all the attention, incorrect though it so obviously is to anyone with the gear to have a listen.
 

Welwynnick

Distinguished Member
Exactly the same as "Detection threshold for distortions due to jitter on digital audio" by Ashihara et al.
It was shown that the detection threshold for random jitter was several hundreds ns for well-trained listeners under their preferable listening conditions.
https://www.jstage.jst.go.jp/article/ast/26/1/26_1_50/_pdf

It looked more scientific and reliable than sighted tests, so people believed what they read online.

Nick
 

lokyc

Well-known Member
Those are in test conditions. In real life, the difference is probably more discernable. Highe res content tend to have a higher premium and are likely to be properly remastered to take advantage of the bit depth. On dynamics alone it will be superior.

16/44 CDs are overly compressed (google loudness war), so the effective bit depth is far less than 16 bit anyway.
 

TomScrut

Well-known Member
Is it the recording and the actual data that has the problem here or is it the way the DAC uses it? Most DACs nowadays need to upsample, will this have a negative effect on the sound?
 

andy1249

Distinguished Member
Those are in test conditions. In real life, the difference is probably more discernable. Highe res content tend to have a higher premium and are likely to be properly remastered to take advantage of the bit depth. On dynamics alone it will be superior.

16/44 CDs are overly compressed (google loudness war), so the effective bit depth is far less than 16 bit anyway.
The original Meyer/Moran paper already says that clearly sacds and DVD audio discs sound better than CDs but put this down to mastering and loudness war phenomenon as you say here.

What they did was take the high res content from sacd and DVD audio and downsample it to 16/44 and then claim no one could tell the difference, which has been disproven and is what is under discussion here.

The Meyer/Moran paper is linked to at reference number 42 in the sacd wiki, note what is said under the conclusions...
http://drewdaniels.com/audible.pdf

In short, the same equipment and the same content, but with the content down sampled to 16/44 and in its original higher resolutions , can the difference be heard?
Meyer/Moran says no...but as said, this paper is free to share, the ones that say different are not, so most will have read this one!

It most definitely can be heard, especially if you are very familiar with the material.
 

lokyc

Well-known Member
No disagreement with you here. Its good to put all that "digital is the same" or "highres doesn't matter" argument to bed; or at least not consider it a done deal.

Of course, we need more than just subjective hearing tests to be fully convinced. And I think Rob Watts approach with the Hugo provides a good scientific argument that its more than sample theorem at stake here. The frequencies may be produced faithfully, but not their discrete time intervals.

AFAIK, back in the day, AD conversion was to DSD format. Now its more like 24/192. Downsampling creates quantization error. So its not a truly lossless conversion. Even if frequency and volume is accurately coded, the timing information is lost. The sampled frequency may even be a "just" an approximation.

If time domain doesn't matter, then the Hugo wouldn't be making waves despite looking like a college science project.

the question is for the "average joe" music lover, of which I consider myself one of, is whether any of these technical mambo jumbo matters. I mean I mainly browse Richersounds and when feeling flushed, venture into Sevenoaks. Furthermore, I only listen to music when I'm say browsing the net and stuff rather than have adedicated room, darkened lighting and focused listening etc. Would it make a difference to me?

I have not done enough tests to answer that question. But even the most basic gear will benefit from better mastering which high res sources will be.

As for oversampling Tom, the idea is to improve noise shaping. By increasing the sample resolution, ilters can be applied much finer and reduce signal loss.. Net result is increased SNR.

The FPGA in the Hugo oversamples by the thousands to improve the resolution of humble PLL jitter correction, such that its pretty much what is required.
 

TomScrut

Well-known Member
I know why it does it, but I am purely speculating as to whether it has an effect on the comparison
 

lokyc

Well-known Member
I see. But it shouldn't affect the test any more than the DACs will do in real life. From what I can gather from all the high end DACs, they also use oversampling to improve interpolation and smoothen the wave for.

On a side track, I might be going to listen to a Exasound DAC later. I'll let you know how it goes.
 

TomScrut

Well-known Member
Ah that is interesting... Must be one of the few in the country! They need to get dealers sorted over here as I aren't going to buy one without a listen
 

lokyc

Well-known Member
Don't know if you picked it up on the thread "London Society of Sound". TB303 and I got in touch with David and turns out he runs one of them. He came over from New York so his gear is mainly N American.

Turns out he not only lives in London, but is walking distance from TB303!. I remembered you expressing an interest. Pity you're all the way in Scarborough, but I'm sure he wouldn't have a problem if you want a listen. I thought I might give you my impressions and you can then see if its worth making the effort over here!

Digression over. (sorry mods)
 

Jota180

Active Member
The original Meyer/Moran paper already says that clearly sacds and DVD audio discs sound better than CDs but put this down to mastering and loudness war phenomenon as you say here.

What they did was take the high res content from sacd and DVD audio and downsample it to 16/44 and then claim no one could tell the difference, which has been disproven and is what is under discussion here.

The Meyer/Moran paper is linked to at reference number 42 in the sacd wiki, note what is said under the conclusions...
http://drewdaniels.com/audible.pdf

In short, the same equipment and the same content, but with the content down sampled to 16/44 and in its original higher resolutions , can the difference be heard?
Meyer/Moran says no...but as said, this paper is free to share, the ones that say different are not, so most will have read this one!

It most definitely can be heard, especially if you are very familiar with the material.
Did Meter and Moran really say no or was it their test subjects?
I think you're confusing scientific testing with an agenda led test and no one who knows the slightest thing about psycho acoustics or bias can or should say they can hear this that or the other without doing it blind.

Comparing 24 bit and 16 bit examples of the 'same' music, the greatest importance must be put on the mastering. It has to be the same or comparisons cannot meaningfully be made between 24 and 16 bit.

And frankly, all this thread and no doubt numerous others around the internet proves is there's believers, non believers and those open to being persuaded by the evidence.

I'm the last of those three and it would seem to be 1-1 in the comprehensive testing sphere which is hardly conclusive either way don't you think? Or are you a confirmed believer?
 

Jota180

Active Member
There's another AES paper here by Pras and Guastavino at McGill University, Montreal:

AES E-Library » Sampling Rate Discrimination: 44.1 kHz vs. 88.2 kHz

This also states that trained listeners can distinguish between digital audio files recorded at 88.2 and those recorded at (or downsampled to) 44.1kHz:



Nick
They also found that the trained listeners - 9 sound engineers and 4 musicians of age ~28 (SD 5.6 yr) A/B testing CD and MP3 the results were for 256 and 320 preference was 50/50.
For 192, 128 and 96 it was 60/40, 75/25 and 80/20, respectively (significant).
 

andy1249

Distinguished Member
Did Meter and Moran really say no or was it their test subjects?
I think you're confusing scientific testing with an agenda led test and no one who knows the slightest thing about psycho acoustics or bias can or should say they can hear this that or the other without doing it blind.

Comparing 24 bit and 16 bit examples of the 'same' music, the greatest importance must be put on the mastering. It has to be the same or comparisons cannot meaningfully be made between 24 and 16 bit.

And frankly, all this thread and no doubt numerous others around the internet proves is there's believers, non believers and those open to being persuaded by the evidence.

I'm the last of those three and it would seem to be 1-1 in the comprehensive testing sphere which is hardly conclusive either way don't you think? Or are you a confirmed believer?
By Meyer/Moran I'm referring to the paper , not the people ! I,m an Electronics Engineer in chip design , I read and write papers all the time, and it's common practice to refer to papers by the authors names, especially where the title is long, and they nearly always are!

The paper is linked too, did you read it ?
The test was blind under strict aes conditions, the material used during the test "was" the same material with the same mastering!

Other tests under the same conditions contradict this paper, with the only variable being the people under test.
The issue with Meyer/Moran seems to be that the subjects didn't know what to listen for! Change that and the results change, as per the other papers linked too here.

For a lot of people, music is just a nice background to a party, they don't really listen to it.
I think if such a test is carried out with randomly selected subjects from the general population, the result will always be 50/50.
Conduct the same test with musicians, recording engineers , and those that love the material, and you always have a significant result.
 
Last edited:

Steven

Senior Moderator
Comparing 24 bit and 16 bit examples of the 'same' music, the greatest importance must be put on the mastering. It has to be the same or comparisons cannot meaningfully be made between 24 and 16 bit.
The problem with the internet is that there is a lot of talk, which does not equate to knowledge. Probably one of the best sites every audio enthusiast should RSS feed bookmark is realhd-audio: 16 vs. 24 Bits | Real HD-Audio

Also investing in "high res" now is probably money down the drain. DSD fad could just as easily disappear like sacd for the masses. The format that wins will be the one Apple adopt for iTunes if and when they get around to it
 

davidavdavid

Novice Member
Ah that is interesting... Must be one of the few in the country! They need to get dealers sorted over here as I aren't going to buy one without a listen
There is but one dealer, Sound Setup (Andrew Tilley) of exaSound here in the UK. He is also the distributor for Janzsen electrostatic speakers from Columbus. Ohio. He's the man. ;)
 

sounddog

Well-known Member
Bob Stuart (of Meridian) also presented a counter to the Meyer/Moran paper at this years AES meeting. (Sorry don't have any links to hand).
 

Trollslayer

Distinguished Member
This isn't a surprise but if it's CD or audio cassette then there is no contest. A big step forward.
 

lokyc

Well-known Member
For pedants and geeks like me (even if I don't completely understand the complex lingo), its the why.

Digital vs analogue should really be Red Book vs analogue.

The camp who say Red Book is more than enough (yes, even compared to Hi res), base their arguments mainly from the Nyquist sampling theorem and human perception of dynamic range.

If it was only test tones, then yes, the theory applies. But in music, it is more nuanced.

There is the time domain factor. ADDA conversion is also more complex than just sampling and interpolation alone.

As we understand more about psychoacoustics, we also learn human hearing is more than just frequency and loudness. Of course, the proof comes from listening tests.

For those who have insisted Red Book vs Vinyl is different, it is good vindication. No longer will they been seen as just deluded Luddites.
 

Trollslayer

Distinguished Member
Nyquist frequency is about general bandwidth i.e. amplitude of a sine wave, it does not look at maintaining a complex signal.
 

Trending threads

Latest News

Disney developing AI based video compression
  • By Andy Bassett
  • Published
Vizio announces Disney+ available on SmartCast TVs
  • By Andy Bassett
  • Published
Audio-Technica reveals Kokutan and Asada Zakura headphones
  • By Andy Bassett
  • Published
Sky announces combined Sky Sports and BT Sport packages
  • By Andy Bassett
  • Published
Qobuz drops MP3s and introduces Studio Premier Plan
  • By Andy Bassett
  • Published
Top Bottom