Was Meyer and Moran debunked by Robert Stuart?

[esldude]

I don't think anyone has suggested we have no memory after 15 seconds. Or that familiarity with an audio scene or stream doesn't enhance your ability to recognize certain aspects of that situation. However, your ear and immediate sensory system has to perceive auditory information for it to be used at all. In the extreme if your ear drum doesn't move then you can't hear it and no amount of additional exposure will enhance that.

 

None of the links indicate sound being perceived longer term that isn't shorter term. They do suggest your ability to recognize patterns or understand context changes. Your ability to consciously relate something like who you hear talking increases with familiarity. Your base perceptual sensitivity doesn't change according to these. It is a regular result to score on blind tests after you no longer consciously recognize a difference. Perhaps with repeated exposure you would consciously recognize it, but it gets perceived either way. Again these longer term effects require you to actually get it above a hearing threshold. The probing of that threshold is best done with shorter test intervals.

 

Longer term would you find you prefer DAC A over DAC B or speaker A over speaker B in some emotional sense or from recognizing a longer term sound signature? You might assuming there is a perceptible sound difference. Your feeling about how well you like it may sour or go up over time. Then again your feeling about it can change over time for many reasons even when there is no perceptible sound difference. Long term listening doesn't reliably separate that out.

[mmerrill99]

I don't think anyone has suggested we have no memory after 15 seconds.

No, but the contention is that comparisons outside of this time-frame are less reliable.

Or that familiarity with an audio scene or stream doesn't enhance your ability to recognize certain aspects of that situation. However, your ear and immediate sensory system has to perceive auditory information for it to be used at all. In the extreme if your ear drum doesn't move then you can't hear it and no amount of additional exposure will enhance that.

Audio streams are, of course, based on perceptible auditory information but the grouping into streams isn't confined to just 15 seconds - it happens on a continual basis throughout an audio piece & is a dynamic process. Taking a 15 sec. audio extract to spot differences between two audio files will not necessarily reveal this aspect of the reproduction & many consider this aspect the most important part. If this doesn't happen correctly & consistently we lose some of this aspect of the music

 

None of the links indicate sound being perceived longer term that isn't shorter term. They do suggest your ability to recognize patterns or understand context changes. Your ability to consciously relate something like who you hear talking increases with familiarity. Your base perceptual sensitivity doesn't change according to these. It is a regular result to score on blind tests after you no longer consciously recognize a difference. Perhaps with repeated exposure you would consciously recognize it, but it gets perceived either way. Again these longer term effects require you to actually get it above a hearing threshold. The probing of that threshold is best done with shorter test intervals.

I contend that the ease with which we differentiate the audio streams within the music is paramount to how we feel about it's realism. In other words it's the consistency (over the full song) in reproducing these audio factors. As said, these factors & their interrelationships are being studied in the ongoing ASA research.

 

Longer term would you find you prefer DAC A over DAC B or speaker A over speaker B in some emotional sense or from recognizing a longer term sound signature? You might assuming there is a perceptible sound difference. Your feeling about how well you like it may sour or go up over time. Then again your feeling about it can change over time for many reasons even when there is no perceptible sound difference. Long term listening doesn't reliably separate that out.

 

I think what SPDIF-USB & I are saying is that some perceptible differences are probably only recognised in long-term listening. In music, this psychoacoustic separation into musical streams is what makes us understand the interplay between musicians & the emotional structure of the piece.

[gmgraves]

What is below is from a thread George Graves started a few weeks ago. It became a mess even though no one had access to the paper at that time. So I am starting this again by quoting George's first post from the other thread. Hoping it can stay a bit cleaner now that the AES paper is out.

 

 

I recently found out that the oft-cited and notorious Meyer and Moran paper on high-res audibility was debunked in a new paper delivered at this month's 137th AES (Audio Engineering Society) Convention in LA by no less an authority on digital audio than Robert J. Stuart of Meridian. I have been unable to find the paper posted online, but I do have this capsule description that was posted by Mark Waldrep, the head of AIX Records, who attended:

 

"I finished the afternoon by attending a few paper sessions. The first was titled, “The Audibility of Typical Digital Audio Filters in a High-Fidelity Playback System”. Although it may not be obvious from the title of the paper, this is the first AES publication that refutes the Meyer/Moran research that has been so often quoted as “proof” that CD specification PCM audio is enough for music reproduction (Meyer and Moran’s research has been widely discredited including by myself because of the lack of real high-resolution content used during the study).

Robert Stuart and his colleagues conclude: “first there exist audible signals that cannot be encoded transparently by a standard CD; and second, an audio chain used for such experiments must be capable of high-fidelity reproduction.” In other words, CDs aren’t good enough. This paper was given the top award by the AES organization. This is a very important finding."

 

If anyone here has access to this paper, or more information on it, please post it as this is critical to a number of strongly held beliefs. Chief among which, is that hi-resolution audio is a high-end marketing gimmick that adds no "real" value to recorded music, and also that there is no difference between DACs, and that one can encode/decode digital audio multiple times with no perceptible change in SQ.

 

 

The problem with the Meyer and Moran study, to my way of thinking, is simply that the premise is ludicrous at the outset. If we are to believe that any musical performance can be digitally encoded and decoded through a daisy-chain of processors multiple times and at the end of the chain still remains audibly indistinguishable from the original digital source, then we might as well believe that all of high-end audio is a fraud! Never mind the subtle changes that people claim to hear between cables, never mind the larger differences we hear between CD players, Red-Book vs high-resolution formats, different amplifiers, different DACs, or even different speakers; musical signals that have been digitally quantized at 44.1 KHz/16-bits and then converted back to audio over and over and over again, results in no audible difference between the result and the original source? I'm sorry, that premise is ridiculous on the face of it, and the fact that Meyer/Moran profess to have proved that premise just stretches all credibility. Something just has to be wrong with such ABX test results! Even if Robert Stuart has failed to debunk Meyer/Moran, I don't buy their results, and you know what a skeptic I am when it comes to audio results that can't be measured, mathematically proven, or generally heard under any conditions.

[beanbag]

The problem with the Meyer and Moran study, to my way of thinking, is simply that the premise is ludicrous at the outset. If we are to believe that any musical performance can be digitally encoded and decoded through a daisy-chain of processors multiple times and at the end of the chain still remains audibly indistinguishable from the original digital source, then we might as well believe that all of high-end audio is a fraud! Never mind the subtle changes that people claim to hear between cables, never mind the larger differences we hear between CD players, Red-Book vs high-resolution formats, different amplifiers, different DACs, or even different speakers; musical signals that have been digitally quantized at 44.1 KHz/16-bits and then converted back to audio over and over and over again, results in no audible difference between the result and the original source? I'm sorry, that premise is ridiculous on the face of it, and the fact that Meyer/Moran profess to have proved that premise just stretches all credibility. Something just has to be wrong with such ABX test results! Even if Robert Stuart has failed to debunk Meyer/Moran, I don't buy their results, and you know what a skeptic I am when it comes to audio results that can't be measured, mathematically proven, or generally heard under any conditions.

 

I dunno, man. That sounds like strawman to me. The M&M experiment, and this current paper under discussion, is only about resampling higher bit-rate down to lower, or digital filtering, and has nothing to do with speakers, amps, cables, etc.

[esldude]

The problem with the Meyer and Moran study, to my way of thinking, is simply that the premise is ludicrous at the outset. If we are to believe that any musical performance can be digitally encoded and decoded through a daisy-chain of processors multiple times and at the end of the chain still remains audibly indistinguishable from the original digital source, then we might as well believe that all of high-end audio is a fraud! Never mind the subtle changes that people claim to hear between cables, never mind the larger differences we hear between CD players, Red-Book vs high-resolution formats, different amplifiers, different DACs, or even different speakers; musical signals that have been digitally quantized at 44.1 KHz/16-bits and then converted back to audio over and over and over again, results in no audible difference between the result and the original source? I'm sorry, that premise is ridiculous on the face of it, and the fact that Meyer/Moran profess to have proved that premise just stretches all credibility. Something just has to be wrong with such ABX test results! Even if Robert Stuart has failed to debunk Meyer/Moran, I don't buy their results, and you know what a skeptic I am when it comes to audio results that can't be measured, mathematically proven, or generally heard under any conditions.

 

Well one of those "aha" moments for me was getting an MSB AD unit to go with a Wadia 12 DAC. I still had phono, FM, reel-to-reel machine etc. I found AD to DA for all of those to be utterly transparent. More so than any pre-amp I had used other than perhaps a Spectral (which I only used a couple weeks). I had a pre-amp on hand initially to do a side by side comparison. I also did the comparisons direct vs AD/DA in the middle. It just didn't do anything you could hear short or long term.

 

I am sure plenty are already thinking I must not hear that well, or have good equipment etc. etc. etc. I ran into some skepticism from friends too. Until I demoed the very effect, and they couldn't hear it in the loop either vs a piece of wire (for a couple I let them bring their favorite wire). This was before M&M. But when I saw their test it wasn't as hard for me to believe as some might find it that an AD/DA loop is transparent. In my case it was a 48/24 conversion. The Wadia was really good for about 20 bits of resolution.

 

Reiterating my opinion on it I suppose, but that would appear the strongest of the things you could claim for M&M. Whether it was fully transparent, and an excellently run test or not, the AD/DA loop in the middle made little difference. If it was a major degradation that many imagine it would have easily been heard.

[tailspn]

I don't think anyone has suggested we have no memory after 15 seconds.

 

I do. More like 15 milliseconds. We're simply not wired that way. Visual experiences, yes. Aural, no.

 

Hearing, and the mental processing of sound is meant to be a defense mechanism. Snake in the grass localization, to then train our eyes upon for recognition. We remember aural cues well, not details. We're actually very easily fooled with memorizing aural details, and believe others, especially those claiming to possess "golden ears", have a special ability to judge for us.

 

That does not mean we can not be trained to recognize the value of particular cues, and how they relate to audio system performance. But A/B ing two 15 second relatively close pieces of audio content, and reliably comparing the two; forget it.

[esldude]

I do. More like 15 milliseconds. We're simply not wired that way. Visual experiences, yes. Aural, no.

 

Hearing, and the mental processing of sound is meant to be a defense mechanism. Snake in the grass localization, to then train our eyes upon for recognition. We remember aural cues well, not details. We're actually very easily fooled with memorizing aural details, and believe others, especially those claiming to possess "golden ears", have a special ability to judge for us.

 

That does not mean we can not be trained to recognize the value of particular cues, and how they relate to audio system performance. But A/B ing two 15 second relatively close pieces of audio content, and reliably comparing the two; forget it.

 

Well actually if it is 15 seconds, you need an instant switch and two 7.5 second segments. 15 seconds is probably the high side of echoic memory.

[beanbag]

 

I think what SPDIF-USB & I are saying is that some perceptible differences are probably only recognised in long-term listening. In music, this psychoacoustic separation into musical streams is what makes us understand the interplay between musicians & the emotional structure of the piece.

 

OK, that's nice and all, but I am looking for some kind of actionable advice or technical metric I can use to tell if CD material can be as good as all that other "hi res" formats.

 

Getting back on topic...

[mmerrill99]

OK, that's nice and all, but I am looking for some kind of actionable advice or technical metric I can use to tell if CD material can be as good as all that other "hi res" formats.

 

Getting back on topic...

Actionable advice - Try listening !!

[spdif-usb]

Actionable advice - Try listening !!

Best piece of advice ever IMO...

[beanbag]

What is the point of research publications if I have to re-do all their work anyway?

[semente]

Actionable advice - Try listening !!

 

It would be easier if one could access 16/44.1 and High-Res files that were actually comparable...

 

R

[mmerrill99]

It would be easier if one could access 16/44.1 and High-Res files that were actually comparable...

 

R

What's wrong with downsampling a high-res file to 16/44.1?

[semente]

Do you mean doing it myself?

What software should I use?

 

R

[Jud]

What's wrong with downsampling a high-res file to 16/44.1?

 

You'd be hearing the performance of the downsampling software, which quite possibly differs greatly from that used when making most CDs.

 

To semente's question, I'd suggest trying out Audiophile Inventory, which I use for offline upsampling, but was originally made for downsampling.

[semente]

Thanks, I'll give it a try.

 

R

[esldude]

There are number of places with format comparison files. The problem is the resampling software. iZotope is probably the Gold standard. However, almost all of them will reduce or alter the volume. They also usually mess with the timing. I believe iZotope has or will soon update their software to maintain timing and level. The level change is usually .2 or .3 db which is just enough to invalidate a careful comparison. The resampled file is lower in level and will sound less good for just that one reason.

 

I don't know about the one Jud suggests. Its performance appears to be good, but I don't know about the timing and level issue.

 

Sox is a free command line resampler that doesn't alter timing or level and does an excellent job. It is also in Audacity if you have version 2.0.3 or newer. I can tell you how to set things in preferences of Audacity for good performance.

 

Here is what is left after converting one of Barry Diament's 192 files to 44/24 and back to 192/24. The FFT is showing residual signal left after taking the original file and subtracting the resampled file. It shows how the files differ. The area above 20 khz is because the 44.1 khz sample rate is limited to around 20 khz. While the original has bandwidth and signal content well above that.

 

If you convert to 44/16 and back to 192/24 this is what you get with shaped dither (the previous one used triangular dither). The noise floor rises a bit, but is still inaudible. I took the original file and subtracted the resampled file to see what was left over.

 

Click to see larger picture.

 

If you don't have any files for converting, Barry offers some excellent sounding free samples.

 

Soundkeeper Recordings Format Comparison

 

You could use those Barry converted, but they don't null out as well as using Sox. No reason you can't do all three.

 

You also can get some samples here.

 

High Resolution Music DOWNLOAD services .:. FLAC in free TEST BENCH

 

The lower sample rate files here also have a lower level. I would get the highest and resample it yourself.

[gmgraves]

I dunno, man. That sounds like strawman to me. The M&M experiment, and this current paper under discussion, is only about resampling higher bit-rate down to lower, or digital filtering, and has nothing to do with speakers, amps, cables, etc.

 

Reread what I said, but first try to read the Myer/Moran paper. They took a SACD recording, converted it to analog, then re-quantified it as 44.1/16-bit, then decoded back to audio and then re-quantified it again, converted that back to audio and the repeated that process several more times. At the end of that daisy-chain they ABX'd the multiple iterations of the original SACD with the actual SACD playback and found that statistically, there was no difference. They are saying that all of those multiple series conversions from analog to to digital and back again created no audible degradation in SQ! And here we are arguing about the differences made to a digital audio stream by different USB cables! Does that sound logical, or even rational to you? If you think this is a straw man argument, then obviously the proposition does soeund logical and rational to you. That being the case, the high-end audio must be a fraud. Right?

[gmgraves]

Actionable advice - Try listening !!

 

 

And while listening, keep in mind that the mastering of a recording is often more important to the ultimate SQ than is the format. We all have (I'm sure), Red Book CDs that actually sound better than hi-res versions (SACD, 24/96 or 192 LPCM downloads or DVD-As or Blu-Ray discs) of the same material.

[esldude]

Reread what I said, but first try to read the Myer/Moran paper. They took a SACD recording, converted it to analog, then re-quantified it as 44.1/16-bit, then decoded back to audio and then re-quantified it again, converted that back to audio and the repeated that process several more times. At the end of that daisy-chain they ABX'd the multiple iterations of the original SACD with the actual SACD playback and found that statistically, there was no difference. They are saying that all of those multiple series conversions from analog to to digital and back again created no audible degradation in SQ! And here we are arguing about the differences made to a digital audio stream by different USB cables! Does that sound logical, or even rational to you? If you think this is a straw man argument, then obviously the proposition does soeund logical and rational to you. That being the case, the high-end audio must be a fraud. Right?

 

Okay so it is a strawman argument. You have just confirmed it.

 

http://www.drewdaniels.com/audible.pdf

 

You talk of all these multiple conversions when that is not what happened. There is a description and block diagram in the paper in the link above. A "professional CD recorder" took the analog output of the SACD/DVDA player doing an AD conversion to 44/16. It then did the DA conversion outputting the result . Level matching was done on the analog signal, and that was it. No multiple conversions you describe. Though often described as an AD/DA loop it actually was a once through AD/DA pass through.

 

As for discussing the sound of USB cables, no it doesn't sound logical or rational.

[mmerrill99]

And while listening, keep in mind that the mastering of a recording is often more important to the ultimate SQ than is the format. We all have (I'm sure), Red Book CDs that actually sound better than hi-res versions (SACD, 24/96 or 192 LPCM downloads or DVD-As or Blu-Ray discs) of the same material.

Thanks but I think most people know that & why I suggested downsampling in a later post. The point of my post was that some people want to be told what to hear & some people listen & make conclusions on what they hear. Yes, the mastering is a huge determinant of what is heard but apart from that there is usually more realism, more sense of the acoustic space, more subtlety in higher-res material (of course it has to have been recorded in the first place)

[Jud]

Dennis, re Audiophile Inventory: Yes, it preserves level. And in fact it can be used to establish equal levels for files that don't start out that way, such as SACD and PCM from the same master, through an adjustable loudness parameter for the converted file.

[Miska]

Okay so it is a strawman argument. You have just confirmed it.

 

http://www.drewdaniels.com/audible.pdf

 

You talk of all these multiple conversions when that is not what happened. There is a description and block diagram in the paper in the link above. A "professional CD recorder" took the analog output of the SACD/DVDA player doing an AD conversion to 44/16. It then did the DA conversion outputting the result . Level matching was done on the analog signal, and that was it. No multiple conversions you describe. Though often described as an AD/DA loop it actually was a once through AD/DA pass through.

 

Funny part in this test was that a multi-format DVD-A/SACD player was used, based on Mediatek chipset doing (known bad) conversion to PCM for all SACD sources...

 

So the comparison tells mostly that A/D/A conversion to RedBook PCM didn't sound worse than original conversion to RedBook PCM. In this case.

[mmerrill99]

Funny part in this test was that a multi-format DVD-A/SACD player was used, based on Mediatek chipset doing (known bad) conversion to PCM for all SACD sources...

 

So the comparison tells mostly that A/D/A conversion to RedBook PCM didn't sound worse than original conversion to RedBook PCM. In this case.

Yes, one has to be careful that "no difference" doesn't mean both RB & High-res are both being veiled by some process such that they both now sound the same. The opposite also applies - one has to be careful that differences heard doesn't mean that a flawed device is revealing it's own shortcomings.

Tricky stuff, this audio hobby!

[beanbag]

Here is what is left after converting one of Barry Diament's 192 files to 44/24 and back to 192/24. The FFT is showing residual signal left after taking the original file and subtracting the resampled file. It shows how the files differ. The area above 20 khz is because the 44.1 khz sample rate is limited to around 20 khz. While the original has bandwidth and signal content well above that.

[ATTACH=CONFIG]15880[/ATTACH]

 

Did you brickwall filter this or something?

 

You also can get some samples here.

 

High Resolution Music DOWNLOAD services .:. FLAC in free TEST BENCH

 

The lower sample rate files here also have a lower level. I would get the highest and resample it yourself.

 

Oh, how sneaky of them.