My Lying Ears

Industry luminaries Toole and Olive disagree.

Floyd Toole and Sean Olive are engineers, not trained psychoacousticians. Bob Stuart has a PhD in neuroscience, and is specialized in psychoacoustics. In fact, the ONLY peer-reviewed AES Convention paper that has EVER won AES Best Peer-Reviewed Paper Award (i.e. the highest award an AES paper can receive) in the category of Perception comes from him so not exactly the kind of guy whose opinion I would easily trade for an engineer's in this particular case: TAS 194: Meridian Audio's Bob Stuart Talks with Robert Harley | The Absolute Sound

They are selling their brand, their expertise, and some of it is questionable.

Things that experts say are often repeated as if they were facts. Some of it is questionable.

I think there are clearly times we should *not* trust our senses. For example, people will virtually always think louder music sounds better, with differences of as little as 1dB or perhaps even less. And unless you're someone extremely well practiced (producer, acoustician), you probably shouldn't rely solely on your ears to set up a room. But the vast majority of the testing I've seen leaves a lot to be desired as well, so I think it's not a great idea to take an extreme position either way.

Very well said.

P.S. Michael, some of Nisbett's work on people not being consciously aware of the reasons for their own perceptions looks very interesting.

Nesbit's book is a great read. I love the cultural studies too. It's funny how Americans tend to self-inflate their reports compared to other eastern cultures.

As a criminal defense lawyer, I'm naturally skeptical of what others say. But that book makes me question my self-observations. He has some interesting self-experiments he recommends. The statistics/research design part of the book is presented much better than I recall in undergrad psychology. It's very readable.

Nesbit's book is a great read. I love the cultural studies too. It's funny how Americans tend to self-inflate their reports compared to other eastern cultures.

 

As a criminal defense lawyer, I'm naturally skeptical of what others say. But that book makes me question my self-observations. He has some interesting self-experiments he recommends. The statistics/research design part of the book is presented much better than I recall in undergrad psychology. It's very readable.

Thanks for the book suggestion Michael! Looking forward to "cracking" it open when I get a chance.

Chris:

Thanks for this post. I think it's great to hear about the different perspectives on audio from folks with experience, especially all the DIY guys out there who have accumulated experience and wisdom in their pursuit for quality sound.

A *realistic* perspective on the limits of human hearing has to be considered and readers of reviews likewise have to calibrate their expectation from what any one of us can be honestly asked to provide. There is much a good subjective reviewer can say about the hardware, functionality, esthetic appeal, convenience, and overall sonic impression. Unrealistic expectations placed on subjective reviewers of sonic quality push reviewers to say things which are unlikely to be within the domain of human perception (eg. sonic differences between good cables, ablility to "hear" jitter in already excellent hardware even if instrumentation suggests no difference, claims of perceptible improvement even when nothing is changing in the digital domain and noise/timing changes unlikely). This of course feeds back to more objective readers who look at these comments with more and more frustration. Sarcastic accusations of "bits are bits" and "flat earth believers" become the unfortunate pithy statements in these expressions of frustration on so many discussions.

Thanks for taking on a more balanced perspective as evidenced by your openness in publishing this post... And thanks to the writer of this post (Michael) for his frank insights.

Confused am I too, and questing. As a disciple of jRiver I was nonplussed at the Sydney guys findings. They did not indicate whether it was night and day, or less. Jriver, admittedly developed for Windows, has a preferred user interface, which to some of us is of the essence, all things being near enough to equal or close enough.

Is a Mac better at handling the bits? It does seem that any differences (if any) would be down to the various different computation background processes.

Many suggestions about reducing Mac processes: was the Mac optimised for JR? Can the sound of a PC be improved by Fidelizer which closes down many background processes?. Seems feasible..

One expert here on ComputerAudiophilehas sweated to discover precisely identical waveforms from jRiver over PC and Mac, and that is it, amen and goodnight.

Other others swear that HQPlayer wins hands down soundwise. Most HQP-lovers use Macs; most also reckon a high-spec machine quadcore is better. Yet we’ve been told that any old PC can coast with its processor doing audio bitstreams. Go figure.

So I Fidelised. And I reckon it 'sharpened up' the sound. I think. I'm living with it on, then going backto try for a clearer opinion -- it takes time to get the feel of these subtle things.

As to 'If it measures identical...' well what about the AudioQuest Jitterbug, with a feasible explanation, which wise ears on Stereophile (and elsewhere) found positive, yet John Atkinson couldn't find any measurable differences. Hmmm.

Confused am I too, and questing. As a disciple of jRiver I was nonplussed at the Sydney guys findings. They did not indicate whether it was night and day, or less. Jriver, admittedly developed for Windows, has a preferred user interface, which to some of us is of the essence, all things being near enough to equal or close enough.

Is a Mac better at handling the bits? It does seem that any differences (if any) would be down to the various different computation background processes.

Many suggestions about reducing Mac processes: was the Mac optimised for JR? Can the sound of a PC be improved by Fidelizer which closes down many background processes?. Seems feasible..

The data bits sent to the DAC by different bit-perfect players are, by definition, identical. Any difference in sound quality must thus come from variations in jitter or noise carried over the USB cable. On the same hardware, USB jitter depends on timing in the host controller, and as long as its buffers never run dry, what the application software does should not matter, assuming no other USB devices are connected to the same host port. Moreover, decent DACs these days use asynchronous transfer, making them immune to jitter, and for those that do not, there are devices like the REGEN and Jitterbug.

Noise is a different matter. The CPU power draw fluctuates quickly depending on what code is running, and this can put noise on the power rails to be picked up by other components. All the hundreds of I/O lines criss-crossing the motherboard also spray electromagnetic noise everywhere. I once had a computer where the onboard "CD quality" sound card picked up so much noise that I could tell quite accurately what the CPU was doing just by listening. It is certainly possible (has been observed and measured, e.g. by Archimago) that USB noise can make its way through to the analogue side of a DAC. Now any noise carried by the USB cable can be removed by an optical isolation device (or an optical USB cable). If sonic differences between players are due to noise, they should go away if an isolation device is inserted in the chain.

One expert here on ComputerAudiophilehas sweated to discover precisely identical waveforms from jRiver over PC and Mac, and that is it, amen and goodnight.

Other others swear that HQPlayer wins hands down soundwise. Most HQP-lovers use Macs; most also reckon a high-spec machine quadcore is better. Yet we’ve been told that any old PC can coast with its processor doing audio bitstreams. Go figure.

HQPlayer can do some rather heavy DSP operations (in which case it is obviously not bit-perfect), and these benefit from a powerful CPU. Simply playing back an audio file as stored on disk requires less than 1% of a modern CPU.

So I Fidelised. And I reckon it 'sharpened up' the sound. I think. I'm living with it on, then going backto try for a clearer opinion -- it takes time to get the feel of these subtle things.

As to 'If it measures identical...' well what about the AudioQuest Jitterbug, with a feasible explanation, which wise ears on Stereophile (and elsewhere) found positive, yet John Atkinson couldn't find any measurable differences. Hmmm.

There's always a possibility that they didn't measure the right thing. I'm convinced that if there are audible differences, there are also measurable differences, provided one measures the relevant parameters.

If noise can be propagated to the DAC (either from the computer through the system, or from the work of the DAC's input circuitry), then this noise itself can create jitter in at least a couple of ways. The noise may operate directly on the clock circuitry; or it may produce variations in the ground level to which the signal is compared in the DAC chip when the momentary value of the bitstream is determined.

If noise can be propagated to the DAC (either from the computer through the system, or from the work of the DAC's input circuitry), then this noise itself can create jitter in at least a couple of ways. The noise may operate directly on the clock circuitry;

Messing with the clock is one way noise might propagate to the analogue side. Others mechanisms are possible as well, but the details are irrelevant. If the noise is removed with an optical isolator, it can't possibly do anything at all.

or it may produce variations in the ground level to which the signal is compared in the DAC chip when the momentary value of the bitstream is determined.

USB signals differential; they are not referenced to ground. Either way, the noise levels are clearly not so high as to interfere with correct recovery of the digital bitstream. If that were the problem, we'd be hearing very obvious pops and crackles, not the rather subtle differences people are experiencing. Moreover, verifying correct data transmission is trivial, and so far I've seen no reports of actual bit errors in the data arriving at the DAC.

USB signals differential; they are not referenced to ground. Either way, the noise levels are clearly not so high as to interfere with correct recovery of the digital bitstream. If that were the problem, we'd be hearing very obvious pops and crackles, not the rather subtle differences people are experiencing. Moreover, verifying correct data transmission is trivial, and so far I've seen no reports of actual bit errors in the data arriving at the DAC.

I wasn't speaking of the USB signal, but about the evaluation of the bitstream in the DAC chip as the bits are clocked out of the buffer. There's an ESS white paper or two floating around the Web discussing the importance of keeping ground noise as low as possible for this step.

Edit: Also, I'm not talking about gross errors to the extent of "bit flipping," but about enough noise to create jitter.

I wasn't speaking of the USB signal, but about the evaluation of the bitstream in the DAC chip as the bits are clocked out of the buffer. There's an ESS white paper or two floating around the Web discussing the importance of keeping ground noise as low as possible for this step.

 

Edit: Also, I'm not talking about gross errors to the extent of "bit flipping," but about enough noise to create jitter.

DAC chips (all the ones I've seen datasheets for, at least) have separate inputs for clock and data. A little noise on the data lines has no effect as long as the correct digital values are still read (no bit flips). Ground noise can of course show up as clock jitter, the extent depending on the rise/fall time of the clock signal. Again, though, the precisely how the noise interferes with the DAC operation is irrelevant to this discussion.

I believe we can agree that variations in sound between different bit-perfect players must be caused by something transmitted through the USB cable (assuming a USB-connected DAC). A USB cable carries data (obviously), an implicit clock signal, and (unwanted) noise. We can verify that the data arriving at the DAC is the same in all cases, so any sonic differences must be caused by clock jitter or noise, both of which can be entirely eliminated (e.g. using an optical cable and a REGEN). Therefore, I predict that a comparison of bit-perfect players in conjunction with noise/jitter suppression devices will yield no differences. If someone believes otherwise, I'd be interested in hearing their reasons.

DAC chips (all the ones I've seen datasheets for, at least) have separate inputs for clock and data. A little noise on the data lines has no effect as long as the correct digital values are still read (no bit flips). Ground noise can of course show up as clock jitter, the extent depending on the rise/fall time of the clock signal. Again, though, the precisely how the noise interferes with the DAC operation is irrelevant to this discussion.

 

I believe we can agree that variations in sound between different bit-perfect players must be caused by something transmitted through the USB cable (assuming a USB-connected DAC). A USB cable carries data (obviously), an implicit clock signal, and (unwanted) noise. We can verify that the data arriving at the DAC is the same in all cases, so any sonic differences must be caused by clock jitter or noise, both of which can be entirely eliminated (e.g. using an optical cable and a REGEN). Therefore, I predict that a comparison of bit-perfect players in conjunction with noise/jitter suppression devices will yield no differences. If someone believes otherwise, I'd be interested in hearing their reasons.

You're still being too simplistic in your model.

Regarding the relationship between noise on ground and jitter, if you have access to AES papers you should start in 1991 with Meitner and Gendron's "Time Distortions Within Digital Audio Equipment Due to Integrated Circuit Logic Induced Modulation Products," presented at the 91st AES Convention, New York, October 1991, Preprint 3105. If you don't have access, look at Figure 1 in this paper, which is from Meitner and Gendron (the rest is worth reading as well): http://www.amr-audio.co.uk/large_image/MAC%20OSX%20audio%20players%20&%20Integer%20Mode.pdf

One reason I said your model is too simplistic is because of the simple question of what we mean by "noise." What is desired at one point in the process and thus should not be gotten rid of may cause something undesired - noise - at another point in the process. You'll get a better idea of what this means by reading several posts by John Swenson in another forum. I mentioned them before, but now here are links, so there's no excuse! (I'd forgotten that not only Gordon Rankin but Charles Hansen of Ayre as well had chimed in on the thread to praise John's explanations.)

RE: It's not relevant - John Swenson - Computer Audio Asylum

RE: Thanks for the response! - John Swenson - Computer Audio Asylum

RE: I'm still not sure I understand - John Swenson - Computer Audio Asylum

These should prepare you to understand Alex Crespi's (Superdad) very good explanation regarding the function of the Regen, why this function does not "manifest" until the input of the DAC, and why there won't be any difference in the bits measured at the Regen's output:

http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/uptone-audio-regen-listening-impressions-24078/index38.html#post463111

Alex links to some nice preliminary testing by audio engineer John Westlake: MDAC First Listen (Part 00101001) - Page 46 - pink fish media

John Westlake's post in turn links to a couple of nice scope pictures:

Now you're right that this shows the Regen will reduce differences among bit perfect audio players. However, if you've been following the discussions so far in the posts and papers I've linked, you'll realize it won't completely eliminate those differences because the normal, correct operations of the programs themselves will present different modulations of the signal and (especially) ground received at the DAC chip, i.e., you'll get Meitner and Gendron's "logic induced modulation."

Finally: You've mentioned "optical" several times with regard to reducing noise. If you mean Corning's optical USB cable, it does not provide galvanic isolation, because the cable also contains electrical wires. If you mean opto-isolators, the operation of those by and large causes more noise in the DAC than is eliminated coming through from the computer. (The conversion from optical to electronic at the DAC end of the Corning cable may also be guilty of this.)

So the hypothesis is summarized like this:

Because we can see a "change" in the eye pattern (digital domain), we theorize there will be some kind of change in the analog outputs with a limited number of unknown USB DACs.

So far, the evidence is that there's no change in the analog outputs of a few well designed usb DACs (from Meridian and PSAudio), notwithstanding the many eye patterns posted.

What is the threshold for a 1? What is the threshold for a 0? Does the eye pattern, in either case, demonstrate that the USB receiver correctly sees the 1 and 0?

From a science standpoint, do you see a flaw with this theory? Isn't it obvious that this is a non-testable hypothesis? Or, at least, a hypothesis with zero evidence?

Jud,

I am all for experimentation. Enjoy it!! But don't post a bunch of scientistic scope shots and try to persuade others to believe a hypothesis which can never be falsified.

You're still being too simplistic in your model.

 

Regarding the relationship between noise on ground and jitter, if you have access to AES papers you should start in 1991 with Meitner and Gendron's "Time Distortions Within Digital Audio Equipment Due to Integrated Circuit Logic Induced Modulation Products," presented at the 91st AES Convention, New York, October 1991, Preprint 3105. If you don't have access, look at Figure 1 in this paper, which is from Meitner and Gendron (the rest is worth reading as well): http://www.amr-audio.co.uk/large_image/MAC%20OSX%20audio%20players%20&%20Integer%20Mode.pdf

 

One reason I said your model is too simplistic is because of the simple question of what we mean by "noise." What is desired at one point in the process and thus should not be gotten rid of may cause something undesired - noise - at another point in the process. You'll get a better idea of what this means by reading several posts by John Swenson in another forum. I mentioned them before, but now here are links, so there's no excuse! (I'd forgotten that not only Gordon Rankin but Charles Hansen of Ayre as well had chimed in on the thread to praise John's explanations.)

 

RE: It's not relevant - John Swenson - Computer Audio Asylum

 

RE: Thanks for the response! - John Swenson - Computer Audio Asylum

 

RE: I'm still not sure I understand - John Swenson - Computer Audio Asylum

 

These should prepare you to understand Alex Crespi's (Superdad) very good explanation regarding the function of the Regen, why this function does not "manifest" until the input of the DAC, and why there won't be any difference in the bits measured at the Regen's output:

 

http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/uptone-audio-regen-listening-impressions-24078/index38.html#post463111

 

Alex links to some nice preliminary testing by audio engineer John Westlake: MDAC First Listen (Part 00101001) - Page 46 - pink fish media

 

John Westlake's post in turn links to a couple of nice scope pictures:

 

[ATTACH=CONFIG]20802[/ATTACH]

 

[ATTACH=CONFIG]20803[/ATTACH]

 

Now you're right that this shows the Regen will reduce differences among bit perfect audio players. However, if you've been following the discussions so far in the posts and papers I've linked, you'll realize it won't completely eliminate those differences because the normal, correct operations of the programs themselves will present different modulations of the signal and (especially) ground received at the DAC chip, i.e., you'll get Meitner and Gendron's "logic induced modulation."

 

Finally: You've mentioned "optical" several times with regard to reducing noise. If you mean Corning's optical USB cable, it does not provide galvanic isolation, because the cable also contains electrical wires. If you mean opto-isolators, the operation of those by and large causes more noise in the DAC than is eliminated coming through from the computer. (The conversion from optical to electronic at the DAC end of the Corning cable may also be guilty of this.)

So the hypothesis is summarized like this:

 

Because we can see a "change" in the eye pattern (digital domain), we theorize there will be some kind of change in the analog outputs with a limited number of unknown USB DACs.

 

So far, the evidence is that there's no change in the analog outputs of a few well designed usb DACs (from Meridian and PSAudio), notwithstanding the many eye patterns posted.

 

What is the threshold for a 1? What is the threshold for a 0? Does the eye pattern, in either case, demonstrate that the USB receiver correctly sees the 1 and 0?

 

From a science standpoint, do you see a flaw with this theory? Isn't it obvious that this is a non-testable hypothesis? Or, at least, a hypothesis with zero evidence?

 

Jud,

 

I am all for experimentation. Enjoy it!! But don't post a bunch of scientistic scope shots and try to persuade others to believe a hypothesis which can never be falsified.

These eye pattern images are nothing more than an updated version of the classic stair-step graph "proving" the inferiority of digital audio.

... a limited number of unknown USB DACs.

 

So far, the evidence is that there's no change in the analog outputs of a few well designed usb DACs (from Meridian and PSAudio), notwithstanding the many eye patterns posted.

Well, I can see that you haven't been paying attention to the feedback in the CA Regen threads !

MANY eye patterns posted ?? Exactly two is many ?? That goes to how much I believe the rest of your posting(s)

These eye pattern images are nothing more than an updated version of the classic stair-step graph "proving" the inferiority of digital audio.

mansr, I cited an AES paper and referenced several posts by a good engineer. Please engage with the substance of these references and show me where I've misinterpreted what they are saying (for which I'd be honestly grateful - I always want to learn), or less likely, where they are wrong. Hurling cheap insults (equating eye pattern shots linked by Alex Crespi for illustrative purposes to the stair-step graph showing a fundamental lack of understanding of the Sampling Theorem) teaches me nothing and does you no credit.

So the hypothesis is summarized like this:

 

Because we can see a "change" in the eye pattern (digital domain), we theorize there will be some kind of change in the analog outputs with a limited number of unknown USB DACs.

 

So far, the evidence is that there's no change in the analog outputs of a few well designed usb DACs (from Meridian and PSAudio), notwithstanding the many eye patterns posted.

 

What is the threshold for a 1? What is the threshold for a 0? Does the eye pattern, in either case, demonstrate that the USB receiver correctly sees the 1 and 0?

 

From a science standpoint, do you see a flaw with this theory? Isn't it obvious that this is a non-testable hypothesis? Or, at least, a hypothesis with zero evidence?

 

Jud,

 

I am all for experimentation. Enjoy it!! But don't post a bunch of scientistic scope shots and try to persuade others to believe a hypothesis which can never be falsified.

Hi Michael. I may have misunderstood what you are saying, but if I'm reading you correctly, you're reading what I've said incorrectly.

If you have not done this already, and you have the time and inclination, I would like to suggest a little game. Treat me like an expert witness for the prosecution. Read and understand my references (excepting the Meitner and Gendron paper if you can't get access); know them better than I do (luckily, this should not be difficult ). Then come back and show me substantively where I've misapplied what they are saying to the Regen, or where I've otherwise got it wrong. I think you know me and my conduct on the forum well enough by now to know that if you do so, I'll happily say I was wrong and thank you for the correction.

Best article I've read here. Thank you for saying what audiophiles don't want to hear: We can't trust our ears.

Thank you for saying what audiophiles don't want to hear: We can't trust our ears.

Thank you for saying what audiophiles don't want to hear: We can't trust our ears.

I never trusted my ears but I trust others!!

Best article I've read here. Thank you for saying what audiophiles don't want to hear: We can't trust our ears.

Then you better sell your listening gear, buy an oscilloscope and stay looking at the screen when fed with music. But beware the eyes are easier to deceive than the ears.

I find the article is rubbish and should never have been published in this forum.

Roch

An interesting story, but I do not see how the bit perfect players relate to crossovers and EQ. In the article, the author got "lost in the weeds" and no longer had a good sonic reference point. Once things are askew, it IS hard to get back, but BP players do not do that. This could almost be read as an op ed for why JRiver lost the shootout.

+1

Roch

Funny, I mentioned on the CA forum the other day something that happened to me regarding one of these apps. I write for Macworld, and some years ago I was asked to look at one of these audiophile players for a potential review. I tried it out for a while, and found no difference between this app and iTunes, at least in sound quality. I contacted the company - who will remain nameless, but let's just say it's one of the Big Names in Audiophile Music Player Apps - and had a phone call with the lead developer. In about a 30 or 40 minute conversation, he was unable to tell me exactly how his app was supposed to make the music sound better. There were some things, such as loading files into RAM, hog mode, and the like, which have (IMHO) questionable value, but aside from that, he simply couldn't tell me why his app sounded better.

 

Needless to say, we passed on the review.

How strange that the developer does not know why his player makes the music sounds better!

The guidance does not rely, at the time of establishing the changes in his player, in the opinion of its many users? And I say many because you talk about "Big Names in Audiophile Music Player".

Regarding the "questionable value" of "loading files into RAM, hog mode, and the like" sounds to me the words expressed by JRiver in this forum

Roch

Then you better sell your listening gear, buy an oscilloscope and stay looking at the screen when fed with music. But beware the eyes are easier to deceive than the ears.

 

I find the article is rubbish and should never have been published in this forum.

With respect Roch... this forum is for a range of views from the pure "I hear what I hear therefore it exists" to "All DACs / Amplifiers sound the same because the measure the same to the known threshold of hearing".

To suggest the article shouldn't have been published in this forum is tantamount to censoring views you disagree with.

Eloise

Hi Michael. I may have misunderstood what you are saying, but if I'm reading you correctly, you're reading what I've said incorrectly.

Jud ... I think what Michael is saying, is that while the eye pattern images show that yes, the signal integrity to the USB input of the DAC is improved with the Regen, this is not able to prove that it will correlate to an improvement in the analogue output of the DAC (which is what is then amplified and turned into sound waves by the speakers).

For a very limited analogy... Its like measuring the temperature of the water you put into the kettle, and because the water is chilled to 2c rather than 15c out of the tap pronouncing that the cup of tea that is produced is going to be vastly superior.

Eloise