JPLAY Responds: An Open Letter

[amirm]

Back in mid 80's, there were people that said the same about jitter. Those who were arguing the most were ... software engineers.

Software engineers? Not sure there were that many software engineers then as there were hardly any universities offering pure software engineering degrees. To wit, even though I loved computer programming, my only choice was to take a few programming classes on top of the degree the university offered: Electrical Engineering. A that time you could get computer degrees but they were some lower quality degree program in "data processing" and such.

That "technicality" aside, no, the analogy is not the same. Said software engineers then had not modified the system, got it to sound better, but were not able to show an electrical difference. Or didn't even bother to measure. And proposed technical theories that violate how the system works in practice.

Now, if you mean the analogy in the highest level possible, yes, that holds. There may be some characteristics of audio performance we don't understand or know how to measure. The job is to then prove this using some objective mean that is not subject to clear placebo as was done with jitter. It shouldn't be too much to ask if such differences clearly exist as the availability of tools today is far better than it was in 1980s. I will even take the most extreme case as I suggested earlier in the thread: take the onboard low quality DAC on any motherboard and at least there, show some difference. Instead of making that effort, I am asked to go and do a bunch of work. What's wrong with proponents of the solution trying to do more than they have done?

Correct me if I'm wrong, but aren't you a software engineer as well ?

Consider yourself corrected per above . I grew up with electronics as my hobby so naturally I pursued that as my degree. I bought a computer in 1978 and so developed a love for programming as well. My career has been a mix of managing software-only group and software+hardware including all aspects of computers from VLSI to system design. My analog experience also allows me to straddle the fence between digital and analog. My career also allowed me to learn such related topics as psychoacoustics and knowing what to listen for by becoming a trained listener. It is through this combined lens that I make the comments that I do. The only pure software guy in this discussion ironically is the developer of Jplay!

 

[barrows]

Barrows, by your description of your configuration, you have just proven that there are probably very few systems that offer the isolation you are talking about. Contrary to what you stated/claimed initially "I am a little bewildered by jkenny's apparent lack of knowledge on how to properly isolate a USB interface though? I thought he was much more knowledgable about this?". You then go on to describe an XMOS based system which only partially isolates the DAc & now have described your system & the lengths you have gone to & are still striving for a better solution!!

I'm confused by this shifting of opinion - is it easy to isolate as you claimed initially or at least claimed my statement that it was difficult was erroneous

Hahaha! Nothing is as "easy" as it seems at first glance. But, I entirely disagree with you, but perhaps this a question of semantics. The USB interface does provide isolation from the computer. There is absolutely no need for the XMOS chip itself to be isolated, what is important is that the oscillators are isolated, and the I2S signal is re-clocked by a clean clock signal, on the clean side of the interface. As long as this is done, you have low jitter (at virtually the intrinsic level of the masterclock) and all the samples are there (if you were dropping any samples you would hear ticks, pops, or clicks). The facts have not changed, and neither has my "opinion" of them.
Now, reducing noise put back on the AC mains is a separate issue from USB isolation-this is a system problem, which exists in any audio system, even one just using an all analog approach. Audiophiles know about these problems with AC line noise already, it is nothing new.
My approach is to reduce the sources of noise as much as reasonably possibly. There is no evidence to suggest that jplay has any influence on computer noise reduction; in fact, a good theory has been put forth in the preceding posts which suggests jplay is likely to increase noise. Since these noise levels would be relatively trivial for someone with a financial motive to measure, I would like to see measurements like this from the developers of jplay…

In these systems we have only three things which go to the DAC conversion stage which matter: the samples, the jitter, and the noise. As long as we address these three things, then the DAC converter can do its job very well. There seems to be this kind of weird belief among some, that somehow the samples themselves are being "damaged" in some way, which results in subtle audio degradation. This just sounds like nonsense. If this were true, or possible, why can no one who believes these things present a single, plausible, mechanism for how this "degradation" might happen (much less measure their effects)? I can accept that noise coupling into the clean side of things could affect clock jitter, I can also accept that noise coupling into the output stage of the DAC could result in sonic degradation, other than that…

I have heard different things from different software approaches, I associate this differences with different noise spectrums. Additionally, usually the differences I hear are not clearly of the type which can prove one software approach is "better", just that there is (maybe) a difference; the kind of difference where one track might sound better, another might sound worse. The more dialed my system got, the less software seemed to matter, at which point I stopped playing with software, as my time was better spent choosing different resistors in my DAC (for example) which makes much more prominent improvements.

 

[jkeny]

Hahaha! Nothing is as "easy" as it seems at first glance. But, I entirely disagree with you, but perhaps this a question of semantics. The USB interface does provide isolation from the computer. There is absolutely no need for the XMOS chip itself to be isolated, what is important is that the oscillators are isolated, and the I2S signal is re-clocked by a clean clock signal, on the clean side of the interface. As long as this is done, you have low jitter (at virtually the intrinsic level of the masterclock) and all the samples are there (if you were dropping any samples you would hear ticks, pops, or clicks). The facts have not changed, and neither has my "opinion" of them.

You state this with conviction but that doesn't necessarily make it correct. I'm humble/stupid enough to say that I don't know all the intimate hardware details of how the different USB receivers work, what the dependencies are & how it effects the downstream reclockers, DACs, etc. I know this sort of exchange becomes a debate between what SHOULD work (in theory) Vs what my ears tell me doesn't work (in practice)

Now, reducing noise put back on the AC mains is a separate issue from USB isolation-this is a system problem, which exists in any audio system, even one just using an all analog approach. Audiophiles know about these problems with AC line noise already, it is nothing new.

All these issues are systems problems which need systems thinking - the sort that is usually lacking in these debates

My approach is to reduce the sources of noise as much as reasonably possibly. There is no evidence to suggest that jplay has any influence on computer noise reduction; in fact, a good theory has been put forth in the preceding posts which suggests jplay is likely to increase noise. Since these noise levels would be relatively trivial for someone with a financial motive to measure, I would like to see measurements like this from the developers of jplay…

Lots of assumptions here - theory of how it SHOULD work, ease of measurement, etc. Remember you are also arguing about playback software that you have never listened to!!

In these systems we have only three things which go to the DAC conversion stage which matter: the samples, the jitter, and the noise. As long as we address these three things, then the DAC converter can do its job very well. There seems to be this kind of weird belief among some, that somehow the samples themselves are being "damaged" in some way, which results in subtle audio degradation. This just sounds like nonsense. If this were true, or possible, why can no one who believes these things present a single, plausible, mechanism for how this "degradation" might happen (much less measure their effects)? I can accept that noise coupling into the clean side of things could affect clock jitter, I can also accept that noise coupling into the output stage of the DAC could result in sonic degradation, other than that…

I have heard different things from different software approaches, I associate this differences with different noise spectrums.

You best tell Amir your theory of how this happens because he is not of the same opinion i.e he claims that PC noise will swamp any possible noise improvement from playback software - all bit-perfect playback software sounds the same in his world view

Additionally, usually the differences I hear are not clearly of the type which can prove one software approach is "better", just that there is (maybe) a difference; the kind of difference where one track might sound better, another might sound worse. The more dialed my system got, the less software seemed to matter, at which point I stopped playing with software, as my time was better spent choosing different resistors in my DAC (for example) which makes much more prominent improvements.

Glad you found a solution which works for you. Others have found Jplay is a solution which gives them better sound - go figure?

 

[barrows]

"Glad you found a solution which works for you. Others have found Jplay is a solution which gives them better sound - go figure?"

No need to "go figure", many people often imagine "better sound". There are plenty of folks who have reported no difference between using using jplay and using jriver alone, are all these people wrong?
I do not use windows, so there is no option for me to try jplay, simple as that. I have already chosen my preferred playback software, and have moved on to more important aspects of getting my system closer to "perfection" (as I define perfection, after all my system is for my enjoyment).

Anyone who bases their evaluation of an audio system SOLELY on listening is making a lot of big mistakes. Audio systems are so flawed, that listening alone as a means of evaluation often leads to the selection of addition flawed components and approaches, to "balance out" the flaws already present. Many, many audiophiles add components (and tweaks, software, etc) to their systems to address shortcomings in their existing components. I prefer to address problems at their source, rather than to add additional problems to cover up the old problems. The only way to do this is to respect measured performance, proper technical design, and to carefully listen critically. Willy nilly adding of software like jplay, which even the developer themselves cannot even offer an explanation as to what could make it sound better, is akin to playing darts blindfolded: you might hit the target, but you are more likely to miss.

Additionally: To digital audio in general, there are some who seem to believe that digital audio is subject to analog like degradations. But, no one, to my knowledge, has ever, even by the wildest speculation, proposed a mechanism as to how these degradations occur. Sounds like voodoo to me. And, if these people were correct, then we are all screwed, as digital audio would be so "broken" as to be an impossible course for recording, storage, transfer, and playback.

 

[jkeny]

I now that theory is useful & I would love to have the theory of everything important in computer audio. I would also love to have the measurements to boot.
But lacking these does not mean that I should therefore deny what I hear. I know there are those who say that hearing perception is easily fooled & therefore I'm fooling myself but they haven't yet tried Jplay so again, go figure!

 

[Julf]

I now that theory is useful & I would love to have the theory of everything important in computer audio. I would also love to have the measurements to boot.

And in this case neither the theory nor the measurements are rocket science. So why don't you just make the measurements?

I know there are those who say that hearing perception is easily fooled & therefore I'm fooling myself but they haven't yet tried Jplay so again, go figure!

Anyone who is working with audio and doesn't take into account perceptual issues is probably ignoring the single most important issue. People who say "I know what I am hearing" are probably in the same class as alcoholics who say "I can stop whenever I want to". Go figure skating.

 

[Elberoth]

There is no evidence to suggest that jplay has any influence on computer noise reduction; in fact, a good theory has been put forth in the preceding posts which suggests jplay is likely to increase noise.

There is also no evidence that Jplay increases the noise. Those are all just theories.

As of today, I can only say that I can hear the difference, and the only logical reason I can think of, is that the Jplay somehow reduces/changes the noise level and/or spectra.

Jplay latest solution, which includes TWO computers, linked by ethernet cable where the computer connected to a DAC is just a renderer, with absolute minimum services running (chence lowering the number of services processes running even further) further sugests that it is the noise that might be the case.

 

[barrows]

Elberoth: I absolutely agree, that having the second computer, running a minimal platform, acting as just a renderer (hence very low power consumption and noise) is indeed a good approach. This approach has similar advantages to what I do, running a relatively minimalistic server in the first place in effort to reduce the noise footprint. Jussi with HQplayer also runs a similar, two computer, set up. Ethernet is also "isolated" by specification (in quotes per our previous remarks) which gives it an inherent advantage.
I consult a little bit with Jesus at Sonore, and the Sonore Rendu Ethernet Renderer is a very good computer audio approach for those who might like to simplify their systems even further, and have Ethernet wiring in place.

As to things being just theories, indeed, that is the truth, and is how I stated it. Until a promoter of jplay makes the relevant measurements all we have are theories, and it appears that jplay's developers are not likley to offer any measurements.

 

[barrows]

hmmm...

John:

"But lacking these does not mean that I should therefore deny what I hear"

I would suggest, that if there is no possible explanation, offered by anyone, for what you hear, then it may be wise to deny what you hear. Without even a a possible theory of how/why something might make a sonic difference, it would be prudent to suspect that the objective reality is that there is no difference.
I have a big problem with any product which claims to offer improvements in performance, where the developers can offer no possible explanation, no matter how speculative, as to the cause of the improvement. Why, well, imagine developing said product, how would the development proceed? Just randomly trying different approaches, and then subjectively evaluating them??? I am sure the problems of such an approach are obvious to anyone. Especially when we are talking about software, there must be millions of possibilities to try, by doing so with no theory as a basis for proceeding, the development process would be totally random and haphazard.

 

[Elberoth]

barrows - I agree. I wish all PC audio went the Ethernet route instead of USB, if only for the added benefit of increased isolation that Ethernet standard provides.

But if what you are saying about the AC ground contamination is true (and I do not have any reasons not to belive you), then we will still have the problem, unless we go with battery power on PC side and take it out of the grid.

 

[jkeny]

John:

"But lacking these does not mean that I should therefore deny what I hear"

I would suggest, that if there is no possible explanation, offered by anyone, for what you hear, then it may be wise to deny what you hear. Without even a a possible theory of how/why something might make a sonic difference, it would be prudent to suspect that the objective reality is that there is no difference.

If I was the only one who heard such improvements then I would agree with you. Did I test myself using an informal blind test - sure I did. If, I also hear many, many people all reporting agreement on the improvements then I take note. I know "mass hypnosis" & suggestibility can be used to explain this, sure!!

I have a big problem with any product which claims to offer improvements in performance, where the developers can offer no possible explanation, no matter how speculative, as to the cause of the improvement. Why, well, imagine developing said product, how would the development proceed? Just randomly trying different approaches, and then subjectively evaluating them??? I am sure the problems of such an approach are obvious to anyone. Especially when we are talking about software, there must be millions of possibilities to try, by doing so with no theory as a basis for proceeding, the development process would be totally random and haphazard.

I think you fall into the trap of assuming that the developers blindly just try "random" changes to see what affect it will have. This is a naive, black & white view of how such developments occur. People often follow their theories/ instincts in developing products, analysing the results with listening tests as they proceed. The end result may not easily succomb to measurements to show these audible differences but it therefore does not invalidate them. So I disagree with your hypothetical picture of the software being developed in a random way just because they have no measurements to show what they can hear.

 

[Vincent Kars]

I wish all PC audio went the Ethernet route instead of USB, if only for the added benefit of increased isolation that Ethernet standard provides.

I wonder if this really helps.
Ethernet is a rather complex interface.
To do so you need a small processor running Linux (or an Anroid stick) with e.g. MPD or DLNA.
Your next question is of course how to get the audio out of this small computer without the noise.
What about an async USB DAC with galvanic isolation?

 

[Elberoth]

What about an async USB DAC with galvanic isolation?

There are some DACs that have galvanic isolation o USB input, but that limits your max USB speed to USB 1.0 (24/96kHx max). And even then, galvanic isolation is not your silver bullet, as it is not 100% efective (contrary to what many ppl think).

 

[barrows]

barrows - I agree. I wish all PC audio went the Ethernet route instead of USB, if only for the added benefit of increased isolation that Ethernet standard provides.

But if what you are saying about the AC ground contamination is true (and I do not have any reasons not to belive you), then we will still have the problem, unless we go with battery power on PC side and take it out of the grid.

Well, as I have explained, it is not impossible to have a USB interface which is as well "isolated" as Ethernet, as these interfaces already exist. But I agree, that for some users, Ethernet is a good solution. Still, I find that audio over Ethernet has more glitches than via USB, and is subject to network variations which can be a hassle to deal with. Nevertheless, and Ethernet based set up can be made to work very well.
On noise on the mains from the server: this is not just a problem for grounds, it applies to hot and neutral as well. And, it applies, even to servers remote from the audio system, as the AC mains are all connected together at the panel (although high frequency noise does drop off with distance through the wiring). These are probably low level effects in the grand scheme of things, still, I really do not like all that noise on the mains-I am lucky to live in the mountains, with a transformer serving only my home. Additionally, I live alone, and so can keep my mains pretty clean from grunge as caused by other appliances, computers, TVs etc. It is not that difficult to power a low consumption server with batteries, and to set it up so that for non-critical listening it can be powered while charging as well.

 

[barrows]

Vincent: The Sonore Rendu runs via Ethernet on a purpose built (for audio, not a computer MoBo) board, using very low power. It runs on a 3.3V supply at about 500 mA or less, and has a much smaller noise footprint than any computer. So there are advantages. Ethernet may not be for everyone, after all, the audiophile will not be spending his time changing the firmware and "settings" on the Rendu in order to "improve" the sonic performance, it is already purpose built for the task. I am not familiar with the all the other Ethernet options available, but I suspect they offer similar advantages.

 

[Vincent Kars]

Hi Barrows

The Rendu is part of the rapid growing family of audio streamers.
I do expect streaming to become mainstream.
What sets it apart is that it is a renderer only.
Most of these boxes are also control point.
Reminds me of the squeezebox duet.

Personally I do think UPnP/DLNA more complex than direct playback. You have to deal with both server and renderer and often you don’t have much options to control what is send to the renderer. ?
http://www.thewelltemperedcomputer.com/HW/DAP.htm

 

[Julf]

Personally I do think UPnP/DLNA more complex than direct playback.

Indeed, and UPnP/DLNA has it's shortcomings. There are a bunch of newcomers (such as MagicPlay) that are trying to provide more open alternatives to AirPlay etc.

 

[barrows]

Indeed, and UPnP/DLNA has it's shortcomings. There are a bunch of newcomers (such as MagicPlay) that are trying to provide more open alternatives to AirPlay etc.

What are the "shortcomings" you are referring to? Just curious.

 

[Phelonious Ponk]

Is it possible that some audiophile company smaller than the secretarial pool at Harmon International's headquarters has discovered and managed to very effectively implement some as of yet undiscoverable, immeasurable and unverifiable quality of audio reproduction capable of raising quality up to the next level but incapable of being quantified and, therefore, verified in development? Sure. The odds of that are as substantial as a bit of vapor on the wind compared to the chance that, "testing" for a difference in sighted listening sessions, the hopeful developers and their loyal customers experienced expectation bias, but yeah, it's possible. The people who hear it will never believe that they're fooling themselves. The people who do not will believe in the substantive evidence that supports them. And that is the bottom line that will never be justified.

Tim

 

[barrows]

John: "If I was the only one who heard such improvements then I would agree with you. Did I test myself using an informal blind test - sure I did. If, I also hear many, many people all reporting agreement on the improvements then I take note. I know "mass hypnosis" & suggestibility can be used to explain this, sure!!"

The above explanation only works if you ignore the users of jplay who heard no difference, surely they are just as likely to be correct as the ones who do hear a difference. When there is no objective evidence of any difference, and there is no plausible explanation given of even the wildest speculation of how there could be a difference, then one should very well question their own observations if they hear a difference. BTW, I care not whether you are blinded, in fact I generally do not believe in blind testing, as it tends to test the observer rather than the DUT.

"I think you fall into the trap of assuming that the developers blindly just try "random" changes to see what affect it will have. This is a naive, black & white view of how such developments occur. People often follow their theories/ instincts in developing products, analysing the results with listening tests as they proceed. The end result may not easily succomb to measurements to show these audible differences but it therefore does not invalidate them. So I disagree with your hypothetical picture of the software being developed in a random way just because they have no measurements to show what they can hear."

I do not have a naive view at all, I consult in product development. The problem I have here, is that the jplay developers offer no explanation of how their SW could improve sound (assuming a working asynchronous interface environment). This is what I am responding to, please go back and read my post because you appear to have not gotten it. When they admit to not knowing how, or why their approach could improve sonic performance, they are also admitting to having no working theory to start with. If one has no working theory to start with, then development can only occur by random chance, trying things, and seeing what happens. Now, if noise reduction (airborne and USB cable borne RF) is all that jplay is affecting then it should be relatively trivial for the developers to pay some RF engineers to make the appropriate measurements