theorist

You really are gunning for me, aren't for you! Just do me a favor and give me some notice before you start distributing torches and pitchforks to the locals. More seriously, can we please tone it down? I've not previously gone after you personally, so why are you lobbing personal insults at me (and this isn't your first)? Why all the schoolyard behavior? As for this particular case, you caught me mid-edit. I initially wrote "...valid objectivist points..," but decided there was too much presumption there, so I softened it to useful, and then for the final edit I wrote: "(because I raise what I hope are useful objectivist points to the subjectivists, and hopefully useful subjectivist points to the objectivists)." And, yes, as you can see from the time-stamp, these changes were made before your post went live. Finally, the reason I added "useful" was to make it clear I'm not doing this to tweak anyone -- I carefully choose points that I hope are useful. I suppose that makes me the opposite of you -- you clearly enjoy writing inflammatory things, likely for the purpose of getting a reaction. Say, there's a word for that....

Jud, thanks for generously taking the time to provide all that detail -- it certainly helps support the argument. In my defense, I looked carefully for somewhere in the paper where their assertion that digital processing modulates the power supply was supported by a reference, and found none (did a word search for "power"). In particular, when they wrote: "And these current waveforms are also going back through the computer PSU, polluting the mains power supply," what they should have written instead was: "And these current waveforms are also going back through the computer PSU, polluting the mains power supply.[ref]" Yes, as you showed, one of the references they supplied does support that, but they didn't attach that reference to the statement in question to show it in fact was supported, which is SOP for a technical paper. If I were a reviewer I'd tell them they needed to address that . But anyways, thanks again for your post. Interestingly, asynchronous USB is supposed to address all this by decoupling the timing of what the DAC sees from that of the computer output, by putting it into a FIFO buffer and then using the DAC's own clock to pull the bits out of the buffer for decoding. Yet some have argued that it doesn't fully eliminate the effect of input jitter, since the timing of the incoming signal modulates the power supply of the DAC, thus modulating the clock the DAC uses to pull the bits out of the buffer (or something like that -- I might not have the argument exactly right). So lots of possible subtleties here. More broadly, and FWIW (since you mentioned my skepticism), I should note that I like to straddle the ground between the the objectivists (those that favor testing via measurement) and the subjectivists (those that favor listening). As a consequence (because I raise what I hope are useful objectivist points to the subjectivists, and hopefully useful subjectivist points to the objectivists), the objectivists can mistake me for die-hard subjectivist, and the subjectivists can mistake me for a die-hard objectivist, where what I'm really trying to do is pull both groups toward the middle. [From my limited postings on HA, for instance, they've probably concluded I'm a "fluffy audiophile" .] Personally, I think the only way to improve audio is if the two camps can talk to either other in a civil, collaborative fashion (really, I don't think there should be two camps). I think the subjectivists tend not to realize how easy it is to think one is hearing a difference when none is there (and hence the value of blind tests). And I think the objectivists tend not to realize that psychoacoustics is (like medicine and nutrition) not a yet science (these are what Thomas Kuhn, the great 20th century philosopher of science, would describe as “pre-sciences”), and that our understanding of the relationship between measurement and hearing is thus not on the same solid foundation as that of core results in the sciences (hence the value of listening). For instance, right now I'm enjoying this brief AES paper, which combines analysis and listening to try to both raise and address some of the limitations of LPCM encoding: https://dl.dropboxusercontent.com/u/59174624/AES_131_e-brief_BAILEY_Playback%20Disappointment.pdf

Ah, so is this your view?: 'Bit-perfect software players would sound identical. However, in practice, the software players do sound different, and the reason is because they don't quite reach the bit-perfect state. Further, to the extent they deviate from this state, it's not something that can be captured as jitter.' If so, could you please explain more what this deviation from bit-perfection is? My coarse-grained view is that there's only two types of errors a player could make: getting the values wrong, or getting the timing wrong ....

Thanks for explaining the limits of diff. Another CA member just kindly PM'd me several links. This one, in particular, appears to indicate that there are no significant differences in the jitter spectra among the various bit-perfect players: Archimago's Musings: MEASUREMENTS: Part I: Bit-Perfect Audiophile Music Players (Windows). Is that essentially what the measurements you recall reading said? Now that I have a sense of what folks are saying technically, my next step is to try listening for myself.

Sorry, I didn't know the etiquette here required an introduction. For all the other sites on which I participate, one simply starts posting and people get to know you through your posts. It's typically only after you've been on a site for a while (say, post count >100) that you start to fill in the personal section, since it's only after you've been on a site for a while that people care about who you are. Of course, my experience on CA has been, ah, unique. So: I've been an audiophile for decades -- I bought my first serious stereo system in college, because I needed something on which to listen to the records for the Music 1 course I took. In spite of being a science major, I also published a paper on music composition in a philosophy journal (it's actually been cited -- once ). Subsequently, I worked part-time in an high-end audio store, and then started my own pro audio business, where I attempted to sell audiophile-grade equipment into the pro market. This was also part-time; as my "day job" (and nights and weekends...) has been science. I also collaborated with a well-known high-end speaker manufacturer on the design of a new type of loudspeaker driver using a beryllium diaphragm (for a diaphragm you want the highest possible ratio of Young's modulus to density^3, and beryllium is an amazing material because it combines a high Young's modulus with an extraordinarily low density -- and note density enters as density^3). As part of this business I carried one of the premier high end D/A converters (this was in the 90s). But as good as that was relative to other products, I found it not even close to analog. Given the passage of nearly two decades, I've decided to take another stab at digital audio in the hope that things have sufficiently improved in musicality.

Thank you for the sincere apology. It's appreciated. So based on what both you and Paul said, I gather there is no clear technical answer to why they could sound different -- it's a bit of a mystery. That helps my understanding. Before, it wasn't clear to me if there was a straightforward answer out there or not, so thanks. But (and please tell me if I'm missing something here): 1) Isn't it technically straightforward to test if the outputs of different players are bit-perfect ( just load them into a file and use the diff command in Linux, say)? 2) Likewise, isn't also technically straightforward to do comparative jitter and noise measurements on the outputs of the various software player outputs (like those John Atkinson does in Stereophile, say)? Given how much uncertainty there seems to be about this topic, would it not clarify things enormously for someone that is technically-inclined to do those measurements? I.e., then we wouldn't need to speculate about possible differences in bits or jitter, we'd know. Of course, the measurements would need to be done using many different computer configurations, etc., but that's a technical detail. And if they haven't been done, any idea why not? I'd do them but, hey, I'm a theorist. We need an experimentalist for this .

Many have said there is a difference in players so, contrary to what you've assumed, my starting position was there well could be a difference. Given this, though, I was puzzled how players could sound different. I really don't see what's wrong with then seeking a clear technical explanation of how such a difference could exist. I'm a scientist, so of course I want to find answers! Sheesh! So yes, your assumption is incorrect. And no, I don't need to approach this issue in the serial fashion you state -- I am free to explore both topics (the technical explanation, and my own personal listening tests) in parallel, or in whatever order I find convenient. To answer sandyk's question, I've got an old but good-quality system (a pair of PSE Studio V monoblock amps feeding a pair of reference monitors custom-built by Bill Dudleston of Legacy Audio for studio use; each features a pair of Focal kevlar mid-bass drivers in a d'Appolito configuration around a Vifa 1" aluminum dome tweeter). Cables are Audioquest Lapis and Clear. Alas, while I have a reasonably high-end analog front end (Sota Vacuum Star turntable feeding a Mod Squad Phono Drive), my digital front end is decidedly mid-fi (an Onkyo SR-805 home theater receiver, which uses 5 x TI PCM1796 DACs). Friends have offered to lend me, respectively, a Blue Note Stibbert Improved CD player and an Ayre QB-9, so I've been holding off doing the limited-period software trials until I get my hands on those (presumably more resolving) pieces. In the fall I'd like to demo the Benchmark DAC2, exaSound e20, and Tentlabs b-DAC+. And of course I'm waiting to see what Neil Young does with PONO. But I digress. You remarked that "There are numerous - very numerous - discussions on exactly this subject on this system." That, plus your extensive post count, suggest to me you are familiar with such discussions, so perhaps you can make up for your false assumption about me by helping me out: I've searched through Computer Audiophile to try to find a link to comparative measurements of jitter output by the various software players, but met with no success. Do you know if such measurements are posted anywhere?

Hunh? My question was clearly "why": "I.e., why should there be any difference in the digital signal output from my computer because I'm using a different software player? "

"Noise" was a reference to a set of earlier inappropriate posts that another had made that you may not have seen, as they were removed by Mr. Connaker. "theorist" is simply an abbreviated description of my profession. Most scientists are experimentalists, but a few are theoreticians (we work with pen, paper, computer, and a very large wastebasket). I'm the latter. I don't see why so many find my original post so threatening. Really, it was an innocent question.

Hi Jud. Thanks for your reply. I think you may be misunderstanding me. First, my reply to bibo01 was provisional -- I thought it more courteous to give a quick reply rather than wait until I waded through the long thread he linked. More broadly, you mentioned that Fig. 1 of the Audirvana white paper "simplifies things about as much as they can be." I should therefore clarify that the issue for me is not that the paper wasn't sufficiently simple, it was that it wasn't sufficiently robust to be convincing. [Though Fig. 1 did make me feel pleasantly nostalgic -- it reminds me of a similar figure I created for a white paper I wrote as a consultant for a digital audio company in the mid-90s, explaining why jitter is an issue and how that company addressed it.] The problem with the Audirvana paper is it tries to give the impression it's addressed the question of player effect on sound, when it hasn't addressed it at all. There's quite a bit of sleight-of-hand here. For instance, in discussing Fig. 1, they write a "slight change in the reference voltage of the source will lead to a slight temporal shift in the value change detection." OK, no argument there. But then in section 4 they write: "Furthermore, as we have seen in section 1, the computer load (and its variations) has an impact on sound quality. Well, no, they haven't shown this -- they just want you to think they have! I.e., the legitimate argument is: A) changes in player -> changes in computer load B) changes in computer load -> changes in reference voltage C) changes in reference voltage -> increase in jitter The only thing they've argued convincingly is "C." They've also made a reasonable plausibility argument for "A." But nowhere in the paper do they show that whatever changes in load are effected by a change in player are sufficient to change the computer's reference voltage so as to significantly increase noise and jitter! I.e. they quietly elide over "B," and hope you don't notice. I'm not saying "B" is incorrect. I'm saying there's no way for me tell whether it's correct or not from this paper. To quote theoretical physicist Wolfgang Pauli, this paper is "not even wrong." [ See http://en.wikipedia.org/wiki/Not_even_wrong] Of course there's an obvious way for them to resolve this question (whether players affect noise and jitter): just have an independent laboratory do the darn measurement! [The next step is to show that any difference in noise and jitter is audible, but I'm just trying to get to the first part here.] So why on earth haven't they done this? It's this measurement that I'd really like to see, preferably from an independent source. I spent some time Google searching and wasn't able to find any such measurement but, given the obviousness of doing this, I suspect it's out there (say from some hobbyist that has the right equipment) (maybe a link to such is buried somewhere in the thread the Musicophile linked). If you happen to know of anyone that's done such a measurement, I'd love to see it.

Sorry, Musicophile -- with all the noise on this thread I forgot to thank you for that link!

Thanks, bibo01. It's a long thread, so it will take me a while to get through it. But upon initial perusal it sounds like they are talking about the effect of noise introduced onto the circuit board, and my initial thinking is that I can't see how that would be affected by the choice of player.

Hi souptin, thanks for your reply. Actually the "bit perfect" issue is exactly where I want to go -- that's the essence of what I was asking: shouldn't all the players be bit-perfect? From your reply, I gather they are, which answers part of my question. Do you have a link to someone that has done this test and shown them to be bit-perfect? Likewise, if the issue is varying levels of jitter/noise, that's easy to measure. Has anyone done this study?

Hi, first post. Many on this forum report sonic differences among iTunes, Amarra, Pure Music, Audirvana, etc., even when none of the added DSP features (volume control, EQ, upsampling, etc.) are being used. Can someone please provide a technical explanation of how this is possible? I.e., why should there be any difference in the digital signal output from my computer because I'm using a different software player? Or to put it another way: If the players are all designed correctly, and you're not asking the players to do any added DSP (beyond that needed to output the signal, which should be standardized), then shouldn't their outputs be bit-for-bit identical? And likewise, shouldn't all these outputs in turn be bit-to-bit identical to the digital output from a CD player used as a transport (assuming the same track, of course)? [i am assuming that, for the same track, the digital outputs of different CD players will output the same sequence of values -- i.e., that the process is standardized. Is this not correct? Then if the software players are all properly designed, shouldn't they also be outputting the same sequence of values as the CD players?] The only difference I can imagine is in the level of jitter, but I don't know if the players would affect that (and, besides, if your DAC is a serious implementation, with proper jitter rejection, modest differences in source jitter should not be an issue). I should add I've not yet myself demoed any software players other than iTunes, though I will soon. I'm simply curious how it is even possible that they could sound different. Thanks!