1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

Some questions about settings (sample rate, latency, etc.) for an external USB DAC running off of ASIO

Discussion in 'Sound Science' started by goodyfresh, Jul 30, 2015.
1 2 3 4 5 6 7
9 10
  1. KeithEmo
    OK - direct answer.....
    If you want to prove that there was an audible difference - then one test subject who can reliably and consistently hear a difference under one test condition will do the job nicely.
    However, if you want to prove the negative, then the only real way to do it is to test everyone on Earth, with every possible combination of equipment and test files, and show that nobody is ever able to reliably detect a difference. Since that is rather impractical, then, failing that, you need to make a "fair try" at covering all reasonable permutations. (Notice how the term "reasonable" again places us back in the realm of opinion - reasonable according to whom?)
    To put it bluntly:
    1) How do we confirm that Foobar2000 isn't somehow making the results inaudible?
    2) How do we know that they didn't simply make a poor choice of sample material?
    3) Assuming that the samples at both sample rates are created from some original, by a conversion process, how do we confirm that the conversion process used didn't produce artifacts that mask the other presumed difference from being audible.
    As I noted, my proposed test protocol is also limited..... after all, even if 10,000 audiophiles fail to hear a difference, we haven't proven that audiophile  #10,001 won't be able to do so. The whole problem in the end is that "audibility" is self-referential. If I wanted to determine, once and for all, if a human being can identify a 0.1 degree difference in temperature, I can take a whole bunch of metal plates with heaters and calibrated thermometers attached to them, set them to precise temperatures, and see if my test subjects can tell the difference. And, if nobody can tell the difference, I can state with certainty that "xxx participants were unable to notice a 0.1 degree difference". The point there is that we do in fact have current technology that will allow us to "create" specific temperatures to much greater accuracy than 0.1 degrees.
    However, we do NOT have ANY combination of equipment that we can use to generate audio test signals where we can state with absolute certainty that the test signals are rendered entirely accurately. We can play your test tones on DAC x but, if nobody hears any difference, we can't know whether the difference is actually inaudible, or whether DAC x is reproducing it imperfectly, and so MAKING a difference that really exists inaudible. (If you wanted to test whether 1 mS of ringing on a DAC was audible, you would first have to find a speaker with less than 1 mS of ringing. If you can't do so, THEN YOU CAN'T RUN A VALID TEST.)
    We're asking our test subject to listen to those files on a "test system" which includes, as you pointed out, speakers or headphones with significant distortion. We also know that there are in fact differences in the files (we're trying to determine if those differences are audible or not). However, we DON'T know whether the specific amount and type of distortion caused by our speakers will mask the differences in those files or not - so we in fact don't know if we've got a valid test situation or not. (We would need to measure the actual sound - in the air - and show that the differences "made it through" the speakers and electronics in our test setup and to the listener's ears.) Even if we succeed in proving that, and find that no difference is heard, we haven't proven that it wouldn't be audible with a different set of speakers, with different distortion characteristics, which may or may not mask the same errors the same way - or with some new and as-yet-undiscovered type of DAC.
    Just so we're perfectly clear here.... I am NOT personally claiming that the differences between different sample rates (assuming that no other differences are introduced by the conversion process itself) are in fact audible. Or that, even if they are audible, they don't rank far down the list of things likely to make a bigger difference.
    What they've done (at the link you've provided) is to provide a simple method by which individuals can confirm whether they find the differences to be noticeable or not - using specific test samples and a specific player. For the average listener, trying to decide "whether it's worth it to them to buy high-res files or not", this will provide useful information. (It's sort of like claiming that "aspirin is effective and totally safe". As a marketing claim, that's reasonable, and it might even pass muster as a claim of liability. However, the reality is that aspirin is not effective for 100.0% of the people who try it, and a very few people have actually died from allergic reactions to it - so, to be scientifically correct, it is neither 100% effective nor 100% safe. That doesn't mean that it isn't "close enough" to be an excellent choice if you have a headache.)
    [Just to be annoying, I'm going to suggest a test which I'm pretty sure will ruin some preconceived notions about how much distortion is audible.... Start with a perfectly clean 50 Hz sine wave. Now generate a pure 500 Hz tone (the tenth harmonic of 50 Hz) and a pure 2 kHz tone (the fortieth harmonic of 50 Hz). Now create a test tone consisting of an alternating pattern of the 500 Hz and 2 kHz tones - alternating between them at 0.1 second intervals. This will give us something that sounds somewhat like a typical ambulance siren - beep-boop-beep-boop. Now reduce the level of our beep-boop tone by a factor of 200x. Now mix that reduced tone with our original 50 Hz tone. We will now have a 50 hz tone with 0.5% THD (consisting of my specifically chosen "annoying harmonics"). My guess is that, if you play that 50 hz tone reasonably loudly, the 0.5% beep-boop tone, which is well above the noise floor, will be clearly audible. Which will then prove that, AT LEAST UNDER THOSE SPECIFIC CONDITIONS, 0.5% THD is in fact audible - which will disprove the generalized claim that "0.5% THD is inaudible". ]
  2. Music Alchemist
    Why all this fuss over something that has already been demonstrated by the laws of physics to be impossible? You cannot hear well over 20 kHz. You cannot hear added dynamic range that isn't there. Case closed.
  3. nick_charles Contributor
  4. KeithEmo
    Awwwww. Marketing is a tricky subject. I can easily imagine a fancy watch company, or one that sells really expensive wine, using a slogan like:
    "Only one in a million people is discerning enough to recognize how much better our product really is; are you one of them?"
    (In fact, in audiophile advertisements, variations on that theme are somewhat common... and work rather well. After all, almost everybody wants to think he or she is one of the "discerning few".)
    I was specifically responding to the very specific claim that the difference was "inaudible". If you want to talk about how MANY people can tell the difference, or how significant that difference is, even to those who do claim to hear it, then that is a separate discussion. I would certainly go as far as to agree that the differences between different speakers, headphones, rooms, and source material are all typically much greater than the differences related strictly to sample rate.
    As I mentioned before, I'm also willing to consider the possibility that different versions of the "same" content offered at different sample rates may well sound slightly different for any number of other reasons - ranging from unintended conversion artifacts to deliberate modification of one or the other version to fit the expectations of their different target audiences. 
  5. arnyk
    I see no direct answer with any substantiation, instead I see  a list of speculative deflections.
    You have a direct challenge - show us that you can hear what you claim.  Costs nothing but a little time. 
    Truth or Dare? :wink:
  6. goodyfresh
    My thread has gone out of my own control, like Frankenstein's Monster!  WHAT HAS SCIENCE WROUGHT!?!?!?!?!?!?!?!?!
  7. arnyk
    The above seems to reflect a very strong bias against the possibility that any difference might be inaudible.
     Talking about how MANY people can tell the difference is a distraction from the true fact which is that most differences (actually almost all of them) are inaudible to everybody.
  8. castleofargh Contributor
    my own experience could never confirm what Keith is saying about audibility. maybe my gears suck, maybe I don't have enough experience and don't know what to look for, and most likely, my headphones aren't good enough(are any headphones good enough for those levels?) to resolve what we're talking about here. so I keep the idea alive that maybe if I had some real cool headphone or speaker, and knew where to look in the music, I might somehow be able to notice "something". but TBH that's mostly my skepticism telling me not to stay too comfy around my own assertions as an ethic principle. nothing more.
    now I really enjoyed reading those pages, Arny trying to be the voice of objectivism, following what tests and statistics are telling us.
    Keith has enough knowledge to challenge some general conclusions without the sad ignorance we usually read on that side of the argument. and I find that he is very careful not to make any major claims, always warning about the magnitude of possible differences and their actual size relatively to the rest of the audio chain. so while I don't share the opinion, I really enjoy reading the rational behind the point of view.
    and Nick"Powell"Charles always comes in with more facts than opinion to my delight.
    there might very well be 5 people on the entire website to find those posts interesting, but know that I'm one of them. thanks for the effort guys.
  9. goodyfresh

    I find them interesting too!  But I still feel like the thread I started has grown into a monster out of my control :p
  10. KeithEmo
    Not at all. The only bias that I will (cheerfully) admit to is that I'm extremely skeptical when someone claims that there is an audible difference, but every conceivable measurement indicates that there is no difference, and I'm somewhat more willing to expend effort to look into claims that a difference that is actually measurable might also be audible under even certain circumstances. In this particular case, every test I've heard claiming to "prove" that there is no audible difference between different sample rates is, to put it bluntly, so badly flawed as to be scientifically meaningless. I've never seen any well thought-out and conducted tests that prove that a difference IS audible either, so I'm perfectly willing to leave it at "inconclusive"....
    However, making claims about what "everybody" can or cannot hear, based on the results of tests conducted using only a few test subjects, using only a few samples of equipment and content, none of which has (in my opinion) been shown to be adequate to the task, seems about as dishonest as claiming that an audible difference HAS been conclusively proven.
    Personally, I can say with absolute certainty that I can hear obvious differences between copies of specific source material, purchased from specific vendors, which are purported to be "copies of the same content at different sample rates". And, since most modern equipment, including most of the DACs I own, can play any of them, I'm perfectly happy to buy the 24/192k version that sounds better than the 16/44k version without worrying about WHY it sounds better (maybe it's because of the higher sample rate; maybe it's because my particular DAC works better with that sample rate, and maybe the vendor deliberately sabotaged the lower-sample-rate version so I'd buy the other more expensive one). And, once I've acquired that 192k version, the amount of time it would take to convert it to 16/44k is worth more to me than the few cents worth of storage space doing so would save. (Therefore, why expend extra effort to make a copy that's "almost certainly audibly just as good as the original" when I have the actual original?)
    As for doing an AB test, or an ABX test, to see if I personally can hear a difference between different sample rates... I have only two problems with that. First is the fact that I wouldn't be at all interested in doing so with source material selected by someone else - for a variety of reasons. For one thing, not all source material is equal, so we would always have the question of whether the quality of the particular source material chosen was adequate to showing the difference we're trying to test. For another, I personally find myself to not be especially "discerning" with music which I am either unfamiliar with, or simply don't especially like. This is a relatively common human trait - you're much more likely to notice whether your best friend has a slight sore throat when talking to him or her on the phone than you are with a total stranger. Second is the fact that, since I'm over fifty years old, even if we were to determine that I was personally unable to tell one sample rate from another in a truly fair test, that would in no way suggest that nobody else can hear a difference either. (Statistically, according to those "widely accepted sources" everyone likes to quote, about half of the humans on the planet should have better hearing than I do - at least in terms of frequency response. So the fact that I can't hear something would in no way even suggest that nobody else can.....)
    To answer the original question (or "challenge") directly.....
    At some point in time I probably will try a proper ABX test to determine if I can in fact hear a difference between familiar music played at various sample rates - using my equipment, and my ears, in my room. If I do in fact do so I will post the results here. However, if it turns out that I CAN reliably tell the difference, then all that will prove is that EITHER the difference is inherently audible, OR that there are audible artifacts from the conversion process, OR that the particular DAC I'm using produces an audibly different analog output when playing files at different sample rates. And, if I CAN'T reliably hear a difference, then it will prove that I (specifically) can't hear a difference, using my selected sample files, and my equipment.
    It would also be interesting to extend the test to various conversion software (to find out if a file converted from 24/96k to 16/44k does or does not sound audibly different when the conversion is done by different programs). 
  11. Music Alchemist
    It's easy to hear a difference between hi-res and CD when there is a different master used. Once you convert the files yourself, those differences disappear.
    Here is what I recommend.
    Convert your own 24-bit files to lossless 16-bit / 44.1 kHz with dBpoweramp. It will do a perfect conversion, but you need to convert to either uncompressed WAV or AIFF. It won't convert to a different resolution when you use compressed lossless.
    Then do an ABX in foobar2000. You can use this guide if you're not sure how.
  12. RRod
    Another thing to try is ABXing Redbook versus even *lower* specs, and realizing how for many (most?) tracks even Redbook is delivering more than you can hear.
    Music Alchemist likes this.
  13. Music Alchemist
    Oh! And here's another idea that can be really fun:
    Convert to various lossy AAC bit rates, from 10 kbps to 256 kbps, and see how the sound goes from completely destroyed to pretty good to as good as lossless. [​IMG]
    (Technically, though, all those bit rates are still Red Book, since they're 16-bit / 44.1 kHz.)
  14. nick_charles Contributor
  15. KeithEmo
    I also noticed that the Foobar ABX test plugin allows the listener to switch back and forth between A, B, and X as many times as they like and in any order. This eliminates the main objection that I have to what I would call "formal ABX testing". (In what I would call formal ABX testing, the listener hears A, then B, then X - and is then asked to make a value judgement about which reference sample matches X, which requires some level of "acoustic memory" of both A and B. With the Foobar plugin, one may instead choose to go directly back and forth between A and X several times, and decide if they are identical or not, then go directly back and forth between B and X, and decide if they are identical or not. This would seem to me to require less acoustic memory, and so to be a more direct and sensitive indicator of "similarity".) 
1 2 3 4 5 6 7
9 10

Share This Page