Untitled Document

Page 2 of 18 FirstFirst 123456789101112 ... LastLast
Results 11 to 20 of 175

Thread: Audible Jitter/amirm vs Ethan Winer

  1. #11
    Banned
    Join Date
    Jul 2010
    Location
    New Milford, CT
    Posts
    1,232

    Lightbulb

    Dammit Steve, now look how much work you made me do!

    Amir, I read both of your posts twice, but I'll hit only the high points in my reply. If you think I'm skipping or ducking something important let me know. And I think it's very cool that you were JJ's boss!

    Quote Originally Posted by amirm View Post
    I am assuming the debate is whether there is scientific research in the audibility of jitter not whether jitter is audible or not.
    Actually, I'm far more interested in knowing if jitter is ever audible in usual amounts. Knowing about research is fine, but as you pointed out a lot of research is suspect. Personally I don't trust anyone's research unless I was there in person! Well, almost, anyway.

    120 db? Even dedicated high-end DACs have trouble keeping their noise level below that which is roughly equiv. to 20 bits of resolution ... The specs published for a soundcard are likewise nonsense even if they rise up to the level you mention. A PC sound card doesnít do anything until you plug it into a PC. As soon as you do, it gets polluted by the noise in there.
    If anything, that proves my point that jitter is a non-issue since it will always be masked by other, louder sounds. Remember, my interest is whether jitter ever matters in practice. For example, it's easy to "prove" that dither reduces distortion when going from 24 bits to 16 bits, but in practice it's rarely / never audible. As further background, my motivation for caring about jitter is the silliness I read in magazines and web forums, where some "expert" tells a newbie the reason his system sounds poor is due to jitter so he should buy a new sound card. Jitter is never the reason someone's system sounds poor. At best, and I'm still not convinced, jitter might be audible in certain contrived situations when played back at near-pain volume levels.

    if I remember right, you used loud synthesized music which based on masking effect, does a nice job of hiding the noise.
    Great, you acknowledge that for pop music recorded at sensible levels jitter is not a problem. We're halfway there! But that was only one of the examples. As I mentioned in Post #5 above:

    one example (starting at 33:28) in the video linked above has gentle classical music with an average level of -30, peaking around -15 when it gets to the loudest spot. By the time the nasty noise is reduced to -70, it's then only 40 dB softer than the music at -30. Even cranking the volume way up it's difficult or impossible to hear the noise. And that's only a 40 dB difference!
    So even with gentle soft music, and a harsh nasty buzzing noise only 40 dB softer, it's difficult or impossible to pick out the noise.

    Folks, all anyone has to do to convince me that a typical amount of jitter is ever audible is to post a Wave file proving the point!

    That's all that's needed. Make an example like my AES video where jitter or a simulation is turned on an off, or however you'd like to present it, and let's hear it. Note I don't mean this at you Amir as much as at everyone who believes jitter is ever audible in practice. And I further mean with normal music or speech, not contrived test signals.

    I can pick music clips where when encoded at 64kbps, untrained listeners cannot tell them from the CD. Does that prove that 64kbps compress is lossless? Of course not.
    Great point, and I agree. But again, all I ask for is one example.

    I want to make clear once more that jitter is not a knob you turn and represent all cases of it. Jitter is not one number. Jitter has amplitude, frequency and distribution. The latter alone will have infinite range let alone all three combined. So even if you had the device you wish, you still couldn't simulate the situation at my home with my DAC and PC for example. Your device and experiment would still represent a theoretical simulation and not representative of all combinations in the field.
    Another good point, and again all I need is one example proving that jitter is ever audible in any form having any spectrum.

    Let me ask you this: which artifact do you think is more audible? The situation where the sidebands stand out like the first example or when it resembles noise? I suspect the answer is obvious and hence the reason we care more about non-random distortion.
    As soon as someone posts an example file I'll be able to tell you which is more audible. Though obviously the spectrum of the noise is a factor, with emphasis around 2 to 4 KHz being far more audible than noise having mostly bassy content. If one believes that correlated artifacts - which are like harmonic distortion - are audible, they need to see typical THD specs for loudspeaker drivers!

    Using my old Dell laptop, I can hear the CD ROM spin up as I access it through the headphone jack. Ditto for the hard disk turning on and off. On my sonís desktop PC, using a discreet sound card, the same thing happens.
    Depending on the sound card, you can often get rid of those noises by muting or turning down the volume of all unused inputs such as "TAD" and microphone in etc. The problem with some software mixers is they by default hide some volume controls. So you don't even see them and know to turn them off!

    We are not debating if someone can hear noise at levels below 100db. We are debating if the other side did a good job of proving jitter is inaudible below certain rate.
    I'm not a trained scientist so I'm not qualified to say if that test was valid or not. It convinced me! But more important, my own experiments confirm that anything that soft is simply not audible no matter the spectrum of the artifacts, or whether the artifacts are correlated or uncorrelated.

    It is not then my job to prove that the distortion is audible but yours and others to say it doesnít! If the math didnít exist, it would indeed be my job but that is not the case.
    I disagree. One basic rule of logic is you can't demand proof of a negative. We all know that green cheese exists, but it's not my job to prove the moon doesn't contain green cheese. So logically speaking, in this case, it's not my job to prove that jitter can't be audible either. All logic and common sense - and experiments about the audibility of soft artifacts - conclude that typical amounts of jitter can't possibly be audible. So the burden of proof is on those who say it can be audible. Again, all that side has to do is make an example Wave file proving their point! I've asked for this for years now, but so far nothing.

    --Ethan

  2. #12
    Banned
    Join Date
    Apr 2010
    Location
    Seattle, WA
    Posts
    16,044
    Thanks Ethan for a measured response . Hitting on a few simple points for now:

    1. The fact that PC music is not of high fidelity doesn't mean for the people in this forum, we get to dismiss low level distortion when majority of the people here spend lots of dollars to do a lot better. Take me. I have an $8K Mark Levinson DAC. The box is getting long in the tooth but that was the only way I could get the best fidelity (combined with a Pro Audio sound card). I agree with you that with the general public, they have no business worrying about things like jitter. CD quality in all forms recreated for them is plenty good. Again, that doesn't fly here were people want to know if they had all the money, could they do better.

    2. I can't give a wav file to hear jitter. Jitter is induced by the timing of the source. The Wave file is devoid of that timing. I could try to mess with the samples as the research paper did but then I still wouldn't know if your setup is transparent to it. Finally, I don't know if you are able to hear jitter. I would say even amongst the high-end audiophiles, only a fraction would be able to detect it.

    3. On the issue of hearing jitter, the best way to get there is to actually listen to compressed music and see if you can hear its artifacts! OK, sit back on the chair. Didn't mean to make you fall off of it . Reason for this is that compressed music is all about reducing the effective bit depth of music samples. We convert music to frequency domain and then selective represent each band at lower and lower quantization levels (resolution). When we do that too much, then you hear the artifacts. So see if you can spot artifacts in 320kbps AAC (not MP3 -- that codec is never transparent no matter what the bit rate). Keep listening until you can hear and spot where it fails. Then use the same clips to test your digital audio gear. No guarantees of course but you will at least be able to take a step in the right direction.

    4. On the last point I am puzzled. You say you don't hear jitter and that is proof enough for you that it must be inaudible. Isn't that proving a negative by itself? If you don't believe in proving a negative, then you shouldn't believe any study that says they couldn't hear jitter either!

    At the end, this is my view regarding jitter and digital audio reproduction. To the extent possible, you should try to get a system that is at least transparent to CD audio resolution. That is not an exceedingly high bar. It means having an honest, 16-bits of resolution without distortion and noise. And a frequency response to 22Khz. The cost of doing that is less than any high-end speaker these days. I can't assure you that your ears can hear the difference for sure. But you don't sit there wondering if you are missing something either.

  3. #13
    Banned
    Join Date
    Jul 2010
    Location
    New Milford, CT
    Posts
    1,232

    Lightbulb

    Quote Originally Posted by amirm View Post
    1. The fact that PC music is not of high fidelity doesn't mean for the people in this forum, we get to dismiss low level distortion when majority of the people here spend lots of dollars to do a lot better.
    I agree that just because some people don't have a good system - and probably less than 1 percent even have minimal room treatment - this doesn't mean we shouldn't strive for the highest fidelity possible. However, I disagree that all computer-based systems are lacking. My Dell computer has an M-Audio Delta 66 sound card, and it's definitely clean for both recording and playback down to the noise floor of 16 bits.

    I agree with you that with the general public, they have no business worrying about things like jitter.
    Yet look at how many people read that jitter is a problem, and then spend ten times more than needed to buy a converter that some ignorant reviewer claims sounds better than usual due to its low jitter.

    2. I can't give a wav file to hear jitter. Jitter is induced by the timing of the source. The Wave file is devoid of that timing. I could try to mess with the samples as the research paper did but then I still wouldn't know if your setup is transparent to it.
    The task isn't to worry about the listener's playback system, only to create a file with jitter-like artifacts at a level that's typical. This is the crux of it. If you create a pair of files and run a poll asking people to choose which file they think has the added jitter, the poll results will hold the answer. If out of 100 people over multiple trials nobody can reliably pick the file with the added artifacts, that proves to my satisfaction that jitter is not an audible problem.

    I don't know if you are able to hear jitter. I would say even amongst the high-end audiophiles, only a fraction would be able to detect it.
    I'm sure you think you can hear jitter, but how do you know that for sure? As I explained in my AES video, the only way to know at what level jitter is audible to you is with a box that let's you add from zero to a lot. Better, someone else would turn the knob while you listen blind, and you say when you think you can just hear it. Do that ten times and see if the thresholds are consistent, and below the noise floor of a CD. If so, you win. If not, I win.

    3. On the issue of hearing jitter, the best way to get there is to actually listen to compressed music and see if you can hear its artifacts! OK, sit back on the chair. Didn't mean to make you fall off of it . Reason for this is that compressed music is all about reducing the effective bit depth of music samples. We convert music to frequency domain and then selective represent each band at lower and lower quantization levels (resolution). When we do that too much, then you hear the artifacts.
    I can't see how reducing the bit-rate of lossy-compressed audio until you hear it is related to increasing jitter until you hear it. One is added noise, the other creates holes in the frequency response. To me, low bit-rate compressed audio sounds like comb filtering. And that's because it is more or less. That's not at all like jitter which is the addition of artifacts. You can express jitter as being at a level equal to a particular low bit (16th bit, 18th bit), but that's not the same as carving out parts of the spectrum!

    You say you don't hear jitter and that is proof enough for you that it must be inaudible. Isn't that proving a negative by itself? If you don't believe in proving a negative, then you shouldn't believe any study that says they couldn't hear jitter either!
    Not quite. I say that a typical amount of jitter is so soft that nobody can possibly hear it. This is a far more sensible way to approach audibility testing! That's why I created the nastiest buzzing noise I could muster for my AES test. Since that noise disappears at levels far higher than jitter, that is what convinces me that jitter is never audible.

    you should try to get a system that is at least transparent to CD audio resolution. That is not an exceedingly high bar. It means having an honest, 16-bits of resolution without distortion and noise. And a frequency response to 22Khz.
    I have that. Two systems in fact. Well, except for the 22 KHz part. But at the age of 61 I can't hear much past 14 KHz anyway so its moot. But my ability to hear detail and distortion is pretty good.

    Not to turn this into a pitch for room treatment, but it kills me when some audiophiles talk about clarity and detail, while listening in a room where the early reflections are uncontrolled and only a few dB below the direct sound. That obscures clarity far more than a cheap sound card, and even cheap speakers. I'd rather have $100 bookshelf speakers in a treated room than [whatever expensive brand] in a room with bare walls and no bass traps. I guarantee the sound will be clearer with the $100 speakers.

    --Ethan

  4. #14
    Member Sponsor [Technical Expert] DonH50's Avatar
    Join Date
    Jun 2010
    Location
    Monument, CO
    Posts
    3,517
    I can set up test files showing the impact of various jitter on various simple signals (mathematical test cases). At least that would let you play with magnitudes and look at the FFTs to see if it makes any sense at all at typical jitter levels. I am pretty sure I can generate .wav files but have to check what toolboxes I have at work (I am off this week, hurray!) Since it is system-dependent, what adding jitter would do is let people hear what it does more so than allowing conclusions about it's audibillity in the test files, since it makes sense (to me anyway) to run the jitter high enough so that you can hear it (maybe a series of runs with various levels?) Might help, but probably won't settle the debate. All it takes is time, and lots of it!

    Small amounts of random jitter can actually improve the noise floor by decorrelating other spurs. Colored noise also helps and does so without corrupting the in-band noise floor (SNR). Deterministic jitter adds discrete tones, as has been noted, and those are (imo) the main culprit/basis for any claim to audible signal degradation.

    I agree deterministic jitter is by far the largest culprit. And that room response swamps other error sources for most of us. I'd still stick with my expensive speakers even in a bare room, though, Ethan -- I can treat later!

    Great debate! - Don

    Sorry -- I missed that this was to be only Amir and Ethan! I tried to delete this post but it still cropped up -- hopefully an admin can wipe it. Senility, or maybe just getting back from vacation blues - Don
    Don Herman
    "After silence, that which best expresses the inexpressible, is music" - Aldous Huxley
    Don's Technical Articles on WBF

  5. #15
    Site Founder And Administrator Steve Williams's Avatar
    Join Date
    Mar 2010
    Location
    Coto De Caza, California on the 13th fairway of the south golf course
    Posts
    26,383
    Its a good addendum and feel it should stand but lets listen from the sidelines going forward.
    Steve Williams
    aka oneobgyn
    There's ALWAYS another Steve Williams BUT there's only "oneobgyn"
    USA Dealer of Center Stage Feet and owner of PitchPerfect Sound (www.pitchperfectsound.com)
    Dealer Lamm Electronics
    My System

  6. #16
    Banned
    Join Date
    Apr 2010
    Location
    Seattle, WA
    Posts
    16,044
    Quote Originally Posted by Ethan Winer View Post
    However, I disagree that all computer-based systems are lacking. My Dell computer has an M-Audio Delta 66 sound card, and it's definitely clean for both recording and playback down to the noise floor of 16 bits.
    Hmmm. The discussion started with you saying "any PC sound" card is that good. To the extent you now are saying that your specific configuration sounds good then we have made a ton of progress!

    That said, how do you know your configuration resolves 16-bits? How did you measure it? And can you post those results? Delta makes really good cards but once you plug it into a random PC, you are never assured of its performance.
    Yet look at how many people read that jitter is a problem, and then spend ten times more than needed to buy a converter that some ignorant reviewer claims sounds better than usual due to its low jitter.
    I would say that the other side has not found an effective way to communicate its message then . Besides your side is guilty too. This very same argument over this paper is going on in another forum as I type this. People are quoting that paper as gospel and when I asked the poster what the conclusion meant, he couldn't even spell it alone know what it meant!

    I suggest we put aside solving for world peace and continue on with the merits of the matter. Whether someone should or should not be worrying about jitter is not something that changes the facts. If you and I do a good job of discussing it, maybe there will be less of it .
    The task isn't to worry about the listener's playback system, only to create a file with jitter-like artifacts at a level that's typical. This is the crux of it. If you create a pair of files and run a poll asking people to choose which file they think has the added jitter, the poll results will hold the answer. If out of 100 people over multiple trials nobody can reliably pick the file with the added artifacts, that proves to my satisfaction that jitter is not an audible problem.
    I think we are still talking past each other. I already explained how Jitter can have infinite variations. Which profile would I simulate and why? And how do you enable people to use their home systems rather than clicking on a file on their laptop and testing that way? Burning CDs and such is a lot of work to ask people in the process.
    I'm sure you think you can hear jitter, but how do you know that for sure?
    Well, let me say that I am more sure of hearing it than you are that you can't hear it . At least I have positive proof for something. You are relying on negative logic that if I have not heard it in my limited testing (or worse, assumptions), then it must not be audible.

    As I explained in my AES video, the only way to know at what level jitter is audible to you is with a box that let's you add from zero to a lot. Better, someone else would turn the knob while you listen blind, and you say when you think you can just hear it. Do that ten times and see if the thresholds are consistent, and below the noise floor of a CD. If so, you win. If not, I win.
    Ethan, this is getting a bit frustrating. I have said repeatedly that there is no one number you can dial this way. It is not frequency response or volume where you could just have one dial to affect it. As a minimum, you would have to have three dials and the last dial, would have to have infinite dials itself to represent all the spectrum distributions.

    Let's put that aside for now. Please explain why I heard what I heard. Aren't there only two explanations?

    1. Jitter is audible.

    2. I imagined it, even though I am schooled in objective and scientific testing of audio and have an engineering background to boot, with an opinion prior to testing that all of this was non-sense as you, and that the test was run blind?

    Remember, my goal here is not to prove audibility of jitter. But rather, giving you a perspective that it is not just ordinary audiophiles who worry about jitter but people with extensive experience in the science of audio (especially the perceptual aspects of it) solid enough knowledge of the engineering behind it. In other words, as long as you accept that the case is not open and shut as you thought, then I am golden . I have no higher aspirations.
    I can't see how reducing the bit-rate of lossy-compressed audio until you hear it is related to increasing jitter until you hear it. One is added noise, the other creates holes in the frequency response.
    That's incorrect in the way you are stating it. If you look at compressed music, it will appear to have full response (at high enough data rates and with modern codecs than MP3). There will be no holes in its spectrum. Let's review again what how we perform lossy audio compression:

    1. Signal is transformed from time domain (PCM samples) into frequency domain. Think of dancing bar graphs in an equalizer. For now, assume if the samples are 16 bits each, the frequency "bins" are also 16 bits.

    2. Those frequencies are analyzed and based on perceptual model of the ear, we reduce the number of bits we assign to each frequency. So for example, if there is a loud sound at 1 KHz, we will leave that band alone but take bits away from another sound at 900 Hz since that is in the "masking" shadow of the 1 KHz sound.

    The reduced bits allocated to the 900 Hz band causes increased distortion which we call "quantization noise." If for example, we chop that band down to 8 bits, then we have created a much coarser signal and added distortion because of it. However, since this sound is likely to not be audible anyway, then butchering it this way is probably not harmful.

    3. We then take all of those frequency bins values which are just a bunch of numbers and apply a *lossless* compression algorithm to it as we would with zipping a file on your computer. This is called an "entropy coder" if you want to impress your friends at the next holiday party .

    4. The output of #3 is your compressed file. The player takes that file, applies the reverse of entropy coder and gets the frequency bins. It then does the reverse transform of that and gets the PCM samples. Data rate is reduced because we have reduced the bandwidth allocated to each bin variably and the entropy coder is optimized in compressing those values further.

    While it is true that if you heavily compress a file, some of the bins may get zeroed out, in the high data rates that I mentioned, that does not occur. Instead, what we get is increased quantization noise at certain frequencies. The increased noise tends to rob the signal its ambiance which is at much lower level than the music and increases its high frequency content (symptom of over quantizing a signal). Both of these occur in a very similar way with jitter. Jitter reduces system resolution and as such, it can also obliterate low level detail and help increase the high level frequency energy of the signal.

    Is the effect identical? No. But to the extent you learn to hear the artifacts of lossy music, you are a long way toward hearing artifacts rather than just hearing music. And that is key: there is a big difference in using your ears to enjoy something than to use it as an instrument, trying to find a flaw. Most people are not good at the latter. By practicing to hear lossy compression artifacts at high data rates, they learn what it takes to do that. If I may use a crude analogy, it is the difference between reading lips to understand someone in addition to hearing their voice or doing the latter alone. It takes practice to read lips, it doesn’t come naturally to people.
    You are ahead of most members here in hearing acoustic flaws. I am sure you don’t take that skill for granted or there or there would be no need to do the demo videos you have done. Likewise, I feel that more experience is needed to hear digital artifacts than what a person may already know.
    To me, low bit-rate compressed audio sounds like comb filtering. And that's because it is more or less.
    You have a tendency Ethan to read everything through the lens of acoustic problems . There are certain artifacts at very high levels of compression that might sound like comb filtering but that is not a correct generalization. Encode some speech and listen to it at 32kbps. It won’t sound like comb filtering but rather, will have a coarseness to it that is created by the effects I mentioned (more commonly called “pre-echo”).
    I say that a typical amount of jitter is so soft that nobody can possibly hear it.
    Have you ever heard jitter at any level Ethan? Do you have an external DAC and does it support multiple interfaces? If so, can you hook them up to your sound card and then switch between them and tell us whether they all sound identical? Try comparing S/PDIF, AES and Toslink. Then for the former two, try both a high quality short cable and long, crappy cable. Again, with every change, do an A/B against the other.

    I have ran the above test and I was shocked to hear the differences. It was totally unexpected for me. I would say you owe doing tests like this way before we owe you getting 100 people to vote whether they hear jitter or not .
    Not to turn this into a pitch for room treatment, but it kills me when some audiophiles talk about clarity and detail, while listening in a room where the early reflections are uncontrolled and only a few dB below the direct sound. That obscures clarity far more than a cheap sound card, and even cheap speakers. I'd rather have $100 bookshelf speakers in a treated room than [whatever expensive brand] in a room with bare walls and no bass traps. I guarantee the sound will be clearer with the $100 speakers.
    Again, that is neither here nor there. People here want the best so you should assume that they have done all they can to their room and are now wanting to understand two things:

    1. If they believe in digital, what is the best system for them.


    2. If they believe analog sounds better, what would explain that.

    I am trying to provide data for both. People natively understand analog concepts of distortion. They hear it readily. Turn up the amp too much and you hear clipping distortion. Get a speaker that has too much highs or is boomy and they hear that too. But where would they even begin to know what jitter sounds like? That is what I am trying accomplish here. Teach them the science as you do with acoustics. Tell them how it *could* be audible. And then let them “go fish.” We don’t control their pocketbook nor can we ever get into their heads to know what they hear. So let’s not keep debating where they should spend their money and what is good for them…

  7. #17
    Member Sponsor [WBF Founding Member] rblnr's Avatar
    Join Date
    May 2010
    Location
    NYC/NJ
    Posts
    1,861
    Quote Originally Posted by Steve Williams View Post
    darn but this is the makings of something good.

    Im watching from the sidelines as is everyone else. Thanks Amir and Ethan
    Yep, thanks guys. Really appreciating this.

  8. #18
    Banned
    Join Date
    Jul 2010
    Location
    New Milford, CT
    Posts
    1,232

    Lightbulb

    Quote Originally Posted by amirm View Post
    The discussion started with you saying "any PC sound" card is that good.
    I did? I thought I said that with any sound card, for whatever reason, jitter will be too soft to hear.

    That said, how do you know your configuration resolves 16-bits? How did you measure it? And can you post those results? Delta makes really good cards but once you plug it into a random PC, you are never assured of its performance.
    I know I can hear down to the last bit by playing very soft music or other sounds at high levels. I just did a test by creating a wave file generated in Sound Forge having one second of silence, followed by one second of 250 Hz at -90 dB, followed by another second of 250 Hz at -84. When I played it back at full volume I could easily hear when each level of 250 Hz started. Now, I admit there was some very soft hum and noise in the background. So you could argue that I was hearing the tones below the noise. I'd have to look further to see if that's true. Unless you'd like to visit me and we'll do it together.

    For anyone else who would like to test their system's playback, I put the file (about 260 KB) on my web site:

    250hz.wav

    Besides your side is guilty too.
    All I'm guilty of is asking for proof that artifacts 100 dB below the music are ever audible while the music plays, regardless of the makeup of those artifacts.

    I already explained how Jitter can have infinite variations. Which profile would I simulate and why?
    You would apply whichever profile you believe best makes your case. Do whatever tests you want, then pick the one you believe shows the (simulated) jitter at its worst. All I ask is that you do this at a level that's typical for jitter, not at -60 or whatever.

    And how do you enable people to use their home systems rather than clicking on a file on their laptop and testing that way? Burning CDs and such is a lot of work to ask people in the process.
    If you post a Wave file, people can make their own CD. In fact, that would be a nice addition for this forum, allowing Wave file attachments. I understand this takes up server space, but even if you limited the files to 1 MB that would be big enough for short tests like this. A 1 MB mono Wave file at 44/16 is about 12 seconds long.

    Ethan, this is getting a bit frustrating. I have said repeatedly that there is no one number you can dial this way. It is not frequency response or volume where you could just have one dial to affect it. As a minimum, you would have to have three dials and the last dial, would have to have infinite dials itself to represent all the spectrum distributions.
    Again, simulate jitter using whatever "dial" settings you like. As long as its truly representative of actual jitter I'll be happy.

    Please explain why I heard what I heard.
    What did you hear, and under what conditions? Was it a blind test? Double-blind?

    I've seen people say they switched sound cards to one that claims lower jitter, heard an improvement, and concluded the lower jitter must be the reason. But that ignores a dozen other possible differences between two brands of sound card.

    my goal here is not to prove audibility of jitter.
    Then I guess we're done!

    Seriously, I thought your point was that in some situations typical amounts of jitter can be audible. If you don't think that jitter is ever audible in usual amounts, then it's just an intellectual curiosity, and maybe an engineering goal, so we're done. I agree that design engineers should always strive for the highest performance possible, even if the result is not audible. For example, some things that are not immediately audible can become audible after several generations of copying or processing.

    If for example, we chop that band down to 8 bits, then we have created a much coarser signal and added distortion because of it. However, since this sound is likely to not be audible anyway, then butchering it this way is probably not harmful.
    I did not know that. I ASSumed that lossy compression was similar to digital noise reduction, where a many-band gate is applied to remove content below a threshold. Though in this case I figured the threshold for any given band was dependant on whatever else is going on in other bands at the time. Noise reduction and low bit-rate lossy compression both have a "swirly" phase-shifter comb filtered type sound.

    While it is true that if you heavily compress a file, some of the bins may get zeroed out, in the high data rates that I mentioned, that does not occur.
    Ah, great explanation. That explains the swirly sound. Thanks!

    Have you ever heard jitter at any level Ethan?
    How could I know? As I said, I don't know of any scenario to test this other than a black box that lets you just jitter from zero through a lot. That's why I asked about the tests you've done, and how you concluded that jitter is ever audible.

    Do you have an external DAC and does it support multiple interfaces?
    My Delta 66 can be slaved to an external clock via SMPTE, but I don't have anything other than an old Sony DAT recorder that can send SMPTE.

    I have ran the above test and I was shocked to hear the differences.
    But how do you know the difference you heard was due only to jitter? And was this a blind test someone else administered? As I'm sure you heard JJ say in our AES video, even he needs someone else to test him blind with stuff like this.

    --Ethan

  9. #19
    Banned
    Join Date
    Apr 2010
    Location
    Seattle, WA
    Posts
    16,044
    Quote Originally Posted by Ethan Winer View Post
    But how do you know the difference you heard was due only to jitter? And was this a blind test someone else administered? As I'm sure you heard JJ say in our AES video, even he needs someone else to test him blind with stuff like this.

    --Ethan
    That specific test was not run blind. I was simply suggesting that as an example of something I thought you could try. The more rigorous tests I ran were all double-blind. For example, I would test the effect of turning off the video circuit in a digital source (e.g. DVD-A) by turning away from the gear, hitting the switch many times in a row so that I didn't know what state it was in at the end. And then toggle back and forth while listening. If I could hear a difference, I would go back and forth a couple of times to be sure. Then I would turn and look at the display to see what mode the device was in. Often, I would repeat the test a second time to be sure there was not a pattern that was effecting me. I had done similar tests by turning off the front panel.

    As to how I know it was jitter, I can't explain it any other way. What other things could cause a source driving a DAC through three separate digital connections sound different? OK, there could be a reason for one of them being different but I let that be an exercise for the reader .

    I am open to an explanation why the above small modifications to the setup would create audible distortion or differences but jitter would not.

  10. #20
    Banned
    Join Date
    Apr 2010
    Location
    Seattle, WA
    Posts
    16,044
    Quote Originally Posted by Ethan Winer View Post
    I know I can hear down to the last bit by playing very soft music or other sounds at high levels. I just did a test by creating a wave file generated in Sound Forge having one second of silence, followed by one second of 250 Hz at -90 dB, followed by another second of 250 Hz at -84. When I played it back at full volume I could easily hear when each level of 250 Hz started. Now, I admit there was some very soft hum and noise in the background. So you could argue that I was hearing the tones below the noise. I'd have to look further to see if that's true.
    Beyond noise, you don't know if what you heard was true to the original. It could have been super distorted and you wouldn't know it. The only way to make sure you are linear down to the last bit is to measure it.

    All I'm guilty of is asking for proof that artifacts 100 dB below the music are ever audible while the music plays, regardless of the makeup of those artifacts.
    I don't know why you keep mentioning 100db. I have explained why that is not the right way to look at this. Let's review what the paper said:

    "It was shown that the detection threshold for random jitter was several hundreds ns for well-trained listeners under their preferable listening conditions."

    Now let's look at the AP graph I posted which was for a periodic jitter of just 7 nanoseconds, not "several hundreds" mentioned in the article:


    So as you see, even at much lower levels, the sidebands peak up to -80db.

    Now let's again look at real music capture using my audio precision:



    So does it not look like even at 7ns the distortion could rise up to the level of high frequency content in the music?

    Do whatever tests you want, then pick the one you believe shows the (simulated) jitter at its worst. All I ask is that you do this at a level that's typical for jitter, not at -60 or whatever.
    Per above, sounds like you want a situation where on purpose, jitter would not be audible. Show me where high frequency content is at 0db and I buy your arguments. Otherwise, we don't have meeting of minds on where jitter causes distortion and could be audible.

Page 2 of 18 FirstFirst 123456789101112 ... LastLast

Similar Threads

  1. Hi from Ethan Winer
    By Ethan Winer in forum Introduce Yourself
    Replies: 8
    Last Post: 07-12-2010, 10:14 AM

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •