Jitter Correlation to Audibility

Sep 7, 2013 at 1:53 PM Post #91 of 361
here is an example of the effect of adding extreme amounts of sine wave jitter (From Benjamin and Gannon, 1998)


Yes, jitter manifests as noise, and unless you add an extreme amount it's just not an audible problem. This graph from John Watkinson's Art of Digital Audio correlates jitter amount to noise level versus frequency.

--Ethan

 
Sep 7, 2013 at 2:31 PM Post #92 of 361
^
Depends on the type of jitter Signal-correlated jitter rather than random jitter will give you discrete distortion sidebands which are more detectable
 
 

 
 
In the above it is conceivable that someone could detect about 3ns, overall the average (better indicator) would be about 10ns. This using a single pure tone. When the jitter was added to a musical signal
 

 
the thresholds were much higher, however in specific circumstances detection was possible at as low as 20ns. Finding a commercial digital device with that much jitter (excluding HDMI)  is a real challenge. Even the crappy McIntosh MS750 Music server http://www.stereophile.com/content/mcintosh-ms750-music-server-measurements  only delivers 14ns which is just conceivably audible but the device is so awful that jitter is the last of its problems !
 
Sep 7, 2013 at 9:16 PM Post #93 of 361
  Hi folks,
 
Audiophile by birth, head-fier by choice. I am new to Head-fi discussions, so I might have missed lots of discussions but this seems like a fun thread to post to.  I am a EE by profession.  Part of my daily routine is clock measurement for gigabit communication systems. That's frequencies in ranges of 10^9 Hz and higher.  So when I see picoseconds and ppm talked about in these threads, my mouth starts to water.
 
What I want to say is that when operating at frequencies that I am working on, where bit times themselves are on the order of few tens of picoseconds, every picosecond counts.  We take this further to breakdown the types of jitter in the clocks in order to understand the causes and possibly mitigate the the problem.  There are numerous ways to measure jitter. There is no industry agreement on how jitter is measured and which parameters are critical for a specific application. For the audio DAC applications I would imagine the more relevant jitter metrics are either cycle-to-cycle and/or period jitter.  The former are timing variations between adjacent clock cycles, where is period jitter is variation of any clock cycle with respect to desired (zero-jitter) clock period. These two metrics are actually interrelated through a difference equation. Furthermore, jitter is further broken down into random and bounded distributions. The random element is interesting because theoretically there is no limit to amount of jitter as more measurements are made over time. In other words, there is always some low but non-zero probability that there will be clock transition increasingly before or after the ideal clock time. Therefore, clocks jitter numbers need to be always measured with a qualifier, such as over how many clock cycles, or to a confidence level for a specific probability. My point I am trying to make here is that descriptions like "5ps peak-to-peak" have little value as a key element of information is missing.
 
I'd like to comment about the frequency deviation from nominal,  typically referred to as ppm of a clock source. This is usually defined as a static frequency offset with respect to the desired frequency of the clock source. As we all know, ppm stands for parts per million, with 1ppm being 10^-6 deviation.  As a simple example, I will choose an ideal 1000 hertz tone out of a DAC sourced by a +/-100ppm clock. A maximum of 100pm deviation will generate a tone somewhere between  999 and 1001 hertz.  In otherwords, the deviation simply manifests itself as a minute shift in pitch. I don't know how many people would be able to distinguish a pitch shift to that detail, but I am personally skeptical that even a fairly loose frequency deviation like this example will create perceivable difference. You can run a simple tone test yourself on your setup and see if you could determine the differences in pitch.
 
I am not going by get into the spectral characteristics of jitter here, which is a fairly complex topic. In itself it is often a cruical piece of information about the jitter. I suspect might be relevant to audio applications, but off the top of my head it's not immediately obvious what the impcat might be. I would imagine the resulting waveform is a function of jitter spectral characteristics, sampling rate, and sampling technology (PCM vs a sigma-delta of DSD).

 
Thanks for chiming in, Digitalchkn. Have you experimented with different usb/spdif converters, usb receivers, and digital cables?  If the easily perceived sonic differences are not caused by jitter then what would be the cause? (I'm being both rhetorical and cautionary)
 
Currently the best explanation I have found on the audibility of jitter is in Robert Harley's book "The Complete Guide to High-End Audio". Basically even the tiniest differences in the timing of the digital input manifest themselves as distortions in the analog output. In other words it's entirely in the  conversion process where jitter is significant. This seems to be where most people get confused b/c they think jitter is only applicable to the digital signal itself not at the precise moment it is converted to an analog signal. THAT is where the audible differences come from. And that is most likely why those listening tests with the added-in artificial jitter have failed to prove anything.
 
Although much of it is over my head, your explanation Digitalchkn seems to be somewhat in line with Harley's. Have you read his explanation? It's a little simplistic as it is ended to be read by audiophiles, not engineers.
 
I am not an engineer myself. It was a usb cable thread that brought me to this "Sound Science" discussion board.
 
Sep 7, 2013 at 10:11 PM Post #94 of 361
How can jitter occur at the moment of conversion when there is a buffer? Are you saying jitter occurs between the buffer and the DAC? I can't believe that wouldn't be simple to avoid.

Have you determined through controlled testing that there is a discernable difference between converter, receivers and cables, or are you just assuming that from anecdotal info?
 
Sep 7, 2013 at 10:27 PM Post #95 of 361
   
Thanks for chiming in, Digitalchkn. Have you experimented with different usb/spdif converters, usb receivers, and digital cables?  If the easily perceived sonic differences are not caused by jitter then what would be the cause? (I'm being both rhetorical and cautionary)
 
Which easily perceived differences would these be and how have they been rigorously verified ?
 
 
Currently the best explanation I have found on the audibility of jitter is in Robert Harley's book "The Complete Guide to High-End Audio". Basically even the tiniest differences in the timing of the digital input manifest themselves as distortions in the analog output.
 
If you are taking Harley seriously as a source for information on jitter you really need to educate yourself. Harley's writings on jitter are so appallingly misinformed, that The Audio Critic even commissioned a paper by Robert Adams of Analog Devices to properly explain jitter and counteract the inane misinformation he continuously peddles ! I'm linking to it below
 
http://www.theaudiocritic.com/back_issues/The_Audio_Critic_21_r.pdf
 
In other words it's entirely in the  conversion process where jitter is significant.
 
OK, as far as it goes
 
And that is most likely why those listening tests with the added-in artificial jitter have failed to prove anything.
 
That might just conceivably be believable if there was any empirical evidence to back it up such as any controlled listening tests between low jitter and high jitter devices. Stereophile and others keep telling us how damaging jitter is but have never given us anything other than poor dog and pony shows, anecdotes and biased tests not the properly controlled listening tests that would show to what extent two devices differing only in jitter are audibly different when rigorously tested.
 
No, the best jitter experiments to date have been done by people like NHK (Japanese Broadcast Body), the BBC and Dolby labs - they know what they are doing, they have had between 40 to 50+ years of this stuff,  you can find their papers online where you will see how their experiments inject jitter into the decoded analog signal giving the distortion as noise or distortion sidebands depending on the type but exactly as it would be from jittery conversions. You can look at Stereophile's jitter measurements and see how jitter creates sidebands exactly like the sidebands Benjamin and Gannon created and at exactly the frequency and amplitude predicted by their models of the devices.
 
But once again the key is that jitter was only audible when it was present in outrageous amounts. Even the extremely cheap Marantz CD5004 has jitter levels so low that they are barely measurable and 
 

 
There is nothing here above -124db - are you seriously going to suggest that this distortion is in any way audible ? Okay lets look at something far worse a 1994 Optimus portable CD player

 
Even this terrible device has distortion sidebands that barely break above -100db - maybe we need something worse ?
 

 
 
Even this unforgivable piece of junk has folks who think it sounds good. This really is as bad as it gets for non-HDMI dedicated digital devices. Maybe this jitter is just audible but even here nobody has even tried to bliond compare this to something better so we have no concrete evidence that the jitter here is actually audible
 
 
 
 
 

 
Sep 7, 2013 at 11:15 PM Post #96 of 361
Which easily perceived differences would these be and how have they been rigorously verified ? 

 
Easily perceived differences meaning the differences you can easily hear w/o any kind controlled testing. The kind that virtually everybody who has tried different USB/Spdif have absolutely no difficulty in hearing. Something you obviously have not tried for yourself.
 
 
Harley's writings on jitter are so appallingly misinformed, that The Audio Critic even commissioned a paper by Robert Adams of Analog Devices to properly explain jitter and counteract the inane misinformation he continuously peddles ! I'm linking to it below
 
http://www.theaudiocritic.com/back_issues/The_Audio_Critic_21_r.pdf

 
 
Hmmm. That article is from 1994 and I'm highly skeptical of it's accuracy. Have you read the Harley explanation that I'm referring to? I didn't spend too much time reading that article and I don't care to spend a Saturday evening doing so, but I did spot a few obvious contradictions. I would guess that almost years 20 later now parts of that article have been thoroughly debunked (again, just a guess). You understand that this topic is still being heavily researched, right?
 
You seem so sure of yourself, Nick. :lol:
 
 
Edit: I wanted to copy and paste part of that article (in the conclusion) that does not in anyway contradict or even attempt to contradict any of Harley's claims. And yet the editor of this magazine tries hard to paint that sort of picture. In other words his interpretation of the article is very different from my own. It's not even a very good article regardless. Maybe I'll look it over more carefully sometime.
 
Sep 7, 2013 at 11:42 PM Post #97 of 361
Thank you for playing... Next!
 
Sep 8, 2013 at 1:02 AM Post #98 of 361
   
Easily perceived differences meaning the differences you can easily hear w/o any kind controlled testing. The kind that virtually everybody who has tried different USB/Spdif have absolutely no difficulty in hearing. Something you obviously have not tried for yourself.

...and there it is.  Darn that controlled testing anyway, and darn the truth it reveals, while we're at it. 
 
Sep 8, 2013 at 1:39 AM Post #99 of 361
It is because it is.
 
Sep 8, 2013 at 3:48 AM Post #100 of 361
   
[  robertsong re:  Bob Adams ]:
 
Hmmm. That article is from 1994 and I'm highly skeptical of it's accuracy. Have you read the Harley explanation that I'm referring to? I didn't spend too much time reading that article and I don't care to spend a Saturday evening doing so, but I did spot a few obvious contradictions. I would guess that almost years 20 later now parts of that article have been thoroughly debunked (again, just a guess). You understand that this topic is still being heavily researched, right?
 
Edit: I wanted to copy and paste part of that article (in the conclusion) that does not in anyway contradict or even attempt to contradict any of Harley's claims. And yet the editor of this magazine tries hard to paint that sort of picture. In other words his interpretation of the article is very different from my own. It's not even a very good article regardless. Maybe I'll look it over more carefully sometime.
 
 
Perhaps you should.  Actually take the time to read it, that is.  And understand it.  Oh, and along the way, figure out who Bob Adams is.
 
 
Quick quiz for all Sound Science denizens.
 
Based on available information, you have the opportunity to learn correct information regarding digital signal processing basics; digital audio theory; data converter architecture and implementation; and concepts of jitter as it pertains to digital audio; from one of the following Roberts:
 
a/  robertsong
b/  Robert Harley
c/  Robert (Bob) Adams
 
Which of these three might you choose?
 
and a ps for robertsong:  suggest you invest in a copy of Pohlman's Principles of Digital Audio.

 
Sep 8, 2013 at 10:34 AM Post #101 of 361
In Rob Harley's mag they also write stuff like this:
Subjective reviewing of audio components continues to generate controversy, [...]
Yet a pair of coincidences surrounding my review of these Marantz  components  served  to  reaffirm  the  validity and necessity of critical listening as an evaluative tool.

 
and here's what they wrote about that Marantz:
- I hear no bogus warmth
- The top end is very extended, but quite natural and easy on the ears
- all  the  high  percussion  emerge  with  truly  crystalline  clarity and glitter  yet without any excess audiophile “sparkle”
- The  midrange  displays  welcome body,  ripeness,  and  dimensionality
- the  bottom  end  has  the kind  of  extension,  weight,  power,  and  definition you’d expect from an amplifier equipped with a transformer the size and weight of this one

 
The coincidences surround that Marantz review are:
- the Oppo might strike some as lean-sounding
- Though I didn’t use the word “lean” in my notes, clearly I was hearing some of the same things as Chris.
- Now anyone who reads Chris’ review will soon notice that he and I do not attach the same values to what we heard.
  But that is not my point, which is rather that our descriptions of the Oppo’s reproduction are remarkably similar quite apart  from  how  we  ultimately  interpreted  what we heard.

 
 
What a bunch of utter ..
 
Sep 8, 2013 at 10:42 AM Post #102 of 361
Originally Posted by robertsong /img/forum/go_quote.gif
 
Currently the best explanation I have found on the audibility of jitter is in Robert Harley's book "The Complete Guide to High-End Audio".

 
If we're talking about Appendix C, Jitter Explained then it's probably the worst I've read. It's purely theoretical, there is no graph, no math, no nothing other than his word that x ps are theoretically low enough not to influence the theoretical SNR of a perfect y-bit converter.
 
Even if those numbers are right it has nothing to do with audibility.
 
 
PS: I cannot recommend the book at all.
 
Sep 8, 2013 at 1:55 PM Post #103 of 361
   
Easily perceived differences meaning the differences you can easily hear w/o any kind controlled testing. The kind that virtually everybody who has tried different USB/Spdif have absolutely no difficulty in hearing. Something you obviously have not tried for yourself.
 
I've done loads of biased sighted listening tests, thanks. They are the weakest possible form of evidence as many others have pointed out repeatedly and which you can also determine if you read a few of the sources out there which compare sighted vs unsighted tests. Even as far back as Masters and Clark (1988) who showed that the obvious, night and day perceivable differences between components often disappear when the visual cues are removed. An esteemed member here Dr Sean Olive can put it more eloquently here based on hundreds of tests at Harman. He refers to it as the dishonesty of sighted listening tests http://seanolive.blogspot.com/2009/04/dishonesty-of-sighted-audio-product.html
 
 
Hmmm. That article is from 1994 and I'm highly skeptical of it's accuracy. Have you read the Harley explanation that I'm referring to? I didn't spend too much time reading that article and I don't care to spend a Saturday evening doing so, but I did spot a few obvious contradictions. I would guess that almost years 20 later now parts of that article have been thoroughly debunked (again, just a guess). You understand that this topic is still being heavily researched, right?
 
 
Hmm - this is an actual patent holding expert in digital signal processing not someone who got his job by writing an essay and has no engineering qualifications at all. FWIW I've read pretty much every published paper on jitter I can find including as far as I can tell everything from the AES library on jitter and going back to 1974. I am constantly on the search for papers attempting to conclusively define the audibility of jitter. One thing however has not changed in all that time. There has not been one single empirical study to even suggest that the quantities and modes of jitter found in competent commercial digital devices approach audibility. Sure we know what the theoretical effect of jitter is and we can easily measure it. The elephant in the room is can we hear it, and for that we need actual experimental evidence not dog and pony shows and not anecdotes. The people who make anti-jitter devices have had ample opportunity to do controlled tests which would be trivial using an ABX box, device in circuit, device not in circuit ask the listener which is which - not rocket science. 
 
 

 
Sep 8, 2013 at 3:05 PM Post #104 of 361
Right, what better validation would there be than scientific proof? The problem is that such tests would require there to be actual audible differences for positive results. That's one of the main reasons we don't see the people making/selling those devices doing such tests publicly. They probably did privately, possibly even plenty, but not with positive results.
 
Sep 9, 2013 at 12:47 AM Post #105 of 361
  Right, what better validation would there be than scientific proof? The problem is that such tests would require there to be actual audible differences for positive results. That's one of the main reasons we don't see the people making/selling those devices doing such tests publicly. They probably did privately, possibly even plenty, but not with positive results.

 
 
I think jitter can manifest itself as different phenomenon. In the context of digital communication, excessive jitter can result in misinterpretation of the data perceived by a receiving device. Further, the error can be classified as recoverable or not. If it is recoverable, the received data is guaranteed to be a precise copy of that transmitted. In this scenario, there is no loss of information by definition. In the case of non-recoverable error, the received data is corrupted in one or more bits. Depending on some factors (e.g. frequency of errors and encoding of the data stream) a loss of digital data should be very audible. It's analogous (pun intended) to corrupt microblocks found in HDTV broadcasts that manifest themselves as squares or strips of miscolored video on your TV.
Streaming media typically does not provide solid recovery mechanisms. I believe USB isochronous protocol is in that category. So in this sense, ensuring data integrity at the receiver is crucial, and hence quality USB cabling could be a factor. However, I want to stress that if the data integrity across the cable is maintained, there is NO benefit in improving the cable. A 100% data transfer (at the intended data rate) is all that is required.
 
In the case of the D/A conversion process, I can see how a shift in individual sample timing due to DAC clock jitter will slightly alter the analog waveform. The amount of deviation highly depends on the amount of jitter and the sampling technology used. This is the point I am trying to convey - namely that there are multiple ways to obtain jitter numbers. The A-B listening tests need to explicit in ways that jitter is being measured.
 

Users who are viewing this thread

Back
Top