Audible Effects Of Jitter

[Wombat]

Basic, old but OK for digital newbs like me

http://www.ultrahighendreview.com/uploads/documents/Benjamin_and_Gannon_on_Jitter.pdf

I think that the effects are included in the THD of analogue outputs.

 

[Jakob1863]

@Wombat,

after all those years still very worth a read, but the reader should keep in mind the special framework conditions of the study (and the listening trials).

 

[oivavoi]

@Wombat,

after all those years still very worth a read, but the reader should keep in mind the special framework conditions of the study (and the listening trials).

May I ask what you have in mind, Jakob? Under what conditions do you think jitter levels could be more audible than what their experiment showed?

 

[amirm]

Basic, old but OK for digital newbs like me

http://www.ultrahighendreview.com/uploads/documents/Benjamin_and_Gannon_on_Jitter.pdf

I think that the effects are included in the THD of analogue outputs.

It is the best paper there is out there. Such research stopped around that time so there is not a whole lot more.

As to THD, I want to make sure there is no confusion about that. Jitter effects are very different than THD. THD creates harmonics of our main signal at multiples of its frequency. Jitter on the other hand, creates two sidebands, one of which is actually lower than our main tone. That can create audibility scenarios that cannot exist with THD. For example if I am playing a 12 Khz tone, its second harmonic will be at 24 Khz which is well above hearing range. The same 12 Khz modulated by jitter of 10 Khz, will create distortion products at 12 -10 = 2 Khz and 12+10 = 22 Khz. The latter is inaudible but the former at 2 Khz, clearly not.

The reference to THD in the paper is to using that test harness to measure effect of jitter. Briefly, analog THD measurements present a tone to device under test, then filter out that tone using a notch filter, and measure everything else. That then becomes our total harmonic distortion as we have taken out our original tone (as we should). The authors used the same test method to take out the main tone that is subjected to jitter. That created the notch in their measurements:

That 20 Khz notch is due to that analog filter. In an ideal measurement that trough would be so small as to not be visible. But we can't build such sharp analog filters so we see the above picture.

Outside of this, there is no relevance to THD in anything discussed in the paper or jitter topic in general.

In my testing, I perform the analysis using digital means and I show the spectrum of the main tone (usually 12 KHz). So this notch effect is not in play:

If I were to do the test like they have done, the center tone in red would not be there (notched out by the "THD filter").

 

[Blumlein 88]

If you are using Audacity it has a notch filter function under effects. Set the frequency and the Q. A Q of 12 works for a fairly narrow notch for these purposes. If your source and recording device are pretty far off in timing to each other you might want to use a Q of 6 or 4 even.

 

[Jakob1863]

May I ask what you have in mind, Jakob? Under what conditions do you think jitter levels could be more audible than what their experiment showed?

Methodologically Benjamin/Gannon´s approach wasn´t purely about the impact of jitter effects but examined it in a way where it is a combined effect of the jitter modulation, the DAC´s jitter supression and the specially selected music samples.
Benjamin/Gannon at first examined the accordance of model predictions with real world DAC measurements and found in general a good accordance although some measurements were differring.
They used for their listening tests not the DAC with the least jitter suppression - otherwise the results of some listening tests could have been provide evidence for audibility of even lower jitter levels -
and were concentrating on examples where the jitter function was a sine and the music samples most similar to "pure" tones.
They used a prismsound jitter modulator to generate the jitter but afair they did not specify the uncertainty connected to the various jitter amplitudes. The modulator offered two ranges with diverging accuracies/uncertainties and most of the examined jitter levels would have been accessible in both ranges but with quite different uncertainties.

If you are interested in more details i´d have to retrieve my notes from the archive or to read Benjamin/Gannon again......