[Vincent Kars]

Can’t say I have found conclusive research about the audible threshold of jitter.
The experiment by Ashihara (2005) indicates that random jitter is not audible below the 250 ns .
But as early as 1974 the BBC research department concluded:
For jitter having a random, white noise, spectrum extending from 30 Hz to 16 kHz, it is estimated that impairment on critical programme would be perceptible to less than 5% of listeners provided the jitter amplitude is no more than 50 ns r.m.s.
Then there is the work by the late Julian Dunn
We can see that the audibility threshold decreases from 500ns at low frequencies to as little as 20ps at 20kHz. Especially when using formats or converters with high sample-rate this will be a major issue.

According to Wikipedia
research by Benjamin and Gannon involving listening tests found that the lowest level of jitter to be audible on test signals was 10 ns (rms). With music, no listeners in the tests found jitter audible at levels lower than 20 ns
A bit more details can be found here: http://thewelltemperedcomputer.com/KB/BitPerfectJitter.htm

[Nuance]

Isn't jitter a dance? "Jiiterbug" baby!   I can hear this jitter any day of the week man!

[firedog]

According to Dan Lavry, famous converter designer, and Hugh Robjohns (SOS writer), that's physically impossible with normal converter designs.

Do you happen to live anywhere near me in Western Connecticut? If so, I'd love to come by for a listen in person.

--Ethan

I don't know what physically impossible means. The equipment was modded to so that the internal source clock slaves to an external master clock. There are other master clock devices for home audio on the market, so apparently it's not impossible.

Sorry, I don't live in the US, so you can't hear my rig. But if you want to hear a master clock system for home audio, try an Esoteric or dCS dealer. I beleive they both market (expensive) master clock components. See if you hear a difference with the master clock in the chain and out.

As far as the DB jitter test I took: A track consisting mainly of guitar, voice and cymbals was made. Then additional versions to the original were made with plus 10ns jitter, plus 30ns jitter, and plus 100ns. Added jitter was random over the entire spectrum, if my memory is correct. I can't tell you the exact methodology used in making the tracks.

The tracks were played at random. I immediately picked out the 100ns track as having the most jitter. It simply sounded less clean and natural, what I would call slightly distorted. I also reliably picked out the 30ns track a large majority of the time as being the one with the "next highest" jitter.  I couldn't reliably differentiate between the 10ns track and the one without added jitter, although I "sensed" some difference between them.

A couple of people could pick out the 10ns track reliably. One got it every time. But he builds DACs, so he probably had an advantage.

I think to note low levels of jitter, one needs to have the experience of hearing what it sounds like. Once you know it is much easier to spot. In my experience, it is most noticeable with percussion and drums (various "banging noises") and with high frequency material like cymbals. I'm not sure it would be so noticeable on a electronics store system, or on one that didn't reproduce from a very quiet background. In short an "audiophile" system. On some types of music it is less noticeable.

[SteveFord]

This is the Jitterbug.

[Nuance]

^ That's actually quite a nice lure, by the way.  I've had great luck with it.

[Ethan Winer]

I don't know what physically impossible means. The equipment was modded to so that the internal source clock slaves to an external master clock. There are other master clock devices for home audio on the market, so apparently it's not impossible.

Yes, it is impossible, at least if the sources I trust are correct. In his article I linked to, Hugh Robjohns at SOS magazine did extensive research using real test gear (versus "I thought I heard") and measured the jitter become worse in all but one master/slave configuration. The one converter that didn't get worse stayed the same. Hugh explained why it's not possible for an external clock to ever improve performance of a converter. It might retain the same jitter levels when clocked externally, but it can never improve. Dan Lavry has also written about this.

http://www.lavryengineering.com/white_papers/jitter.pdf

As far as the DB jitter test I took: A track consisting mainly of guitar, voice and cymbals was made. Then additional versions to the original were made with plus 10ns jitter, plus 30ns jitter, and plus 100ns. Added jitter was random over the entire spectrum, if my memory is correct. I can't tell you the exact methodology used in making the tracks.

The key question is how the controlled amounts of jitter were created. This is extremely difficult to do! Did an engineer design a special box for the test with a "Jitter Amount" knob? I'm not aware of such a box.

--Ethan

[firedog]

Actually, neither article you mentioned says it is impossible. The SOS article is basically talking about using certain off the shelf master clocks for studio use, and again, doesn't say it is impossible or can't work, it just points out difficulties involved. In fact, it specifically points out a few situations where it can work or is a good solution.

I'm not a digital engineer, but from  my understanding the SOS article isn't relevant to my situation. On my equipment the source was modded specifically to work as a slave to the master clock device. The two work as a unit. The source unit (clock) actually doesn't work anymore without being connected to the master unit (or another device like it);I think this is a very different situation from the one described in the article.

My source unit is a modded Logitech Touch. The jitter was measured before the mod and compared with the jitter from the output when using the master clock after the mod. In addition, we measured several other "off the shelf" Logitech Touch units, just to see if their output was similar to mine before the mods. And it was.

We also attached my modded Touch to a master clock unit identical to mine except with a much more expensive, more accurate master clock placed inside. The resulting output from this combination lowered the jitter even further.

But in any case, I sent the SOS article to the designer/technician for comment. I'll let you know if I get a reply.

[firedog]

Here's the response to the SOS article I received from the person who did my mods:

Studio master clocks are not the same as how my products work.  They are generally used for synchronizing events for editing.  They are generally not used for playback.  His article confuses the word-clock with the frequency I call master clock, which is generally 128 or 256 times the word-clock.  Some of the things he says about master clocks are true, like with the A/D converter.  However, all of the "master-clocks" that he tests are actually word-clock generators.
 
The reason that I use the term "master-clock" is that this is the term that is used at the D/A chip.  Unfortunately, the word also describes a system word-clock in a studio environment.  Totally different.
 
The tests that he performs are interesting primarily for studios.  IMO, providing even a superb low-jitter word-clock is not very effective for reducing system jitter.  The clocks need to be low-jitter true master clocks at 256X the word-clock frequency.  This is how my devices work.  When you provide only word-clocks, then the device must synchronize its internal clock to this and then the jitter is a function of the internal high-frequency clock, not so much the external word-clock.

So, my source and the article are in agreement, but as I tried to say, it's comparing apples and oranges. The article is about studio word clocks, my master clock setup is a different use of clock technology. If I understand it correctly, my system is a slave/master system of the clock oscillators.