[time-nuts] Clocks for Audio gear

[Heinzmann, Stefan (ALC NetworX GmbH)]

MailLists wrote:
> Ashihara's tests were with music/voice, taking into account
> psychoacoustics, for an average group of music savvy listeners, and
> even music professionals.
> As uncorrelated jitter is practically raising the noise floor, most of
> it was masked by the signal, making it more difficult to detect.
> Benjamin and Gannon used sinusoidal jitter, which isn't appearing
> normally in signal chains (badly designed ones excepted).
> In a real case, with higher probability (added) jitter would be
> correlated with the digital content transmitted over a path - S/PDIF,
> and AES/EBU are prone to jitter induced by the signal path
> characteristics, ISI - PSUs, and even external noise sources.
> A more realistic simulation would take those into account.
> OTOH there where tests on pure sine tones, with sine jitter, detectable
> by trained ears at even lower levels of jitter, which might indicate
> the lowest threshold of hearing, but using artificial conditions.
> Who would listen to pure sine tones?

Ashihara et.al. wanted to find out what level of jitter was likely to be audible under real-world conditions. Those conditions would likely include music as the main signal, and random jitter.

Benjamin/Gannon wanted to find out what levels of jitter could be detected if the conditions were as favorable as possible for detecting jitter. That is not the real-world situation, of course, but it can establish a baseline where you may legitimately say that if you stay below this line with jitter of whatever type, the effects are very unlikely to be audible. And, to add a comment towards Attila, one of the results by Benjamin/Gannon was that training matters a lot, and the best sensitivity was by trained listeners. Your comment is therefore warranted, but already accounted for.

Hence, even though their results appear to be very different, they are both valid, because it depends on the exact question asked. I would dare to say, that no matter how you set up your "realistic" simulation, the results are likely to be somewhere between the values by Benjamin/Gannon and by Ashihara et.al.

So, for the purpose of this group, I'd say the psychoacoustic stuff would lead too far, but it might be helpful to know at which jitter levels one can assume to be on the safe side in an audio system, regarding audibility of jitter effects. Judging from the mentioned studies, I concluded (for myself at least), that this boundary is somewhere in the single figure nanoseconds, until someone comes forth with hard evidence that it needs to be set lower.

Cheers
Stefan