[quote author="clintrubber"][quote author="jdbakker"]Jitter only really matters when it comes to converter clocks, or their sources (like word clock, in a primitive PLL like the ADA8000 seems to use). For the impact on audio, the phase noise (the frequency transform of the jitter) appears to make quite a difference on the 'sound'.[/quote]
The relation between phase noise & jitter (translating figures into each other) seems a bit vague and/or requiring proper accompanyment(sp?) by the relevant conditions but which often lack.
Formula from section 3, page 4:
http://eesof.tm.agilent.com/pdf/jitter_phasenoise.pdf
(link down for some reason today)[/quote]
Thanks for the link. They seem to use a rather theoretical and simplified model for most of the calculations;
these folks have some real-world measurements (and I am inclined to believe them).
[quote author="clintrubber"]Thanks for the clear maths. The +/-20ns as 'allowed' by the AES/EBU standard is the actually quite 'bad' (things may have become quite worse by then) but as you said it's a worst case calculation here.[/quote]
They have a different purpose in mind: locking onto the signal at the receiver end. IIRC an AES link can have pulse times on the order of 100ns; 20ns would be 0.2UI, and that's a reasonable upper limit to jitter for a digital receiver to lock onto, even when the
eye pattern is closing due to noise, reflections and group delay.
[quote author="clintrubber"]And FS-20kHz-signals in audio are rare to say the least.[/quote]
Sure, but keep in mind that jitter impacts fast slewing signals most, and music has plenty of those (hi-hats, anyone ?). Jitter will distort the same signals that are degraded by slow-slewing amplifiers, albeit in a different way. The 20k sine wave is just a convenient example for calculations, and nowhere near the worst signal that a converter can get to see.
[quote author="clintrubber"][quote author="jdbakker"]So yes, for a CD-player 3ps jitter could be considered overkill, but not by very much.[/quote]
I can see that now, thanks. It's the 'best' figure I've seen so far. Surprised a bit that other 'masterclocks' seem to be worse by a few orders of magnitude.
For instance the RME Word Clock Module MKII states:
Low Jitter Design: typical 2.5 ns (PLL / Test), < 1 ns (Masterclock)
I'm not sure though I get fooled by the frequency this all is happening. The RME-figure is say at 44k1Hz and the Tent-spec at say 11.2896 MHz
(256*fs). I guess we need to correct the jitterspec [ps] for these different frequencies, right ? (Just like we need to do when comparing oscillator phase noise specs [dBc/Hz] at different frequencies).[/quote]
Sort of, yes, but it's not that clear-cut. That 48k word clock with 1ns jitter will need to be passed to a PLL to generate an MCLK for the converters. The PLL's loop filter will improve the jitter some, on the other hand the PLL's noise may make matters worse, not to mention the reference spurs.
An important point here is that
no PLL can improve a good low-noise crystal oscillator. Dan Lavry makes this point much more eloquently than I could, but the bottom line is that (a) an electrically-tuneable XO
must have lower Q than a fixed-frequency one, and lower Q = more phase noise, and (b) pretty much any PLL will have more noise than the amplifier in the XO. So if you want best jitter performance, don't use that word clock input.
[quote author="Rochey"]I assume that jitter noise level would be similar to the THD+N number on an ADC?[/quote]
Mostly, yes.
[quote author="Rochey"]So, an ADC that usually has a -110dB THD+N would be fine until you hit a jitter level of 10ps or so?[/quote]
I would have to say no, since psycho-acoustics play a role too. Those -110dB THD+N are dominated by THD (all modern ADCs that I've seen have a THD level of 6..8dB over the noise floor). As far as I can tell, harmomic distortion sounds different than the non-harmonic distortion caused by jitter. Being a conservative designer, I try to keep the theoretical jitter noise below the converter noise floor.
Then again, some people appear to like the effects of jitter. There is a 14-page flamefest on Dan Lavry's forum on this subject which is well worth a read. On one side there's Dan Lavry and Bob Katz, who say that jitter will change your recordings in a non-reversible way, and that that's why you'd want as little jitter as possible. On the other side are some folks from (mostly) Apogee, claiming that some of their customers prefer the
sound of external clocks like the Big Ben, even though they might increase jitter.
It's like back when I still shot a few dozen rolls of film every month. I'd gone through great pains to ensure that the process I used to develop my negatives was as consistent and benign as possible. Others would experiment with funky chemicals and (shudder) light sources during development.
Some people like tube sound, others like the sound of iron, both of which distort the sound in one way or the other. Maybe some people like the sound of jitter distortion for certain kinds of program material. I think I'm on Dan's side here. Especially for live recordings I'd rather not have any distortion mechanism present that I can't turn off.
[quote author="Rochey"]Thats very interesting - the PLL inside the ADAT chip (AL1402G decoder) is specified as 1.5nS which would give an equivelent noise figure worse than -75dBFS...
have I got my dB's mixed up?[/quote]
See above, no telling how much impact that 1.5ns jitter has on the converter clock. The fine folks at Wavefront might have included some more context, too, like the phase noise spectrum. Even so, 1.5ns jitter (or 1.26ns jitter on BCLK) isn't much to be proud of.
JDB.