bobkatz wrote on Thu, 16 December 2004 21:32 |
Dan Lavry has been extraordinarily correct technically in all his responses about clocking and how it works, the principles of internal and external clocking, including one recently to Nika today. Since you cannot defeat the laws of physics, and therefore it should be Apogee's onus to demonstrate, via blind listening tests and published technical measurements how an external clock could possibly make a properly-performing converter perform "better".
|
Bob,
I hope you do not mind me posting a few segments from this archive here. The following is a segment from a dialogue that took place from November 1 through November 7 of 2001 between David Josephson, Julian Dunn, yourself, and many others on this very topic. The evidence that it
would be possible to best the performance of an internal clock with an external clock was compelling enough that you changed something in your book: Responding to Ben Gilsdorf's reason that an internal clock might be subject to more problems than it's accompanying PLL you write: "...And so do I... I'll have to revise my book....
BK"
This is the dialogue that led up to that:
Bob Katz:
> I've started a running argument in the Digital Performer maillist. My
> basic contention is that converters will inevitably perform better
> (jitter-wise) on internal crystal sync unless the designer of the
> converter was brain-dead or made some severe design compromises. No
> one denies the contention that a phased lock loop can never exceed
> the performance of a perfect crystal oscillator, the best it can do
> is equal that performance. And typically, be worse, at least
> measurably.
>
> However, many critical listeners report better sonic performance from
> converters such as the Digidesign 888 under stable external WC sync.
> I find it very hard to believe that the internal crystal oscillator
> of the 888 is so bad that it can't do better on internal! Perhaps
> this is just a matter of "the lesser of two evils".
David Josephson:
Bob, you of all people should be careful about ignoring critical
listener reports in favor of supposed theory! A crystal oscillator is
no magic device; the crystal is simply a very high Q mechanically
tuned circuit that passes signal predominantly at one frequency. Put
it in a feedback loop around an amplifier and it oscillates, same as
any other tuned circuit. But the bandwidth is not infinitely narrow,
the Q is not infintely high. Design of a low phase noise crystal
oscillator is not terribly difficult but it isn't as simple as connecting
it around some CMOS inverters like you see in most digital implementations.
Different crystal cuts have different bandwidths too; sometimes at the
expense of frequency stability over temperature. I can very easily
believe that a good phase locked loop would perform better in terms of
phase noise in the audio band than an ordinary crystal oscillator
would.
Bob Katz:
> "In general, a typical converter will likely perform better on
> internal sync than external".
David Josephson:
In general I think that's probably so. But the reader should be
cautioned that if it seems that it sounds better on external
sync, believe your ears.
Bob Katz:
> It just has always stood to reason.. it makes so much engineering sense.
David Josephson:
As someone who has designed and built many crystal oscillators over
the years, mostly for radio applications, I will refrain from commenting
on whether it makes so much engineering sense.
later:Bob Katz:
> Then let's talk practical engineering at the price/performance levels
> of a Pro Tools 888.
> Now I don't know shinola about what's inside an 888, so I could be
>off base...
>
> If, as opposed to the crystal-around-inverter-trick, Digidesign were
> to buy an off-the shelf DIP-socket type rectangular metal can master
> crystal oscillator, feed it from a clean, bypassed power supply,
> grounding, etc. This is "engineering by Xerox" or at least buying
> someone else's module, but it is doable at this level of dollars. But
David Josephson:
A quick look at off-the-shelf crystal oscillator modules in the Digikey
catalog reveals only one with a phase noise or jitter spec, and it's 30
psec typical, 50 maximum. It would be safe to assume that the more
generic parts are worse than that.
Bob Katz:
> to my knowledge, you can't buy a good PLL off-shelf... it has to be
> weaned and wined and dined and tweaked and double-PLLed until it's
> blue in the face.
David Josephson:
If you are in the business of producing such equipment, it is customary
to have on your staff at least one competent analog designer with some
background in PLL design. Techniques for combining off-the-shelf VCOs,
phase detectors and reference oscillators to produce a predictable level
of PLL performance have been well documented for at least 30 years.
Bob Katz:
> Anyway, the question, based on the above is: Which is more likely to
> perform better, an ordinary PLL designed with a few hours of
> engineering, or the off-the-shelf crystal oscillator module... which,
> internal, or external, is likely to perform better?
David Josephson:
While I agree that in my experience, it is more likely that internal
clock will produce lower jitter at a given converter, there are enough
variables to make a blanket statement that it "will sound better" unwise.
There is also the question of the spectrum of the jitter; while the
PLL will almost certainly have higher jitter at low frequencies, a
crystal oscillator may well have significant phase noise (which becomes
jitter) at high frequencies in the audio band, which may mean accentuated
jitter artifacts at certain frequencies of program material (this is
possible with PLLs too).
later contribution by Julian Dunn:>Another explanation is that the phase alignment under external WC sync
>reduces some crosstalk effect. If the comparison of "sonic" qualities
>is made with the WC connected and a switch selecting between internal
>and WC sync then on internal there would be the potential for crosstalk
>from the WC input.
There was more. I picked what I thought was probably critical to your revision on this topic in your book. Of course, I made the same point in my own book, but I think you and I both allowed for the notion that in some conditions boxes like Big Ben COULD provide better results - or at least one should be open to it and do their own listening tests. This notion is directly contradictory to the point made so emphatically by Dan when he writes, regarding whether or not Big Ben could improve on an internal clock, "No."
Cheers!
Nika