bobkatz wrote on Sat, 23 April 2005 11:29 |
Daniel Weiss wrote on Thu, 21 April 2005 23:55 |
Well, the added noise does not "clean up" the wordclock in the usual sense, you may even see higher jitter at your DAC with the "DAC enhancment" on. But the spectrum of that jitter is (ideally) flat. This causes the SNR to rise slightly but does not cause the nasty effects which discrete jitter frequencies do. It is something to the effect like dither, but in the time domain.
Daniel www.weiss.ch
|
I've seen approaches like that before, including the way that Meridian manipulated the spectrum of the jitter in one of their processors. It gets down to psychoacoustics and noise masking. But it seems to me you have to know the entire system real well and predict how the particular PLL/DAC system will react to the "pre-emphasized jittery" incoming signal.
But this discussion does get down to questions of "is less jitter always better sounding?" And I certainly have experienced audible tradeoffs... relating to the fact that if you reduce the jitter to the lowest you can make it, there will be residual jitter with a particular spectrum that translates to a certain noise floor or distortion floor within the DAC. Which if it hits the ear's sensitivity region could sound worse than if you add a carefully controlled film of white noise to cover that up. Doesn't matter if that white noise was created in the time or frequency domain. Think about it.
BK
|
First, a lot of people here are mixing external clocking of AD and Clock issues of a DA.
Second, I find the analogy between jitter and dither to be very poor. We
dither because we HAVE TO confront reduced accuracy due to word length limitations.
The most common example is when word length reduction from 24 bits to 16 bits for CD. But if the standard allows 24bits, we do not need to dither. In the case of jitter, there is no “standard” forcing us to accept more jitter. Less jitter at the converter is better signal transparency. Period.
Third, a “noise shaped jitter” does not arrive directly at the DA device. It gets altered by the circuitry between clock input and the DA, so the outcome is device dependent. Also, noise shaping and spread spectrum can not eliminate all the non random source of jitter, only jitter due to some limited causes. There have been other schemes to eliminate or help fight against systematic (non random) jitter, but they are all for DA’s.
Fourth, the cleanest way to clock an AD, and to avoid systematic jitter is to use INTERNAL crystal. When that is not possible, the next best thing is to focus on a real good jitter reduction INSIDE the AD (PLL circuits and alike). You can have an internal crystal provide few to a few tenth of psec jitter. By the time you are going for external clock, there is NO POINT in spending tons of money on a few ps (or tenth of ps) because the rest of the damage will be so much larger. Better to put your money on a AD with good clocks and PLL’s, and get a cheap clock, than to spend a lot on a clock that drives a poor AD (in terms of jitter rejection).
For more details regarding my thoughts on clocking, I am going to post a lot of the information about proper wordclock implementation on the forum on my site at:
http://www.lavryengineering.com/forum/phpBB2/index.phpYour comments are appreciated.
Regards
Dan Lavry
www.lavryengineering.com