bushwick wrote on Thu, 02 June 2005 21:16 |
Very kind of you Mr. Lavry. Very kind. Well that is on par with what I have read in your column yet the question still remained that I am clocking three AD8000's off of one master (4 converters) and would the overall benefit remain, assuming that perhaps the big ben is a better clock for external driving than the one in the AD8000. Clearly if I had one converter only, this would not be in question, but where do the tradeoffs begin if you catch my meaning?
If you can answer feel free to be concise. I feel badly taking up your time.
Thanks, joshua kessler bushwick studio brooklyn, ny www.bushwickstudio.com
|
The issue boils down to details including specifications and numbers. A funny thing, if there is one single specification number I would want to see, it would be the jitter number. It is the MAIN ISSUE. After all, what do you expect from a clock? The signal goes up and down and up and down and…. It is a simple square wave, not a complex thing… You just want to know how repetitive that “up down” is, not much to ask for. But don’t hold your breath. Some clock makers say next to nothing, others tell you about cures for the jitters, testimonials, features, subjective claims about sonic improvements…
So I will do my best to answer “without the numbers”. In my opinion and experience, there is no way on earth that you can have a high speed cable driver, 6 foot non rigid coax cable of the highest quality, termination (for the cable) and a clock receiver IC and not pick up some jitter. Let us be optimistic and agree that you pick up only 50psec jitter (that would be great). Let’s call it the “link jitter”.
Now we need to combine the jitter coming out of the “Clock box” of your choice to the link jitter. But I said “combine” instead of add because when dealing with random phenomena, you do not go for a direct addition. Adding 2 equal sine waves doubles the signal, but adding 2 equal amplitude (on average) random signals increases the magnitude by only 40%. Sometimes the random signals enforce each others, other times they oppose each other…. When you take a random source A with amplitude 1, and B with amplitude 2, the combined outcome is 2.236
For A=1 and B=1 the outcome is 1.414
For A=1 and B=2 the outcome is 2.236
For A=1 and B=3 the outcome is 3.162
For A=1 and B=5 the outcome is 5.099
For A=1 and B=10 the outcome is 10.05
You can see that when one random source (B) is only twice as big as the other (A), the outcome is already almost completely up to the noisy one (B). By the time B is 3 times noiser than A, the contribution of A is "near negligible" ….
So having stated the above, lets get back to jitter. The big jitter is the “link jitter” (this is our B) and we "agree" it is 50psec (on a good day). So how much jitter can you afford to have from our clock source (this is out A)?
A clock source with 1psec jitter and a 50psec jitter link combine to 50.01psec
A clock source with 10psec jitter and a 50psec jitter link combine to 51.0psec
A clock source with 20psec jitter and a 50psec jitter link combine to 53.9psec
A clock source with 30psec jitter and a 50psec jitter link combine to 58.3psec
In other words, the difference between a 1psec clock and a 10psec is a change of 1psec. I would not pay for a 1psec clock source.
The difference between a 1psec and a 30psec clock is 8.3psec. Is that worth spending money for?
Keep in mind that most crystal oscillators, including the lowest cost clocks on the market will yield better than 30psec.
I would simply buy a low cost clock source, and put the attention into a WELL SPECIFIED cable and a good termination.
Regards
Dan Lavry
www.lavryengineering.com