R/E/P > Dan Lavry
Proper word clock implementation
Big Bri:
Dear Dan,
Man, I really appreciate you joining the forum. Thank you so much for your time.
I've been experimenting with different word clock scenarios in our facility and am dumbfounded by the sonic differences of just changing which unit is the clock master. One unit gives me a great clear 3d image while another is quite smeared and more 2D.
Can you please tell me why this is happening and what the best possible scenario is to set up a central word clock distribution and how to get the best results. Also, how do you control digital units with no dedicated word clock input?(i.e. Dat machines, stand alone cd writers, outbord fx)
Once again, thanks for your time and your knowledge.
Brian
bobkatz:
Big Bri wrote on Sat, 02 October 2004 03:08
Dear Dan,
Man, I really appreciate you joining the forum. Thank you so much for your time.
I've been experimenting with different word clock scenarios in our facility and am dumbfounded by the sonic differences of just changing which unit is the clock master. One unit gives me a great clear 3d image while another is quite smeared and more 2D.
Can you please tell me why this is happening and what the best possible scenario is to set up a central word clock distribution and how to get the best results. Also, how do you control digital units with no dedicated word clock input?(i.e. Dat machines, stand alone cd writers, outbord fx)
Dan is the man! And this is what I bet he'll tell you!
If you hear differences when you change clocks "controlling" your converters, then you have a defective converter design! A well-designed converter should contain internal phase locked loops whose performance reduces any incoming jitter artifacts to inaudibility. An external clock is a bandaid for a "cure" which can only be done properly within a good converter design. In fact, any converter which does not perform equally as good or BETTER on internal clock than external is also defective.
BK
danlavry:
If you hear differences when you change clocks "controlling" your converters, then you have a defective converter design! A well-designed converter should contain internal phase locked loops whose performance reduces any incoming jitter artifacts to inaudibility. An external clock is a bandaid for a "cure" which can only be done properly within a good converter design. In fact, any converter which does not perform equally as good or BETTER on internal clock than external is also defective.
BK
Yes indeed! Well said.
The best way to clock a converter is with internal clock, using a good fundamental frequency crystal (third order types are more jittery), and locating the crystal properly (good ground to the AD ample hold and so on). You now have a low jitter clock inside the machine.
What happens when you get a stand alone “almost no jitter clock”? You look AT THE OUTPUT CONNECTOR of that “super clock box” and it generally can work as well as the internal crystal clock Now take a cable and hook it to the AD chassis. Now you have to go through some electronic circuit to receive the clock. At this point, you have accumulated a lot more jitter (I can list half a dozen causes).
Well, this is not the end of the road. The big one is the PLL circuit. Unlike the internal clock (fixed crystal case), you have a crystal that can be pulled up or down by some amount, we call it a VCXO (voltage controlled crystal oscillator). There is some circuitry in there that keeps comparing the incoming external clock rate to the VCXO, and makes the proper adjustment on an ongoing basis…
What is more steady? A mediocre internal crystal implementation is going to outdo even a good external clock implementation.
But there are times and reasons to use external clocks. For example, if one needs to sync many chassis…
It is true that the PLL does better when fed a less jittery clock, but that is just a tiny portion of the overall issue. As Bob stated, most of the burden is on the PLL. A Good PLL, inside the AD chassis should clean most of the jitter out.
Why do you get such different results with different sources? I am not there to probe. I would not start with comparing how much jitter each source provides. I would look into issues such as driving coaxial lines, and proper termination impedance. Make sure the clock lines have no “branches” – Driver to point A, than to point B, than to C all in series.
I am no fan of distribution amplifiers either. You can not beat:
Driver to point A (with a BNC T), than to point B (with BNC T)… at the end the BNC T is terminated with the proper line impedance (if the cable is 75Ohm, so is the termination). It is a cost effective solution that yields the best results.
BR
Dan Lavry
bobkatz:
danlavry wrote on Sat, 02 October 2004 17:54
Why do you get such different results with different sources? I am not there to probe.
Dan said, modestly! While adding much of Dan Lavry's designer-level substance to my short reply to the question.
I'd like to add: It is possible that you can't EVER stop getting different results with different sources unless you throw out the entire poorly-designed converter (it's mostly due to the poor PLL design within the converter, though power supply and grounding is all part of that).
Quote:
I would not start with comparing how much jitter each source provides. I would look into issues such as driving coaxial lines, and proper termination impedance. Make sure the clock lines have no ?branches? ? Driver to point A, than to point B, than to C all in series.
I am no fan of distribution amplifiers either. You can not beat:
Driver to point A (with a BNC T), than to point B (with BNC T)? at the end the BNC T is terminated with the proper line impedance (if the cable is 75Ohm, so is the termination). It is a cost effective solution that yields the best results.
I'm sure Dan would like to point out that while this is the best way to do it, each of the devices that is to receive the wordclock must have a termination on/off switch. The termination must be turned off until the end of the line, or the source will be overloaded. In other words, the last device in the line must have the 75 ohm input. Unfortunately, the VAST MAJORITY of devices and converters I've encountered do not have such a switch, or else you have to go inside and remove a jumper or in worst case desolder a termination resistor.
Dan's points are very well taken, though, and you can save a lot of costs (both hardware and performance costs) by avoiding that distribution or clock amplifier circuit and just using his BNC-T approach. But you DO have to know how to measure input impedance if in doubt whether the input word clock is truly 75 ohm.
I'd like to point out that while wordclock has potential jitter superiority over AES/EBU "black" (there's an official AES number for AES black, I forget the number) there are many problems with using wordclock when synchronizing multiple channels and/or multiple devices. There is NO STANDARD for wordclock phase! You can end up with left and right channel reversal, or even a phase shift on the output of some of your supposedly "synchronized" devices. This is why AES/EBU sync is much preferred. However, there is a much higher jitter penalty with AES/EBU sync and even more attention has to be paid to the PLL on the input of any converter which is receiving AES/EBU sync.
BK
heinz:
bobkatz wrote on Sat, 02 October 2004 21:40
An external clock is a bandaid for a "cure" which can only be done properly within a good converter design. In fact, any converter which does not perform equally as good or BETTER on internal clock than external is also defective.
danlavry wrote on Sat, 02 October 2004 22:54
A mediocre internal crystal implementation is going to outdo even a good external clock implementation.
Wow! This thread has been very illuminating, kind of shatters some preconceptions I had. Thanks to all for the great information.
Navigation
[0] Message Index
[#] Next page
Go to full version