R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 [3] 4 5 ... 15   Go Down

Author Topic: Proper word clock implementation  (Read 102412 times)

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: Proper word clock implementation
« Reply #30 on: October 13, 2004, 08:04:29 pm »

“Great thread. Its confirming things I knew and adding to my understanding in some areas.

As always, the devil is in the details.
When judging the possibility that an external WC might yield better conversion than an internal crystal the specific implementation of the internal clock must be understood.”


Well, it is possible that some fixed frequency crystal oscillator badly designed or poorly constructed is going to be absolutely terrible. But as a rule, a crystal oscillator tends to be very stable, and the crystal itself keeps things very constant.
On the other hand, external clock call for an adjustable crystal – the crystal is cut in such a way as to be sensitive to the amount of capacitance across it, and we are talking tenth of pF’s (small capacitance).

That capacitance change is driven by a voltage change. The capacitance is in fact a reversed biased diode called varactor, specially made to change capacitance with reverse voltage. All diodes do some of that (the less DC the more capacitance). Varactors are diodes made to enhance that effect.

So where is the voltage across the diode, which changes the capacity across that voltage controlled crystal, comes from? It is an error signal being made from the difference between the input frequency and the variable crystal…. Shell I go on? A lot of factors are at play here. Just a bit noise on that control voltage and you have some jitter. Can you filter the noise out of the controlling voltage? Yes, but only down to some frequency (be it 100Hz or 1KHz), but anything below that frequency becomes jitter.

And now you still have to distribute the clock… So which one is worse? The internal clock is much better.    

“So here's a question. Are the sonic artifacts of jitter random or highly program dependant OR can they be characterized in some way. Is it the case that jitter always produces a harshness via unintended high frequency components generated OR is it possible that jitter can manifest itself as a warmer/muddier sound beacause the highs get "smeared out". If there is some inherent characteristic of the sonic impacts of jitter then I could better judge what is most accurate in the context of my system.”

Random noise makes random jitter. The noise that is correlated to music or to the data that carry the music – in fact any non random noise can modulate the jitter. Go to my web at www.lavryengineering.com and look at a paper called ON JITTER (under the support section). Look at the plots and you will get much more detail.

BR
Dan Lavry
Logged

Roland Storch

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 406
Re: Proper word clock implementation
« Reply #31 on: October 14, 2004, 01:04:55 pm »

danlavry wrote on Thu, 14 October 2004 01:04



So where is the voltage across the diode, which changes the capacity across that voltage controlled crystal, comes from? It is an error signal being made from the difference between the input frequency and the variable crystal?. Shell I go on? A lot of factors are at play here. Just a bit noise on that control voltage and you have some jitter. Can you filter the noise out of the controlling voltage? Yes, but only down to some frequency (be it 100Hz or 1KHz), but anything below that frequency becomes jitter.

And now you still have to distribute the clock? So which one is worse? The internal clock is much better.    
BR
Dan Lavry




So what is the best solution if I need 24 AD converters for 24 tracks? Cascading 8-channel ADC with the first as master? Do the external clocked second and third 8-channel converters have a worse performance, because they are externally clocked?

Or should I prefer a 24 channel modular ADC system?
Logged

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: Proper word clock implementation
« Reply #32 on: October 14, 2004, 01:52:47 pm »

Roland Storch wrote on Thu, 14 October 2004 18:04

danlavry wrote on Thu, 14 October 2004 01:04



So where is the voltage across the diode, which changes the capacity across that voltage controlled crystal, comes from? It is an error signal being made from the difference between the input frequency and the variable crystal?. Shell I go on? A lot of factors are at play here. Just a bit noise on that control voltage and you have some jitter. Can you filter the noise out of the controlling voltage? Yes, but only down to some frequency (be it 100Hz or 1KHz), but anything below that frequency becomes jitter.

And now you still have to distribute the clock? So which one is worse? The internal clock is much better.    
BR
Dan Lavry




So what is the best solution if I need 24 AD converters for 24 tracks? Cascading 8-channel ADC with the first as master? Do the external clocked second and third 8-channel converters have a worse performance, because they are externally clocked?

Or should I prefer a 24 channel modular ADC system?


For 24 tracks, you probably need to go for an external clock source and send a cable (for a clock signal) to each chassis.

My statement is: when possible, use internal clock. But when you have more than one chassis it is not possible.

My first reaction is to suggest that all clocks arrive at the same time (or nearly so) and the proposed BNC and T connection is a good one.

Regarding having one chassis as master and doing daisy chain? Probably not a good idea, because there may be some time delay from the master AD to the clock output, and additional delays in the other chassis. Does it matter? Read the thread I started yesterday about time delays.

BR
Dan Lavry  
Logged

Roland Storch

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 406
Re: Proper word clock implementation
« Reply #33 on: October 19, 2004, 05:58:02 pm »

danlavry wrote on Thu, 14 October 2004 18:52


My first reaction is to suggest that all clocks arrive at the same time (or nearly so) and the proposed BNC and T connection is a good one.

Regarding having one chassis as master and doing daisy chain? Probably not a good idea, because there may be some time delay from the master AD to the clock output, and additional delays in the other chassis. Does it matter? Read the thread I started yesterday about time delays.

BR
Dan Lavry  



I went through the time delay thread and think time delay is not an issue in my situation, where I have a main pair/array (AB, ORTF, Decca Tree,...) with maximum 5 microphones plus spot mics.

The main pair/array will be converted by an 8-channel converter, so there is no time delay. The spot mics follow the 3:1 rule anywhere. Some are even delayed a few miliseconds (Haas-effect).

Wouldn?t it be better in this case to take the internal clock of the first converter and and take the wordclock out to clock the ADC of the 8+ channels?

The idea is that, as you said, a converter always works best with its internal clock. And I would like to have the best performance on my main pair/array.
Logged

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: Proper word clock implementation
« Reply #34 on: October 19, 2004, 07:12:37 pm »

Roland Storch wrote on Tue, 19 October 2004 22:58

danlavry wrote on Thu, 14 October 2004 18:52


My first reaction is to suggest that all clocks arrive at the same time (or nearly so) and the proposed BNC and T connection is a good one.

Regarding having one chassis as master and doing daisy chain? Probably not a good idea, because there may be some time delay from the master AD to the clock output, and additional delays in the other chassis. Does it matter? Read the thread I started yesterday about time delays.

BR
Dan Lavry  



I went through the time delay thread and think time delay is not an issue in my situation, where I have a main pair/array (AB, ORTF, Decca Tree,...) with maximum 5 microphones plus spot mics.

The main pair/array will be converted by an 8-channel converter, so there is no time delay. The spot mics follow the 3:1 rule anywhere. Some are even delayed a few miliseconds (Haas-effect).

Wouldn?t it be better in this case to take the internal clock of the first converter and and take the wordclock out to clock the ADC of the 8+ channels?

The idea is that, as you said, a converter always works best with its internal clock. And I would like to have the best performance on my main pair/array.


I do not have much experience with placing mics and the 3:1 rule. The math shows that one can get into trouble when adding 2 or more signals that share a portion of the same signal. Clearly if mic A picks up say a clarinet, and mic B, pointing elsewhere picks up the same signal but with -20dB attenuation, having non equal delays will make the addition and cancellation impact to the tune of -20dB (the impact is dictated by the weaker signal). So the more you isolate the mics, the less the impact of time delay...

The time delay stuff is what happens after you have taken care of all the acoustic considerations. If you are sure that there is no significant sharing of sounds between channels of different delays, you could indeed be free to run your "mains" at an internal clock, and the others as slaves. That is of course better from jitter standpoint.

BR
Dan Lavry


Logged

Albert

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Proper word clock implementation
« Reply #35 on: October 21, 2004, 11:11:41 am »

Okay, to continue regarding chaining many WC clock devices together as opposed to using a distribution box...

I've got all my BNC t-bars, terminators, and short WC cables, and have found chaining WC devices to be very fiddly. In short, I haven't been able to get it to work "as it should" although I have been able to get it to work with some fudging.

When chaining through a number of units using the t-bars, some problems occur:

A) at some point the WC signal seems to get lost and all the rest of the devices on the chain don't lock.

B) when "A" happens, often the lock will be lost all along the chain, or a good part of it.

C) WC lock can be lost by the first unit in the chain (which will then revert to internal), but everything else later in the chain locks to it fine.

D) WC seems to be very sensitive as far as which devices it "likes" and "doesn't like". Even different version of the same basic model can cause problems.

E) Without fail, WC lock is lost when using an external terminator on the last device rather than its internal termination. Let's just say I've found the chain "hates" external termination.

The units being chained are two digital mixers, Tascam DM-24's, one Akai S6000 sampler and five Akai S5000 samplers. All these units allow WC termination to be turned off. My clock source is an Aardvark AardSync II, the best clock I have.

I have been able to work around these issues by doing the following:

The Tascam's have WC in and out. So I go into the WC in on the first Tascam, then use its WC out to feed the next Tascam. That Tascam WC in has a t-bar connector, and the five Akai S5000's are fed in a chain from there. The WC output of the second Tascam is used to feed the S6000. I did it this way because while the S5000's locked together happily, they didn't when the S6000 was involved. So I had to send it signal separately.

So it's not a pure chain using t-bars only, the WC signal goes through the Tascam's in order to feed units further down the line. I've found this works very reliably. However, I have a question: Do I lose the advantages of chaining the units when using the WC "thru" on the Tascams? In other words, is that basically the  same as using a distribution box?

My next challenge is to figure out how to chain my second rack of Akai's.

Incidentally, I do think the Akai's sound better using the AardSync II clock rather than their own internal clocks. Even with my bastardized version of a wordclock chain, I've been able to lock these units to a better clock, something I wasn't able to do otherwise without spending a bundle on another distribution box.

I'll be interested in hearing any suggestions for further troubleshooting. Thanks again for this very informative thread!
Logged

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: Proper word clock implementation
« Reply #36 on: October 21, 2004, 11:36:33 am »

Albert wrote on Thu, 21 October 2004 11:11



When chaining through a number of units using the t-bars, some problems occur:

A) at some point the WC signal seems to get lost and all the rest of the devices on the chain don't lock.





I'm sure you are running into multiple terminating word clock inputs. You will have to remove the 75 ohm resistor inside the box(es) for this to work.

BK
Logged
There are two kinds of fools,
One says-this is old and therefore good.
The other says-this is new and therefore better."

No trees were killed in the sending of this message. However a large number of
electrons were terribly inconvenienced.

Albert

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Proper word clock implementation
« Reply #37 on: October 21, 2004, 11:48:58 am »

All the units I'm chaining have switches to turn WC termination on/off. In all cases I've turned it off, except for the last unit in the chain. That's why I'm a bit puzzled.
Logged

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: Proper word clock implementation
« Reply #38 on: October 21, 2004, 01:09:34 pm »

"So it's not a pure chain using t-bars only, the WC signal goes through the Tascam's in order to feed units further down the line. I've found this works very reliably. However, I have a question: Do I lose the advantages of chaining the units when using the WC "thru" on the Tascams? In other words, is that basically the same as using a distribution box?"

I can not answer question regarding all the gear you have. But in fundamentally, the impact on a WC signal due to a short coax cable is less then going through a circuit. The circuit is made out of semiconductor device which will a some jitter. Maybe not much but a little here and a little there and it does add up. The mechanisms of increased circuit jitter are many. One is Power supply noise (to the "through" circuit). Another is due to the fact that the detection threshold is in fact not fixed, it moves around slightly (semiconductor noise). There may be other issues, such as grounding.

You do not always get to optimize everything. All it takes is one device to "mis behave". But if you can it is a good idea.
I know hooking things up and setting up gear can be complicated and it is tempting to "just getting it to work"...

It is difficult to guess what is going on there. Removing the termination at the end does increase the amplitude.
1. Do you have enough WC amplitude?
2. Is the WC source back terminated? If so, end termination will cut the signal by 1/2
3. Is one of the units stuck at internal termination?  

If your combined cables are not say thousands of feet you may just try it without termination. Read what I said about WC and reflections. At those low speeds, with pretty limited rise times and and reasonably short cables, termination just does little good, while reducing the amplitude some.

BR
Dan Lavry
Logged

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: Proper word clock implementation
« Reply #39 on: October 21, 2004, 01:50:49 pm »

Albert wrote on Thu, 21 October 2004 11:48

All the units I'm chaining have switches to turn WC termination on/off. In all cases I've turned it off, except for the last unit in the chain. That's why I'm a bit puzzled.



Oh dear. But that may be a good sign. Maybe one of the termination switches is bad or sticky. It's real easy with those BNC T's to pull them off the box while watching the scope and find the culprit.

BK
Logged
There are two kinds of fools,
One says-this is old and therefore good.
The other says-this is new and therefore better."

No trees were killed in the sending of this message. However a large number of
electrons were terribly inconvenienced.

Albert

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Proper word clock implementation
« Reply #40 on: October 23, 2004, 12:35:30 pm »

Thank you for your suggestions, I'm going to troubleshoot along those lines. I don't have a scope, but I can try everything else. Thanks again, I'll report back.
Logged

jcoyle

  • Newbie
  • *
  • Offline Offline
  • Posts: 5
Re: Proper word clock implementation
« Reply #41 on: October 24, 2004, 04:00:06 pm »

Hi Dan,

Some of what you are saying is contrary to what others have told me, and I am starting to feel like my mind has been filled with CRAP to make me think I have to buy a GOOD clock.

I am a home recorder.   I send my stero mic pre signal directly to an Apogee PSX-100 stereo converter, and then send the Apogee Lightpipe signal to an Alesis HD24 Hard Drive recorder.  The HD24 locks to the PSX-100 signal of course.

This is a pretty simple setup.

I was told the best thing I could do to improve my sound to tape (HDrive) would be to clock the PSX-100 with a real clock!!. The Lucid GenX6-96 was recommended.

But you are saying the internal clock of the PSX-100 will be better than externally clocking.   As I have no multiple convertors issue, requiring an external clock.

Am I a sucker for believing others up to this point???  Is everyone just clueless, has marketing hype led everyone astray?

thanks,
joey

Logged

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: Proper word clock implementation
« Reply #42 on: October 24, 2004, 05:44:19 pm »

jcoyle wrote on Sun, 24 October 2004 21:00

Hi Dan,

Some of what you are saying is contrary to what others have told me, and I am starting to feel like my mind has been filled with CRAP to make me think I have to buy a GOOD clock.

I am a home recorder.   I send my stero mic pre signal directly to an Apogee PSX-100 stereo converter, and then send the Apogee Lightpipe signal to an Alesis HD24 Hard Drive recorder.  The HD24 locks to the PSX-100 signal of course.

This is a pretty simple setup.

I was told the best thing I could do to improve my sound to tape (HDrive) would be to clock the PSX-100 with a real clock!!. The Lucid GenX6-96 was recommended.

But you are saying the internal clock of the PSX-100 will be better than externally clocking.   As I have no multiple convertors issue, requiring an external clock.

Am I a sucker for believing others up to this point???  Is everyone just clueless, has marketing hype led everyone astray?

thanks,
joey




This is a technical forum. I stand by my statements that when possible to do the AD conversion with internal crystal, you are much better off doing so. Is everyone else wrong? Yes they are. I already elaborated about the difference between an internal crystal that is MADE FOR A FIXED FREQURNCY, vs. all the jitter sources when using  a VOLTAGE DEPENDENT CRYSTAL (+/- 100ppm at least), with a PLL circuit that hunts the average frequency (going up and down constantly), with a cable (pickup of external signals such as 60Hz), thus a semiconductor transmitter, and a semiconductor receiver… both very suceptible to supply noise...I did not say everything, there is more!

One can never argue with subjective opinions. But if less jitter is what you are after, use internal clock whenever possible.

BR
Dan Lavry

Logged

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: Proper word clock implementation
« Reply #43 on: October 24, 2004, 07:16:24 pm »

jcoyle wrote on Sun, 24 October 2004 16:00




But you are saying the internal clock of the PSX-100 will be better than externally clocking.   As I have no multiple convertors issue, requiring an external clock.

Am I a sucker for believing others up to this point???  Is everyone just clueless, has marketing hype led everyone astray?

thanks,
joey





Dan is saying that, "all other things being equal, in a well-designed converter, internal clock will always perform better than, or equal to an external clock."

You have to prove whether the Apogee is a well-designed converter! If it's a poor converter, it may very well perform better on external clock. A left-handed "compliment" if I ever saw one!

And yes, everyone is clueless... very very few people have grasped the implications of what Dan has to say. John Watkinson, among many other competent writers, myself included, have been haranguing this topic for a long time. Manufacturers, it seems are the most clueless.

The BEST article I've seen on this topic from a competent manufacturer is TC Electronic. Go to

http://www.tcelectronic.com/TechLibrary


and download the article: "Clock and Synchronization in the TC System 6000". It demonstrates, measures, and summarizes the issues for all to see that when it comes to poorly-designed converters, the Emperor has no clothes. Every external clock that makes (or seems to make ) a converter "sound better" is a bandaid, not a cure.
Logged
There are two kinds of fools,
One says-this is old and therefore good.
The other says-this is new and therefore better."

No trees were killed in the sending of this message. However a large number of
electrons were terribly inconvenienced.

Joe Bryan

  • Newbie
  • *
  • Offline Offline
  • Posts: 19
Re: Proper word clock implementation
« Reply #44 on: October 25, 2004, 01:54:01 am »

danlavry wrote on Sun, 24 October 2004 14:44

This is a technical forum. I stand by my statements that when possible to do the AD conversion with internal crystal, you are much better off doing so. Is everyone else wrong? Yes they are.


Dan is 100% correct. It's extremely unlikely an external clock can outperform an internal clock unless there are serious problems with the design.

I wrote an article about this for our webzine a while back. It's available here:  http://www.uaudio.com/webzine/june/content/content2.html

-Joe

Joe Bryan
VP Engineering
Universal Audio
Logged
Pages: 1 2 [3] 4 5 ... 15   Go Up