R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 [2] 3 4 ... 15   Go Down

Author Topic: Proper word clock implementation  (Read 100219 times)

punkest

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 63
Re: Proper word clock implementation
« Reply #15 on: October 06, 2004, 10:57:44 am »

bobkatz wrote on Wed, 06 October 2004 03:51


   I think it's a good idea you're using the Apogee's as the master clock. They're certainly going to do better on internal sync than locking to any of the other current devices in your system. And only a measurement and listening test of the Apogee's would reveal if they would do better when fed some external "high-end" clock. And if they did do better on external sync, that would be a denigration of the Apogees, not a praise of the external clock. As long as you are not using the converters of the Yamaha and mixing in the box, you are doing fine from the point of view of clocking but you certainly should expand your horizons and look into a good set of high quality outboard processing gear.




 We?ve got some nice analog compressors and we have a lot of pre?s. We usually get the sound as close as we want BEFORE A to D. We are deciding if we are going to get an analog mixer and a 16 channel D to A,  (since we?ve got two 2408 MKIII and each have 8 outputs) and go the mix out of the box route, but then we would need some more compressors, eq?s and time/modulation processors, besides extensive cabling and patch bay.

Quote:



  Your weakest link in this setup is actually your monitoring, and you can solve that by using a high quality jitter-immune external  D/A converter and monitor level control for monitoring.




  Yes, for the actual setup, I have been thinking of getting something like the Benchmark DAC-1 or Lavry D to A, and feed it from the SPDIF out of the Yamaha mixer, and get a Coleman audio to control level and speaker sets.

Quote:



  The 2408 goes digitally into the Yamaha, and is processed digitally, so jitter is irrelevant there, and thus only the monitor DAC is in trouble, and the Yamaha's DAC is probably suffering there from point of view of jitter.




  I guess the new D to A would take care of this

Quote:



  Where you are going to run into trouble is when you introduce external analog processing gear. And for that, I would recommend that on each analog insert you use the digital outs of the Yamaha to feed higher quality external converters. A good approach would be to replace the 2-channel Apogee with a single stable 8 channel A/D/A system running on internal sync and proven to perform best that way.




  The apogee we have is 16 channel A to D, and we have as well 16 channel D to A from the 2 MOTU 2408MKII, I guess it is not fancy, but it'll have to do while we get better, and I prefer the MOTU D to A than the Yamaha 03d, so when I go external I go out of the MOTU to processor to Apogees and back to the box.

Quote:


 
That's my take on the affair!


Bob, again,  thanks for the insight.

Hans Mues
Logged

Albert

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Proper word clock implementation
« Reply #16 on: October 06, 2004, 12:20:40 pm »

Bob and Dan, thank you for your very helpful and interesting replies. I've learned a lot from this thread and there is a lot more to think about.

Perhaps I missed this, but how many bnc t-bars can you put in a row before the signal degrades, or is this a concern at all?

As luck would have it, some of my gear does indeed have WC termination on/off switches, so I am going to try out this bnc daisy chain thing. I just bought a bunch of bnc t-bars and terminators on eBay! Smile

Bob, of particular interst to me was your comment regarding daisy chaining anything with a converter, but using a dsitribution box for digital to digital units. I had not heard that before, and in fact believed that jitter affected all digital gear equally. Apparently (or obviously) I was incorrect in that.

Incidentally, I am still enjoying "Mastering Audio", it has become one of a handful of essential reference books for me. Thank you again for sharing your experience and knowledge.
Logged

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: Proper word clock implementation
« Reply #17 on: October 06, 2004, 03:04:10 pm »

Albert wrote on Wed, 06 October 2004 17:20


Perhaps I missed this, but how many bnc t-bars can you put in a row before the signal degrades, or is this a concern at all?



As I stated earlier in the thread, the issue of termination is much more impotent for higher frequency digital data (such as AES) than lower frequency (such as WC).

Lets me show an example with some numbers:

Say we have 1000 feet of coaxial cable (very long). That is 800nsec delay. Say we are working with a square wave WC at 96KHz. So a cycle time is about 10uSec. The time between transitions is about 5uSec (assuming a square wave). So how may reflections will occure between 2 transitions?
10usec/800nsec = 12.5 But since we need to account for twice the cable length (reflections travel from source to load and back to source), so we have over 6 round trips.

Now, lets get practical and assume a 5% cable to termination mismatch. So the first reflection yields 2.5% (half the mismatch), the second reflection is 2.5% of 2.5% = .0625%. The third round trip yields 2.5% of .0625%... and at 6 round trips we have 2.4 pico (a millionth of a millionth) of the original step. Say the step was 5V, the reflection activity is truly tiny 1.2nV (nano volt). And that is for 1000 feet of cable.

So the point is, proper termination is almost a joke at WC speed. The same 1000 feet cable with say 1MHz frequency (instead of 96KHz) will be a different story. The same exercise with 1uSec/800nSec = 1.25 - a new clock transition
is being sent down the cable before the first reflection had enough time to go through a round trip! So the first reflection which is 2.5%. or .125V "rides on" the 5V signal... That will much things up... That is at 1MHz, imagine faster rate such as AES signal.

Reflections and proper termination are about cable length vs period (frequency). The whole concept of proper termination plays no role at very low frequencies, where audio folks look at a cable as a capacitive medium. 44.1KHz-48KHz WC is still very low period about 20uSec. The signal contains high frequencies (fast rise and fall time square wave), but it is the basic period that counts for reflections.

So what is the bottom line? Don't worry about WC reflections. If you want to "be pure", go for short interconnects say 3 feet between units, and the first cable leading to the units can be long. That is all for WC, not for AES. How many units? A pile of units! Assume that each unit input looks like 10KOhm in parallel to say 20pF. Say you have 100 units hooked, thus a distributed load 100 Ohms and that will cost you some in amplitude, but still work fine. The capacitance will muck the purity of the 75Ohm line and so your starting point is no longer 2.5%. Say it is 25%, than the 1000 feet cable example will yield 4mV activity on the next edge. So yes 100 units is potentially on the hairy edge. I would not hesitate to have 25 units.

You could not hear the difference between the system with or without the termination resistor, that is in a double blind test. You would hear it with AES signal.

BR
Dan Lavry

   
Logged

Albert

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Proper word clock implementation
« Reply #18 on: October 06, 2004, 04:17:36 pm »

Thanks for that explanation Dan. Especially the last couple paragraphs, which were simple enough for me to understand. I'm a composer/performer not an engineer, though I do try to educate myself as much as possible. The whole concept of reflections is new to me, so that will require some more thought for me to understand it.

I'll probably have around 15 pieces on one WC chain, so that seems to be well within the safety margin. The daisy-chaining approach simplifies things in my studio and actually saves me some money I was going to spend on another WC distribution device.

Dealing mostly with midi devices for so many years, I just assumed *any* daisy-chaining would be a bad thing. For midi devices, it certainly is. You can't go more than a few units and then you are in trouble.

Great thread, thanks again for the advice.
Logged

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: Proper word clock implementation
« Reply #19 on: October 06, 2004, 06:42:07 pm »

Albert wrote on Wed, 06 October 2004 12:20



Bob, of particular interst to me was your comment regarding daisy chaining anything with a converter, but using a dsitribution box for digital to digital units. I had not heard that before, and in fact believed that jitter affected all digital gear equally. Apparently (or obviously) I was incorrect in that.





Hi, Albert! Thanks for your very wonderful comments on my book. Well, to reiterate, jitter is ONLY important with converters! And a good converter is supposed to be immune to incoming jitter, as Dan has eloquently pointed out. So this means that you can use Dan's T-Bar strategy to minimize jitter in the feeds to critical gear that contains converters, and use a cheap distribution box to feed WC to gear that doesn't matter.

Hope this confirms your thoughts!

By the way, I'm not trying to hijack this forum away from Dan, it's just that on the subjects that I know a little something about, I do like to contribute! And at 130 words per minute typing, my thoughts tend to go virtually direct from brain to fingers to Internet   Smile


Take care,


Bob
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.

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: Proper word clock implementation
« Reply #20 on: October 06, 2004, 06:46:56 pm »

Albert wrote on Wed, 06 October 2004 16:17



Dealing mostly with midi devices for so many years, I just assumed *any* daisy-chaining would be a bad thing. For midi devices, it certainly is. You can't go more than a few units and then you are in trouble.

Great thread, thanks again for the advice.



Well, it's not exactly "daisy-chaining". Unlike MIDI-through, which does add a delay, all you are doing with a BNC-T is putting a Y cord with a little load on it. Think of it like a big river flowing downstream, and little pipes on the side of the river taking a tiny tiny little drop out of the river at each pipe. That's effectively what the high-impedance connection is doing.

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.

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: Proper word clock implementation
« Reply #21 on: October 06, 2004, 07:06:54 pm »

"I'm a composer/performer not an engineer, though I do try to educate myself as much as possible. The whole concept of reflections is new to me, so that will require some more thought for me to understand it."

I understand that in a forum one has a mix of interests, from the very technical to the very musical and all the in between. It is hard to talk to all at the same time, so I am just trying my best, and hope that it is understood.

There are a lot of EE's that do not know about reflections. Certainly the computer science types have little reason to get into it. On the other hand, designers of very high speed gear are heavily into transmission line theory. Imagine a 5GHz digital data transfer - that is 200pSec (pas is a billionth of a second). An electrical signal travels about 1.3 inch in 200pSec...
It is a bit "funny" for me to see how a concept such as proper line termination is followed at say 44.1Khz, but a line level cable with 22KHz does not call for impedance matching between audio cable and some temination. The practice is almost "brainless impulse", you see a coax, and you terminate.
The coax is good because of the nice shield, and the locking connectors are nice, but that is all, for audio WC.

I'll probably have around 15 pieces on one WC chain, so that seems to be well within the safety margin. The daisy-chaining approach simplifies things in my studio and actually saves me some money I was going to spend on another WC distribution device.

Save money and improve jitter performance at the same time.

Dealing mostly with midi devices for so many years, I just assumed *any* daisy-chaining would be a bad thing. For midi devices, it certainly is. You can't go more than a few units and then you are in trouble.

Of course midi is a different thing all. I do not know where midi basic clock rate is these days, it used to be 500KHz, than 1MHz... and it takes many clocks to construct a midi command (all the data is in series)... So you end up with serious delays, not to mention the ability to put too much data on the buss (especially with continuess controllers such as after touch, pitch bend and so on). So midi problems can be in the milliseconds. WC over cable is about 1.5 nano second per foot, or 2/3 the speed of light!

BR
Dan Lavry
Logged

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: Proper word clock implementation
« Reply #22 on: October 06, 2004, 07:50:24 pm »

[quote title=Ozzie Bostic wrote on Sun, 03 October 2004 15:58
Hello and Welcome Dan,
I am currently enjoy the wonderful sound and clarity of your Blue series converters great product.
My question is since your not a fan of distribution clocks; what is the maximum number of devices that should be connected in series via the BNC T method mentioned above before the clock source is degraded?
Thanks in advance.[/quote]

Hi Ozzie,

Well, it took me a couple of days to try and figure how to answer it. I hope my latest posts with all the details are not "too much".

Assuming you are talking WC sync:

It is difficult to put a number on it, but I do not see a any reason why one can not hook 25 devices in series. You have to be sure that all the devices (except the last one)are not internaly terminated.

If the last device is internal terminated, do not terminate the last BNC T. In fact you do not need the lase BNC T and can plug directly.

If the last device is not internally terminated, you can put a BNC-T at the end and terminate the line there, at the last BNC.

If you are doing AES sync, I would go for a lot fewer devices. At that point I would want to know some things (such as line length, what sample rates are used...)

BR
Dan Lavry
Logged

12345

  • Full Member
  • ***
  • Offline Offline
  • Posts: 140
Re: Proper word clock implementation
« Reply #23 on: October 07, 2004, 08:52:31 pm »

Hello All,

Great topic!  

When do you recommend using a grounded (with a grounding chain/wire attached) BNC T, and when do you recommend using an open one?  

Regards,
My World
Logged

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: Proper word clock implementation
« Reply #24 on: October 08, 2004, 12:48:36 pm »

My World wrote on Fri, 08 October 2004 01:52

Hello All,

Great topic!  

When do you recommend using a grounded (with a grounding chain/wire attached) BNC T, and when do you recommend using an open one?  

Regards,
My World


BNC is for coaxial cables. The cables have one "center conductor" and a shield (typically braided). So it is aimed at sending signal on the center and the shield is for ground return path.

The question is which ground to connect the cable to. That is a system grounding question.

The BNC female connectors (what you plug the coax cable to) come in variety of configurations. You can get an all metal connector, designed to short your return to chassis ground if mounted on a metal chassis. Or you can get a BNC connector that does not short to a chassis. I am not sure why some come with a grounding chain and cover. My guess is for protecting the open connector in harsh environments. It is only a guess.

BR
Dan Lavry    

Logged

Ozzie Bostic

  • Newbie
  • *
  • Offline Offline
  • Posts: 38
Re: Proper word clock implementation
« Reply #25 on: October 08, 2004, 02:35:28 pm »

[quote title=danlavry wrote on Wed, 06 October 2004 18:50
Hi Ozzie,

Well, it took me a couple of days to try and figure how to answer it. I hope my latest posts with all the details are not "too much".

BR
Dan Lavry
[/quote]

Yes Dan, the previous post was exactly what I was looking for. I'm actually an EE and IT Major by degree and find your posts absolutely astounding. I am kinda sorry I didn't do more EE work, but more IT related. At any rate, great post.

Thanks again,

Logged
Ozzie Bostic
Sandstorm Entertainment, Inc.

Albert

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Proper word clock implementation
« Reply #26 on: October 09, 2004, 12:32:28 pm »

Dan and Bob, thanks again for your very illuminating posts, I appreciate it very much.

Okay, from my non-technical point of view, how would the pictured arrangement work as far as distributing WC? My question being, if it is possible to string 25 units in a WC chain, what if those same units were spread out on 3-4 shorter chains that originated from the same place?  The BNC t-bar configuration at the originating WC device would look like this:

http://www.albypotts.com/images/bnctbarx3.jpg

Crazy? Does that put too much of a load on the signal? Let's say there are the same number of devices as in a straight WC chain, would there be four times the load with this setup?

As you can see, I got my shipment of BNC t-bar connectors. Smile
Logged

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: Proper word clock implementation
« Reply #27 on: October 09, 2004, 04:36:50 pm »

Okay, from my non-technical point of view, how would the pictured arrangement work as far as distributing WC? My question being, if it is possible to string 25 units in a WC chain, what if those same units were spread out on 3-4 shorter chains that originated from the same place?  The BNC t-bar configuration at the originating WC device would look like this:

Crazy? Does that put too much of a load on the signal? Let's say there are the same number of devices as in a straight WC chain, would there be four times the load with this setup?

As you can see, I got my shipment of BNC t-bar connectors. Smile[/quote]


From a purist stand point, don't do it - avoid branches. A transmission line (the coax cable) is a constant impedance line. Say you apply a 1V sudden step on a 50Ohm line. The “good old electrons” do not know what is ahead. They “can not know” how long the cable is, before they got to the end. So, the line at first look like a resistor. That 75Ohm line looks like a 75Ohms initially. So the current initially is 1V / 75 Ohm = 13mA and that current wave is traveling down the cable, and wherever it goes, what is ahead is “the rest of the cable”. Again, the electrons just see 75 Ohms.
At some point soon the wave (goes at about 8nSec/foot) gets to the end of the line. What does it see?  If the termination is proper, it sees a 75Ohm resistor. All that current (13mA) now goes through the resistor, thus 13mA*75Ohms=1V.
You can now see how a reflection can happen with wrong termination. Say, you terminated with 100Ohm resistor. The 13mA will “meet” an 100 Ohms thus we have 13mA*100Ohms=1.3V !!!  Now the load has .3V higher voltage than the source, so you have 4mA current going backwards…. It ends up as a lot of back and forth reflections.

The point is: The pure way to deal with transmission line is to have everything in series, and a correct termination at the end. If you split the line with a T into 2 long cables, that same 13mA current will divide with 6.5mA going to each cable. For “a while” (till the waves get reflected from their respective coax ends) each line has only 1/2V… The whole concept of impedance matching is disturbed. Where is the termination? On line A or B?...

For WC you can get away with some short branches if it is really necessary, a few feet here and there. I suspect it will work fine for WC whatever you do (say less than 1000feet), but why “push it” if you can go in series or as close to series as you can? I like to do things as cleanly as possible, when I can.

BR
Dan Lavry  
Logged

Kendrix

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 988
Re: Proper word clock implementation
« Reply #28 on: October 13, 2004, 12:32:45 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.

In my case I have a Yamaha Aw4416.
It comes with 16 AD's on board and I added an I/O card wtih an additional 8 AD's.  How the "internal" clock gets distibuted to all these converters is not clear to me from the schematics provided with the unit.  However, it is clear that to transmit the internal clock to the add-on I/O board requires the clock signal passing over some distance and through an interface to the board.  

To be honest when I A/B  the Yamaha internal clock with a genx WC supplied via BNC ( I use each clock for both AD and DA conversions for the comparison) I do hear differences.  However im unable to clearly judge which is most accurate.  Ive used challengeing signals, such as a bell recorded via the same chain cept for varying the clock- and trying to judge which clock yeilded a sound closest to the actual bell ( difficult to do).  When converting and playing back a well recorded piece I have no objectivre reference of what the "correct" sound should be.  Depending on the material there can be plusses and minuses to each clock.  Generally the external clock seems a bit more defined and 3D but my monitoring environment (KRK7000's in a smalll room with some treatment for reflections) is marginal for making such fine judgments.  Reconciling this observation with the statement that an internal clock should almost always be better than external clock is bothering me.  Again, the details of how the internal clock gets distributed to the 24 individual converters in my system is unclear.

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.  

(I Hope all the above was coherent enough to allow for meaningful responses).

Logged
Ken Favata

bobkatz

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2926
Re: Proper word clock implementation
« Reply #29 on: October 13, 2004, 07:49:09 pm »

Kendrix wrote on Wed, 13 October 2004 12:32



To be honest when I A/B  the Yamaha internal clock with a genx WC supplied via BNC ( I use each clock for both AD and DA conversions for the comparison) I do hear differences.  However im unable to clearly judge which is most accurate.





You might be trying to choose the lesser of two evils, or between two "goods" each of which have some minuses, Kendrix.

I doubt that measurement would settle the issue; you'll probably see jitter artifacts at different frequencies with each implementation. But then again, you might think that one is more euphonic than another or hiding something better. If the music has a lot of distortion and the jitter is random noise, the additional noise could be masking the distortion! In that case you may be preferring "more jitter". Regardless, I tend to NOT try to let clocking drive my thinking as this will prevent you from having your ultimately good sound with a good source. Instead, I would try to find the clocking which is the most accurate with a very pure and good-sounding source. Then, if you think you need to "degrade" your sound with noise or whatever when you hear a lesser source through the better clocking, keep the better clocking and try to fix your source up with some other technique (equalization, distortion generation, compression, etc.)

Let me try this:

1) Which one has more depth?

2) Which one is warmer, or tonally more accurate?

3) Which one is wider and deeper?

I'll bet you dollars to donuts that one will be the clock with the least jitter!

Hope this helps,


Bob
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.
Pages: 1 [2] 3 4 ... 15   Go Up