Topic: Syncing a V3 Word Clock output with a DMIC-20 AES REF input

[gratefulphish]

As it was explained to me by Edirol's techs, as long as all four channels are being recorded on the same deck, they are going to be in sync.  I don't believe that the resampling, or not, has any effect on that, since the four channels are being recorded on four tracks, on the same machine, at the same time. I could be wrong, but that is what I was told.

[trajhip2000]

funny, I was kinda wondering about that - since the files all start and stop at the same time you'd think that would be true. however,  I just remembered a thread on the Sound Devices board where I was asking about feeding two channels digitally from a V3 to a 744T and also using the WC from the V3 to sync the 744T (I even bought a WC cable). The SD support guy said that even though the clocks would be digitally in sync, there is a difference in something called the "group delay" (iirc) between the A/D chips in the V3 and SD that would mean they won't be exactly time-aligned - so even tho the files would start and stop at the same time and their clocks would be synced, the data streams wouldn't hit the recorder at the same time. my recollection is that we're talking about a very small difference, I believe he gave us a formula based on sampling rate, I never did any calculations - but I think we're talking about low msec at most. I guess this principle would apply to most of the previous discussion on this thread. certainly the most important thing to have is the clocks synced, that way you only have to worry about aligning the whole files as opposed to individual songs (potentially) if the clocks are mis-matched.

Steve

[Todd R]

As it was explained to me by Edirol's techs, as long as all four channels are being recorded on the same deck, they are going to be in sync.  I don't believe that the resampling, or not, has any effect on that, since the four channels are being recorded on four tracks, on the same machine, at the same time. I could be wrong, but that is what I was told.

Was he just talking in general, or was he referring specifically to the R4 or specifically to the R4 Pro?  This ^ makes sense when talking about the R4, but doesn't seem right about the R4 Pro.  If you feed 2 digital channels into the R4 along with 2 analog channels, the R4 will resample the digital channels using the same clock as is being used on the analog channels -- so all should be in sync (though not perfectly time aligned).  But in the case of the R4 Pro, the digital channels are not being resampled and will be sampled off the clock chip of the V3, not the clock of the R4 Pro, which will be sampling the analog inputs.  Bottom line, in the case of the R4 Pro, the digital channels are being clocked by one wordclock, and the analog channels are being clocked by another.  So it doesn't seem to me that in this case the digital channels and analog channels will be in sync.

[TNJazz]

Bottom line, in the case of the R4 Pro, the digital channels are being clocked by one wordclock, and the analog channels are being clocked by another.

I don't believe this is possible - you can only have one master clock in a digital system.

From what I've been reading it seems that word clock is being confused with sampling rate.  Word clock does not carry audio, it merely emits a signal based on the sampling frequency.  It's a speed reference (sync) that tells the recording device how fast to run.  Resampling the bits on digital input doesn't really apply - they are still running off the same clock so they will be at the same speed - they have just been resampled but their speed has not changed.

So if a device is set to receive word clock input (external) the analog channels should run at the same clock speed as the signal coming in.  Whether they are time aligned or not is another matter entirely, but they will be in sync.

[Todd R]

Bottom line, in the case of the R4 Pro, the digital channels are being clocked by one wordclock, and the analog channels are being clocked by another.

I don't believe this is possible - you can only have one master clock in a digital system.

From what I've been reading it seems that word clock is being confused with sampling rate.  Word clock does not carry audio, it merely emits a signal based on the sampling frequency.  It's a speed reference (sync) that tells the recording device how fast to run.  Resampling the bits on digital input doesn't really apply - they are still running off the same clock so they will be at the same speed - they have just been resampled but their speed has not changed.

So if a device is set to receive word clock input (external) the analog channels should run at the same clock speed as the signal coming in.  Whether they are time aligned or not is another matter entirely, but they will be in sync.

You might be right, but that is not my understanding of things.  I'm not sure you're following me, but I'm not confusing wordclock with sampling rate -- the wordclock is what the system is using to generate its timing in the internal circuitry -- it is an actual clock: rising clock edge makes things happen with the circuitry, falling clock edge is ignored, 1/44,100 of a second later and there is another rising clock edge and the logic chips react accordingly.  Your sample rate can be at 44.1k, but you still need to have a system clock to make that happen.  Gear will have that clock onboard, and depending on design, can take the wordclock of an external system, which is a lot of what we were discussing.

From the initial discussion, the Dmic can take a wordclock reference from an external source and will use this for clocking information.  In this way, the DMic will have exactly the same clock timing as the reference gear (say a V3).  This is akin to your final sentence -- getting an external reference wordclock will insure that all channels are being recorded from the same clock. (I'm sure you know it, but 44.1k is not 44.1k, is not 44.1k.  Each sampling of 44.1k will be dependent on the particular system clock that generates that sampling rate.  Even though two different digital streams have a nominal sampling of 44.1k, they both will be ever so slightly different unless the sample rate was clocked from the exact same reference clock.)

The issue about the R4 is different though than the DMic, since the DMic accepts an external wordclock reference and the R4 does not.  So this might be where my confusion comes in, but this is my understanding:  Digital recorders taking a digital signal can either accept that digital signal exactly (including the inherent timing information embedded in that signal) -- this is the bit perfect digital recording we're talking about, or the recorder can re-sample that digital stream, which makes the recorder not bit perfect.  When the recorder does this, it resamples the digital stream based on its own internal clock, so the now resampled digital stream no longer has the exact timing of the original digital stream -- it is now timed by the internal system clock of the recorder.

So this gets back to the issue with the R4 vs the R4 Pro. The R4 will resample the incoming digital stream using its own internal clock, the same clock that is used to generate the timing of the A/D conversion on the analog inputs -- thus all 4 channels have now been clocked by the same clock.  In the case of the R4, it does not resample the incoming digital stream.  So this stream has been clocked by the V3 that fed the digital stream.  The analog inputs though will have been clocked by the system clock of the R4, so they will not be completely synced to the digital input, which was clocked by the V3.

Reading this over, I think this might be part of the confusion:

Word clock does not carry audio, it merely emits a signal based on the sampling frequency.  It's a speed reference (sync) that tells the recording device how fast to run.

The wordclock is not just a reference to tell the recording device how fast to run.  It is an actual digital clock (think of a squared off sine wave) that is required in almost all digital systems using ICs.  The clock doesn't provide a just a speed reference telling something how fast to run, it actually provides the exact timing of running the system.  So it isn't a syncing mechanism, or at least I find that language pretty confusing.

[TNJazz]

OK, maybe I'm not understanding this completely.

Here is what MOTU says about word clock:

"Word clock is a time base reference only, providing a measurement of the passage of time and the speed at which samples should go by."

They are referring to time and speed, but resampling of the digital input should have no effect on either of these variables to my way of thinking.  It simply changes 1's to 0's and vice versa.

I'm trying to understand why there is a need to sync two devices to a single clock when the third device supposedly utilizes its own clock anyway (the R4, according to all the statements above)?  Would clocking the V3 and the DMIC not be useless in that case?

I'll go back into lurk mode for a while now...

[trajhip2000]

resampling is really not much  different than what happens in your original A/D conversion, so it has to be tied to a clock. when you resample that's controlled by the clock in the resampling device, so you lose the incoming clock and the timing basis is now the new clock - unless the resampling device can take it's clock from the input stream, which the Edirol may or may not do, I don't know.

this is giving me a headache...

Steve

[Todd R]

I'm trying to understand why there is a need to sync two devices to a single clock when the third device supposedly utilizes its own clock anyway (the R4, according to all the statements above)?  Would clocking the V3 and the DMIC not be useless in that case?

There is no need.  There are two different discussions going on in this thread.  One is how Chuck can use the V3>recorder to record two channels, and then record 2 more channels of mics>DMic20>second recorder, with the DMic externally referenced to his V3.  This will enable him to easily make 4ch mixes.  The R4 discussion has nothing to do with any of this.

The other is a discussion about how the R4 and R4 Pro handles clock timing, particularly in regards to recording 2channels digitally and 2ch analog.

On the other point, Steve has got it -- when we say a digital input resamples, it has to resample based on some clock.  What happens is that it will be resampling based on its internal clock, which will be different from the external clock of the original A/D that created the digital feed.

So all of this is a total aside from Chuck's original thread, but may be very important to users of the R4 Pro.  That is, it sure looks like they're going to be totally screwed if they want to do a 4ch mixdown to 2channels if they use 2 digital channels on input and two additional analog channels.

[TNJazz]

There are two different discussions going on in this thread.

 

That's why I couldn't figure out what was going on.  I'll go back and re-read it with that in mind.

[Chuck]

I'm trying to understand why there is a need to sync two devices to a single clock when the third device supposedly utilizes its own clock anyway (the R4, according to all the statements above)?  Would clocking the V3 and the DMIC not be useless in that case?

There is no need.  There are two different discussions going on in this thread.  One is how Chuck can use the V3>recorder to record two channels, and then record 2 more channels of mics>DMic20>second recorder, with the DMic externally referenced to his V3.  This will enable him to easily make 4ch mixes.  The R4 discussion has nothing to do with any of this.

The other is a discussion about how the R4 and R4 Pro handles clock timing, particularly in regards to recording 2channels digitally and 2ch analog.

On the other point, Steve has got it -- when we say a digital input resamples, it has to resample based on some clock.  What happens is that it will be resampling based on its internal clock, which will be different from the external clock of the original A/D that created the digital feed.

So all of this is a total aside from Chuck's original thread, but may be very important to users of the R4 Pro.  That is, it sure looks like they're going to be totally screwed if they want to do a 4ch mixdown to 2channels if they use 2 digital channels on input and two additional analog channels.

I'm really enjoying this discussion. I'm learning and re-learning things I've only, really, had a general grasp of. This stuff is important to know, especially when working with several digital devices. Thanks guys.

[Chuck]

Good news!

I got the DMIC-20 this week and did some modifications to it. Some to improve performance, some are meant to make it more usable in the field, at least from my perspective.

- Replaced the phantom power supply switch (back panel) with a lower profile switch
- Added low profile, power switch (front panel) above power LED
- Removed the BNC connector (front panel) and replaced with gold RCA jack
- Removed the AES Ref BNC connector (back panel) and replaced with gold RCA jack
- Removed the AES XLR output (back panel) and replaced with Toslink (optical) output*
- Added (2) 5k ohm log taper pots for variable gain between 5 db steps (front panel)
- Replaced the (2) SSM2017 original instrumentation op-amps with (2) INA217

* this made it possible to make a recording from the DMIC-20 into an iHP-140 optical input
My V3 already has the optical output modification.

Today, I got to test it. 

This is what I did...

- While I was doing the modifications to the DMIC-20, I tested it along the way, to ensure that it was functioning properly. So I tested the box after I was done and was happy to find it worked properly.
- I ran a 75 ohm coax cable (RCA to RCA) from the V3 SDPIF output to the DMIC-20 AES Ref input (Thus "syncing" the V3 sample rate reference to the DMIC-20)
- I set the V3 to 44.1 kHz, 16 bit
- I connected a CD player analog output into the DMIC-20 XLR inputs (with attenuator cable)
- I recorded, with 2 microphones ~ one foot away from my monitor speakers, the same audio into the V3, XLR inputs
- I played several compact discs and recorded for ~4 hours (the iRiver's automatically split the files at 2 GB) into an iHP-120 (V3) and iHP-140 (DMIC-20) using the optical inputs
- I then imported both stereo files into CuBase. I aligned the beginnings of the audio files visually (and sonically)
- I then looked at the end of the files and they still lined up perfectly (in phase too!) There must be a phase reversal in either my playback system or one of the pre-amps? I listened to the beginning and end of the files, muting and unmuting each stereo recording listening for delay or any other anomolies. I heard none. 

I can conclude, from my test, that running the SPDIF output into the AES Ref input on the DMIC-20 does make it possible to make two recordings (through the analog inputs on both pre-amps) at the exact same sample rate, thru the two pre-amps to two separate recorders. The files matched up visually and sonicaly after 4 hours record time.

This is very good news, for me. Now, I have a four channel recording system, on the cheap.


BTW, I really want to thank Richard (poorlyconditioned) for his generous sharing of information on doing the optical modification to the DMIC-20. With his help I was able to do that mod in less than an hour with less than $3 in parts!

[rowjimmy]

+t
Did you take any pics?

[Chuck]

I'm not done yet, but I can take some pictures...

[rowjimmy]

Just curious how everything looks. I wondered how/where you would fit the toslink connector in there but, seeing that you removed the AES-XLR output, there's probably lots of room.

If only one could toss those dip switches...

[Chuck]

Here are some photos I took this morning. I originally had the optical output mounted on a Neutrik DBA cover plate, but I botched the cut out on it and have reverted to a piece of scrap aluminum I had until i can order another Neutrik cover.