DACs with word clock sync

[Andy Peters]

What is the current thinking about syncing a two-channel DAC to incoming word clock? Is it even necessary?

I'm doing my own two-channel ADC and DAC box and the schematic is mostly complete. It has word clock in for the ADC section (which is necessary, I think, for situations where you want to sync two different AD C units) and I've been going around trying to decide whether it matters for the DAC side. I asked a couple of studio friends, including a mastering guy, what they thought, and they all said, "we don't bother." This is obviously a small sample, and it's also something I won't need.

(Why my own DAC/ADC box? I suppose the better question is, "why not!")

 

[[silent:arts]]

I can't think of any situation where you need it technically.
The audiophile may want to hear the better clock they paid $$$$ for.

 

[JohnRoberts]

The analog and digital glue circuitry around the codec can surely make a difference (but analog more than the digital I suspect)., what are you using for convertors?

JR

PS: I dabbled with the notion of making one from scratch, but I'm not smart enough (or rich enough). I wanted to float a high performance A/D up to 48V for a DC coupled mic preamp, just to see how good it could be done... of course this ignores the crap circuitry inside microphones.  :

 

[Andy Peters]

The analog and digital glue circuitry around the codec can surely make a difference (but analog more than the digital I suspect)., what are you using for convertors?

The top-end TI parts. I don't think I can get the ESS Sabre things in quantities less than 1,000.  It's line-level in and out, no preamps or any of that.

The thing is, I have a use for a good stereo DAC and ADC, but there's no middle market between the $100 cheap things (without WC sync on the ADC, which I need) and the $1,000+ products. And I know I can spin something for $100 in parts that'll be equal to the $1,000 units. (Because that's what my BOM seems to be right now, sans enclosure.)

I figure that the cost to add WC sync for the DAC side is minimal, but I don't see a use case for it.

 

[abbey road d enfer]

What is the current thinking about syncing a two-channel DAC to incoming word clock? Is it even necessary?

I would think there are cases where it would be detrimental. Syncing two DAC's that receive digital signals from two different (non-sync) sources is like having no sync at all.

 

[Andy Peters]

I would think there are cases where it would be detrimental. Syncing two DAC's that receive digital signals from two different (non-sync) sources is like having no sync at all.

No doubt true.  But is it common?

Thinking a different way: you have two stereo DAC boxes, both fed with AES3. If the sources are different, then I think sync at the sample level might be beside the point. What is skew between them? This is an interesting question. I suppose I could do a test in Logic, with two audio devices aggregated (a great feature of Core Audio on the Mac), and send the same signal to the two devices and look at the analog results on an oscilloscope. 

The cost to implement WC sync isn't much, just some muxes.

 

[abbey road d enfer]

No doubt true.  But is it common?

One instance I can think it would happen is just comparing a test CD with the original mix; one DAC must be synced to the mastering chain, the other to the CD player.

 

[Monte McGuire]

Each stereo digital audio feed has its own framing and clocking, so each is able to be independently clocked without imposing the need for a master clock. So, one could simply switch among independent digital signals with some sort of muting to prevent switching glitches as the PLLs adjust to the new signal. Sure. it's nice to have a synchronous room, but if that were the case, simply switching among synchronous sources is really the same thing as switching among free running sources, except for how much glitch energy could be expected. Since one would have to handle that in any case, there should be no need to expect or require separate stereo digital sources to be locked, and thus require a word clock.

As long as clock recovery is good, word clock is not needed. However, one could argue that it's easier to lock onto WC than it is to lock onto a digital signal, but I think that problem is not so difficult now.

 

[ruairioflaherty]

I've been a mastering engineer for 15 years or so and I've yet to feed a DAC with word clock.  For 2 channel work I want the DAC to follow whatever I feed it, my mastering rig or Spotify for example.  Assuming the clock recovery is good and/or you are reclocking locally then the practical difference is zero in my world.

Naturally in a multitrack recording scenario we do prefer to have everything locked.

 

[[silent:arts]]

There are two channel DACs converting the incoming sample rate to the "high accurate" internal clock too.

 

[Rogy]

Hi,

Extracting embedded bit clock from an AES or SPDIF stream does contain more jitter than using a dedicated external word clock, unless there is no audio modulated over the AES/SPDIF interface.

Look up 'pattern dependent jitter' or 'intersymbol interference', eg page 5 of PDF in link:

http://www.audiophilleo.com/zh_hk/docs/Dunn-AP-tn23.pdf

Best Rogy

 

[Andy Peters]

I suppose I should clarify:

I know that S/PDIF receivers have fairly yecchy recovered clocks, so I will use an SRC to decouple the receiver from the DAC chip, clocking the converter from a quality local oscillator which also provides the SRC reference clock.

So it's really not a question of using word clock because modulator clocks synthesized from word clock are "better," it's a question of "does anyone sync more than one reference DAC together," for example to give six outputs for 5.1 surround? (If I was doing surround, I would use a six- or eight-channel DAC box with USB or FireWire or Thunderbolt input and not futz around with synching three stereo DACs together.)

 

[ruairioflaherty]

So it's really not a question of using word clock because modulator clocks synthesized from word clock are "better," it's a question of "does anyone sync more than one reference DAC together," for example to give six outputs for 5.1 surround? (If I was doing surround, I would use a six- or eight-channel DAC box with USB or FireWire or Thunderbolt input and not futz around with synching three stereo DACs together.)

I've never seen it, as you say if people are doing more than stereo they'll have dedicated 8 or 16 channel D/A setups.

Which SRC are you using?

 

[Andy Peters]

Which SRC are you using?

TI SRC4392, with a Silicon Labs ARM for configuration and other stuff. When the ARM is as cheap as an 8051 ... Also, the ARM has an I2S port, so the converted data from the ADC can drive it and I can do metering digitally.

 

[ruairioflaherty]

TI SRC4392, with a Silicon Labs ARM for configuration and other stuff. When the ARM is as cheap as an 8051 ...

Sounds good!  I'm in the middle of a huge lab clean up and purge, I came across an eval board for the 4192 and it went on the maybe go pile...

 

[Andy Peters]

Sounds good!  I'm in the middle of a huge lab clean up and purge, I came across an eval board for the 4192 and it went on the maybe go pile...

The main difference between the 4192 and the 4392 is that the latter includes the AES3 transmitter and receiver. Saves a couple of chips.

 

[Matador]

Hypothetically:  you have 32 DAC chips, all running a local clock.  All are outputting a reconstituted analog waveform to 'outside'.

Even if there were drastic differences in the local PLL reference clocks inside those chips, how could you possibly tell from the output waveforms on those channels?  How could anything downstream of them ever figure out the difference?

Or did I completely misunderstand the issue? (which has happened countless times before)

 

[Andy Peters]

Hypothetically:  you have 32 DAC chips, all running a local clock.  All are outputting a reconstituted analog waveform to 'outside'.

If those 32 DAC chips are all in the same box, they'd all run on the same clocks. All synchronous.

Even if there were drastic differences in the local PLL reference clocks inside those chips, how could you possibly tell from the output waveforms on those channels?  How could anything downstream of them ever figure out the difference?

if each DAC was running at the same sample rate, with only phase difference between them, then worst case your variance is one sample. I think you'd be hard pressed to notice that with music sources, but you could contrive a test case and send the same signal to two different DACs and sum them you could see the result on a 'scope.

Or did I completely misunderstand the issue? (which has happened countless times before)

I think you did

 

[abbey road d enfer]

If those 32 DAC chips are all in the same box, they'd all run on the same clocks. All synchronous.

Matador made the assumption that those DAC's would run each on a different "local" clock; it would be a strange architecture, but it is quite thinkable. Unless these DAC's receive un-sync'ed digital signals, these "local" clocks are not totally uncorrelated, though. They have to be sync'ed to each individual incoming digital flux.
The possible degradation to the reconstructed analog signal is not more than significant than when only one DAC is involved, which means that a possible difference may exist between two versions of an identical digital signal. Indeed, the way the MCLK is extracted will have significant effects on this difference, particularly SRC.
In practice, these sync issues matter only when processing digitally.