Word clock distribution amp propagation delay
- Thread starter audiox
- Start date
Help Support GroupDIY Audio Forum:
Walrus
Well-known member
If your WC DA is the first device after your WC, then it doesn't matter how long the delay is, as everything being fed by the DA is getting the same timing.
Walrus
Well-known member
Yes, you are quite correct. I'd overlooked that scenario. (spoilt by years in big studios with separate master clocks)
I would have thought that any decent WC DA will have neglible delay anyway. Even cheap ones would be hard pushed to get it wrong!
I would have thought that any decent WC DA will have neglible delay anyway. Even cheap ones would be hard pushed to get it wrong!
jdbakker
Well-known member
Interesting question.
Have a look at AES11, which discusses synchronization of digital equipment in studios. According to AES11, a synchronized device which transmits digital data may have a phase offset at its output of no more than +/- 5% of the cycle time of the sync clock, and a synchronized device which receives digital data must accept this data with a phase offset of up to 25% of the sync clock. For 48ksps sampling this translates to 1us and 5.2us respectively. So if the clock distribution device is the only device in the chain that causes phase shift, it isn't allowed to introduce more than ~4.2us delay @48ksps sampling.
It is unlikely that a clock distribution device has this much delay. A simple HCMOS buffer has a propagation delay of a few tens of ns, and an analog distribution amplifier will have delays in the same order of magnitude. Commercial-grade coax cable (as used in WC distribution) has a velocity of propagation of ~20cm/ns, so a typical distribution amplifier will be comparable to 10-20 meters of cable. Of course the story is different for a PLL-based clock recovery device, where a naive PLL design can introduce much larer phase differences between input and output, but I'd judge such a device to be defective.
One of the issues with WC is that it isn't standardized, and different manufacturers tend to apply their own interpretation. See Appendix B of AES11 for more on that.
JDB.
Have a look at AES11, which discusses synchronization of digital equipment in studios. According to AES11, a synchronized device which transmits digital data may have a phase offset at its output of no more than +/- 5% of the cycle time of the sync clock, and a synchronized device which receives digital data must accept this data with a phase offset of up to 25% of the sync clock. For 48ksps sampling this translates to 1us and 5.2us respectively. So if the clock distribution device is the only device in the chain that causes phase shift, it isn't allowed to introduce more than ~4.2us delay @48ksps sampling.
It is unlikely that a clock distribution device has this much delay. A simple HCMOS buffer has a propagation delay of a few tens of ns, and an analog distribution amplifier will have delays in the same order of magnitude. Commercial-grade coax cable (as used in WC distribution) has a velocity of propagation of ~20cm/ns, so a typical distribution amplifier will be comparable to 10-20 meters of cable. Of course the story is different for a PLL-based clock recovery device, where a naive PLL design can introduce much larer phase differences between input and output, but I'd judge such a device to be defective.
One of the issues with WC is that it isn't standardized, and different manufacturers tend to apply their own interpretation. See Appendix B of AES11 for more on that.
JDB.
Digital clocking and which one (WCK or AES "Black") can cause friends to fall out forever!! In my experience of film dubbing stages and digital music studios, the most reliable systems have a house master clock and EVERYTHING slaves to that, console as well.
Use AES reference where possible, WCK if you have to, and some things still need video syncs (black'n'burst) as well. This may mean you have to purchase an AES DA as well, most master clocks will not have enough outputs.
In theory, the whole point of AES is that it is self-clocking, and the AES reference can be extracted from within the incoming signal. The purists maintain this is not the case., ie the clock is less stable this way. Can't say either way, but the most stable systems I know of follow the AES/WCK/video order resulting in less drama than those that just use WCK. The point of AES as well is that there is a sync word within the AES signal, so apart from going at the correct speed, the units are going at the correct "now". Timing is everything in digitland, so in answer to your original question, they are going to have to be seriously crappy DAs for the propagation time to be part of the discussion, but no daisy-chaining clocks please, unless you know it's just a loop-through on the units.
When I was researching all this a few years ago, basically the musos thought WCK was the best, broadcasters thought AES was a no brainer, but they the broadcasters did have to have a single-ended BNC based AES reference signal that would run down 75ohm video cables, because their installations were already flooded with this.
Use AES reference where possible, WCK if you have to, and some things still need video syncs (black'n'burst) as well. This may mean you have to purchase an AES DA as well, most master clocks will not have enough outputs.
In theory, the whole point of AES is that it is self-clocking, and the AES reference can be extracted from within the incoming signal. The purists maintain this is not the case., ie the clock is less stable this way. Can't say either way, but the most stable systems I know of follow the AES/WCK/video order resulting in less drama than those that just use WCK. The point of AES as well is that there is a sync word within the AES signal, so apart from going at the correct speed, the units are going at the correct "now". Timing is everything in digitland, so in answer to your original question, they are going to have to be seriously crappy DAs for the propagation time to be part of the discussion, but no daisy-chaining clocks please, unless you know it's just a loop-through on the units.
When I was researching all this a few years ago, basically the musos thought WCK was the best, broadcasters thought AES was a no brainer, but they the broadcasters did have to have a single-ended BNC based AES reference signal that would run down 75ohm video cables, because their installations were already flooded with this.
