[time-nuts] disciplining sound card

Magnus Danielson magnus at rubidium.dyndns.org
Sat Jul 7 10:45:25 UTC 2012


On 07/07/2012 04:10 AM, Chris Albertson wrote:
> There is a profesional "standard" for clock distribution for computer audio
> interfaces.  They call it a "Word Clock" and it is usually distributed over
> 75 ohm coax cable.   It is common for a studio to have a master word clock
> generator and to use audio interfaces that accept an external clock.  Some
> of these master clocks have rubidium or OCXO inside.  Most can also PLL to
> any external clock
>
> The trouble is that lower priced audio interfaces lack a "word clock" input
> and you'd need to get into something like this as a minimum to have that
> feature
> 828mk3Hy<http://www.sweetwater.com/store/detail/828mk3Hy/>
>
> So if you are going to hack a cheap consumer interface it might be a better
> hack to install a word clock input.  At some point inside the interface
> there MUST be a sample rate clock running at 44.1, 48, 96, 192 kilohertz.
>   Find then cut that trace and bring it out to a 75 ohm BNC connector.  Now
> you have a standard pro level feature.
>
> Now when you divide down your 10Mhz lab standard divide it to "word rate"
> and you only need to build the divider once and you can use it with any
> audio interface that has word clock I/O.
>
> Yes of course you can send 14.4356MHz or whatever but that is a one time
> design and it will be different with every audio interface depending on
> whatever TTL can oscillator the engineer used.
>
> All that said.  I have a "cheap" Presonus firewire audio interface that has
> S/PDIF input and it has the option to accept word clock over the fiber
> s/pdif.  Many low priced interfaces can do this.   Then you happen to have
> a quality s/pdif device around you are set.

In professional audio settings, you distribute clock over AES/EBU (AES 
3), which is the original interface that was budgetized into the 
S/P-DIF. The preferred way is to distribute it over a quite channel 
(much like the black burst video signal) and then you get a Digital 
Audio Reference Signal (DARS) which is specified in AES 11.

A particular trick of receiving the clock is to use the end of the Y 
symbol, as this would have the Y symbols known pattern as a preamble and 
hence lowering the inter-symbol-interference quite dramatically.
The X/Z symbol still identifies the start of the frame, but being 
isochronous it's predictable from the Y frames as well.

Use AES/EBU whenever you can. It's been designed to handle long 
distances, something which was compromised out of S/P-DIF as it was only 
intended to wire together consumer equipment on short distances.

Cheers,
Magnus



More information about the time-nuts mailing list