On Fri, Apr 07, 2006 at 01:09:07PM +0930, Tom Cook wrote:
> Timestamps are not necessary. Audio is a synchronous data stream, so as
> long as one sample follows another, it's all good.
>
> But still not good enough for live work. Any latency or skew will be a
> killer.
>
> What about the constant bit rate circuits of ATM as a transport for this
> type of thing? You could make each input or output on the box appear as a
> different ATM circuit endpoint.
That's pretty much what I was thinking of.
I hadn't thought about feeding the stage monitor speakers or studio
headphones from the bit stream, but you're right. It requires short latency
and channel sync in real time.
OK, the various digitizer boxes on the cable need to have manually
selected unique IDs, so the recording engineer can control time slot
assignment and make each channel come up in the right place on the mixing
board or the hard disk. Suppose they multiplex their channels in order of
ID, and the box with ID 0 transmits a frame sync pulse that basically says
"SAMPLE NOW and count time slots until it's your turn to transmit"? This,
of course, means that NAK and retransmission are impossible, so forward
error correction is the only kind that can be used.
The simplest approach to error-free transmission is to base it on
physics instead of protocols -- EMI-harden the living hell out of the cable
to make it ignore cell phones and ESD, or use optical fiber.
But some of the remarks on the thread indicate that the recording
industry is way ahead of us. It may not be worthwhile for us to bother
reinventing these wheels. Maybe it would be better to discuss what a sound
card for computers with an open ABI should be like. It sure wouldn't ge a
professional recording device.
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
