Subject: Re: [linux-audio-dev] [OT] Digital mixing console and soundcards
From: David Olofson (david_AT_gardena.net)
Date: Sun Mar 10 2002 - 01:09:35 EET
On Friday 08 March 2002 21.19, Paul Davis wrote:
> >Sorry that this is a bit off topic, but I think there are some
> > people here who can give some advice.
> >I want to build a digital mixing console, using a Pentium and a
> > soundcard. The PC won't run an OS, so I can optimise I/O (there
> > won't even be a monitor connected to the PC).
>
> i've said it before, i'll say it again: this is a mistake. there is
> no reason not to run an OS. absolutely no reason. you are going to
> make a lot more work for yourself by doing this. probably a lot
> more than you can imagine.
I totally agree.
If you have extreme latency requirements enough that Linux/lowlatency
doesn't cut it (that is, you need < 3 ms input -> output digital
latency), I have news for you;
Bad news:
You will have trouble finding a sound card
that can use small enough buffers to support
this. In fact, you'll hit the PCI DMA burst
size limit *long* before you hit the limit
of your average PC with a good RTOS.
Good news:
If the 64 or 32 byte DMA burst limit isn't a
problem (ie you're using 24 bits at 96 kHz
and/or a multichannel card that uses a single
"DMA stream" for all channels), you can cut
the worst case scheduling latencies down into
the µs range, using RTLinux or RTAI.
Good news 2:
By programming the sound card driver to share
the DMA buffer, and then "hooking" an RTAI or
RTLinux thread to the sound card IRQ, you can
even make use of RTL/RTAI without porting the
sound card driver!
As to latencies and not using an OS - indeed, there's no reason not
to use an OS. Taking it to the extrem, RTLinux and RTAI commonly
deliver worst case latencies below 10 µs on P-III and Celeron
systems, and even better on PPC systems. (PPC has much better IRQ
handling than x86 - mostly because it's not dragging around that
badly designed legacy garbage that is the "PC chipset".)
If you need to cut latency below that point, please tell me where you
found your AD/DA chips! ;-)
("What's so funny about that!?" Well, the latency in your average
converter is in the range of milliseconds, due to oversampling,
digital filtering etc. The fastest I've seen so far has a latency of
well over .5 ms.)
So get a real OS, and hack away - no hairy cross-compiling setups
need! :-)
//David
.- M A I A -------------------------------------------------.
| Multimedia Application Integration Architecture |
| A Free/Open Source Plugin API for Professional Multimedia |
`----------------------> http://www.linuxaudiodev.com/maia -'
.- David Olofson -------------------------------------------.
| Audio Hacker - Open Source Advocate - Singer - Songwriter |
`-------------------------------------> http://olofson.net -'
This archive was generated by hypermail 2b28 : Sun Mar 10 2002 - 19:21:23 EET