Subject: Re: [ecasound] Simultaneously recording many tracks
From: Kai Vehmanen (kai.vehmanen_AT_wakkanet.fi)
Date: Wed Jun 16 2004 - 22:06:38 EEST
Hello,
first a quick general note: I've been offline for nearly a month due to
problems with my cablemodem net access. Now I'm back online at last, had
to switch to ADSL in the end...
On Tue, 8 Jun 2004, Cassettes2CDs wrote:
> I have this application where I need to record many
> many 32-bit/44100 stereo channels at once with some
> M-Audio delta 1010LT sound cards.
[...]
> ecasound -z:db -r -b 8192 -f 32,2,44100,i -i
> "/dev/dsp$dsp" -f 32,2,44100,i -o "$outputfile"
I recommend:
ecasound -f:32,12,44100 -i alsa,default -f:32,2,44100 -o foo.wav
... that avoids routing all the audio through the ALSA
plugin layer.
> The Thing Is that the process seems to be very
> fragile. If the computer is doing anything else that
> involves the disk, then it'll screw up and
> occasionally miss some audio and give me a "skip" or
> repeated snippet of sound here and there. This is a
> commercial enterprise (converting cassette tapes to
> CD), so this is bad bad bad.
That is fixable. You can get far just by running Ecasound as root.
Please read Dave's excellent article about this issue at:
http://www.linuxdevcenter.com/pub/a/linux/2000/11/17/low_latency.html?page=1
... for more information.
> I've got a throughput bottleneck in there somewhere,
> but I'm clueless as to where it could be:
[...]
> o disk?
> o disk controller?
> o the OSS driver?
> o PCI bus?
> o IRQ's?
> o the sound card's internal buffer
> (M-Audio denies this)?
> o memory buffers in linux?
Probably none of the above. More likely:
o kernel scheduling problems (latency spikes to
application scheduling)
-> caused by either problems in core kernel code (low-latency
patches help) or device drivers (more difficult,
changing hw is often the quickest workaround)
o spikes in interrupt-wake-up latency
o ...etc
> And on the same note, if I could -detect- when a
> recording hiccup has happened, that could let me dodge
> the problem too.
When using ALSA for audio i/o, ecasound will report about each xrun. If
xrun (i.e. buffer underrun/overrun) should be considered as an error
condition, add -z:xruns to command-line and ecasound will abort the whole
session if xruns are detected.
> I'm putting this out there to see if anyone can give
> me some hints as to what my issues could be. The
> M-Audio people insist that I should be able to do 16
> stereo channels at the same time with no problem.. in
> WINDOWS! Green with envy I be!
This is certainly possible. Recording for example 32bit-12ch-44.1kHz does
not cause any problems to my humble Celeron466 system.
--- http://www.eca.cx Audio software for Linux!
This archive was generated by hypermail 2b28 : Wed Jun 16 2004 - 21:57:26 EEST