Subject: Re: [ecasound] full-duplex sync
From: John Denker (jsd_AT_research.att.com)
Date: Fri Oct 12 2001 - 15:50:42 EEST
At 03:16 PM 10/12/01 +0400, CLOTILDE Guy Daniel wrote:
>Yes
>Ecasound does it.
That's encouraging!
>About the hardware, the best you have, the best it will be but even with a
>low-end card you can achieve non-noticeable latencies.
That's confusing! We are not primarily worried about latency. We have
already achieved reasonably low latency; further reductions are not critical.
Things that may be "non-noticeable" to you are noticeable to us. What we
care about is !exact! sync between input and output.
>Above all, tuning your kernel can lead you up to a 1,5 millisecond of latency.
To repeat: We need the sync to be reproducible down to the last sample, to
within a few microseconds. Yes, microseconds. Latency is not the same as
sync. A millisecond or two of latency is fine with us, but we really,
really need good sync.
>A google search with 'morton low-latency kernel patch' should lead you to
>the one that match your kernel version.
To repeat: That won't solve the problem of interest.
>On Thu, 11 Oct 2001 15:18:59 -0400
>"John S. Denker" <jsd_AT_research.att.com> wrote:
>
> > Hi --
> >
> > We need to do full-duplex. Simultaneous play + record.
> > It is absolutely essential to get the timing right.
> > We have hardware that is documented to support this; see
> > http://www.midiman.net/support/manuals/pdf/DELTA1010.PDF
> > (Look in the middle of page 24.)
> >
> > The alsa driver is supposed to support this, too; see e.g.
> > http://www.alsa-project.org/alsa-doc/alsa-lib/group_pcm.html#a46
> >
> > HOWEVER.... When I try to do this with ecasound, I find that there is
> > considerable run-to-run variability in the timing relationship between
> > the played and recorded signals.
> >
> > I read the documentation and even grepped for terms like "sync" and
> > "full duplex" without success.
> >
> > I looked through the ecasound code and grepped for likely terms
> > without success.
> >
> > I searched the mailing-list archives. I found a lot of discussion
> > of MIDI-synch (which is !not! what I'm interested in) but not much
> > about full-duplex analog sync.
> >
> > A secondary goal is to have minimal delay between play & record.
> > The primary goal is that whatever delay there is should be
> > absolutely stable and reproducible.
> >
> > So....
> > -- Does ecasound have this capability?
> > -- If so, how do we invoke it?
> > -- If not, does anybody have any suggestions for what we should do???
> > (There are several crude workarounds, but doing it right ought
> > to be easier than kludging it.)
-- To unsubscribe send message 'unsubscribe' in the body of the message to <ecasound-list-request_AT_wakkanet.fi>.
This archive was generated by hypermail 2b28 : Fri Oct 12 2001 - 15:47:29 EEST