From owner-freebsd-multimedia  Fri Jan 19 08:57:42 1996
Return-Path: owner-multimedia
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id IAA26635
          for multimedia-outgoing; Fri, 19 Jan 1996 08:57:42 -0800 (PST)
Received: from miller.cs.uwm.edu (miller.cs.uwm.edu [129.89.9.13])
          by freefall.freebsd.org (8.7.3/8.7.3) with ESMTP id IAA26626
          for <multimedia@freebsd.org>; Fri, 19 Jan 1996 08:57:34 -0800 (PST)
Received: (from james@localhost) by miller.cs.uwm.edu (8.7.3/8.7.3) id KAA15680; Fri, 19 Jan 1996 10:56:03 -0600
Date: Fri, 19 Jan 1996 10:56:03 -0600
From: Jim Lowe <james@miller.cs.uwm.edu>
Message-Id: <199601191656.KAA15680@miller.cs.uwm.edu>
To: hasty@rah.star-gate.com, schulzrinne@fokus.gmd.de
Subject: Re: FreeBSD and VAT
Cc: multimedia@freebsd.org, multimedia@rah.star-gate.com
Sender: owner-multimedia@freebsd.org
Precedence: bulk

> From: Henning Schulzrinne <schulzrinne@fokus.gmd.de>
> 
> > 
> > Any suggestions, as to how to solve the problem that certain cards
> > don't run at given frequency and that they tend to oscillate.
> > 
> If you know the "real" rate, just configure your card to run at a 
> different nominal frequency (it sounds like they can be adjusted to any 
> rate). If the card can't keep its frequency stable, it will sound like 
> a tape deck which can't move the tape at constant frequency. For voice, 
> pauses between sentences and/or turn taking will take care of even 
> large frequency deviations. For music, you probably don't want to 
> listen to a card that can't keep pitch...
> 
> Given how cheap crystal oscillators are, I'm amazed that people build 
> cards that can't stick to a nominal frequency. (All useful frequencies 
> are multiples of either 8000 or 11025 Hz, it seems.)

Yes, and maybe this is the real solution.  Just come up with a
device that one can seperatly calibrate.  Maybe something like
/dev/caudio (calibrated audio).  Then run a calibration program
that would sample against the time of day clock and set the frequency
based on sampled results.  The calibration program could be run at
boot up of the system and determine actual speeds for the different
sound cards and tell the driver what these speeds are.  This way when
the user would ask for 8khz, they should be able to get something within
.1% of 8khz from /dev/caudio (or for Amancio /dev/cdsp :-).  I don't
know why I didn't think of this before (a year or two ago).  Thank you.

I suppose one could make caudio/cdsp a little more complicated and add 
features for full duplex utilizing the oddities of the different
sound cards.  Such as pas-16, 2 SB's, SB-16, gus, gus-max, etc...
This way one could have a calibrated full duplex interface for 
whatever application they wanted to use it for.  The BSD audio interface
seems to have all the stuff needed.  It would certainly solve a lot of
problems.  Anyone care to write it and add it to the VOXware distribution?

	-Jim