Subject: [linux-audio-dev] rtlinux, MIDI timecode
From: Paul Barton-Davis (pbd_AT_Op.Net)
Date: Thu May 04 2000 - 18:22:02 EEST
Well, with a heavy heart, I am beginning to think that I have to use
RTLinux after all. With ardour.
As I've mentioned before, MIDI timecode seems very problematic for
general purpose computers. Sending 2 bytes every 1.3 msec (for 30fps
SMTPE) is just not something that Linux (or Windows, or MacOS) was
really designed to make it easy to do. There is no way to use the RTC
to generate periodic interrupts with the correct frequency, because
the frequency is not an integer power-of-2.
RTLinux, on the other hand, uses the 8253 timer, which be reprogrammed
all the way down to 0.9usecs by RTLinux. Its apparently fairly easy to
get RTLinux to do the right thing here, by getting an 8253 interrupt
thats fairly close to 1.3msec.
I can see only one potential way to generate a correct MIDI timecode
data stream with regular Linux: handling data from the soundcard in
roughly SMPTE-frame units. However, since doing so relies on the use
of the RTC as a periodic interrupt, and the RTC cannot generate the
correct periodic interrupt, I don't really see how this can work.
I am very reluctant to make this transition, but given that I have a
US$10K mixing desk that doesn't work fully without proper MIDI
timecode arriving from an attached "tape" system, it seems inevitable.
Does anyone have any last-minute thoughts on this ? :)
--p
ps. SMPTE pisses me *off*. what a totally stupid, absurd standard to
have to base timecode/sync systems on. they could have at least
used audio frames instead of video frames as the reference, or
better yet, some absolute time unit like usecs. It makes me
sick just thinking about the various SMPTE fps values. Barf.
This archive was generated by hypermail 2b28 : Thu May 04 2000 - 19:22:36 EEST