Subject: Re: [linux-audio-dev] MTC, SMPTE, etc.
From: Bill Gribble (grib_AT_billgribble.com)
Date: Sun Jul 23 2000 - 22:45:35 EEST
Paul Barton-Davis <pbd_AT_Op.Net> writes:
> >i'm confused. isn't this 1/30 sec (or 1/29.xyz or 1/24 or
> >whatever) supposed to be a "frame" (= one image of the
> >film)? what kind of frames are you talking about ?
>
> an audio frame (1 sample per channel).
In my experience with equipment out there now, the terminology isn't
generally used this way. When talking about device synchronization,
"frame-accurate" sync means SMPTE frames; "sample-accurate" sync means
what Paul is calling frame sync.
> the idea that an audio stream's current "position", which changes at
> least 44,100 times a second, should be represented by a timebase
> that changes 30 times a second is just completely ludicrous to me.
Well, for many applications that's either all you need or all you can
get. For example, if you are syncing to any kind of analog tape,
you'll be *thrilled* to get reliable SMPTE-frame synchronization. A
lot of the recording I've done has been partially-digital,
partially-analog tape, and the SMPTE clock is a reasonable time
standard.
Think about what you are using it for: basically to ensure that audio
sources don't drift with respect to each other over macroscopic time
scales. 1/30 of a second resolution for the sync-time-stamp is
perfectly fine for that.
Let's say we're at 44100 Hz sample rate, trying to sync two sources,
and their clocks differ by half a percent. How far out of sync are we
in a single SMPTE frame? Well, we're off 220.5 samples/sec, so that's
about 7 samples per frame. If you can stay synced at that level of
clock skew, you are off by an average of 3.5 samples. Not too
terrible.
Bill Gribble
This archive was generated by hypermail 2b28 : Sun Jul 23 2000 - 23:24:31 EEST