[FFADO-devel] MOTU: Current MIDI status

[Jonathan W. <jw...@ph...>]

Hi everyone

Over the weekend I spent some time with MIDI transmission/reception and
actually tested this myself for the first time in a long time. :-)  The
practical upshot of this work is that I now have MIDI transmission working.
I haven't tested it with really long streams, but jack-keyboard seems to
happily drive my MIDI keyboard now and gmidimonitor reliably sees events
from the same external keyboard.

It turns out that the MOTU MIDI hardware is fairly sensitive to the timing
of MIDI bytes being sent to it.  If it's much faster than the "standard"
MIDI rate of 3125 bytes per second (10 bits per byte on the wire, giving
31250 baud) then bytes will be either dropped or corrupted in interesting
ways.  The simplest way to fix this, at least for now, is to implement some
local MIDI stream buffering in the MOTU transmit processor.  I have done
this.

As as been argued elsewhere over the weekend the long term solution may be
to add this sort of functionality to the ffado jack backend.  I would still
argue for its retention in the MOTU driver though since it is an inherant
property of the MOTU hardware that the MIDI byte rate is observed.  The
new code makes this requirement explicit, and the overhead involved is
extrememly small.

Having said that, having a more intelligent rate limiter "upstream" as well
is probably a good idea since it could act on whole MIDI messages - if the
rate is too great a MIDI message could be dropped for example, rather than
individual bytes.  That way the upstream limiter could work on MIDI
message level while the MOTU limiter ensured that the byte stream rate
within a message during playout remains within spec.

There is still a small issue I have to deal with regarding MIDI reception to
ensure we don't write beyond the end of the buffer.  I'll probably get to
that tonight since it's pretty trivial to fix.

Regards
  jonathan