From experiment, not because of reading and comprehending many pieces of
official and unofficial info on the Mac serial ports, some dating back
almost to Genesis, I've found a modification of the DM CONTROL parameter
block which allows FB 1.03 to shoot MIDI bytes out the printer port of
my PPC (604, 7.5.3) into an Opcode Translator II set to MIDI (not THRU).
Your mileage may vary.

My current guess is that it works not by way of external clocking but by
dividing the 3.672MHz SCC (originally 3.6864?) clock to something
*close* to 31250 baud. I've tested it with some long, very fast
sequences, and 3 channel sequences, and didn't audibly notice any
glitches/framing errors. Almost too good. If I'm right, this oughta work
at least back to the early Quadras.

I've uploaded a simple single-channel random-note FB source (some
customizing constants included) to
<http://www.oz.net/~ars/midif/mrng.hqx>. The program sends only note-ons
and note-offs (including random velocity bytes), which ought to test
with any MIDI box. It backgrounds okay too. If it does/doesn't work for
you, I'd be interested in knowing <ars@...>, but remember it's black
magic, not technical.

Eventually it'd be nice to know why it works. Now I can get on to
learning FB oooeee. Thanks again to list members who offered sources and
ideas, and to Keith Daniel for the lightbulb.

-- Tony
<http://www.oz.net/~ars/> NF MIDI Rock Mac SciTek