source file: m1369.txt
Date: Mon, 30 Mar 1998 09:23:12 -0500

Subject: Re: when truncated

From: wauchope@AIC.NRL.Navy.Mil

By "higher bits" I meant the higher precision (finer tuning) bits, so
all but one of the remaining bits will always be correct after truncation.
It's that highest bit that can be compromised if you don't round off,
putting you one tuning step away from the closest possible value.

For example tuning an 8/5 calls for a pitch bend of +561 (13.7c) which
QT would truncate to +544 (32*17), giving you 13.28c.  If you rounded
the bend up to +576 (32*18) yourself, you'd get 14.06c which is slightly
closer to your goal.  So the worst that truncation can do to you is one
tuning step, or 0.78c in QT.

I don't bother rounding off bends in MIDI files since they're intended
to be played on different machines with different PB precisions, but
if you're targeting one particular synthesizer it might be useful.

Ken Wauchope


>>I'm not a MAX user, but assuming that xbendouts are the same as MIDI
>>pitch bend messages (+-8192 = 14 bits), you can round off to the
>>nearest multiple of 32 for best results since I imagine QT just
>>truncates the higher bits.

>You mean I have to recalculate by myself?
>Or does QT automatically take me to the nearest value?
>In other word ... does higher bit does the 'more' fine tunings?...I mean 
>when 5 bits truncated ... the remaining 9 is not "not correct" but just 
>robust?...hope so...