Hermann, could you send me the score and or the MIDI? I'm curious to know whether the mismatch is audible.
Just for the record, I misstated the length of the guarter in MIDI tics in PMX...it's 240, not 120. But I think everything else I said is accurate. I did a few more tests since then, and reminded myself that the special corrections I had programmed apply only to half-note and quarter-note septuplets, but not eighth-note septuplets or any 9-tuplets (unless the 9-tuplets are on dotted notes, e.g., cd05x9 ... ). I wonder if Score/Finale/Sibelius have any of these same issues. --Don Simons Hermann Hinsch wrote > Am Donnerstag, 14. Juni 2007 16:39 schrieb Don Simons: > > I've just realized that there is a way to allow MIDI 128th notes without > > PMX completely choking. But the resulting MIDI file will be flawed. The > > fundamental quantum of MIDI time is a "MIDI tic" which PMX > defines once and > > for all as 1/15 of a 64th note or 1/120 or a quarter note. > Since that makes > > a 128th note equal to 7.5 tics, which is not an integer, > there's a problem > > right away. But that's not the reason PMX was choking. Rather, > it's because > > PMX puts in a fixed-length gap at the end of every note, subtracting the > > gap from the length of the note. The default is 10 tics, which > is 2/3 of a > > 64th, but longer than a 128th, so the logic was breaking down. You can > > forestall the locigal problem by shortening the gap to > something less than > > 7.5. The command is e.g. Ig5 . With that the code will run. But for one > > thing, you're stuck with that tinier gap at the end of every note. (I > > haven't tested whether you can change it on the fly.) And for another > > thing, a flaw now comes about because a 128th cannot be represented > > exactly. As things now stand, PMX will always round 7.5 in the same > > direction. So if there are more 128ths in one voice than the > others, they > > will get out of phase. Here's an example where you can hear > that things get > > out of whack at the end: > > > > Hi Don, > > Ig5 solves my problem, thanks for your reply > > Hermann > > > ===================================== > > 2 2 2 4 2 4 0 0 > > 1 5 20 0 > > > > > > bt > > .\ > > Ig5 > > b43 d / > > g84d g1x8 a b c d c b a g84d g1x8 a b c d c b a / > > b43 d / > > g84d g1x8 a b c d c b a g84d g1x8 a b c d c b a / > > b43 d / > > g84d g1x8 a b c d c b a g84d g1x8 a b c d c b a / > > b43 d / > > g84d g1x8 a b c d c b a g84d g1x8 a b c d c b a / > > d24 d d d / > > f24 f f f / > > ====================================== > > > > If there were the same number of 128th in each voice, then they > would come > > back together: > > > > ====================================== > > 2 2 2 4 2 4 0 0 > > 1 5 20 0 > > > > > > bt > > .\ > > Ig5 > > b43 d / > > g84d g1x8 a b c d c b a g84d g1x8 a b c d c b a / > > b43 d / > > g84d g1x8 a b c d c b a g84d g1x8 a b c d c b a / > > g84d g1x8 a b c d c b a g84d g1x8 a b c d c b a / > > b43 d / > > g84d g1x8 a b c d c b a g84d g1x8 a b c d c b a / > > b43 d / > > d24 d d d / > > f24 f f f / > > ======================================= > > > > At this point someone might be wondering if the quantum problem > comes up in > > more mundane situations with xtuplets, and the answer is well, sort of, > > depending on your definition of mundane. Since a quarter note > is 120 tics, > > you can divide it evenly by 2,3,4,5, or 6, but not 7. I anticipated this > > one, and tried to build special logic for just that case (quarter-note > > septuplets) to keep them lined up. If that works the way it should, you > > won't get the quantum disconnect until you divide a quarter by > 9. That one > > will not be fixed. > > > > --Don Simons _______________________________________________ TeX-music mailing list [EMAIL PROTECTED] http://mailman.daimi.au.dk/mailman/listinfo/tex-music
