source file: m1576.txt Date: Sat, 07 Nov 1998 14:49:58 +0530 Subject: The Perils of Penelope Pitch Bend From: Drew Skyfyre Hi, John wrote : >>Then again, I suppose this to be a problem with just about any >>commercial sample-playback device. Can anyone confirm my suspicion >>that a synth that is actually a synth, synthesizing it's own waveforms >>(i.e., FM, PM, Additive, etc.) is much more reliable at delivering >>more stable/precise pitches. >> >Not necessarily. Most synthesized waveforms in commercial synthesizers are >still based upon the phase accumulator oscillator just like sample playback >machines. A waveform is read from a table and the phase increment adjusted >to change pitch. It is the fractional part of the phase accumulator that >determines the pitch resolution. By this you mean the following (oui ?) : 1- A buffer stores a soundfile 2- the sound file is read by an oscillator 3- the precision of the oscillator determines the precision of the pitch resolution The MSP tutorial (I have the run-time demo) shows the ability of MSP oscillators to use floats upto 6 decimal places, so this would be a good thing to achieve quite precise pitch bends, non ? Along with being able ot set the bend range to any range, even 1 cent. >It's a matter of how the oscillators are >implemented, and the current crop of of pseudo-analog synthesis really don't >provide any details on this. Certainly the older Yamaha FM, Casio PM and >Kawai Additive synths have similar pitch limitations to the sample playback >machines. Regular Dick Dastardlys, are these synth manufacturers. I wonder how precise and stable the pitches and how thorough (using all 14-bits of Pitch bend data, etc.) some of the new higher end synths are. i.e., Nord, Yamaha's new FS-1R, Korg Z-1. Graham wrote : >I should have asked at the time, but how deviant are your results from a >linear response? I've done some quick tests with my AWE64, and it seems to >be roughly 1 cent steps but not accurate to much more than a cent. Also, >now I've had quite a lot of experience of playing with pitch bends, the >results do sound pretty good. Certainly better than not using pitch bends. > I needed to try and make a proper chart (of pitch bend value equivalents for cent values 0 through 100) for the project I'm working on. So I spent the entire evening playing with spreadsheets, a sequencer, frequency meter, MIDI data monitor, Yamaha PSR-530 XG compatible keyboard, & calculator. Gotta tell you, I nearly lost it for a while. First of all, the Yamaha accepts bend data in the -8192 to +8191 range. Unfortunately, it ignores the MSB, so what you're left with is -64 to +63 (0-127). Setting the bend range at 1-semitone (the smallest possible), that gives 63 different pitches within a semitone (100 cent) range. It's pretty pointless paying too much attention to the results, since it just returns cent approximations of 63 out of the 100 possible values (and I don't mean 0-63 , just all over the place). My solution : bugger it all, I'll use it only for semitone, quarter tone, and random bends. I'm working on a demo of somewhat commercial pop-rock material and for now, the inaccuracies are acceptable. BTW, I used the "sine lead" sound, which happens to be about A = 439.8Hz, not 440Hz. C is 261.6Hz, though. Most pitches are off by at least 1-cent, and this for the sine wave sound, which means it's pretty much a crap shoot using the tuning tables to achieve good microtunings. All slight discrepencies in the std. 12-tET world, but IMHO quite unacceptable to make significant progress in the Xenharmonic. >If you're doing _serious_ work, tuning tables are usually better than pitch >bends because you don't have slurring which can be a problem with some >synths. The exception is where you want more notes than you can get into a >tuning table, so you stick in a small correction from the nearest note that >is in the table. Or, of course, if you don't have tuning tables and you >want to choose the scales you hear. Is assuming a linear response good >enough for these purposes from what you've measured? I am using the -64 to + 64 cent 12-note octave tuning tables in PSR-530. I use them to set up the base tuning for a given section of music, for the present mainly just & 1/4 comma meantone. However, I've been playing with IRCAM's PatchWork program (soooort of like Max, but LISP, dull to look at (black and white) not really real time, and not very MIDI savvy. It's better for writing Csound scores. Anyhow, what I did is put together a patch functioning as a sort of strange random-Xen arpeggiator, using pitch bends to randomly and constantly vary the intonation of each note within an approximate (make that "very" approximate) +/- 5-cent range. The resolution in PatchWork is 1/100 second and I place the bend at "0" and the note on 1/100th of a second later. Just a wee little "trick" brings things to life, animating dull and very artificial sounding step-sequenced lines. I've used the patch to process "guitar solos" and it rocks ! Also, works wonders on lead lines with a violin sound. It works like the pitch equivalent of rhythmic quantization. Common Music might be a much better tool for this sort of thing, compared to PatchWork. And, CM is free. I haven't worked with CM yet, though. So, John L., how's your K2500 at handling pitch bends ? Later, Drew ------------------------------ End of TUNING Digest 1576 *************************