CompuServe Thread

well, with the far project there was only one wave, 450K long which was pre-amplified and phased over itself. to get a more sophisticated deal, you can use a 32k table to do byte multiplies (symmetry cuts in half, rescaling to a byte cuts it in half again….) or, since amplitude is not so important, you can use shifts and adds, via a jump table. since time is important, it's foolish to trust anything but assembler. I do some nasty tricks involving loading all the D-registers with partial sums to compress an unrolled loop, etc. etc. Incidentally, the excessive use of register tricks mpre than doubles speed if you run FAR on an '020. I get 27 voices at 15KHZ on Mission Graphic's '020 as opposed to 8 on my 'fast-Rammed' 1000. On the other hand, RGS/DRW/DRWJ just uses a mulu to do amplitude, but it has a different kind of problem. RGS takes drawn information and compiles a sample while you draw its spectrum. (Or erase it or modify it or….) and does a host of other audio-related processes. If you make the sample length short , it seems instantaneous. On a '020 it nearly is. that's the trouble wwith Amy – it's about 16th as powerful as the problem demands (i.e. interactive additive synthesis on large numbers of voices) however, there may be some way to drag the blitter in to help by having it "add" samples bit by bit in chip ram. BTW, true granular synthesis should be possible on AMY, with good results. J H H Lowengard – audiot