Hi,

i'm working on a patch that switches dynamically between several blocksizes by using the switch~ object. i'm switching quite fast between e.g. 512 and 2084 block sizes in realtime, while an audio file is processed with rfft~.
i experience some very unpleasend glitches between the switching. has anyone an idea how i can somehow smooth the transition between the blocksizes???

regards,

matthias

i tried initiating the switch~ on block boundaries by using [bang~], but that didn't seem to work,

so the only idea i have is to ramp your volume down to zero, with a ramp time of 5ms or so (using [vline~]), do the switch, and then ramp volume back up to 1.

you'd have to have the FFT stuff in a subpatch, and the volume ramp outside the subpatch, otherwise the ramp itself would also get messed up with the block switch glitch.

if you need to switch VERY rapidly between block sizes, and CPU is not too much concern, then you could always just turn your FFT process into an abstraction, and load different instances of it for each block size, and then crossfade between those.
admittedly, not very elegant, but you could even do stuff like switching the abstractions off (with a 0 message to [switch~]) once they have crossfaded to 0.

I've also experienced problems with switch~ in fft applications. It gives me distortion in fft subpatches that seem to be running fine with block~.
I also had the problem with glitches when i change the blocksize with both block~ and switch~, together with several seconds of dsp breakdown everytime i switch.
The only way to get things to go smooth was to have a different subpatch for every blocksize, so that i switch the blocksize by turning the corresponding subpatch on. If it's still too glitchy for your liking after that, you could implement some crossfading mechanism to keep things even more quiet.
i hope this helps!

hi,

thanks for the suggestions. i'm going to test both of them and let you know the results.

does anyone else had some experiece with fast block switching within a pd patch??

here's a test patch that switches between a very large block and a small one. the results are a bit surprising.

when i recorded pd's output using a different application, and then checked it in a wave editor, there were some audio dropouts, which i expected to see - but there were other glitches too - i saw (and heard) random peaks and drops in overall volume while switching. It's a bit weird.

http://www.pdpatchrepo.info/hurleur/switchtest.pd

i recognized the random glitches too. thats weird, but could be sound nice when you time stretch the glitches