Hi all !

I'm trying to use [switch~] to fill-in an array instantly.

first thing I did is a test patch with a simple ramp [line~], but it doesn't work as I expected I've tried a lot of different stuff, changing the block size, and so on ... nothing goes.

My test patch is attached, so that you can test yourself ! The ramp should go from 0 to 500 in 3 seconds. The array is 3*44100, so it should contain the whole ramp exactly. I want to compute the whole thing at once, so I've set the block size to 3*44100 as well. What I have instead, is a full array, but the ramp only goes up to ~250 instead of 500

Can someone tell me what I am doing wrong ?

Thanks

Sébastien

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

Yop,

right click on the [switch~] object to get help about how it should be used.

Here is a corrected patch. [vline~] is said to be more accurate than [line~], this can be easily seen checking the values near 500.

Normally you should be able to speed up everything using for instance [switch~ 64 1 16] to compute everything 16 times faster (oversampling). The 'time' argument for the [vline~] should then be (3000/16). Unfortunately I couldn't find a way to get accurate computation this way yet.

Anyway this kind of ultra-fast computation can be achieved using [rpole~1] objects as samplerate counters, as mentionned by Maelstorm here : http://puredata.hurleur.com/sujet-6194-sample-played-automated-varying-speed-detect-completion . Precise ramp computation at 64 times oversampling becomes then possible, and ... fast

Nau

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

sebpiq, note that blocksize can only be a power of two, if you set a different size Pd will just force it to a power of two.

Katja

Thanks for the info Katja ! I had kind of understood there was something like that, but it's good to know what happens.

Hey Nau ! Thanks ! Actually I spent quite a while trying to read, and find some docs about [switch~], and didn't find anything valuable. I actually tried to reproduce the third example (which computes instantly a random array), with my ramp, but didn't succeed. Also, the options are not really explained ! This oversampling thing, I didn't understand so I thought "yeaaaah ! Probably useless ".

[rpole~] seems to be a bit more technical ! I guess I'll try it tomorrow !

My final goal is actually to build a time machine for a looper : basically it would record continuously the audio in a ~2 second array, overwriting at each passage, and when you send [stop(, it would stop recording, and generate 2 ramps (built with a modulo), as a control signal to read into the array with [tabread~4]. That way you get your loop + the 2 seconds before, in order to adjust it live !
I managed to generate the ramps without pb, but I was thinking they should be generated instantly, otherwise it doesn't make sense

But ... also, can somebody explain me why if I set the block size to a bigger block, then the ramp is computed really weirdly (each point is roughly at Y/2, where Y is the expected y coordinate) ? I tried to find some explanation, but couldn't find any guide, any article on PD's block size, and thus no answer to this "unexpected" behaviour