[PD] DSP cycle - block size & overlapping

[Pierre Guillot]

Hi all,

I'm experimenting with different block sizes and overlapping factors for
subpatches and I encountered 2 issues (or perhaps my mind is completely
scrambled, tell me...):

1st: the output of [bang~] is limited by the default block size (64
samples). I understand why it works like that - the DSP method of [bang~]
uses clock_delay() and the clocks frequencies are limited by the main DSP
cycle - and that  avoiding the use of clocks can break the sequential
behavior (signal-message). But I think it's strange. Do you think that we
can find a way to use [bang~] with smaller block sizes than 64? If not,
shouldn't we add something in the help file (it took me a while to figure
out this limitation)?

2nd: I don't really figure out how the overlapping is performed. Following
the [block~]'s help file and I03.resynthesis from the audio examples, I
understand that with a block size N and an overlapping factor O, the patch
computes blocks of N samples at intervals of N/O samples.
Now, I suppose if the patch has an [inlet~] that directly feeds an
[outlet~] the resulting signal would be the input signal multiplied by O.
Considering a square wave (between 0 & 1) with a period of N for the input
signal and an overlapping factor of 2, I supposed that the resulting signal
would also be a square wave multiplied by 2 (between 0 & 2) with the same
period. But the resulting signal is a unitary signal of 1, like if the 2nd
overlapping block has been shifted by N/2 and is now synchronized with the
1st one. I guess there is something obvious that I don't understand but if
anyone can help me I would be very grateful!

Here a patch that describes the 2nd problem:
https://gist.github.com/pierreguillot/5e775b39860330739e3efbb6adb8ad1a

Cheers,
Pierre
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20170718/e2e826f6/attachment.html>