Hello,

I tried to do the patch described on this imaeg but it sound like a ring modulator instead of a filter

http://pd-tutorial.com/english/images/kap3/3-8m.jpg

so I simplified it to simply do rfft->rifft, and it stills sounds like a ring modulator

how can this be ?

regards

Because all you are doing is amplitude modulating the signal twice. In the attached picture there is overlap and a different window size. Are you implementing that part? that should cause less of a ring-modulator sound I think

@seb-harmonik.ar

I just do fft then feed it back to an rfft, I should get the original signal

where do you see that I was multiplying the signal twice ? I dont

but indeed I missed the overlap part

I just cant make it work, the sound is quiet ok now but once I add the convolution part it goes apeshit and sounds again like a ring modulator. I used a [block 512 4] like in the example, changed the buffer size to 512 too

also I dont understand why my hanning window requires 2048 bins to show the whole hanning bell while in the example it only needs 512

You need to stick your fft stuff into a subpatch and then use [block~] or [switch~] to specify the blocksize/overlap

yeah I know...I finaly used 8 vcf filters and sliders instead, but I am not convinced

I can never get these fft to work

Maybe could you compare it with I03.resynthesis.pd that you will find in the help files, in the directory pure data/3.audio.examples ?

maybe if you post your patch we could figure it out.
Btw just using a hanning window with no overlap is amplitude modulation with a cosine wave, but it does work to time-limit a window so there are fewer artifacts when non-bin frequencies are treated as periodic (like in an fft)

@seb-harmonik.ar I reedited the original pd and it works

the only difference with mine is that the hanning window is generated in the parent patch
still investigating