phase vocoder module user

[headphoner]

Hi,

I tried to modified the fft size, but when i compiled i had a problem with the transform.cpp and the pvoc.cpp.

What is the problem?

I missed something

[senso]

Sorry I can't help you for the moment because I'm very busy.
I've planed to implement a way to change the FFT size with a parameter but you have to be patient.

[dts350z]

Hi,

I'm looking for the source code. In the add-ons section it says it's "on the wiki page", but the only thing I've found is a link to "sorry this document doesn't exist.

Would you please point me in the right direction?

Thanks!

I'm very excited to have found Usine!

[nay-seven]

hello and welcome !
by default this patch is a part of the addons ( which are reserved to gold members ) but as it's for educational purpose I've add a link on the wiki page :
http://www.sensomusic.com/wiki/doku.php ... :phvocforw

[dts350z]

OK thanks. FYI I Ordered Usine today and am just waiting for my Gold Membership.

FYI the link: http://www.sensomusic.com/download.php? ... er_0.4.zip on the above still requires Gold membership.

Thanks for the quick response.

[nay-seven]

oops, you're right
Waiting for your registration ( thanks for your support ) i've updated the link..

http://www.sensomusic.com/wiki/doku.php ... :phvocforw

[dts350z]

Thanks got it downloaded.

Maybe I'm not understanding the purchase flow. I bought it, payed on paypal, got email from paypal. Now what?

Sorry if I'm taking this thread off topic.

[nay-seven]

yep, it's still manual ....
done

[senso]

registration is made manually.
As I'm also very busy musician it can take 24h (time to find an Internet connection)
anyway, done now!

[dts350z]

I tired changing the hard coded FFT_SIZE in the source. Compiled OK but resulted in choppy audio (even though the CPU meter was still low).

Is this code substantially different or more complex than the Usine native FFT/iFFT? I ask because I seem to be able to run those at 8192 with an overlap of 16 without a problem.

What window type(s) are used in the phase vocoder and native FFTs?

Thanks.

[dts350z]

Also, is there some sort of "fan out" limitation with arrays?

While gearing up to do array manipulation between forward and inverse phase vocoders I connected to phase vocoders to five inverse phase vocoders, to check the cpu load and audio quality. One PV connected to three iPVs and the other PV connected to 2 iPVs. No mater what I did anytime there was a 3rd iPV it seemed to "lock up" and put out some fixed value like +246dB. If I add a 3rd PV and connect that same iPV to it, then things are fine (no PV having more than two iPVs connected).

Also the cpu meter in the patch doesn't seem to have any correlation to the one on the layout window. I have a dual core machine so shouldn't a patch cpu load of 30% show up as at least 15% on the main display? The main display only shows 3% Guess I need to do some serious RTFM!

[senso]

hello,
many questions in your post!

1) the PV user module starts to be old now. And try to improve the SIZE is a good idea. I'm sure it can be achieved easily but I can't remember exactly the way to proceed ... I have implemented a PV lib "as is", and you have probably noticed that the code is very complex.
There is no limitation of array size, but be careful big arrays take more CPU.

2) the CPU meter is complex in Usine

- unlike most DAW the dashboard displays the real CPU load, included Graphics, audio, core, etc...

- the CPU indicator inside patches displays the % of calculation time of the patch relative to the total BLOC duration.
for example with BLOC=128 samples: its duration is 128/44.100 = 2.9 ms at 44100Hz sample rate
15% of cpu means that your patch is calculated in 2.9*0.15 = 0.43 ms.
Graphic stuffs are not included.

- Also the patch CPU is estimated every seconds to avoid extra calculation. In the case of the PV, if you have a bloc size of 1024 the PV is calculated only every 8 BLOCs.
It means that 7 time on 8 your patch does almost nothing and 1 time every 8 your patch makes a huge calculation.
So the CPU estimation is not necessary reflecting the real CPU load of your patch.