pfm2 : 2.21

[Xavier]

Kuzma_p, can you try the attached firmware ?

Let me know if it fixes the pitch problem.
It's also able to load 600 (or any) samples waveform such as the one you posted first. It sounds great BTW

And please report if it crashes.

Thanks



[attachment deleted by admin]

[Xavier]

Oh... and don't forget to delete you usr#.bin files once you installed 2.02 so that they are recreated with the new interpolation/normalization algorithm.

[kuzma_p]

Great. Gonna check it today, at the evening.

[kuzma_p]

I messed with this fw a bit and didn't spotted any problem yet.
Gonna test further now.

[solipsvs]

LFO Phase kicks major butt!  Thank you sir!  I dont know if you realize what you have unleashed...  Im so excited wheeee!! 

this is my favorite fm synth evar!

(any chance of getting them on the modulation destination list?)

[kuzma_p]

2.02 did not crashed yet. Gonna go on with more complex patches though.

[kuzma_p]

i guess TX's lookup tables have less detalisation

FYI: By the way, i was wrong here.

It turns out that the sine is 4096 samples per period as well

https://www.gearslutz.com/board/9790791-post252.html

[kuzma_p]

Also, while patching yesterday i've reached interesting sound territories i've never heard from any FM/PM/PD synths before. PFM2 sounded like some additive/spectral generator — amazing. Gonna try to recreate and record this next time.

[Xavier]

Thanks for all the feedback 

As no major problem occured, I'm going to make this 2.02 official.

[norbim1]

Cool! Push to Git please. I can't wait to try it.

[Xavier]

Just pushed 2.02 to git.
https://github.com/Ixox/preenFM2

[SirPrimalform]

Carrier's? pitch starts to drift when you modulate it with my custom waveforms 1, 3, 4 and tweak IM with rest waves pitch stays static.

Sounds like a DC offset. Try processing it with some kind of DC kill process in your favourite wave editor before parsing it to text? I've been making a 'semi-modular' FM synth in Reaper's Jesusonic scripting language and ran into this when using real FM (it supports both true FM like the Preen and PM like the Yamaha 'FM' synths).

EDIT: Ah, I see Xavier has already built offset removal into the import code.

Xavier, does the new import code interpolate short waveforms up to 1024 or use them as is?

[Xavier]

The preenFM engine needs a "power of 2" number of samples. The goal of this interpolation is to make it accept any size of waveform.

If your number of sample is a power of 2 (32, 64, 128, 256, 512, 1024), there is no interpolation.

For the rest it interpolates to the closer bigger power of 2.
Source between 33 & 63 => processed waveform will contain 64 samples.
Source between 65 & 127 => processed waveform will contain 128 samples
...
Source between 513 & 1023 => processed waveform will contain 1024 samples

Just realilzed that in 2.02 waveform smaller than 32 are not accepted. 

[SirPrimalform]

The preenFM engine needs a "power of 2" number of samples. The goal of this interpolation is to make it accept any size of waveform.

If your number of sample is a power of 2 (32, 64, 128, 256, 512, 1024), there is no interpolation.

For the rest it interpolates to the closer bigger power of 2.
Source between 33 & 63 => processed waveform will contain 64 samples.
Source between 65 & 127 => processed waveform will contain 128 samples
...
Source between 513 & 1023 => processed waveform will contain 1024 samples

Just realilzed that in 2.02 waveform smaller than 32 are not accepted. 

Aha, I expected something like that.
I just wasn't sure whether you changed the engine to accept non-powers of two or just added some processing to the import code. The latter seemed more likely, thanks for the confirmation.
Is it a bad thing that waveforms less than 32 are not accepted? I would think that such an undetailed wave would produce a lot of aliasing, but maybe that's what some people want.
Actually, a bitcrushed (NES style) triangle wave might produce interesting (and nasty) results when used for an FM operator...

[Xavier]

Is it a bad thing that waveforms less than 32 are not accepted? I would think that such an undetailed wave would produce a lot of aliasing, but maybe that's what some people want.
Actually, a bitcrushed (NES style) triangle wave might produce interesting (and nasty) results when used for an FM operator...

I'll eventually add back the waveform with less than 32 samples.
But in the meantime, just copy and past each sample values in your text file to reach 32 samples. Should take you 1 minute