MSF: why three Wavetables?

[Tagirijus]

Hey there,

I am just watching the new MSF tutorial. Thanks for this!

A question came up here: there is the WT8, WT40 and WT256 module. In the tutorial it is said that they use a different amount of CPU, while I could not recreate this. With all modules set to the lowest or to 8 or so, they seem to use the exact same amount of CPU.

So am I missing something here, or wouldn't it be more intuitive to have only the WT256 module and let the user decide how many waves they want?

Thanks for clarifying!

[Hexspa]

I'll hazard a guess. Are you testing with multiple voices playing? Since Melda plugins 'sleep', try activating 32 voices with unison and sustained chords. That'll probably show a difference, if any.

My second guess would be that it's a negligible difference.

[werzel]

My guess would be too obvious, at max wavetable number, 256 uses more than 40 which uses more than 8. Thinking that it wouldn't make much sense to use 8 in either 40 or 256.

[Chandlerhimself]

I don't think it uses more CPU, but I believe it uses more RAM. You might want to ask Vojtech about that just to be sure though.

[Tagirijus]

I'll hazard a guess. Are you testing with multiple voices playing? Since Melda plugins 'sleep', try activating 32 voices with unison and sustained chords.

I tested it with many voices playing at the same time, of course. Also enabled unison and so on. I mainly wanted to test it with 8 wavetables, since this is the max for the smalles WT module. I also tested it with 32 for just WT40 and WT256, no difference in CPU.

My guess would be too obvious, at max wavetable number, 256 uses more than 40 which uses more than 8. Thinking that it wouldn't make much sense to use 8 in either 40 or 256.

Well ... of course there might be a difference, with all the modules maxed, but that would be a nonsense test, since I wanted to compare the modules with the same settings, while questioning, if it wouldn't be enough to have the 256 module only, which can go down to 2 as well, like the other modules. That's why I am asking, since it seems to me counter intuitive to have different WT modules, while, when looking at the features, the WT256 module seems to be able to do everything the other modules can. So the question: why are the others there?

I don't think it uses more CPU, but I believe it uses more RAM. You might want to ask Vojtech about that just to be sure though.

Hm interesting thinking. This might be it, while I could not see an obvious difference there either.


Thanks for joining the discussion, everyone! (=

[WasteLand]

I'll hazard a guess. Are you testing with multiple voices playing? Since Melda plugins 'sleep', try activating 32 voices with unison and sustained chords.

I tested it with many voices playing at the same time, of course. Also enabled unison and so on. I mainly wanted to test it with 8 wavetables, since this is the max for the smalles WT module. I also tested it with 32 for just WT40 and WT256, no difference in CPU.

My guess would be too obvious, at max wavetable number, 256 uses more than 40 which uses more than 8. Thinking that it wouldn't make much sense to use 8 in either 40 or 256.

Well ... of course there might be a difference, with all the modules maxed, but that would be a nonsense test, since I wanted to compare the modules with the same settings, while questioning, if it wouldn't be enough to have the 256 module only, which can go down to 2 as well, like the other modules. That's why I am asking, since it seems to me counter intuitive to have different WT modules, while, when looking at the features, the WT256 module seems to be able to do everything the other modules can. So the question: why are the others there?

I don't think it uses more CPU, but I believe it uses more RAM. You might want to ask Vojtech about that just to be sure though.

Hm interesting thinking. This might be it, while I could not see an obvious difference there either.


Thanks for joining the discussion, everyone! (=

perhaps when 'morphing' the wavetable, how is it called in MSF? blend, mirror, etc. with high rates.. and when you have more WT generators. there can be more CPU overhead.
RAM?? a wavetable has a low footprint, very low, does use MSF, FFT? i guess. also low footprint.

but i am guessing, i wondered myself too, i use the 256 version, perhaps, i think now at this very moment!, if you use the WT8, interpolation, between 8 slots, is perhaps less demanding, but, it is adaptive or not.

i have no clue. haha.

[dmbaer]

if you use the WT8, interpolation, between 8 slots, is perhaps less demanding

I doubt that interpolation speed is affected by wavetable size. Only two waveforms are involved in the interpolation, no matter how many the wavetable holds. Unless, of course, interpolation points are pre-calculated - that's a possibility, I guess. But I'd bet on intermediate waveforms are calculated on the fly by averaging frequency domain data (the results of an FFT process). The frequency domain descriptors of the waveforms in the table probably are calculated ahead of time, but (again I'm speculating - I certainly haven't see the code) that's probably done when the wave table is saved while in edit mode, not when the instrument loads it to make some sounds.

[MeldaProduction]

The interpolation is always done between 2 slots (though more cache misses can do a lot). The main reason is memory - the more wavetables, the more "oscillator engines", the more parameters to be automated and stored in various places (and there are MANY for each oscillator).