Spiral Generator fails at FM, what am I doing wrong? about rs-met

[elanhickler]

hmm oh maybe it's because I use * isntead of + for pitch offset. I need to setup a ratio system / multiplication system rather than use addition.

[RobinSchmidt]

hmmm - actually, adding in the (logarithmic) pitch domain sounds right to me. this is mathematically the same as multiplying in the raw frequency domain as logarithms take multiplication to addition: log(a*b) = log(a)+log(b). so when you add a constant in the pitch domain, and then convert to frequency after addeing, this should yield constant frequency ratios.

[RobinSchmidt]

maybe i need more details and/or look at your actual code

[elanhickler]

it's heavily influenced by amplitude of modulation as well.

[elanhickler]

hmm I think it's working correctly (just played with another FM synth to see how it sounds), just the results are not as I want them to be. Not sure what the solution is other than to have per-key settings to fine-tune each key in terms of mod depth and frequency. When I do that I might see a pattern that I could use to have more musical FM.

[RobinSchmidt]

maybe some amount of (negative) keytracking of the modulation depth? this would make higher notes more muffled

[elanhickler]

naa, that's not the issue. will close!

[elanhickler]

re-opening this issue. I'd like to fix this before releasing:

https://puu.sh/wbMeI/9bd9aa25b8.mp4

It's the pitch offset that is bothering me. If you have pitch offset set to -24 -12 0 12 24 etc it works as expected. If you do something like 12.5 or even 13 or ANYTHING that is not a perfect multiple of 12 you get a frequency that is more and more off depending on the note you play.

How would I get a consistent sound with a pitch offset?

[elanhickler]

ok here's a clue

.1 semitone offset playing C3 equates to playing C4 with 0.05 semitone offset. So every octave up needs to halve the semitone offset.

[RobinSchmidt]

aaahh! i think, i see. maybe...how to explain....you are getting some kind of beating effect there which is perceived in the time domain - like an lfo. and you want the "lfo-frequency" to stay the same throughout the keyrange. but when you let the pitch-offset between the two oscillators stay the same throughout the keyrange, the frequency ratio of the 2 oscs stays the same - and the beating frequency goes up as well. so when you play a note an octave higher, you have also twice the beating frequency...oh my god...i think what i just said sounds very convoluted. but yeah - i think, for constant beating frequency, you need the frequency difference between the two oscs (not the frequency ratio) stay the same throughout the keyrange. ...which is what happens when you divide the pitch-offset by two for each octave you go up

[RobinSchmidt]

you could also just add something to the frequency of one osc (after conversion from pitch to frequency), i.e. introduce an additional absolute frequency offset

[elanhickler]

ok, yes, a simple frequency offset works.

[elanhickler]

also, you have to use to use LINEAR frequency modulation not exponential for FM, if you want things to stay in tune. In other words, I shouldn't be doing pitchToFreq(lfo_mod * depth + tuning); I need to do pitchToFreq(tuning) + lfo_mod * depth

[RobinSchmidt]

oh - yes - i think exponential FM creates inharmonic spectra even for simple carrier/modulator ratios (where linear FM produces harmonic spectra)

[RobinSchmidt]

or...does it? not sure

[RobinSchmidt]

ah: https://www.kvraudio.com/forum/viewtopic.php?t=310508

"The main problem with exponential FM is, as you increase the FM index, the perceived fundamental of the pitch shifts."

[RobinSchmidt]

here is a paper about exponential FM: http://recherche.ircam.fr/pub/dafx11/Papers/60_e.pdf it has equations that could probably be used to compensate for this detuning, making exponential FM playable in tune. ...could be interesting to explore....

[elanhickler]

I would be musical exponential fm, will return to this topic in the future.