The first commit feels plausibly correct, the second one feels like it would create possible segfaults (in my mind at least). Removing the fmodf() means that tw[i] shifts from being bounded 0..oscillator-size, to being unbounded. Without that constraint then the integer version (the 'high' variable) is again unbounded, so is the code correct with that bound removed?

Also, since there was a flurry of discussion around the FM modulation, let me know if this PR became out-of-date.

this PR is meant to fix
#261

the unbounded value goes through this chain:
fmold -> tw[i] -> FMmodfreqhi -> carposhi
where some other values are added,
and is bounded then to the correct boundaries.
carposhi &= (synth.oscilsize - 1);

the bounding of fmold, i want to remove leads to discontinuity in the modulator -> frequency jump

ah, as long as carposhi is bounded, then I think this is fine (assuming it improves the FM issues that were being experienced)

That bug that I mention here is already fixed by my Envelope fix (which I thought was part of Friedonlino's PRs, but I cannot find it right now). You remember that fix with the weird constants 0.01 and 0.99, where I also asked Paul Nasca (no reply yet)?

I think and hope the latest commit fixes the fmamp envelope problem of #261.
it seems to be a typo at the definition of MIN_ENVELOPE_DB that affects only amp envelopes at values around 0.

i removed the special treatment for small values in amp envelopes because they start with 0 even without it.