Just messing with gen~ and I tried to make a cross fm patch - pretty simple but when I put it at certain settings the modulation disappears. I am not sure why. Would somebody be able to explain why it happens. There is an example below.
Feedback FM is very hard to predict... but it may be that at low frequencies the interaction between the oscillators never totally gets over the hump so to speak. Hmm, not a very good explanation, sorry... but the point I'm trying to make is that this is an expected(-ish) result of the mathematics - there's nothing wrong with your patch... :-)
I'm guessing this could have something to do with it (the phase relations of the partials).
Some things to investigate:
- Is it the range around 0? (maybe try some sort of inverted clip in a neighborhood around 0?)
- Is it negative frequencies? (use abs to rectify the modulators)
- What about exponential feedback FM, where there's no through-zero?
- Could you look for low output via some sort of envelope following and goose it back into life by injecting energy into it temporarily? (This is probably by far the weirdest, but could be pretty interesting)
Both of those solutions will change the spectrum, but since you're interested in feedback FM, I'm assuming a little mayhem is okay...