Okay Tom.  I’ve taken your thoughts to heart and spent much of my H90 time in the last three months on organs and Leslies.   I too have struggled with Rotary on occasion but I also have some vague memories of success too.  Here’s what I think I have learned (and this is all embodied in my 2-14-24 Leslie/organ post on PatchStorage):

Leslies have a &$%@ ton of bass.  So you may need to take that out.  Josh Smith did that by turning the Rotary rotor/horn control waaaay up (40/99).  That’s like applying a highpass filter.  Then he turned the rotor speed off.  I think he nailed it.

Before I discovered the Josh Smith approach, I tried to do essentially the same thing:  I used EQCompressor as a highpass filter and a sweepable mid-boost.  Not bad, but someone with better ears could prolly do better.  Also, that takes both halves of the H90.

SpaceTime doesn’t seem to have this subwoofer bass issue so it can fake a Leslie and provide some reverb and delay (all in one preset slot) to boot.  It’s not bad.

PolySynth itself has a modulator and sine/triangle pulse width modulation is in the ball park of a Leslie.  I, however, couldn’t get it to be that convincing. Maybe it’s good enough in a pinch but …

I find the TriceraChorus to be the best of the lot for both guitar and organ.  It’s prolly no coincidence that the tone control can be a pretty serious highpass filter.  So I dialed that back a LOT (like -60 or more). I also dialed “Depth C” down to 10-20 (as you suggested) but (so far) I kept right and left depth at 9/11 respectively.  IMHO it sounds great.  I wish it had a bit of verb or delay, but IMHO it still sounds really great.

Relatedly, you may recall I had a crackle problem using PolySynth for tonewheel organs.  Through trial and error I learned to find some pretty sweet sounds with the -oct volume turned down (from 100 to about 80).  That significantly reduced the crackle in my context.  I don’t understand it, but the crackle seems to arise internally to the algorithm and it depends on a number of things.  And, by “internal to the algortihm” I mean there was no way I was overloading the input.  In any event …

It’s been a long slog but I think I’ve made some progress.  If you are interested, you can find the resulting workproduct on PatchStorage.

Thanks again for your help.