A little progress on caps, or at least on what doesn't exist.
Until that's more ready. I did have a thought. It looks like there are still quite a few spare pins on that MCU.
It is actually possible to program frequency in the same way we program iadj. It would require a second buffer cap that Cf2 charges from. The buffer cap gets charged through a resistor (or possibly divider) from the mcu. I say possibly divider because Roff is always taking current from the buffer cap anyway serving as the second half of the divider, pulled to ground during on time (by the IC, to reset Coff) , and never above 1.24 V during off time. The fet duty cycle is very fast and coff discharge cycles won't impact the buffer cap. Coff just sees a constant voltage source with an Roff impedance (exactly as it should), and the buffer cap just sees an average resistive connection to ground as if the second half of its voltage divider, as it needs to fix its voltage. That average does depend on fet duty cycle, and so working out exactly what voltage and off time a given pwm makes has to be calculated still, but this doesn't need high precision either.
This probably isn't necessary for normal uses, but it could
a) offer some frequency vs power correction if it ends up useful (not clear)
b) simply allow frequency to be optimized without resoldering which if nothing else is useful while first figuring things out.
c) allow software reconfiguration if re-purposing the driver for a different voltage output.
Anyway, I'm still working on a shopping list for how things are now. Programmable frequency could be a nice touch though.