> It was a couple of the large diodes on the back of the board connected to the resistors, see pics in link, that fried. Not sure why they fried on the 36v.
this means that the bus voltage went over about 62V with enough energy to fry the TVS'es, even when using the 36V battery. there are two possible causes that could act at the same time contributing to the bus voltage during regen:
-the batt can't accept the charge current. this means that the voltage on the batt terminals will temporarily go up and less current will flow in. this effect has to do with charge distribution and mobility of ions inside the cells. lead-acid cells are particularly bad accepting charge and that's why they can't be fast-charged.
-impedance (resistance and inductance) seen between the bus conductors at the FETs. it comes from the battery (internal batt impedance), the wiring, the shunt, the PCB traces, etc. (bad wiring (like a momentary batt disconnection) can send the voltage through the roof during regen if the controller doesn't abort quickly enough.) the long circuit to the battery is necessarily inductive and so high impedance will be seen in the high frequency range. this means that when the FETs switch, for a very brief period the voltage will go up in a spike. to avoid this the bus is decoupled (caps are added across it).
the problem with that controller is that the bus seems to be decoupled with an alu electrolytic cap (from what I see on the posted pics) and those caps are inductive in the high freq range. so a high voltage spike is to be expected on the bus with every switch cycle, which is a bad thing for TVS'es and FETs alike.
I wouldn't run that controller in regen without better decoupling. with proper decoupling, the TVS'es could be considered optional (but better if they are there nonetheless).