All the (few) larger EV controllers I've seen opened up, like my Curtis, have the sensor on the main mounting bar for the FETs. Some use obvious thermistors, one used a different kind that appears to be a modified TO92 case but as I could not see the face of it in the pics I dont' know what it was; it does have three leads, into a PCB of it's own with support electronics. I can't recall the brand of controller that was on.
I also have seen them on two home stereo amps from the TO3 case days (probably designed in the 1970s) where a thermistor encased in a crimp-style ring connector is bolted to *every* TO3 case in there, and wired into a PCB that had a bunch of can-style op-amps on it, presumably to mix all the signals together and/or compare them to the trigger point (I never looked at the wiring, though).
If I were really worried about the temperature and which parts were overheating, I'd probably do it the latter way, at least for data collection to find out where the best place to put them is. Log all the data and compare graphs to see which got hottest when. But knowing me, I'd just find a bunch of identical little thermometers with the outdoor remote sensor, and array the displays so I could see them all, and see which spikes first if any.
Maybe video them so I can see stuff I might not have noticed (poor man's data logger :lol
.
But probably if I were doing it right now, I'd just stick a single sensor in the center of the FET array on the mounting bar, glued or bolted to it, and set up a comparator to monitor that sensor, with it's open-collector output wired into the ebrake line to disable the controller if it got past something like 130-140F at most, since I figure the cores of the FETs could be a lot hotter than that in a surge situation.
