|
Paul suggested we investigate the sources of heat generation in the controller by changing the switching frequency in half. The result should reduce the switching losses by half at the expense of an audible whine and higher ripple current (I think?). It'll be easy to implement as well - just change the micro frequency to 8mhz using the internal oscillator.
The effect on temperatures, however, should be less than half - that is, the temps should be slightly more than half of what they were before due to the other sources of heat (diodes, Rds_on) that aren't changing.
Anyway, this would give us an idea about what to expect from reducing the gate resistance to speed up switch-on/off times, if we decide to pursue that.
Some observations:
1 - The ~8khz whine is not very loud, but it is very high pitch and I found it quite annoying. I don't think I could take it if I had to leave it like that.
2 - I increased Kp in the PI loop to account for the slower processor, but I forgot about the ramp rate. It took what seemed like forever to ramp up in current.
3 - I froze the micro once. Stepped on the pedal too hard - not sure why it locked up, but it did. It was very reliable at 16Mhz, so not sure why it would change - maybe something to do with the overcurrent reset or the response of the PI loop at the slower processing speed? I plan to change it back to 16 Mhz, so I'm not going to worry.
4 - As expected, the 'feel' of the controller didn't change (aside from the slow ramp).
5 - Temperatures. Conveniently, the ambient temp was about the same as in the video a few pages back - 42C. The only difference was that the sun was going down tonight while the video was in the sun, but since nothing is exposed, I think only the ambient temperature matters. I went on the highway pulling the same amps until the temps stabilized. Mosfet and diode were around 61, capacitor was at 65 (had just clicked up to 65 when I got to my exit - might've gone higher?). The heatsink probe was reading 52. If you happened to watch the part of the previous video where the camera is right in from of the probe display, you might remember the mosfet reading 67, diode 63, the capacitor a little lower than that (at that point), and the heatsink probe at 56C.
*Remember, these are just probes that are touching the case of the mosfets/diodes to get a ballpark temp reading.
So, Before at 15khz:
Mosfet - 67
Diode - 63
Cap - 60ish
Heatsink - 56
Now:
Mosfet - 61
Diode - 61
Cap - 65
Heatsink - 51
6 - The switching components ran cooler, but the caps did not. Previously, I had never seen the caps rise higher than the mosfets/diodes. Now, I suspect with the lower frequency and higher ripple current, the caps are generating more heat and are rising in temp.
7 - Previously, the mosfet ran hotter than the diode, but at half frequency, they were roughly the same temperature. Physically, they are fairly well thermally coupled due to the copper heat spreader, so I'd think a close temp is to be expected. The difference in temps before is likely due to the extra switching heat that is now cut in half. When the mosfets were generating a lot of heat before (more than the diode), they ran hotter.
8 - It's always interesting to compare the mathematical models to real measurements. Using the models for a mosfet and diode presented quite a few pages ago, the heat generation at 15 khz, half duty cycle, and 200 motor amps is 449W. Cutting the frequency in half reduces this to 273W, or a 39% decrease. If we assume that the thermal resistance stays the same, we can then say the component temperature-above-ambient should decrease by 39% as well. This hypothetically would give:
Mosfet - 57
Heatsink - 51
diode - should stay the same? It's producing the same heat as before, so...
However, this is not what was measured - the predictions are too low. The actual temperature changes are closer to a 26% decrease in temp, significantly less than the 40% prediction. Of course, there are so many assumptions that go into the model (which itself is just an approximation), so maybe these results are actually pretty good.
So, here are some other model heat gen results to chew on:
15 Ohm Rgate - 37% less heat generation
10 Ohm Rgate - 50% less heat generation
5 Ohm Rgate - 64% less heat generation
|