Quote:
Originally Posted by mindsensors.com
|
It would be best if Mindsensors can present time/temperature charts that show the temperature stabilizing. It would be even better if you can present such charts with different (continuous) loads (10%, 20%, 40%, 60%, 80%, 100%). In FRC, it is rare that a motor controller would be run for the duration of the match at the same current level but these extra load lines can give the users some indication of what kinds of temperatures they can expect by estimating their average current over a short time period. The time/temperature charts will also give some indication of what the short term thermal capacity of the motor controller is; i.e. it may withstand an 80A load for 1 second better than it can withstand a 50A load for 2 minutes or longer). Your chart ends at just past 4 1/2 minutes which is much longer than a standard FRC match. In practice sessions, we have often run our robots continuously for much longer than 4 1/2 minutes, often repeating the most stressful actions. Considering how quickly the temperature is still rising at 270 seconds, I would be concerned that the transistors hit thermal-runaway and self destruct.
It is also of concern to me that your time/temperature chart is reaching almost 100 degrees at 150 seconds and about 125 degrees at 270 seconds, with your initial temperature of 25 degrees. It appears from your photo that these temperatures are measured on the top of the heatsink. If this is true, your transistor junction temperatures (taking into account the thermal impedance through the base plate of the your heatsink and the junction-to-case thermal impedance) will be much higher, possibly leaving you with very little (or no) margin from the maximum operating temperature of your transistors, depending on your device type. These high junction temperatures will lead to reduced life of the transistors. Have you done extended life testing on your motor controllers? Batch-to-batch variations in the transistor characteristics can have a large effect on your temperature performance so a test result showing little or no thermal margin raises red flags in my mind.
Lastly, the high heatsink temperatures are a concern, as Asid has also pointed out. Various safety standards specify different "maximum touch temperatures" but they will all be well below 100 degrees C (UL/IEC 60950-1 specifies a maximum of 75 C). With the thermal mass of your heatsink, temperatures at 100 degrees and above will most likely cause serious burns to anyone touching the heatsink. America has far too many lawyers. Some of them have children on FRC teams. I would not want to see Mindsensors dragged into court and many students be deprived of a great resource.
I am sorry if I am being hard on Mindsensors. They have a pretty good history in educational and hobby robotics where the power levels are low (NXT/EV3, Raspberry Pi, Arduino) and the energy sources have limited capacity (AA's or equivalent). They have now entered an arena where the energy levels are much higher and the hazards and consequences are much more serious. My comments and warnings are based on my experience developing power electronics products for mass production over the last 20+ years.
Quote:
Originally Posted by asid61
The stalling doesn't matter as much as the current; IIRC the heat generated is proportional to current.
I would be worried about getting a burn or burning something at that temperature. Plus, CIMs often run at high currents for the duration of the match; it's certainly possible to average 40-50 amps for a match.
|
The heat generated has two components; the switching loss in the output transistors and the conduction loss of the output transistors. The switching loss is roughly proportional to the switching frequency (output frequency) of the controller. The conduction loss is calculated by multiplying the on-resistance of the output MOSFET (at the instantaneous operating current) with the square of the instantaneous output current (P = R x I^2).
I have not seen actual efficiency numbers for any of the motor controllers currently in use or for the new ones from Rev and Mindsensors so it is not clear how people in this thread are saying that the new controllers are more efficient.