Where to locate thermistor in controller?

[spinningmagnets]

ES member Hillyterrain lives in hilly San Francisco, and has agreed to perform multiple uphill runs to test heat buildup at various gears and road-grade steepnesses. The GNG motor has proven that it can shed heat better than the stock 9-FET controller, so he will initially put a thermistor in the controller for heat data gathering.

He will be using a "10k NTC" thermistor (very cheap, $1 + shipping) that plugs into the cycleanalyst, but...where to locate it for best data gathering?

Perhaps the heat-sink bar that the hot FETs are attached to? I feel this is worthy of general discussion because all motors and controllers should have heat data gathering as an option, and the motor temp probe issue (which probe and where to locate) has already been discussed.

"Temp sensor that's too cool not to share"
http://endless-sphere.com/forums/viewtopic.php?f=1&t=25502

 

[dogman dan]

That would be my inclination, the heat sink. Possibly even outside the controller, with a scrap of foam keeping it from wind chilling.

If inside, on the heatsink, but not too near any particular fet would be my approach.

 

[zombiess]

The best location is inside directly in contact with a high side FET. I usually wedge my temp probe between the FETs and hold it with a dab of rtv sealent. Directly in the get is important for accurate readings. High side because it is under pwm and gets hotter than the low side. I suggest stopping when the fet reaches 80c.

 

[dogman dan]

That shows how dumb my comment was. I was thinking average heat. You knew which fet gets hottest. That would be the only one that matters.

 

[spinningmagnets]

Thanks to both of you! I must confess and expose my electronical ignorance. Here's the stock GNG 48V controller, could you please describe which FET is the high-side?

 

[dnmun]

the hi side FET is between the battery plus B+ and the phase wire. the drain, middle leg, is attached to the red wire through a trace on the pcb.

the hi side is hottest because the inductive voltage spike that occurs when the mosfet turns off the current in that phase wire it is connected to will forward bias the body diode of the hiside mosfet and cause the body diode of the hiside mosfet to conduct and that little blip in current through the body diode will add heat to it above the heating from the normal conduction.

 

[zombiess]

Look for the FETs (should be 2 of them most likely) that are connected to the B+

Those are your high side FETs. With a 9 FET controller they usually use 2 for the high side and 1 for the low side. 2 because the high side is usually the switched side and has higher losses due to switching.

If that's how your controller is setup, I'd stick the probe between the 2 high side FETs.

 

[chucho]

Another good post! Thank you for the good info!!

 

[alsmith]

Really it depends on what you want to know. Hottest temperature, board temperature, FET, other componentry? Possibly the best to monitor would be the one most likely to fail due to heat?

 

[friendly1uk]

I have 2 part epoxy thermal glue here. You can fit heatsinks to cpu's permanently with it. It is commonly used with peltier devices. I would stick the 'whatever' in place with this. It would be easier than making a bracket and using thermal paste.


Does the ca do good logging? I ask as temperature loggers are not expensive. Mine does something like 56,000 samples over just about any time scale I please. Then graphs it out on my pc, where the time stamps can be compared to journey data. It also has a display for min/max recorded value's so it won't melt before you get it to the pc.

 

[cal3thousand]

Wouldn't a thermistor on the traces also help understand something about the heat development and transfer?

 

[friendly1uk]

A track has so little mass. Sticking a sensor too it will change how it performs. It might be better to measure volts drop across it, and minimise that. If you feel your tracks a little lacking.

 

[cal3thousand]

A track has so little mass. Sticking a sensor too it will change how it performs. It might be better to measure volts drop across it, and minimise that. If you feel your tracks a little lacking.

Yeah for practical use I understand, but it seems like it might provide useful information in this "test" situation where data gathering is the goal rather than fault mitigation