Need help troubleshooting a burned MCP236.

General discussion of the MCP motion controller product line
User avatar
Basicmicro Support
Posts: 1436
Joined: Thu Feb 26, 2015 9:45 pm

Re: Need help troubleshooting a burned MCP236.

Post by Basicmicro Support » Tue Feb 04, 2020 10:08 am

Email me at support@basicmicro.com to setup the time.

James
Posts: 15
Joined: Thu Oct 24, 2019 3:38 am

Re: Need help troubleshooting a burned MCP236.

Post by James » Mon Jul 13, 2020 8:21 am

Update on status after a really long time (coronavirus and some personal issues...)

Thank you for all your help. I have managed to repair the controllers, replaced all burned MOSFETs and also one of the gate drivers on the older MCP230A.
I also rigged up two really rugged voltage clamp boards based on the BUK965, which work like a charm.

During idle testing (no motors connected), while configuring the controllers, I've noticed that the newer controller (MCP236, using the TPW4R50 FETs) idles at around 55 Celcius, while mounted on a heatsink, while the older MCP230A (using the BSC070N10NS FETs), mounted on an identical heatsink, idles at 44ish degrees. Is this normal? Is there a chance that something shorts out within the newer controller? I get no indication or behavior that would support this...
Go tell the Spartans, stranger passing by,
that here, obedient to their laws, we lie.

User avatar
Basicmicro Support
Posts: 1436
Joined: Thu Feb 26, 2015 9:45 pm

Re: Need help troubleshooting a burned MCP236.

Post by Basicmicro Support » Mon Jul 13, 2020 8:47 am

yes, agreed on the factor 10 problem. That is just way to convenient to not be some scope setting. Let me know. I currently get in at 8am PST and leave around 7pm PST. I can push the leave part a little. Prefer not to try making it in early than 8am though.

James
Posts: 15
Joined: Thu Oct 24, 2019 3:38 am

Re: Need help troubleshooting a burned MCP236.

Post by James » Mon Jul 13, 2020 2:38 pm

I went through the wiring and measured some current values. The measurements are made at the controller BAT+ connection.
The older MCP230A runs at ~113 mA idle (no motors connected, motor PWM set at 0).
The newer MCP236 runs at ~121 mA idle (identical conditions).
Supply voltage is 53.7 V (13-cell Li-Ion battery, fully charged). Both controllers consume roughly the same amount of electrical power (~6-6.5 Watts).
It doesn't seem to be a short, and if it is, it's a very minor one. The minor difference in supply current, however, in no way justifies the difference in temperatures. There's a 12-degree (Celcius/Kelvin scale) difference, and with the attached heatsinks (~4 degrees/Watt each), it would require a very large current difference.
Ambient is at ~26 Celcius.
I will re-check if the thermal paste is still properly applied, but the heatsink does get hotter, so something else must be amiss. Is the temperature measured at the same spot on the older and newer boards? (And where exactly is that?)
Go tell the Spartans, stranger passing by,
that here, obedient to their laws, we lie.

User avatar
Basicmicro Support
Posts: 1436
Joined: Thu Feb 26, 2015 9:45 pm

Re: Need help troubleshooting a burned MCP236.

Post by Basicmicro Support » Tue Jul 14, 2020 8:36 am

I assume you mean a 12deg difference when running the motors. The idle current could indicate a bad mosfet driver circuit. This could indead cause a large temp difference if one mosfet isnt being tunred on correctly while driving a motor.

Yes, temp is measured from the same spot. Note we dont use thermal grease/paste. We use something more like putty that comes in preformed sheets. Its one use only. If you take off the heatsink you have to replace it since it deforms to fit the space.

If you have a thermal camera you could check for relative hot spots to get a better idea where the problem is. Its one of the ways we diagnose stubborn problems on RMAs.

James
Posts: 15
Joined: Thu Oct 24, 2019 3:38 am

Re: Need help troubleshooting a burned MCP236.

Post by James » Tue Jul 14, 2020 1:19 pm

Nope, I'm referring to an idle state.
Measured currents and temperatures are on idle, with motors disconnected and the motor PWM set to 0% via the Studio app.
I know of the thermal pads, very convenient, very not messy :) I'll see if I can replace the one on the controller I repaired (though I was careful not to touch or damage the pad).

However, the heatsink on the higher-temperature controller also gets hotter, meaning heat *is* conducted off the board rather effectively.
The second, older controller (the one running at 44 Celcius on idle) has not been tampered with, it never got burned or anything, and I never even dismantled it. It has almost the same idle current as the newer, repaired controller.
I will be keeping an eye on the temperatures during an upcoming live teleoperation test, see how they evolve over time with motors running. A thermal camera is on our TODO list as payload for the robot, but not available right now.
Go tell the Spartans, stranger passing by,
that here, obedient to their laws, we lie.

User avatar
Basicmicro Support
Posts: 1436
Joined: Thu Feb 26, 2015 9:45 pm

Re: Need help troubleshooting a burned MCP236.

Post by Basicmicro Support » Thu Jul 16, 2020 9:24 am

You can assume the difference in power dissipation is 8ma * input voltage(power = I*V). Assuming 13.5v supply that would be .108watts.

Your measured idle currents are basically identical. 113ma vs 121ma isnt anything. 8ma difference.

Did you measure the temp using a probe or Motion Studio? If Motion Studio the reading could be wrong.

But anyway that really doesnt matter right now.

Anything causing a 12C temp rise across the controller with the heatsink installed is going to be HOT. You may be able to find it by touching the components until you find the hot one. Be carefull.

James
Posts: 15
Joined: Thu Oct 24, 2019 3:38 am

Re: Need help troubleshooting a burned MCP236.

Post by James » Wed Jul 29, 2020 3:51 pm

Turns out that controller was not repaired properly...
We attempted a live test today. The clamps worked, nothing got fried.
But the (repaired) controller failed. I used the Studio app to test the controller, with the motors connected but the wheels of the UGV in the air. Minimal mechanical load.
The controller was set for a max current of +/- 10A. I synced the two motor channels, and had to remove the L/R entries because it would show a yellow OC mark for both motor channels.
During the forward test, Studio reported wildly different amps on each channel. At a duty cycle of 30%, Ch1 read 3A, and Ch2 read 17A, with both values fluctuating a bit (+/- 1A).
During the reverse test, the controller immediately tripped the circuit breaker, indicating a short circuit.

I'm guessing a FET is still burned out, even though earlier no-load tests with the oscilloscope showed no signs of trouble. I will take a look at it tomorrow, but if you have any ideas, plz let me know.
This one did not brownout like the first burned-out controller did under load, but still, it seems the overvoltage that burned it in the first place, damaged more than just a couple of FETs...

Edit (TH 301918J Jul 2020): Two more FETs turned out to be completely shorted (gate-drain-source). I replaced them, test will be performed next week...
Go tell the Spartans, stranger passing by,
that here, obedient to their laws, we lie.

User avatar
Basicmicro Support
Posts: 1436
Joined: Thu Feb 26, 2015 9:45 pm

Re: Need help troubleshooting a burned MCP236.

Post by Basicmicro Support » Fri Jul 31, 2020 8:34 am

Drivers are as much a problem as FETs.

1. Take all the FETs off the channel with a problem. Then check the input pins on the driver while driving the channel forward and backwards. Make sure the input PWM is operating correctly. You can check the low side driver outputs then as well.

2. Put the low side fets back on but leave off the high side fets.

3. Test the channel again but this time you should see the highside gates actually getting a driven signal at their correct voltage(low side fets charge the boost caps to power the high side gate drivers).

4. Assuming all that looks ok, put the high side FETs back on and retest again.

Note I'd recommend replacing all 4 FETs on the damaged channel just to elliminate that possible cause of failures.

I'd also check for resistance between the Source and Drain pads BEFORE re-installing the mosfets. If the resistance is lower than 50K you probably have internal damage to the PCB(burned PCB causes carbon which is conductive).

Post Reply