2
\$\begingroup\$

I am using CAN bus to communicate between this BLDC motor with a built in driver and a real-time control computer which uses a galvanically isolated Microchip MCP2561FDE/SN CAN transceiver. Unfortunately I don't have any details on the CAN transceiver built into the motor driver board. Here is a diagram showing the isolation setup for the MCP2561 transceiver.

Block diagram of control computer CAN transceiver galvanic isolation showing that two digital couplers are used to isolate Rx and Tx

The CAN comms works perfectly when powered with 48V by a QPX1200SP power supply. CAN high and low sit at the right voltages and dominant states are clear with a differential measurement. Things however get very weird when I try and power the motors with our 48V 10kW power supply which is on a 3-phase plug (400V 50Hz). I was getting a ton of passive errors on the bus and because of this messages would only get through intermittently, so I decided to scope the CAN lines and this is what they looked like.

Image of oscilloscope screenshot of 11V peak to peak 104kHz sinusoidal signal on both CAN high and CAN low. The differential measurement is unaffected and stable at 0V

Sorry for not showing this in the screenshot but the frequency of these is about 104kHz. The 11Vp-p really concerns me. The purple signal is the differential measurement of both probes. When messages come through the differential signal looks fine (sorry I don't have screenshot of this) which is probably why communication is still happening albeit with a lot of passive errors. Even weirder is that when I increased the time scale there was also this 175Hz modulation happening.

Image of 175Hz sinusoidal modulation visible on previous scope signals with time scale increased to 1ms per division

I immediately tried disconnecting the common ground wire on the motor end and the sinusoidal shape dissapeared but both of the lines still had random signals at 11Vp-p and sometimes even 16Vp-p. After this I turned everything off. I'd really like to provide more scope screenshots I just don't know how much protection is on the motor's CAN transceiver and I can't afford to break one.

I really have no clue what aspect of the power supply could be causing these weird voltages so absolutely any help would be so appreciated!

\$\endgroup\$
12
  • \$\begingroup\$ Check your power supply, that appears to be the source. Also be sure the ground connection is stable and at ground. \$\endgroup\$
    – Gil
    Commented Aug 7, 2022 at 20:56
  • \$\begingroup\$ One quick question, did you add 120 ohm resistors at both terminating points or one 60 ohm resistor somewhere on the bus? \$\endgroup\$
    – Jonathon S.
    Commented Aug 8, 2022 at 1:43
  • \$\begingroup\$ It’s hard to tell what the scope is using for a sample rate, but the modulation could be aliasing. For running at 50 Hertz you may expect to see modulation at 150 Hz (3rd harmonic), depending on the modulation scheme used by the drive. \$\endgroup\$
    – Jonathon S.
    Commented Aug 8, 2022 at 1:49
  • \$\begingroup\$ Another note. The MPC2561 does not say anything about being galvanically isolated. In the past I have seen circuits use a chip separate to the CAN transceiver for isolation. Adding isolation will increase the latency (or effectively propagation delay), requiring a lower baud rate. In my experience, isolation would be needed in any case that devices on the bus do not share a common ground. \$\endgroup\$
    – Jonathon S.
    Commented Aug 8, 2022 at 3:09
  • \$\begingroup\$ The issue here is that your scope may or may not be misleading you because you are trying to simulate a differential probe with two single ended probes. When you do that, there is an implication that your ground is stable. Given that you have a strong power supply connected to 3 phase, that may not be true, at all. Your scope has a ground and it connects to the grounds on the probes. The point on the circuit that you connect the probes to also has a ground. Finally, the power supply is likely connected to ground. You need to be very clear what is connected to what. Good luck. \$\endgroup\$
    – Andrew Lentvorski
    Commented Aug 8, 2022 at 3:41

0

Reset to default