I’ve looked through other posts with the same issue but have not found a solution that works. I have a stock setup BlueROV2 Heavy that uses a stock pixhawk and Blue ESCs. A couple of thrusters spin slowly in a couple of second bursts as soon as I arm in manual mode. Right now, that is thrusters 2 & 3 but that doesn’t seem to be consistent. I have also seen thrusters 1 & 3 spin when armed. I have run the calibration of my Logitech F310 controller and correctly configured the directions of my thrusters. Not sure it matters but I am running BlueOS firmware on the latest version. I’m using QGC V4.1.4.
The main potential sources for that which I can think of are:
Physical joystick deadband is larger than calibrated for, so the motors are receiving a non-neutral signal without you pushing the joystick
should be detectable by seeing non-neutral signals (not 1500µs pulse-duration) for the affected motors in the SERVO_OUTPUT_RAW message in QGC’s MAVLink Inspector while the relevant joystick is released
should be fixable doing deadband calibration in QGC
Relative clock inaccuracies between the Pixhawk and the ESC(s)
detectable by thrusters rotating while the relevant outputs are set to 1500µs
in theory this should be fixable by adjusting the output trim (and extent) values via the firmware parameters
ArduSub does not currently support changing the trims, but it should be possible to implement via some refactoring (since the relevant parameters are already available, they’re just not being used)
alternatively, the Pixhawk timings can be checked by measuring with an oscilloscope
if this is the cause, it should be fixable with a replacement flight controller
if the Pixhawk outputs are not the issue then the ESC inputs most likely are
which should be fixable by replacing the misbehaving ESCs
Electrical noise / cross-talk causing a signal on one wire without the relevant output being commanded to that value
this seems unlikely to me, purely because the accepted signal range is reasonably specific, so it would be hard for this to appear with any kind of consistency
should technically be detectable with an oscilloscope, but it may be challenging to do so
“fixing” would require increasing electrical isolation and/or insulation of the affected signal wire(s) from the source(s) of noise
Thanks @EliotBR. I did figure out the issue today. It was just a bad controller. I bought a new Logitech controller and it fixed the issue without any additional modification. I do appreciate your detailed response and I think this will certainly help someone else with the same problem. Thanks!