I have been doing some research and measurements today. The first thing I did was remove the high current return path from the groundplane and wired them directly to the 24V power terminal on the PCB. This should keep the current from flowing through to the control-pcb. To no avail.
Next i did some measurements and it seemed that the following happens. The input voltage is 5v, which drops about 0.4v over the polyfuse. The digital curcuit draws about 180mA and has its return over a 26AWG wire of about 40cm long.
It seems that when the current flows through the power PCB it raises the ground potential by about 0.5v on the controller plain. This is no problem for the digital logic, which runs on a 3.3v regulated from the 5v by an LDO, but the CAN drivers run of the 5v and have a cut out voltage of about 4.3v.
As you guessed, the voltage drops below the 4.3v and the can drivers stop functioning.
I have been playing around with the ground leads on the power supply, as you mentioned this could get me into trouble. When i do a current measurement between the two ground leads at the end of the leads, there runs about 2.5A of current ( all testleads are 2,5mm² or 4mm² ) so this really surprises me. When i disconnect the coupling groundlead from the power supply and reconnect it at the end of the testleads the voltage drop disappeares and all works as expected.
In the original device, the 24V is the main power supply and a 5V is created using a buck convertor.
The situation inside the device will be the following :
24V PSU -> 1,5m wire 8x22AWG ( 4x 24V + 2x GND ) -> buck convertor -> 1,5m wire 8x 26AWG ( containing 2x 5V and 2x GND along with the CANbus ) -> Controller PCB -> flatflex cable of 20cm -> Power PCB <- 24V PSU
As all can see there is a ground loop over there, because to be able to control the mosfets and read the current back the 24v and 5v grounds meet at the flatflex cable. One thing in the test that also improved the situation was when i connected the testlead of the 5v directly to the pcb, effectively bypassing the 26AWG wire. Which to me means that there is to much impedance in the ground returnpath of the 5v circuit.
I will attach images of the two circuits and the two PCB's. Feel free to laugh but please point out the most obvious errors, since I have no real clues about designing high power electronics. All other faults are also welcome as i am open to learn all i can.
With regards,
Michel
P.S. All the capacitors on the power pcb have been removed due to "current limiting" issues of the mosfet drivers. ( datasheet says that they limit the current, reality says the chip disables itself )