Continuing on the last entry I am in the process of routing ECUMain so that it have CAN + Power on the same M12-5 connector. PSU Schematics is shown below. It is nothing special about this – D4 protect against wrong polarity, U9 measure the current usage and since this connect to the MCU it needs to be galvanic issolated. This is not the best current sensor, but is suficcient and have isolation on the chip. FL1 is a choke trafo protecting against EMI pulses and PS1 is a 24 to 5V 2A module with isolation. U6 is the 3.3V regulator for the local board. This isolation is important because the 5V also feeds up to 10 IO boards. PS1 can be replaced with a 4A version if needed.
This means I have two issolation barriers to route so far – PS1 and U9 -and CAN interface add 3 more chips on the barrier. CAN tranceiver isolation is 5000V, while PSU isolation is ca 1500V + some extra help from the choke on pulses. If I was to increase PS1 isolation I probably would need a lot of extra cost – these modules are expensive.
TMC1101A3 is Hall Sensor based. They work, but they are inaccurate and drift a bit with temperature, so I will in general not recommend these current sensors for everything, but in this situation they are usefully to save space and I only want a shortcut detection on the modules and a capability to measure module current profiles.
Routing itself this is easy , but making space for proper galvanic issolation space is an issue. The picture below illustrate the PCB and the yellow areas are the galvanic isolation components. An old handrule is that you need ca 5mm space between the sides.
Looking above at left the lower chip is ca 3mm and this is the DC/DC for the CAN isolated at ca 1500V, so the actual isolation is 1500V on PSU, but trancient protection is usually much higher. A common mistake is to not respect the space and as you can see I lack space to put this in so I will have to reorganize a lot to get a proper galvanic isolation on the PSU connector. I bring this up because I recently looked at a CAN Tranceiver from an European company and was shocked to see how they placed the components – the issue is that a EMI spike can be thousands of volts and jump from lane to lane if they are to close – and if that happens with anything connected to the MCU it will risk being dead afterwards. On CE/FCC testing we will be pulsing the lines with 2KV spikes for test and SW need to run during the test.
The actual result can be seen above. I diveded ground into three areas – the larger is 3.3V/MCU, the small bottom/left is Ethernet and the PSU is thetop-left. This should work well – thought I will look for replacements for C33 and C34.