1

I'm considering designing a CAN bus (CANopen) where part of the bus, closest to each node, does not use twisted pair. Partially the bus will run in a DSUB, but partially through stackable header or single wires / traces. The length of this sections might be 10-50cm and required speeds are around 120 kbit/s but potentially up to 1 Mbit. Number of bus nodes will be countable on two hands.

Is such design feasible and reasonably robust? How long non-twisted traces / connections can still be considered good practice for CAN?

Other useful threads:

CAN Bus Physical Layer Twisted pair or not

NoobPointerException
  • 730
  • 4
  • 18
  • 1
    This all sounds like fairly standard use for CAN, it's not all that picky with twisted pair unless you plan to use very long distances (tens to hundreds of meters), all mostly depending on baudrate. I'm not sure there's a general recommendation for how long you can use non-twisted. – Lundin Jan 03 '23 at 14:31

1 Answers1

2

There is no requirement of a CAN wiring being twisted.

However, the bus wiring needs to have a characteristic impedance of 120 ohms.

A twisted pair cable is just by far the easiest way of providing that characteristic impedance, at a large range of frequencies, and the twists also make it more robust to external electromagnetic interference. Shielded twisted pair cable is even more tolerant to disturbances.

So you can freely use untwisted wiring, and there is no data how long untwisted segments you can have, because the point really isn't if the wires are twisted or not. But it will not be very good unless you can provide the characteristic 120 ohm differential impedance, at the frequencies used for data transmission.

Also it is a bit unclear if the untwisted distances you gave are loop-throughs or just stubs, but recommended stub length of buses is only 30cm on a 1 Mbps bus.

Justme
  • 127,425
  • 3
  • 97
  • 261
  • It will be stubs. Do you have any advice or best prsctice to keep close to 120 ohm when using connection such as wire-to-board connectors or headers? – NoobPointerException Jan 03 '23 at 15:27
  • 1
    With stubs, it always comes down to how long they are relative to the rise and fall times of the signal. – SteveSh Jan 03 '23 at 15:54
  • @NoobPointerException Some general recommendation for CAN stubs is to keep them < 300mm. – Lundin Jan 04 '23 at 10:37