Normally, I'd advocate use of shielded twisted pair for CANBUS networks but there have been some questions recently about using unshielded cabling to achieve the same aim.
From what I can see there are some options available on how to "construct" unshielded twisted cabling and these are:
a. Three conductors. CANH, CANL and "drain", which are all twisted together, with 30 twists per metre. e.g. Get three wires in a hand drill, attach the other ends in a vice and twist them so that they interleave.
b. Three conductors. CANH, CANL and a "drain"; CANH and CANL are twisted together as for a normal twisted pair but the drain wire runs down the centreline of the twisted pair.
c. Three conductors. CANH, CANL and a "drain"; CANH and CANL are twisted together and the drain runs external to the twisted pair.
Which is the best option of these 3 in terms of being closest to match the shielded twisted pair in performance?
Are there any pros and cons of these 3 approaches?
Is termination of this CANBUS type via the normal 120R?
For the drain connection, should this be connected to a single star point only or is it OK to connect it to one or more ECUs if they have a capacitively coupled path (1R + 68uF) to ground?
I imagined that option b might be the best approach to allow for common mode noise in the twisted pair that would be subtracted at the CAN transceiver but I'm open to being corrected.
Is there anything to particularly look out for when using unshielded cabling, such as limiting the number of nodes, nodes spacing, stub lengths, etc.? I believe that there's a notional limit of 10 nodes where shielded twisted pair is not used but I'm not sure what kind of validation is required to be performed if more than 10 nodes is required.