Posted on March 11, 2015 at 18:35This sounds like a generic question on assigning node IDs in a network, rather than something specific to CAN. There are many ways to do this. A static approach uses some information available to the controller to calculate a unique node ID. This can be an address switch, an address stored in flash at the final acceptance test during assembly, or some hash derived from a unique identifier in the controller. A dynamic approach is to use an address manager node to assign IDs when requested, along the lines of DCHP in TCP/IP nets or LSS and flying bus master in CANopen networks.
You can listen for active nodes, assuming each node sends some kind of heartbeat message. Build an address list and then assign an unused node. This has several problems. First, it takes time to collect active nodes. Second, there may be inactive nodes with fixed address that come on line later and collide with your apparently unused address. And third, what happens if EVERY node listens first and then all pick a node at the same time?
Since you mention ECU (engine control unit) are you using the J1939 CAN protocol? Many CAN protocols incorporate the node ID into the message ID rather than in the data section, since you can't prioritize or use collision detection in the data bytes.
Jack Peacock
