FBL2360 RoboCAN failure?

We are using an application with seven FBL2360 controllers linked via RoboCAN. We had things configured where they were all communicating with each other, but within a single day RoboCAN communciations has failed on three of the seven units. As far as I can tell, the units are working in all other respects, they just can no longer communicate via the CAN bus with the RoboCAN protocol. The afflicted units cannot query other units over the CAN bus and the other units cannot query values from the afflicted FBL2360s.

I've tried resetting the configuration to a factory configuration and reflashing the firmware on these units, but nothing seems to get them into a mode where RoboCAN is functioning. I've also verified the CAN cabling with spare FBL2360s we have on hand and it checks out. These units just can no longer communicate via RoboCAN.

Is this a known issue with RoboCAN or CAN communication on these units? Is there anything I can do to diagnose this further? Without understanding what happened, I'm worried we'll lose another unit soon.

Any help, suggestions or insight into diagnosing and fixing this issue would be very gratefully appreciated.

Thanks,

Mike

Investigating the issue in more detail, it appears that different FBL2360 units have different sensitivities to the specifics with regards to how the CAN bus is wired and terminated. For our chain of seven FBL2360's we were terminating each end of the bus with 120 ohm resistors which seems to be a pretty standard CAN bus configuration, but communication with certain controllers were failing. By swapping the controllers around on the bus and with our spares, we finally arrived at a bus configuration that seems to work. Rather than being terminated at both ends of the bus with a 120 ohm resistor, we instead have it terminated at the far end (opposite of the master controller which queries of the other units) with a single 120 ohm resistor. We also had to cut the data rate down from 1000 KHz to 500 KHz to get reliable communication.

I'm wondering what other people have found with regards to the specifics of creating a CAN bus for a half-dozen or more RoboteQ controllers. Or, if there are good resources on the web for on creating a CAN bus that seem particularly applicable to networking RoboteQs. We aren't experts in the specifics of CAN communication and obviously our naive approaches based on our limited knowledge are the root causes of our problems.

Are you using a twisted pair (preferably shielded) for your cabling? Is the cable's nominal impedance 120 ohms? How long is your daisy chain? Where does it fit in the length vs. bit rate spec for CANbus?

Yes, we are using twisted pair, but unshielded. The total length is just under four feet. I'll check with our electrical engineer regarding the nominal impedance and validate the length vs. bit rate.

At four feet the system should work with any twisted pair and probably even with gross mismatches at the termination resistors. I have a 5 node rawCAN system oif about the same length running at 500k with shielded 100ohm impedance cable (not 120) and termination resistors at both ends. It will also work without dropouts with just one termination resistor, and with only some dropouts with none at all. Only one node is a Roboteq controller (with TI transceiver), the others all use MCP CAN controllers and transceivers. At 1000 kHz it is less reliable, and the scope shows a lot more noise on the bus.