Save as PDF

CAN FD

Plugfest in Detroit

Published 2015-03-31

CiA has organized a CAN FD plugfest in close cooperation with General Motors (GM). About 14 companies tested their products for interoperability.

About 50 participants and observers joined the plugfest in Detroit (Photo: CiA)

TWO TEST ENVIRONMENTS WERE EMPLOYED: a bus-line topology by Bosch and another one by GM. The Bosch wiring harness was made of single cables with a total length of about 23,5 m with very shorts stubs. GM’s approach was a pre-configured single 24-m cable with stubs up to 1,7 m. In both test scenarios, an arbitration bit-rate of 500 kbit/s was used. The dataphase bit-rate was increased starting with 1 Mbit/s to 2 Mbit/s, 4 Mbit/s, and 5 Mbit/s up to 8 Mbit/s. The connected nodes supported the non-ISO CAN FD protocol. Products from Bosch, Copley, Dspace, Etas, GM, HMS/Ixxat, Intrepid, Kvaser, Melexis/IHS, Microchip, NXP, Peak, Renesas, Spansion, STM, and Vector were connected to the bus-lines. Keysight monitored the network with an oscilloscope.

Intrepid and Melexis/IHS used CAN FD micro-controllers from Freescale and Melexis provided its own ISO 11898-2 transceiver qualified for bit-rates higher than 1 Mbit/s. Another transceiver specified for data-rates up to 5 Mbit/s was provided by Microchip. Kvaser used it, too. On the other hand, Microchip implemented the Kvaser CAN FD core in its stand-alone controller. It was shown for the first time at a CAN FD plugfest. Another newcomer was Copley. For some time, the servo controller specialist has implemented the Classical CAN protocol in an FPGA. The company now also provides a CAN FD implementation.

NXP participated with its FD shield transceiver. The product has been improved to overcome the data inconsistency issue of the CAN error passive mode. The transceiver now maintains an error counter, so that the compound of FD shield transceiver and shielded Classical CAN controller behave as defined for “CAN FD tolerant” nodes in the upcoming ISO 11898-1 standard. In case the FD shield detects that the shielded CAN controller does neither give acknowledge (which indicates a incorrectly received CAN message) nor send an active error flag, then the FD shield transceiver sends an active error flag on behalf of the Classical CAN controller. This is done in the end-of-frame field, to enforce a retransmission by the sending node.

Five companies (Bosch, Ixxat/HMS, Kvaser, Peak, and Vector) tested their ISO CAN FD implementations using the GM cabling. They achieved data-rates up to 10 Mbit/s. This seems surprising, but there were fewer nodes connected than in the other tests. In addition, they all used the same bit-timing including the sample-point.

For the next plugfests it seems advisable to use a CAN ground and to test the CAN FD products against the evaluation board by Nakoinz in advance. Of course, the bit-timing as specified by CiA should also be respected in detail. Even little differences in the bit-timing can cause malfunctions.