Do not confuse customers!
At SPS IPC Drives 2017, visitors asked at the CiA booth on the availability of industrial micro-controllers featuring on-chip CAN FD modules.
During the 3-day fair for industrial control, CiA staff was asked about the availability of small-sized micro-controllers with CAN FD interfaces. “We are still missing small industrial µC with ISO CAN FD,” was an often-made statement. Such parties need those MCUs for integration into sensors and tiny motor controllers. Unfortunately, there are mainly powerful MCUs with multiple CAN FD interfaces for sophisticated applications available. Most of them are qualified for automotive applications. Of course, those products are oversized for simple devices. And they are too expensive. Infineon, NXP, and ST-Microelectronics have such small-sized products with limited resources on their roadmaps. According to information given to CiA, such MCUs are planned for late 2018, beginning of 2019.
An exception is Microchip: The company inherited from its Atmel acquisition the SAM family of MCUs. These products provide up to two CAN FD interfaces. According to Microchip’s website, these modules comply with CAN-FD 1.0. This is also declared in the related datasheets and handbooks. But the term CAN-FD 1.0 is misleading. The CAN FD standard, ISO 11898-1, was published two years ago. CiA and the chipmakers agreed on a common terminology: ISO CAN FD for implementations compliant to ISO 11898-1:2015 and non-ISO CAN FD for controllers not providing a bit-stuff counter (not foreseen in the original CAN FD 1.0 specification by Bosch). By the way, the two versions are incompatible. Unfortunately, early birds have already designed-in non-ISO CAN FD controllers for some embedded applications. This is, why those products will be available also in the future.
There is no problem, when using just non-ISO CAN FD conformant chips in one single network. The only disadvantage: The non-ISO CAN FD protocol is academically less reliable than the ISO CAN FD protocol. But if you mix both protocols in a network, no communication is possible anymore. In the meantime, there are just a very few non-ISO CAN FD implementations on the market. When asking Microchip (formerly: Atmel) what they mean with CAN-FD 1.0, the company answered the MCUs support ISO CAN FD. The products have implemented the new version of Bosch’s MCAN core.
CANopen FD and the CiA 602-2 application layer (mapping of J1939 parameter groups to CAN FD data frames) require ISO CAN FD compliancy. Therefore, it is necessary to distinguish clearly between ISO CAN FD and non-ISO CAN FD. No new term such as CAN-FD 1.0 should be introduced. Also the term CAN 2.0 (A/B) is not more state-of-the-art: Legacy CAN is the better term. Since 1993, the CAN data link layer protocol is standardized in ISO 11898, and since 2003 in ISO 11898-1.
The number of configurable time-quanta per bit should be higher than in the ISO 11898-1 standard specified, if you like to achieve a robust communication. This means, the chipmakers should not refer only to the ISO 11898-1 standard, but also to the CiA 601 series. Part 2 recommends, for example, the number of time-quanta for different bit-rates at different oscillator frequencies. The larger the number of time-quanta in the arbitration phase is, the smaller is the quantization error (plus/minus one time-quanta).