iCC 2013: CAN FD
No surprise, CAN FD was the main topic at the 14th international CAN Conference (iCC). About 130 participants discussed the opportunities and challenges of the improved CAN protocol in detail. In particular, the robustness and use cases were debated in detail.
KEYNOTE SPEAKER DR. MARC SCHREINER FROM DAIMLER STARTED with an OEM's point of view. He reported about internal research activities and stated that CAN FD matches future requirements. Because the German carmaker doesn't only want to use CAN FD for software download purposes, but also for the normal operation, he requested small micro-controllers with CAN FD. They will be used to connect simple ECUs to CAN FD in-vehicle networks. He made it clear that from an OEM's point of view, the 64-byte extension of the data field is as important as the higher speed.
Peter Decker (Vector) and Dr. David Gary Hickman (Etas) discussed the usage of CAN FD for calibration purposes. The Asam association, who specifies the XCP higher-layer protocol, has already extended version 1.2 in order to support CAN FD. Vector's paper showed a realistic throughput of about 1,2 Mbit/s, when using a 2-Mbit/s data-phase bit-rate. Hickman also presented some measurements of software download times and busload figures. For the XCP protocol, he measured a busload of less than 5 % when using a 5-Mbit/s data-phase bit-rate, compared to a 30-% busload for legacy CAN. Within ISO a task force has already adapted the transport layer protocol for diagnostics on CAN (ISO 15765-2) for the CAN FD data link layer protocol. It is intended to support software downloads of up to 4 GiB. This standardization activity is supported by all major carmakers.
Dr. Martin Merkel (Ixxat) reported in his presentation the status of the CANopen application layer in respect to the CAN FD adaptation. He showed that PDOs will be limited to 64 mapping objects and that SDO services will make use of the 64-byte payload provided by CAN FD. The adaptation of the CANopen layer setting services has not yet been started.
In order to support CAN FD, several providers of CAN high-speed transceivers announced products qualified for 2 Mbit/s. In addition to NXP, Bosch, Freescale, Infineon, and Texas Instruments will provide those transceivers, respectively system base chips that also integrate voltage regulators. Partial networking transceivers compliant with ISO 11898-6 tolerating CAN FD frames are also under development by different suppliers.
Most of the manufacturers of automotive micro-controllers will provide engineering samples supporting CAN FD next year. Unofficial pre-announcements were made by Freescale, Infineon, Microchip, Renesas, Spansion, ST Microelectronics, and Texas Instruments during the conference. Most of them will start using the M-CAN core from Bosch. But there are also other IP cores, e.g. from Kvaser, Peak, and Vector. At the iCC, Daniel Leu from Inicore introduced their CAN core, which supports CAN FD. The presented implementation currently uses an Amba APB3 bus interface and will be updated with an AXI interface.
Several iCC papers focused on the CAN FD physical layer, in particular the bit-timing requirements and how to achieve sufficient robustness. Although the CAN FD protocol features the same or even better reliability, the communication should not be disturbed due to a poor physical layer design. In his paper Florian Hartwich (Bosch) presented the bit-timing requirements in detail. He concluded that both bit-times (for arbitration and data-phase) should be coordinated. This means, it is recommended that the time-quanta should be exactly the same. He also discussed in detail the transmitter delay compensation. Dr. Arthur Mutter (Bosch) talked about the rules to calculate the oscillator tolerance and the bit symmetry (edge deviations). Because there are several error sources which can't be calculated in all details, the presented formulas are rules of thumb. This means, the robustness needs to be proved by prototype installations. There are two classes of errors: accumulating errors and non-accumulating errors. The accumulating errors include frequency errors (clock tolerance) and bit-rate configuration errors (different nominal bit-rates in sender and receivers). The non-accumulating errors may be caused by quantization errors, by bit-rate switch errors, or by bit-symmetry errors.
Scott Monroe (Texas Instruments) presented and evaluated several options of mixing classic CAN and CAN FD controllers in one network system. This included the concept of silent mode with current high-speed transceivers, the use of partial networking transceivers, and managed CAN hubs. In addition, he discussed the idea of transceiver circuitry blanking CAN FD messages. His conclusion was that while the future looks bright for CAN FD, in the short term (the intermediate time before all CAN controllers provide CAN FD active or passive functionality) compromises have to be accepted.
News and reports