search

OpenCyphal / public_regulated_data_types

Regulated DSDL definitions for Cyphal (standard and third-party)
https://nunaweb.opencyphal.org
MIT License
72 stars 96 forks source link

Define a new standard register for CAN sample point configuration #110

Open pavel-kirienko opened 3 years ago

pavel-kirienko commented 3 years ago

I do concur with @thirtytwobits about having a sample point option as well. Or at least thinking about minimal sample point or something like that. MCU's can indeed calculate timings (e.g. https://github.com/apache/incubator-nuttx/blob/master/arch/arm/src/s32k1xx/s32k1xx_flexcan.c#L393). However only having the timing from just the bitrate isn't enough you might end up with a low sample point e.g. 60% which is non-ideal.

Regarding DS-015 we participated in the network topology discussion and stated if that if there's a well defined configuration of nodes and it's topology we could help analyzing the impact of different bitrates on such a topology.

_Originally posted by @PetervdPerk-NXP in https://github.com/UAVCAN/public_regulated_data_types/pull/109#discussion_r591181563_

I presume it should be named uavcan.can.sample_point_pct and typed uint8[2], for arbitration and data segments, respectively.

PetervdPerk-NXP commented 3 years ago

Hi Pavel here some background information

The sample point is the location, typically given as a percent value, inside each bit period where the CAN controller looks at the state of the bus and determines if it is a logic zero (dominant) or logic one (recessive). All CAN controllers allow this point to be configured and it is always specified as a percentage from the start of the bit period. The location of the sample point is a trade-off. An early sample point decreases the sensitivity to oscillator tolerances and allows lower-quality oscillators. A late sample point allows for a longer signal propagation time and therefore a longer bus. A later sample point is useful for non-ideal bus topologies. The CAN-in-Automation user’s group makes recommendations of where the sample point should be for various bit rates, with 87.5% the most common. "FAQ contribution from Embedded Systems Academy, experts in CAN bus. For more information visit www.esacademy.com"

Sample points recommendations

(87.5 % is the preferred value used by CANopen and DeviceNet, 75% is the default value for ARINC 825).

Furthermore this website gives a nice overview of bitrate calculation http://www.bittiming.can-wiki.info/

pavel-kirienko commented 3 years ago

Well the case of Classic CAN seems obvious enough, but for CAN FD there were no well-formalized recommendations last time I checked. https://forum.uavcan.org/t/uavcan-can-can-fd-profile/867