can_timing.h File Reference

CAN (Controller Area Network) bit timing: a wanted bit rate turned into segment counts. More...

#include <stdbool.h>
#include <stdint.h>

Go to the source code of this file.

Data Structures
struct	can_timing_limits_t
struct	can_timing_t

Macros
#define	CAN_SAMPLE_POINT_DEFAULT   875
	The sample point CiA (CAN in Automation) recommends for rates above 800kbit/s, in permille. Later is better for propagation delay and worse for oscillator tolerance; 87.5% is the usual compromise.
#define	CAN_SAMPLE_POINT_LINK   750
	The sample point this link uses, at both ends.
#define	MCP2515_CNF1   0x2A
#define	MCP2515_CNF2   0x29
#define	MCP2515_CNF3   0x28

Functions
bool	can_timing_solve (const can_timing_limits_t *lim, uint32_t bitrate, uint16_t target_permille, can_timing_t *out)
uint32_t	can_timing_max_bitrate (const can_timing_limits_t *lim)
void	can_timing_limits_mcp2515 (can_timing_limits_t *out, uint32_t crystal_hz)
void	can_timing_limits_twai (can_timing_limits_t *out)
	The panel's TWAI on the ESP32-P4, off the 40MHz crystal: even prescalers from 2 to 32768.
bool	mcp2515_encode_timing (const can_timing_t *t, uint8_t cnf[3])

Detailed Description

CAN (Controller Area Network) bit timing: a wanted bit rate turned into segment counts.

A CAN controller divides its source clock into time quanta and splits each bit into one sync quantum plus two programmable segments. A bit rate that is a fraction of a percent off, or a sample point in the wrong place, works on a short cable with one other node and logs errors when the bus is longer, colder or busier.

The arithmetic is here, in pure C with tests, rather than in a table of register values. Two controllers use it: the panel's TWAI (Two-Wire Automotive Interface, the ESP32-P4's CAN controller) and the IOmcu's XL2515. They have different clocks and segment limits and the same arithmetic.

The rule this module enforces: the bit rate comes out exact. CAN has no framing to resynchronise against beyond its own bit stuffing, and two nodes that disagree by 1% agree on short frames and fail on long ones. A rate that cannot be hit exactly is reported as impossible; an 8MHz crystal cannot make 1Mbit/s.

SPDX-License-Identifier: MIT

Macro Definition Documentation

CAN_SAMPLE_POINT_LINK

#define CAN_SAMPLE_POINT_LINK   750

The sample point this link uses, at both ends.

The IOmcu's XL2515 has one way to make 1Mbit/s from a 16MHz crystal: the smallest divisor and 8 quanta per bit, the fewest a bit may have, which puts the sample point at 75%. The panel has slack and matches the IOmcu, so both ends sample the bit in the same place.

Function Documentation

can_timing_limits_mcp2515()

void can_timing_limits_mcp2515	(	can_timing_limits_t *	out,
		uint32_t	crystal_hz )

The IOmcu's XL2515 (MCP2515-compatible), given its crystal.

can_timing_limits_twai()

void can_timing_limits_twai

(

can_timing_limits_t *

out

)

The panel's TWAI on the ESP32-P4, off the 40MHz crystal: even prescalers from 2 to 32768.

Parameters

out	Filled with the limits; NULL is ignored.

can_timing_max_bitrate()

uint32_t can_timing_max_bitrate

(

const can_timing_limits_t *

lim

)

The highest rate this controller can reach exactly, or 0 if none.

can_timing_solve()

bool can_timing_solve	(	const can_timing_limits_t *	lim,
		uint32_t	bitrate,
		uint16_t	target_permille,
		can_timing_t *	out )

Solve for the segmentation closest to target_permille that hits bitrate exactly.

False when no combination does, which means the source clock cannot make the rate.

mcp2515_encode_timing()

bool mcp2515_encode_timing	(	const can_timing_t *	t,
		uint8_t	cnf[3] )

Encode a solution into CNF1..CNF3.

cnf receives three bytes in register order CNF1, CNF2, CNF3.

tseg1 is split into propagation and phase-1 segments here, because that split is the MCP2515's business: both are 1..8 quanta, and the bus does not care where the boundary falls. Returns false if the solution cannot be expressed; a tseg1 of 17 or more has no legal split.