I am trying to understand the CAN1 BaudRate. Why can't I have a baud rate of exactly 1 Mbps?

[KiptonM]

I have a STM32F446 core and the Peripheral clock is 45 MHz.

I am looking at the RM0390 Rev 6 Section 30.7.7 Bit Timing

I am looking at the documentation Figure 395. Page 1063/1347

I want a bit rate of 1 million bits per second (Mbps)

In my calculation using the definitions in Figure 395, BRP = 14, TS1 = TS2 = 0 gives me exactly 1 Mbps.

But the Configuration Tool says the prescaler should be 16 (BRP = 15) which gives a Time Quantum of 355.555555555554 and with Time Quanta in Bit Segment 1 = 1 Time (TS1 = 0)

and Time Quanta in bit Segment 2 = 1 Time (TS2 = 0) The time for 1 bit is 1066.66 ns and the Baud Rate is 937500 bit/s

When I try to change the Prescaler (for Time quantum from 16 to 15 (BRP from 15 to 14) The Time Quantum is 333.3333333333...

the time for 1 bit is 999.99 ns (there is a truncation as opposed to rounding) and the Baud Rate is 1,000,000 bit/s

Am I getting hit with a floating point rounding error? And it thinks it is more than 1000000 bit/s as opposed to exactly 1000000 bit/s?

Can someone explain why it does not work with 15 (BRP 14) for the Prescaler (for the Time Quantum?

Is there more documentation on this in addition to RM0390?

[KiptonM]

[KiptonM]

I think it is a rounding error.

[Tesla DeLorean]

Pretty sure the goal is not to have a 1+1+1 solution, but rather that you prescale by 3 and then have the timings add to 15 (SJW + BS1 + BS2)

[Javier1]

I had the same problem, i fixed it by rising the peripheral clock frequency.

I also think this is a java rounding bug from cubemx.

You can always try to set the time quantas you want after the code is generated by cubeMX.

Overwrite them before can_init and take your chances....

[KiptonM · Accepted solution]

Found the solution. there was a Java bug in the calculation. I upgraded STM32CubeMX to the latest version from 6.1.1.20201209 to 6.2.1.20210324 and it now gives me Time for 1 bit 999.99 ns and Baud Rate 1000000 bit/s.