F28035: CAN communication stops working at -5°C and below when using internal oscillator 1 as the clock input

It also does not work at 60°C and above.

To determine the cause of “CAN stop working” problem we monitored CAN_RX and CAN_TX lines on oscilloscope while the DSP was exposed to the low temperature only (the rest of the circuit was at about 25°C - just regular ambient temperature). We also used a PC application that sends a request to DSP first and then it receives a response to the request from the firmware.

1. When CAN is working: there is one TX message as a response to each RX message (confirmed by oscilloscope and by the PC application – the data received from the DSP were correct)
2. When CAN is NOT working: after one RX message comes to the DSP, TX messages keep coming out of DSP. All TX messages have same signal pattern and it is different than the one from the test #1. It looks like the message from RX line is just being resent over and over again. Also, the PC application never receives correct response (the TX message is not a response to the message on RX line)

The solution: After switching to the external crystal clock input, both problems disappeared and CAN works reliably in temperature range from -40 to 85°C.

Although, we’ve solved the problem we would like to understand how the temperature affects the internal oscillator 1 so the CAN stops working. The CAN is run at 125 kbps.

Thanks,
Slavica