TMS320F28069: CAN behavior in Low temperature

Luiz,

         You have not attached any waveforms. Please attach good and “bad” waveforms. The “bad” waveforms may not be valid CAN frames.

Luiz Ramelan1 said:

The CAN circuit consists of an MCU F28069, Transceiver TCAN1042-Q1, and TVS diode NUP2105L, all of which have operating temperature up to -55.

What you say is true for NUP2105L, but not for the F28069 and TCAN1042-Q1 which are only rated until -40'C.

Luiz Ramelan1 said:

1. Why does the CAN module behave like this in low temperatures?

What is the exact low temperature at which you start to see this issue? Are you using INTOSC as the device clock source? If so, are you performing oscillator compensation as outlined in www.ti.com/lit/SPRABI7?

Luiz Ramelan1 said:

2. Is there any way to cancel the retransmission, such as setting the CANTRR bit or CANTA bit?

You can try CANTRR. If that doesn't work, you can do a soft reset of the module. But, you need to understand the root-cause of the issue first.

Hi Hareesh,
Thank you for your response.

Hareesh Janakiraman said:

You have not attached any waveforms. Please attach good and “bad” waveforms. The “bad” waveforms may not be valid CAN frames.

Please check the attached file below. Signal and CAN log are captured at -23 degC. The CAN routine is 1-second Tx with 8 bytes of data. Good signal shows consistent 1-second intervals. Bad signal shows very dense intervals. CAN log shows a similar pattern. Good signal shows uniform 1-second interval. bad signal interval is around 500usec.

Hareesh Janakiraman said:

What you say is true for NUP2105L, but not for the F28069 and TCAN1042-Q1 which are only rated until -40'C.

yes. Agree. the symptoms start to arise when the temperature is around -20degC and become worse at -30degC and below where i can not find any good signal.

Hareesh Janakiraman said:

What is the exact low temperature at which you start to see this issue? Are you using INTOSC as the device clock source? If so, are you performing oscillator compensation as outlined in www.ti.com/lit/SPRABI7?

I use INTOSC and do not perform osc compensation. To check on the clock matter, I put a counter on the CAN_Tx function and a timer. the counter is still running as normal, only increasing 1-time in 1 second. So, I confirm that the CAN_Tx function is called every 1 second.

The timer value (in seconds) is also consistent when it is compared to a stopwatch.

In fact, the MCU is running a SCI too, which has no problem in the same condition.

Hareesh Janakiraman said:

You can try CANTRR. If that doesn't work, you can do a soft reset of the module. But, you need to understand the root-cause of the issue first.

I did try using CANTRR method. Checking 10ms after transmission and CANTA, if no CANTA is received then trigger TRR. Theoretically, the bad signal can only last for 10ms. in reality, the bad signal can still last more than 10ms, with almost no improvement whatsoever.

bool check_can_ack(uint16_t mbox_num){
    struct ECAN_REGS ECanaShadow;
    uint32_t ack_bit;

    ECanaShadow.CANTA.all = ECanaRegs.CANTA.all;
    ack_bit = (ECanaShadow.CANTA.all >> mbox_num) & 0x01;
    if(ack_bit == 0){
        return true;
    }
    else{
        return false;
    }
}
void clear_can_ack(uint16_t mbox_num){
    ECanaRegs.CANTRR.all = ((uint32_t)0x00000001) << (mbox_num);
}

main(){
...
    if(systemTime.usec - time > 10000 && check_can_ack(CAN_TX.MBox_Num)){
        clear_can_ack(CAN_TX.MBox_Num);
    }
...
}

The problem when Bad signals appear continuously, the CAN module cannot receive messages so I lose control of the device.

.TI-CAN-Issue-20230503.pptx

Hope my problem statement is clear.

Best regards,
Luiz

I did try to use INTOSC compensation. From a quick test, I didn't see any bad signal on CAN so far.

Thanks for the suggestion.

best regards,
Luiz