TMS320F28069M: SCI Break Detection interpreted as Framing Error

Hey Cody,

I have been using miniterm as a serial terminal which allows me to toggle on a tx-break, I can leave it on indefinitely(and verified that this just causes rx line to be low at the micro) with a stream of framing errors but never a BKDT flag. I had figured that after one framing error, and a reset of the flags, that I would recieve a BKDT flag, but it never ends up getting triggered.

Thanks

Daniel Rush

Daniel,

please confirm my understanding.

1. You set the RX line low
2. You see a FE.
3. You clear the FE, and RXERROR flags
4. My guess at this point you get stuck in a loop
   1. Clear the error flags,
   2. Receive another FE flag
   3. Clear again and repeat

The only way to clear the FE/RXERROR flags is to reset the SCI, this will reset the logic that sets the BRKDT flag. Assuming that you are handling FE flags in a timely manner (less than 10 bit times) then you will reset the SCI module before a BRKDT condition could be detected.

Are you triggering an interrupt on a FE and resetting the SCI? I would expect this to happen much faster than the BRKDT detection.

Regards,
Cody 

Cody,

1. I send a break(RX line low)
2. Interrupt Triggered
3. Acknowledge Int
4. See an FE
5. Reset SCI
6. Re-enable SCI
7. Return to main execution
8. Subsequent interrupts while line still low are also framing errors.

Here is a code snippet from my interrupt:

   if(char_index >= MAX_MESSAGE_LEN) {
      char_index = 0;
      memset(CMD_buf[write_index], '\0', MAX_MESSAGE_LEN);
    } else if(sciA->SCIRXST & SCI_SCIRXST_BRKDT_BITS) {
      // Break Detect: Handle Reset when break detected
      sendDebugMsg("Break");
      return;
    } else if(sciA->SCIRXST & SCI_SCIRXST_OE_BITS) {
      // Overrun Error
      // TODO(DJR): This just discarding characters might change this in future
      SCI_reset(sciA);
      SCI_enable(sciA);
      return;
    } else if(sciA->SCIRXST & SCI_SCIRXST_FE_BITS) {
      // Framing Error
      SCI_reset(sciA);
      SCI_enable(sciA);
      // TODO(DJR): This just discarding characters might change this in future
      return;
    } else if(sciA->SCIRXST & SCI_SCIRXST_PE_BITS) {
      // Parity Error
      // TODO(DJR): Discard Char
      SCI_reset(sciA);
      SCI_enable(sciA);
    }

Daniel,

a FE always occurs 10 bit times before a BRKDT. With the way your code is written you will likely reset the SCI with every FE well before you would receive a BRKDT.

To prove this theory: comment the SCI reset and SCI enable lines in the else if statement for the FE, you should begin to see BRKDT errors.

Regards,
Cody 

Cody,

This was one of my thoughts as well. So, after testing this, It does trigger the break detect after 10 bits, but, the reason that the SCI_reset is there is to clear the error bits so that another interrupt can be triggered, if they aren't cleared the first FE error is detected, and serviced, then all subsequent interrupts are ignored. As outlined in the technical reference, the bits are only cleared once a SW_Reset is set.

When running the debugger, I can step through the interrupt then once it returns to the main code keep stepping through and watching the RXST register, after some time the BKDT goes high but there's no way to know because it's not causing an interrupt. Not sure how I can keep getting interrupts without resetting the interface each time I receive an error.