Post by thread:[PIC]: Lost Interrupts using TMR0 receiving serial

I am attempting to use a pic 16f84a to receive serial data at 9600 baud and display it onto a 4*20 LCD. I use the rb0/int function to detect the falling clock edge (indicating the beginning of a start bit). When this is detected I disable the rb0/int and start TMR0 to cause an interrupt during the middle of the first data bit. When this interrupt occurs I detect the data bit and store it into serialData. I then restart TMR0 to cause an interrupt during the middle of the next data bit. I do this until I have received an interrupt during the stop bit. During this interrupt I pass the received byte(InByte) up to the main level code. I then disable TMR0 and re-enable the RB0/INT function to detect the next start bit. The main level code just loops looking for a non-null value in this passed byte. When one is detected it is displayed on the LCD panel. This will be changed to a 32 byte circular queue to hold data until it can be displayed. This all seems to work great, UNTIL you send serial data to the PIC faster than some(undetermined) rate. 20 to 30 bytes per second is handled fine. However if data is sent to fast, the pic seems to get into a state where the TMR0 is not generating interrupts. I have gone so far as to use external LED's to display the current status of my state machine logic. The LED's are clearly indicating that the state machine is waiting for TMR0 interrupts, but not receiving them. I fully realize that data (at 9600 baud) can come in much faster than the LCD can display it. But, I should just see lost or garbled characters, not a complete apparent lock-up of the pic. This definitely seems like some kind of timing issue, since it will receive thousands of bytes without error (as long as they are not too close together). Yet it will fail almost instantly when back to back bytes are received. I have spent many hours trying different things like setting the INTCON bits within a loop, testing to see if it really was set. I have used nops, switched code around etc... Currently the code does not disable the RB0/INT interrupts, they are just simply ignored by my state machine implementation. Since they are not disabled, I can tell by the oscilloscope that the pic is still alive and running, responding to the incoming falling clocks, but not getting its TMR0 interrupts. Below is my interrupt handling routine. Please don't laugh as I am new to this architecture. Although I have been programming in various assembly languages for about 15 years. I have been doing this long enough to know that just when you think the hardware is not working properly, you need to look harder at your code. However, I am aware that the pic does contain timing quirks when it comes to the RMW. Am I going crazy or what? I will be eternally grateful to anyone who can help me with this. #define SERIAL_IN PORTB, 0 #define INTSCOPE PORTA, 2 ; This pin is monitored with an Oscilloscope to verify interrupts are occurring #define RED_LED PORTA, 1 ;LED's to indicate state of the state machine #define YEL_LED PORTA, 0 org 0x004 movwf W_TEMP ;Save everything swapf STATUS, W movwf STATUS_TEMP ; Interrupt handler right here bsf INTSCOPE ;Show heartbeat pulse 5us wide nop nop nop nop bcf INTSCOPE btfsc INT_FLAGS, L4DB ;Are we looking for data bits goto CK4DB btfsc INT_FLAGS, L4SB ;Are we looking for a start bit goto CK4SB goto IntExit ;Who knows how we got here CK4DB btfsc INTCON, T0IF ;Is it timer0 goto DataBit bcf INTCON, INTF ;Just in case goto IntExit CK4SB btfsc INTCON, INTF ;Is it a start bit goto StartBit ;Yes bcf INTCON, T0IF ;Just in case goto IntExit StartBit bsf INTSCOPE bcf INTSCOPE ; bcf INTCON, INTE ;turn off falling edge interrupt bcf INTCON, INTF ;clear int occured flag bcf INT_FLAGS, L4SB ;Found the start bit bsf INT_FLAGS, L4DB ; now start looking for data bits bcf RED_LED bsf YEL_LED clrf DBCount ;Init Data bits count movlw d'170' movwf TMR0 ; delay till middle of first data bit nop bcf INTCON, T0IF ;always do this before enabling nop bsf INTCON, T0IE ;turn on the timer interrupt btfss INTCON, T0IE ;make sure it is really on goto $-2 ;if not, try it again goto IntExit DataBit incf DBCount, F movf DBCount, W xorlw 0x09 ;Do we have 8 bits yet? btfsc STATUS, Z goto StopBit ;Stop bit received ; This is Data bit to keep bsf INTSCOPE ;1 us heartbeat pulse bcf INTSCOPE movlw d'175' ;Re-load Timer0 for next data bit movwf TMR0 bcf INTCON, T0IF ;Clear Int occured flag nop ;something to try, but it doesnt work bsf INTCON, T0IE ;Re-enable Timer0 btfss INTCON, T0IE goto $-2 ; Read in the bit and pre-pend it to what we have so far ; clrc ; assume clear carry btfsc SERIAL_IN ; if receiving 0 setc ; set carry rrf serialData, F ; roll serialData bits directly into the field goto IntExit StopBit bsf INTSCOPE bcf INTSCOPE bcf INTCON, T0IE ;Turn off timer0 bcf INTCON, T0IF movf serialData, W ;Get the byte received and movwf InByte ; pass it to main level loop bsf INT_FLAGS, L4SB ;Looking for a start bit now bcf INT_FLAGS, L4DB ; not data bits bsf RED_LED bcf YEL_LED bcf INTCON, INTF ;clear interrupt occured flag nop bsf INTCON, INTE ;Turn the rb0/int back on for falling edge detection btfss INTCON, INTE ;make sure goto $-2 ; Restore everything and return IntExit swapf STATUS_TEMP, W movwf STATUS swapf W_TEMP, F swapf W_TEMP, W retfie Michael Brown Instant Net Solutions http://www.KillerPCs.net -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

Thank you for your reply. I really need all the help I can get. It turned out that I definitely was losing my mind. I was switching to bank 1(to flip portb between input and output to do scrolling on the LCD and to check the busy flag on the LCD) in the main level code, but leaving interrupts turned on. Sure enough, with a short enough pause between bytes, I would end up in the interrupt routine while switched to bank 1. I couldn't think of better way to work around this than to load FSR with the address of TRISB (since this was the only field I needed in bank 1) then I just used INDF to access the TRISB leaving the main level code running in bank 0 only, and not ever switching. I saw a web page(http://www.phanderson.com/PIC/16C84/interrupts/interrupt_1.html) where he was saying that you could just switch to bank 0 in the interrupt routine and not worry about the bank that was being used. Since the bank pointer would be re-established when you restored the STATUS register before executing the RETFIE. This doesn't seem right to me since in his example he is blindly saving the W reg and the STATUS regs into whichever bank was currently selected then restoring them from bank 0 before exiting. This would seem to lead to unpredictable behavior. :-o But thank you very much anyway. Wouldn't you know I would struggle for days on this and then finally ask for help right before figuring it out. It seems to work fine now, only I lose characters at high speeds. That's OK though, now to implement that circular queue. Michael Brown Instant Net Solutions http://www.KillerPCs.net {Original Message removed}