Dennis wrote:
>>Hi all,
>>
>>I have a project which requires low power and serial communications.  I
>>am able to send and receive messages to and from my pic project to my
PC
>>at the beginning, but the problem comes when I put the pic to sleep.
>> When the device wakes up, I read the status of one of the ports (B)
and
>>send the info back to my PC.  This data gets corrupted somehow before
it
>>reaches the PC (but only about 20 percent of the time).  The pic is
woken
>>up by a change on rb 4 to rb 7.   When the pic has sent it's info to
the
>>PC, it goes back to sleep again.  I am using a max232 for the + and -
12V
>>and it is powered constantly (not though the pic).  If I cause the pic
to
>>interrupt 2 or 3 times a second, this makes the fault worse.  For those
>>of you that might think I don't have the decoupleling capacitor....I do
>>(on both IC's).
>>
>>Does anybody have any suggestions ?
>>
>>Brian.
>>
>>
>
>
>Humm, you don't give much information. One question I would like to ask
is
>which part of the message is being corrupted? If it is the last byte,
then
>the chances are that the PIC is being placed into sleep mode before the
byte
>is fully sent out-> Corruption. You may which to check this, as often
the
>transmit data register empty flag is set while a byte is still being
sent.
>Please provide more information, like the type of PIC, how you're using
it,
>and which byte or bytes are in question
>
>Dennis

I think I have found AN answer to the problem but I don't know why it
should be this way (see bottom of mail).

Ok here we go....   The PIC is a 16C74 and I am using the internal usart
I set the usart up with a subroutine at the beginning of my code like
this:

INIT_USART:     ;FOR 9600 N81 (TRANSMIT)
 BSF  STATUS,RP0  ;BANK ONE
 MOVLW B'00100000'  ;SET TXSTA FOR ASYNC OPERATION
 MOVWF TXSTA   ;AND TO A LOW SPEED
 MOVLW .05   ;SET SPBRG VALUE (9600)
 MOVWF SPBRG   ; ''
 BCF  STATUS,RP0  ;BANK ZERO
 MOVLW B'10000000'  ;ENABLE PORT
 MOVWF RCSTA   ; ''
RETURN     ;RETURN FROM ROUTINE

I am using is a   3.6864 MHz crystal.  The data corrupted is all of it
while the pic is awake .  I do check for the txreg being clear before I
send the next character.  If I transmit any amount of chars before I put
the device to sleep, they come out 100 percent OK.  If I put the device
to sleep and wake up on the port interrupt, sometimes the whole message
displayed is corrupted and the rest of the time, the message is perfect.
 The fault happens about 20 percent of the wake ups.

I did some experiments last night which were:

1:)  Instead of putting the PIC to sleep, I put it round in a loop (still
having the interrupt to break the loop)
 This fixed the problem.

2:) I also disabled the port by clearing the register RCSTA just before I
put the device to sleep and then enabling the port again when it wakes
up.
 This also fixed the problem.

So, I can get round the problem but I wander why the usart needs to be
reset after a sleep... and I can't help thinking that someone else out
there has encountered the same bug (or feature).

Thanks

Brian.