On Sun, Jan 13, 2002 at 03:30:08PM -0800, Rick Mann wrote:
> Am I just dense?

Nope. Not dense at all.

 I've read the Microchip docs, a fat book by Myke Predko,
> and I've looked through the FAQ. But I have what appear to be bizarre
> problems trying to write to the various port latches on the '877.
>
> I've been working on interfacing a parallel LCD to a PIC 16F877. I
> succeeded, but only after noticing a (possible) problem with port A. If I
> execute the following instructions (all of port A is configured as output):
>
>         bsf     porta,0
>         bcf     porta,1
> hold    goto    hold
>
> then both RA0 and RA1 will be cleared (RA0 will be set momentarily). This
> was causing the assertion of RS on the LCD to fail (or at least not last
> long enough), because I was asserting RS via a port a pin, then pulsing E
> via a different port a pin. I solved the problem by using port B pins
> instead, and the same exact code (with the exception of the different port)
> started to work.

This is a side effect of doing a read/modify/write operation on a port in
successive cycles. It's a confluence of a lot of issues:

- bsf/bcf operations read/write the entire port, not just the bit in question.
- read operations on a port always reads the actual pin state, not the output
  latch. This occurs even when the bit is configured for output.
- reads occur in the first quarter of the instruction cycle while writes
  occur in the last quarter of the instruction cycle.

So here's what happens:

- The bsf sets the output latch of bit 0 in the last quarter cycle of the
  instruction. The output driver starts to drive the pin high.
- Then in the first quarter cycle of the bcf, the PIC reads the entire port.
  But because the output drive hasn't had sufficient time to drive the pin
  high. So when the PIC reads the pins (not the output latches), it sees
  bit 0 still as low.
- In the write cycle in the last quarter cycle, the PIC writes the entire
  port. But since it read a 0 for bit 0, it writes a 0 for bit 0, therefore
  wiping out the effect of the bsf.

>
> At first I had RS connected to RA4, an open-drain output, so I moved both RS
> and E to RA0 and RA1, but I kept having the same bsf/bcf problem.

Actually it probably have worked a bit more consistently on RA4 with a pullup.

But the solutions are simple and well known:

1) Simply give the output driver time to drive the pin. Usually this is done
   by putting a NOP between successive RMW instructions to the same port. I
   do this with fastedious (sp) devotion.

2) Use shadow registers where maintains the state of the port. So in this
case you'd set the state of the shadow register, then transfer the register
to the port. It gets rid of the RMW on the port, thereby alleviating the
problem.

>
> Is there something simple I've overlooked? I also tried two different '877
> chips, in case one was damaged, but they both fail the same way.

This is completely consistent across the PIC line. It'll happen in every
part.

>
> I also tried the following. I've put all the code so you can see exactly
> what I've done. I have a bunch of LEDs wired to the pins of port C, through
> a 100-ohm resistor to Vcc (each), so that the LEDs would light if a pin goes
> low.
>
>         list    p=16f877
>
>         include    "p16f877.inc"
>
>         org        0x000
>         nop
>
> start
>         bsf      status,rp0
>         movlw    b'00000000'
>         movwf    trisc
>
>         bcf      status,rp0
> why     movlw    0xFF           ;   Can change this to 0x00
>         movwf    portc
>
> prob
>         movlw   0x00
>         movwf   portc
>
> circle  goto    circle
>
>         end
>
> Up to the label "prob", the LEDs remain unlit. I would expect them to light
> after "prob" executes. They don't. However, if I change the 0xFF at "why" to
> 0x00, then the LEDs are lit (and I can then change the 0x00 at "prob" to
> 0xFF, and they'll remain lit).

Not sure about this one. The code looks right. There's no RMW issues/

Anyone?

BAJ
>
> Thanks for any help,
>
> Rick
>
> --
> http://www.piclist.com hint: The list server can filter out subtopics
> (like ads or off topics) for you. See http://www.piclist.com/#topics
>

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics