I paraphrased directly from microchip documentation.

-JL

> -----Original Message-----
> From: pic microcontroller discussion list
> [mailto:PICLIST@MITVMA.MIT.EDU]On Behalf Of Byron A Jeff
> Sent: Tuesday, January 29, 2002 12:04 PM
> To: PICLIST@MITVMA.MIT.EDU
> Subject: Re: [PIC]: BSF and BCF use in port pins
>
>
> On Tue, Jan 29, 2002 at 11:23:47AM -0500, Jennifer Loiacono wrote:
> > Because of the read-modify-write structure, a write to
> PORTx actually writes
> > to LATx.
>
> Yes.
>
> > A write to the PORT register first triggers a read of the present
> > port values,
>
> Only on a read modify write cycle. Just clarifying.
>
> > then alters the value in the LAT register, then writes back to
> > the port.
>
> Correct.
>
> >  If another pin on that port is used for bidirectional I/O, it
> > could cause problems.  For example, if a pin is set as an
> input when the
> > data is read, the value would overwrite what is already in
> the register.
>
> True. However this will only cause an effect when you turn
> that bit back
> into an output. Presumably you'd rewrite the latch before doing that.
>
> > Thus, it is safer to perform bit operations in the
> appropriate LAT register.
>
> ???. You don't have direct access to the LAT register.
>
> >
> > The TRIS bits (direction registers) should not be changed
> individually, as
> > certain peripherals override the TRIS settings.
>
> A completely separate point. And since the periperals
> override the TRIS
> setting, but do not in fact overwrite the TRIS bit, usually
> it's irrelavent
> what value those overridden bits are set to.
>
> >
> > More information is available in the mid-range manual, in
> the I/O Ports
> > section (section 9).
>
> Very true. An invaluable resource. Everyone who programs PIC
> should download
> a copy from Microchip's site.
>
> Jen, while you described part of the problem, you missed the
> real point.
> I covered it in detail here (PASSWORDED SITE. login and
> password are both
> 'piclist'):
>
> http://www.infosite.com/~jkeyzer/piclist/2002/Jan/1069.html
>
> RMW cycle can be trouble because even when a pin is set to an
> output, the
> pin is read instead of the latch. The problem becomes obvious
> when attempting
> two consectutive RMW instructions to the same port. The write
> of the 1st
> instruction and the read of the second instruction happen one
> cycle apart.
> So it's possible for the 1st write of a bit not be able to
> change the pin
> before the read of the bit in the next instruction. Result is
> that the latch
> can be reset back to the value of the pin, nullifying the 1st write.
>
> Hope this helps,
>
> BAJ
>
> --
> http://www.piclist.com hint: The PICList is archived three different
> ways.  See http://www.piclist.com/#archives for details.
>
>

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.