On Feb 9, 2007, at 2:30 PM, Vitaliy wrote:

> The reason handshaking exists, is to protect the buffer of
> the receiving device from overflowing.

Well, no.  "hardware flowcontrol" as commonly used doesn't really
exist according to any specification.  RTS/CTS is supposed to be
more about signaling the modem which direction data will be flowing,
so that the modem can do its half-duplex things.

"deep fifos" (ie more than one byte) are recent enough that they're
not well handled by traditional rs232 protocols.  It's very hard to
implement "hardware flow control" in software, for instance, because
the software is on the wrong side of the FIFO status.  UARTS with
DMA capability are even worse; they're in serious need of status
bits that mean "a DMA is in progress but hasn't hit a termination
condition yet" that usually doesn't exist.

If you want to transmit a string of data and then cause some other
externally visible signal to change AFTER the data has been
transmitted, you have to monitor your own software FIFOs, any
hardware FIFOs, the transmitter holding register AND the transmit
shift register to make sure your data is fully transmitted.  This
is usually possible on a hardware uart; many have a "transmitter
empty" status bit in addition to a "space available in tx fifo"
bit.  I think it's more difficult to retain the proper timing of
RXdata and RXsignals; I don't think I've ever seen the necessary
"partial character received" status bit (although I think some
UARTs claim to only poll the modem signals "in between" characters;
One potential difference between "real" uart status bits and GPIO
bits used as such...)

To implement a closely timed half-duplex sort of system, it might
make sense to handle all your signal changes in the same SW FIFO
and ISR as data, so that "rts_off()" ends up queuing a special
data sequence (00 01 ?) on the software FIFO, and nothing actually
happens until the TX ISR dequeues that command from the FIFO, at
which point it would toggle your signal.

The receiver ought to be easier.  Just because you see a signal
that says you can stop receiving doesn't mean you have to fully
disable the receiver.  You can theoretically poll for the last
dribbles of characters a bit later (even after you start transmitting.)

BillW
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist