Mark Willis wrote:

>   Schottky diodes are good here as they lose less voltage (~0.4V vs.
> 0.7V.)  If that makes a difference, apparently it can sometimes.

>  For serial ports, just make sure you have DTR (for example) held high
> as much as possible, a very efficient & low dropout regulator as
> needed, the more diodes conducting in parallel, the better!

  The trick is, why use a regulator?  Parallel all the diodes, Schottky
if you're being critical, to the 5V rail and use a stiff Zener.  Two
aspects of this; RS-232 drivers are by definition current limited
(definition is, if they aren't, then they aren't RS-232 you see!) so
you don't need current limit resistors (you might add a small one).

  Secondly, you don't use an RS-232 receiver chip and you *certainly*
don't use a transmitter.  RXD is routed to the PIC via a limiting
resistor, but the diode "sucking" power from this line to power the chip
also limits this voltage to 5.6V.  A second diode or the appropriate DS
(Dallas) chip can actually "suck" negative voltage to use for the TXD
line, which otherwise uses the logic voltage direct from the PIC.

  Using a shunt regulator has the advantage that you get just that
little extra bit of current from the RS-232 line, though exactly how
much will depend on the individual port implementation.  You can get at
least 10mA however.
--
  Cheers,
        Paul B.