Must work over a distance longer than the RS232 spec
allows (i.e. the
answer is probably not 'A long RS232 cable' :-)).
Don't give up on long cables until you try them. Especially if the
drivers approach the spec's maximum voltages, serial lines work at
distances far greater than the spec says - the spec is pessimistic.
Few environments are as electrically noisy as what the spec is designed
for, and keeping the driving voltages near the maximums permitted
increases the noise margin substantially.
If a simple serial cable doesn't work, measure the voltage produced by
the drivers and add additional circuitry, if necessary, to jack it up
to the 12-15 volt range. It would probably help to decrease the output
impedance as much as the spec allows, too. (I'm sure you, of all
people, don't need to be told how to do either one. :)
At one point, years ago, my house netlink was run over a four-wire
"metallic circuit", which was then-telco-speak for dry copper the whole
run, without any loading or equialization or such. While I did use
modems for normal use, I fould I could run it at something like 50 baud
without any modems at all - and that was probably two miles or so of
(POTS quality) wire.
/~\ The ASCII der Mouse
\ / Ribbon Campaign
X Against HTML mouse at rodents.montreal.qc.ca
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B