Rogerio Odriozola wrote:

> Thanks to all, I do have a couple of 9600 modems which I'm willing to
> let die in the test. I'l let you know if something good comes out.

>> Split out the send and receive optos separately, arrange bias to
>> match the original and you should be away (though there is another
>> trick here).

  Let me just mention the "trick".  Modems generally contain a "hybrid"
to at least reduce, if not eliminate "sidetone", that is, transmitted
signal included in the received.  On a "two wire" circuit, this must be
done, while Rogerio is obviously considering "four wire" links with
separate transmit and receive circuits.

  Where a transformer is used for coupling, it is (nearly) always bi-
directional, and the hybrid (two-wire to 4-wire split) is performed on
the "modem" side of the transformer.  In older modems this is a separate
circuit using an op-amp or discrete components, but on more recent ones
it may well be integrated (as in a common Telecom SLIC device).

  The situation where optocouplers are used can be either way.  The DC-
requiring line circuit may provide a "hybrid" function to maximise the
noise and linearity performance of the receive optocoupler, in which
case you have a 4-wire interface provided cleanly at the optocouplers.

  The alternative is that receive and transmit optocouplers are simply
in series (or indeed, parallel) and the receive circuit cops 100%
sidetone.  In this case, the modem-side circuit is balanced to expect,
and cancel the sidetone, so that if you connect to it pure incoming
signal without sidetone, it will actually re-introduce it (hybrid out of
balance)!

  Disabling the hybrid for 4-wire feed in a modem *not* designed for it
is the real challenge in this application.
--
  Cheers,
        Paul B.