Thanks for the note...  The ideal design having zero parts is
pretty much my idea.  I compare this to "brute force engineering" where
we throw parts at the design until it works.  I finally figured that I
should add the resistor between the serial output and the RS422 driver.
I also tie RC4 to the line after the resistor.  I can then force the data
high or low, or tristate RC4 letting the serial data go through.  This
also works out nicely because I can use the serial transmitter for timing
the break.  Would've been nice to have the break capability in the chip,
but as you point out, the resistor isn't THAT expensive...


>Have you tried setting RC.6 to 0 and not letting the TX side say
>anything
>for 22 bit times?


        RC6 uses SPEN to determine whether the pin is driven by the
output latch or the serial port.  So, if the serial port is enabled (and
I need to keep the receive side running), the pin is driven by the serial
port and ignores the settings of the output latch.  I also tried dumping
a 0x00 into the TxReg and waiting for RC6 to go low (we can still read
the state of the pin), then clearing TXEN, which, according to the block
diagram, should kill the clock to the serial port shift register, leaving
the line low.  However, as pointed out in section 12.2.1, "Clearing TXEN
... will reset the transmitter as a result the TX pin will revert to
hi-impedance."  My checks indicate that the line just goes high and stays
there.  So... that didn't solve it.  I added the resistor!
        Again, thanks for the comments!

Harold