Were they thinking about having the C(extra) at the compressor motor, simply
because of the compressor contactor? That way when the compressor wasn't
running converter voltage would not tend up.


Adding a second potential relay, at the compressor, with the hysteresis changed
such that an additional run-type capacitor (A-B), or capacitors (A-B and C-B)
were placed on-line whenever the B voltage dropped on account of the compressor
starting would be a way to go.

Another way to go would be to add a state machine at the converter, one which
has a starting state for the converter alone, a running state where normal
loads would be supplied, and an overload sate which would handle the special
case where the compressor was placed on-line.

This overload state would place extra caps, hence extra VARS, on the converter,
thereby temporarily boosting the B-phase voltage.

(If the A- and C-phase voltages are sagging while the compressor is starting,
then the motor feeder is too small).

Another possibility is adding an autotransformer for this case.

(In fact, one of the first published rotary phase converters did in fact use an
autotransformer on the B phase).