Mike Gann wrote:

> As for values, I don't recall what I used, something like 100k R and
> .1uf C.

  No!  Actually, 100 *ohms* "R" and .1µF "C".  Note that the resistor is
a 1W type and the capacitor rated at 250VAC, not 600VDC.  A critically
rated fuse in the main supply should be provided to protect against
failure of the capacitors or relays etc.

  The capacitor between forward and reverse windings of the motor is to
provide the quadrature (phase shift) to cause the motor to *turn*.  If
it is shorted *or* open, the motor will stall (and draw excess current).

Douglas Reid:
> The system is constructed on several separate PCBs connected together
> within the same enclosure.

  As to the interference, after "snubbing" with the above components,
the *most* important requirement is *lead dress*; that is, making sure
all returns both in the motor circuit and the PIC circuit and
particularly the relay coils, travel with their actives and the control
wiring is separated from the mains wiring.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

> DC power for each of the other boards is derived from here on a star
> wired basis.

  Fine as long as all the signal connections from each board travel
*back* along the spokes of the star.  You must avoid anything resembling
a loop where two corresponding current- or signal- bearing wires have
ecome separated.  This applies equally to the wiring of the power
transformer (for the PIC system) secondary and the design of the PCB
itself bearing the relays.

  Some relays are made to provide isolation by mounting only the coil
end on the PCB while the contacts are brought out to "spade" lugs at the
other.  SSRs would be a nice idea in the same fashion, especially if
they contain zero-switching logic.

> The incoming 17.5V (the cable from the PSU is about 9 inches in
> length)

  That's a big worry!  It's called an "antenna".

> is filtered by a 220uF (since increased to 4700uF! - which improved
> things) at the input to a LM340 5v reg.

  Even more improvement - put a 100 ohm resistor in series with the +ve
to this board! 

> Subsequently, the Dallas DS1233-15 reset controller has been added to
> control the RST* line to the PIC.

  I have to be a trifle amused at the concept of installing a reset
controller when the problem is spontaneous/ spurious resets!  OK, so I
know it enables you to gracefully recover after brownouts...

> In fact, after I added the DS1233 the PIC was reset by the Dallas chip
> on every switching of the motors!

  Predictable!

> In a further attempt to improve the situation, I have fitted two mains
> input filters. ...  This does not appear to have resulted in any
> improvement at all.

  Two reasons:  1) These filters filter RF, not impulses.  2) The mains
supply is already a low impedance to minor (and yours *are* minor)
surges.

> The mains supply seems to be good in that only a small voltage drop
> (about 1 volt) is noticed on the incoming mains to the system when the
> motors for both wings are driven

  Goes with the above.

> (perhaps the transient voltage drop as a motor switches is greater?).

  Trivially.


> I had imagined that that mains filter would filter the mains
> interference resulting from the motors via the relay.

  Wrong sort of filter.

>  Is there any way that extra filtering could be added to each of the
> motor coils?

  Yes, but ... the motor coils aren't really the source of the
interference!

> I sometimes switch the motors directly between open and close - is it
> okay to do this

  No.

> or should I introduce a delay? 

  Yes.
-- 
  Cheers,
        Paul B.