Quentin, thanks for the reply.. great advise, lets go through it:

> First use relays from a manufacturer that has been around for a couple
> of years and are well represented, so that you know you can trust it.
Currently using Iskra - not very well known, but been around for a while.
Fairly generic relays so second sources are available.

> Here's some catch 22's
> Coil failing? I would say I've had the same amount of driver transistors
> failing as coil failing.
> MOV's? Yes they fail on long exposure (a few millisecs) of over voltage,
> but what you don't see is the job they do on the other hundreds
> occurrences of spikes that they've trapped. And if you get a few
> millisecs or more of over voltage, then there is defiantly something
> else wrong with the circuit.
There would be (something wrong with the circuit).  I have now eliminated
any visible spark using a snubbing cap.

> Using RC filter. I've had a lot more failures with these than MOV's. And
> even worse, the cap has gone short, keeping the circuit on.
mmm. That worries me.  I have decided to go for a 600 VAC WIMA Polypropylene
cap. Should last I thought.

 > My point I am trying to make is that if you are going to look for a
> fault, you will find it. But here are a few suggestions:
>
> Use two pole relays to drive the contactor. Wire both poles in series to
> drive the contactor. If one contact burn, you have a very good chance
> the other one would not be.
That is a very good idea!

> On your contactor use a spare contact to monitor the state of the
> contactor. So if your PIC says the contactor must be off, but read that
> its still on, the PIC can: Send a warning. Shut down the main supply via
> solenoid controlled isolator or separate monitoring contactor on the
> main supply.
Already using the spare contact to implement a forward/reverse interlock,
sadly.  There are other ways and means of detecting whether the contactor is
in or not, and I am torn beween the gain in reliability (failing towards the
safe side) if components are introduced to do this and the loss in
reliability should they fail.  Basically the question is whether it is in
this case worth it to sacrifice (some) reliability in 1000 units to ensure
that one unit in 1000 (10000? 100 000? numbers are not very forthcoming
around here - stats are impossible to come by) will fail towards the safe
side.  These decisions are the fun in the job, after all.

> Use FET's or TRIACs to drive the contactor. over the past few years
> manufacturers of these parts strive to make the component fail open
> circuit, i.e.. if it pops, it will stay open circuit. Never seen a
> guarantee though. but you eliminate a driver transistor, coil and
> contact. Plus you got the advantage of solid state to take care of
> spikes etc.
I would have rated solid state far more succeptable to damage by spikes
etc. - has it come this far over the last few years? (so many companies have
gone under as a result of bad reliability of solid state power electronics
such as inverters... Maybe this field has finally grown up?)


Roland