Jim- I got my start and run caps from surplus center (
www.surpluscenter.com). They've got listings for Motor Run AND Motor Start caps... I think my array consists of three 40's in parallel for A-B, and two 40's in parallel for B-C. Starting cap array is somewhere around 5 or 6 (don't remember). For the motor runs, I had the option of using smaller form-factor caps at lower working voltages, but opted instead to go for the larger units. These Motor Run caps are take-outs, and they're used... some are a tad leaky, dented, etc, but work fine, and are non-PCB. They're also CHEAP. Get a few extra... and a few different sizes, so you can tweak the levels. It's also a good idea to have two voltmeters (one for A-B, the other B-C, for balancing).
I prefer to use voltage ratings for both start and run caps which exceed the system voltage by at least 25%... that's just a preference I've gotten from building Ham Radio equipment. Many say that oil-filled Run capacitors (being rated in 'working volts') are much more tolerant to over-zings, but I prefer not to ride the edge of ANY ratings. Motor start caps are of the electrolytic type, so 'ya really gotta heed the voltage ratings... they take a helluva beating for a short period of time.
There's three types of transients that'll appear in an RPC arrangement like mine that could cause a voltage excursion risking the caps:
1) Contactor coils... when you cut off the coils, they generate a nasty zap... {around 1kv) due to the collapsing field... suppressors take care of that, for the most part.
2) Idler motor windings... SOME amount of EM kickback occurs as a result of dropping out the windings... I believe more occurs in dropping out the start-caps than any other condition
3) Load machinery. Unpredictable, but with the LOAD contactor dropping out, it does grant some peace of mind.
In my design here, everything is very ooops-tolerant... I'm pretty conservative about ratings 'cause I hate fixing my own creations later!