In a twin rotor pma the magnets are placed on the rotors so that the rotors are attracting each other. What would be the result of the pma output if the magnets are placed so that the rotors are repulsing each other.
Essentially zero output. (You'd get a tiny bit of output due to asymmetries in the magnets and the winding of the coils.)
The voltages induced by the fields of the two rotors would be opposite in sign and about equal in magnitude and waveform, and would thus cancel out.
Another way to look at it is to look at the lines of the composite field.
With the rotors attracting the fields are tight bundles of lines from the magnet in one rotor to the magnet in the other, and this tight bundle cuts the windings of the coils. Lots of lines cutting means lots of induced voltage.
With the rotors repelling the fields from opposite poles "splat" off each other. Think of two hoses spraying at each other and hitting a glass wall between them. It's actually more gradual and spaced-out than that - the dead surface is there but the field gradually weakens as you approach it. The lines from the pole on each rotor bend over and connect to the adjacent poles on the same rotor. Much less field, cutting less than half the windings of the coil, which an equal-but-opposite field is doing the same to the other half of the coil to cancel out the little bit the field on this side DID generate.
