Jerry rigged, at a given RPM, gives you the same generated voltage as delta but less current.
- In jerry rigged the coils only conduct when they are individually putting out a higher voltage than the battery plus diode drop. A coil will either be conducting in the one-coil path or turned off at any given moment.
- In delta you get that current from one coil, plus half that current passing through two coils in series - 1.707 times the jerry rig heat losses in the genny for 1.5 times the charging current. A coil will be in the one-coil or the two-coil path at any given moment (unless things are spinning so slowly that ALL the coils are off part of the time - which happens when jerry-rigged would have each coil conducting for less than a third of the cycle.) In a situation where jerry rigged would have each coil conducting for no more than half the time the current waveforms in both the one-coil and two-coil parts of the delta's waveform will be the same as the jerry-rigged cycles (except that the two-coil waveform will be half-amplitude). It gets more complicated when jerry rig coils would have more than 50% duty cycle but I think the results come out about the same.
- In Y all the current goes through two coils' resistance (though you get a higher voltage so you wind things differently).
For a given amount of copper in the core, though, your power limit occurs at the same amount of heating - and heating goes with the square of the current. Jacking up the power through the jerry rig (say, by raising RPM) to match the resistive losses in the core for delta leave jerry rig with more losses than delta. (Delta conducts on two parts of the cycle, jerry rig on one, so delta is less uneven and the highs are less efficient than the lows.)
Multiplying jerry rig heating by a factor of 1.707 multiplies the output current by 1.307 or so, so jerry rig gets 1..307/1.5 = 87% of the output of delta for a given amount of heat. Less than delta, but a 13% shortfall is not too shabby.
A Y/wye/star connection gets about 1/1.732 = 58% of delta's power or 67% of jerry-rigged for a given amount of heating - assuming a resistive load. I'm not sure if the current distribution among the coils would throw it off that ratio for a charging application - but even if it did it has a long way to go to match either jerry rigged or delta.
Finally: Delta may have additional heating from circulating currents driven by waveforms that don't add up exactly to zero around the circle. This will reduce its advantage over these alternatives.