I went outside yesterday and Y2K proofed my electric generator. Took some data with my fancy o'scope while loading the generator with incadescent light bulbs, halogen light bulbs, and a hair dryer. The graph to the right shows the results.

The bottom 5 lines (including the light blue one that shows 53 Hz at 1500 watt load) show frequency droop as the generator loads. The top lines (all on top of each other except the purple one) show the voltage drop as the generator is loaded. Okay, quiz time: what is the resistance of the copper wiring that is in the generator? Or, how can I raise the frequency about 6 Hz without raising the voltage?

For the bottom 4 lines in each case, I adjusted the governor feedback spring to give just under 120 volts with no load. For the line pair that is higher than the others (light blue and purple) in their respective sets, I adjusted it to be about 5% high with no load. I think that's how I'll leave it.

Conclusion: More droop than I imagined, and the feedback adjustment (which is why I did this whole thing) just doesn't matter very much!