>I have no hard data on that, but a 50% cycle seems to work, for a few >cylces at least.
Does that mean 20 sec cooling for each 20 sec burst of cranking?