I suspect that the problem is that IRL there would have been a lag between application of reduced or negative g and engine misbehaviour, such that flying around in bumpy air wouldn't cause problems. The model that we had in the last general release version had no lag, so going below 0.5 g for 0.01 second or something would cause engine trouble.
So at the simplest level what we need is the g limits from the previous patch, plus a lag so that misbehaviour only starts if the limits are exceeded for a length of time.
Since the lag was a function of the degree to which the limit was exceeded, the best solution would be to model the rates of fuel flow into and out of the carburettor as a function of g and just let the simulation sort itself out; that way you'd naturally generate both the lean and rich cuts without the need to hard-code behaviour...
|