[Diy_efi] Timing Calculations

[Steve Ravet]

> I was thinking of checking the trend of the last x number of 
> cycles and 
> trying to anticipate whether the engine is accelerating or 
> decelerating.  It would never be perfect, but it would average the 
> timing out better.  Or maybe using the throttle sensor to anticipate 
> accelerationg/deceleration, kind of like a throttle pump, but 
> for timing.

Say your engine accelerates in first gear from 1000 rpm to 5000 rpm in 2
seconds.  That's an acceleration of 2rpm/ms.

Revolution time at 1000 rpm is 60 ms

If you start at 1000 rpm, at full acceleration, you will finish that
revolution at (60 ms * 2rpm/ms)=120 rpm faster, or 1120 rpm

Revolution time at 1120 rpm is 53.5 ms

Under full acceleration, you will begin the revolution at 1000 rpm and
end the revolution at 1120, with a revolution time of (60+53.5)/2=56.8
ms.  The error comes in not anticipating the acceleration, and
calculating a spark timing based on a 60 ms revolution when the motor
actually finished in 56.8 ms.  Or, to put a different way, your spark
will be (60-56.8)=3.2ms too late.  3.2ms, at 1120 rpm, is
(3.2/53.5)*360=21 degrees.  That doesn't seem right...  maybe I made a
math error above.  Anyway, you get the idea.  As the rpms increase
there's less time for acceleration from rev to rev, so the error gets
smaller.  This'll give you an idea how important rpm prediction is.

--steve