Tom Cloud's self tuning EFI Idea

[Woodd, Michael]

Excuse me? Anyway,

>I do see potential problems though. For starters, if the system had all the
>sensors needed to determine the "state of ultimate tune", how long would it
>take the cpu to find it?? People that dyno tune engines on a regular baises
>(TODD) would have a very good idea of this. I don't, but I'll guess any 
way.
>Let say 20 full throttle quater mile runs just to get maxium dead stop
>acceleration. There are lots of things to change and adjust, right?. I've
>done this manually in my early hotrod years. You could spend a whole 
weekend
>adjusting and evaluating things.

All a matter of software - the algorithm used.  Actually I was thinking more 
of a
circuit racing application, cos thats what I'm into, being as a car should
be nothing more than a tool for the driver, right? ;-)

And thats why I stressed the importance of having accurate measurements,
cos if they aren't, then your "perfect" algorithm is going to do funny 
things, like
take longer, or maybe even head in the wrong direction.

I plan on starting by simply investigating relationships between one sensor 
and
another reading, e.g. acceleration vs vertical g, to account for road bumps,
difference tween linear g vs wheel derived acceleration should give us the
gravity vector that tells us we're going up/down hill (I think, like I said, 
don't
know the physics well yet).
My end system will be a combination of these little black boxes, and for
prototyping they may be just that.

Mike Woodd
(wooddm at akcity.govt.nz)
Come on, gimme just one drive, I've been on a track before without pranging 
it...