Hey, I posted a link to the PDF version of that dissertation a while back. I agree the IGS system would be ideal for WI.
My Saab uses the Trionic 7 engine management (considerably more complex and more powerful than the Trionic 5 they were using in that test) ... I hope to eventually be able to modify it as they have, only internally. I'd love to be able to market a chip for WI using an Ion Feedback signal.
A few other cool things that Ion signal can do:
1. It can measure (if calibrated) the EGT for each cyllinder, for each individual combustion. That could make tuning much easier as you could more easily spot differences in EGT between cyllinders.
2. The ion peak represents the burn rate multiplied by the cyllinder pressure, and is a VERY good knock descriptor. Almost unwaveringly a cyllinder will detonate when the BRxCP is equal to some fixed amount for any given octane. If the air temp goes up, coolant temp goes up, or ignition timing is too far advanced it will all show up on that curve, and as long as the peak is kept below the knock threshold for that octane there won't be any detonation.
3. It can detect the A/F in each cyllinder for each combustion. That could be a very useful thing for the ECU to know, as it would give it immediate feedback about the results of its injector duration and throttle position etc.
4. By far the most amazing thing it can do is spot combustion differences. Differences in combustion of up to 20% are normally seen. Only the very highest end of that 20% causes knock. If you could use a feedback signal to design a piston/combustion chamber that caused say 10% variation instead of 20% you could theoretically increase the power 10% without detonation.
Sorry if I sound a bit over-enthusiastic ... I've been sorely temped to just whip out my ECU and play with it ... but obviously that would be a fool's errand as I simply don't have to tools, or software, to do anything with it.
Oh well.
I may have also thought of a quicker way for the ECU to get a feedback schedule for ignition timing without resorting to the computation of two gaussian curves to match least-squares fit scenarios. It'd be really wicked to see if I could impliment that with a good WI setup for mileage purposes.
Adrian~