Has anyone played with Ion sensing ?

[dnb]

I was reading about this last night and the circuits look simple enough to make, even if the signal processing required afterwards is a bit of a brain beater...

I wondered if anyone else had looked at it?

In case you don't know what I'm writing, ion sensing is where you apply a bias voltage to a spark plug after combustion and measure the current. This is supposed to let you know what's going on in the cylinder.

[Foolish Dave]

Nice idea, but I'd have thought the current levels will be low, esp compared to the voltage. Then you need to process it all, in real time and knowing what angle the engine is at to compare to... and that's just to get the data - what do you do with it?

[dnb]

Currents are around 1mA, so we're not talking about anything & silly undetectable.

Engine angle isn't too much trouble. I can sample the data (at high speed) along with engine RPM from the ECU to give context.

This datalog and one from the ECU can then be combined to give a complete picture using RPM and timestamps as an index. (It's difficult to merge data like this, but I do more difficult data fusion activities at work...)

I can then construct an average chart of peak cylinder pressures using the ion data for the whole ignition map of my ECU and thus derive MBT for the areas where the engine is not limited by knock.

The angle of peak cylinder pressure remains nearly constant for all the MBT limited areas of operation, so I just need to find one area where I know the advance is at MBT and then I can set all the other table cells to move the point of peak pressure to the correct point via an iterative algorithm.

The processing doesn't have to be real time - only the acquisition.

[GreenV8S]

I think Saab have done this haven't they? Perhaps there's something you could nick. I seem to remember Bowling and Grippo were looking into this too. It looks like a very useful way to detect misfires, knock, cylinder-to-cylinder variations and cycle-to-cycle variations.

[Foolish Dave]

dnb said:

Currents are around 1mA, so we're not talking about anything & silly undetectable.

Engine angle isn't too much trouble. I can sample the data (at high speed) along with engine RPM from the ECU to give context.

This datalog and one from the ECU can then be combined to give a complete picture using RPM and timestamps as an index. (It's difficult to merge data like this, but I do more difficult data fusion activities at work...)

I can then construct an average chart of peak cylinder pressures using the ion data for the whole ignition map of my ECU and thus derive MBT for the areas where the engine is not limited by knock.

The angle of peak cylinder pressure remains nearly constant for all the MBT limited areas of operation, so I just need to find one area where I know the advance is at MBT and then I can set all the other table cells to move the point of peak pressure to the correct point via an iterative algorithm.

The processing doesn't have to be real time - only the acquisition.

Easy levels to detect then, just cycle voltages for the sparks.

I was thinking you'd use the data to driving timing on the fly.

Hope it works.

[350Matt]

I've used it before and SAAB have been using it for knock detection for last 10-12 years, Its a good way of controlling Det as once you've calibrated the current / pressure rise you write the software to look for spikes or too steep a gradient on the curve and retard appropiatly.

It can also be used to replace cam sensors

Matt