Twin Wideband Lamba sensor code for MS
Turbo Mini Forums

I've been in recent talks with Jean Belanger about what would be required to change the MS code to accept the data from two wideband lamba sensors for the sake of 'auto-tuning'.

This of course can be split in to many different areas, such as the type of siamese application, be it staged siamese injection with more than 2 injectors, or the 2 injector approach with the base siamese code using specially timed injection events. Also, depending on what aspects need to be tuned, whether it be pulse width, injection timing, or a combination of both.

Jean has already told me that the pulse width correction should not be too much trouble to add in, but it is the injection timing adjustment that is the tricky subject. I will be studying the MS Extra code myself, along with the Siamese specific code and work out for myself how the current system works, and whether it can be adjusted to efficiently alter the injection timing based on lamba data, as well as any other relevant sensor data.

I am only really a programmer so I will need some input from those who have more knowledge of the specifics of the siamese setup, but I currently have a fairly good understanding so it is only really the details that illude me.

So, I have made this thread to hopefully start the development of dual lamba sensor code for use with siamese injection. Anyone please feel free to input related ideas.

OK,

Firstly it would be ideal to just tell the ECU to work out the fueling and timing, based on the two widebands. Good idea, but in practical terms very difficult.

With EGO control on conventional engines, it is relatively straightforward to set up a closed loop control system with PID parameters to match a target AFR.

This could be carried forward to the Siamese code in theory. The AFR Tables 1 & 2 could be made to adjust to a target AFR for inner and outer cylinders with a dual closed loop system. But unless the injection timing is spot on and you are sure that the first squirt is going into the inner cylinder and the second squirt into the outer cylinder, then the two control loops will fight each other. Add to that some closed loop control on injection timing, then mayhem will ensue.

My experience with testing the sequential siamese code in its intended mode on a real engine is that if you make one adjustment, VE or injection timing for inner or outer cylinders, it will affect the AFRs on both inner and outer cylinders.

The only way that I have been able to get the current code to work is to implement the "hybrid" mode with a single pulse that delivers to both inner and outer cylinders. That way, I have one VE table and one timing table to deal with and all adjustments have an understandable impact.

Must go, dinner is served....................

EDIT: My advice would be to get a working system with a close set of tables, then develop the code to provide some precision.

No point in writing code without the practical experience of setting it up in the first place, otherwise it will be idealistic and theoretical.

Edited by Paul S on 4th Nov, 2009.

Saul Bellow - "A great deal of intelligence can be invested in ignorance when the need for illusion is deep."
Stephen Hawking - "The greatest enemy of knowledge is not ignorance, it is the illusion of knowledge."

I mostly agree with Paul that it would be very difficult to have both fueling and timing controlled in a closed loop and I've already mentioned something along those lines to Carl. However, it might be possible to slightly adjust fueling in dual pulse/table mode but it might require some refinement to take into account the interaction between the 2 cylinders. And it might be possible to do something about timing or fueling in single pulse mode but that would also need some thinking and experimenting.

I do agree that there is a need to have more data points before even attempting some code improvements because at the moment while Paul's work is very good and well documented, it's still only one setup. Rod should bring some additional data in the near future which will help but the more data we have the better we can understand all the constraints, limitations and overall behaviour.

Having said all that, there might be one area that would help. If the data log were to be post processed to try and establish how the timing and fueling should be modified to get a better AFR and fuel distribution, that would be very helpful. The static nature of the data and the fact that it is done off line on a more powerful processor (a PC) would mean more data can be looked at and in more details with much more complex modeling.

Of course, the static part also means that it would have to be then uploaded and validated but it would also mean that you have a fixed set of data to analyse and you'd be able to see immediately if some data (fueling or timing) is completely wrong.

I don't know if Carl is up for such a project but I think it could be more useful at this time that trying to have some close loop control.

Jean

Edited by jbelanger on 5th Nov, 2009.

http://www.jbperf.com/

Sorry to reply to this so late, completely forgot about the thread!

I think Paul has a good point about contructing an actual system first, and then try experimenting with the code to see if some good results can be achieved.

As soon as I have the time to adapt my own mini for the dual pulse siamese setup, which I hope will be within the next year, I will then look in to what can be done for the coding on dual lambda as I will be able to test it all myself and adjust accordingly.

So this will be on hold for now, but I will still examine the code and get an understanding for how the MS code is structured so I have a head start.

Jean, your idea of analysing the wideband data from logs sounds very good indeed, but also sounds like a mammoth task!

I think if I was to attempt to try something like that, I would need to meet with either Paul or Rod and discuss it in person, and make notes on it, as I dont think I have enough understanding currently to apply to the data collected by the logs.

Edited by Carl S on 12th Nov, 2009.

http://www.jbperf.com/