interpolation of mixture map / autotune values

[Davidv]

Hey, 

just a thought for the mixture map / autotune.

When the ECU is deciding fuel values, it interpolates between rows to come up with a number.

However autotune / mixture map just adjusts the closest cell.
So what happens sometimes if you have a lean row and rich row next to each other, the lean row is influenced by the overly rich row. So it decides it needs to lean out the mix more by adjusting the already lean cell that it's closest to. 

Then when the load value changes to be closest to the rich cell, it now detects that it's running too lean so richens the cell closest to it, making it richer. and then you get a big seesaw / wave in the fuel map.

So it's a balance between getting enough samples to autotune with, and having your sampling criteria too "loose" to be useful.

I reckon it would be cool if the mixture map and autotune could interpolate between the relevant cells. So if it's 45% away from one cell and 55% away from the other, it will adjust richer or leaner by that % between them.

This would have a few benefits:

1. It would avoid see sawing by adjusting all relevant cells instead of just the closest, when this is done 100s of times over at various points surrounding a cell, all values become more accurate. Even if you never reach the centre of a particular cell, it doesnt matter if you've got values "around" it.

2. You wouldnt need to try get the car near the centre of a cell for autotune to work

3. If you dont need to be near the centre of a cell for it to start working, the number of recorded values that can contribute to the tune increases HUGELY. which will mean a lot less time spent to refine a tune really awesomely, and autotune would sort itself out really quickly with minimal supervision.

Instead of having the criteria being near the centre of a cell, you could set a threshold for engine speed rate of change. 
So that it's only adjusting during times with lots of load, or when you're cruising along.
Rather than rapid transitions where the wideband etc cant necessarily keep pace accurately or there are TPS enrichments etc.

Edited December 9, 2015 by Davidv

[Dave Kriedeman]

Hi,

are you referring to the MIXTURE MAP only or Quick tune also.

If referring to Mixture MAP tuning, or any of the tuning the ECU does the interpolation on all cells.

The mixture map X and Y values are derived from your main fuel table,so if you want more accurate resolution you can add more load sites to the fuel main table X and Y axis's.

Same as QUICKTUNE, the setup parameters are super critical to the system working correctly.

Turning tune ahead on or off has a big effect.

How many % added to the tune ahead,

Tune ahead position, across, across and right, across and down etc.

Cell centre accuracy cell centre target time etc , etc ,etc.

These functions only work as good as the setup information you put into the software.

Like the MIXTURE MAP, making sure you have setup limiting parameters etc.

EXAMPLE, TPS and MAP, if you lift your foot off the throttle the TPS value will drop and the MAP will increase as in you may have been cruising at 60 Kpa , take foot off the throttle and you have 20 Kpa (increased vacuum).

Obviously the AFR will show up as a lean condition, you do not want this data effecting or showing up on the mixture map or you will end up with incorrect fuelling due to incorrectly deciphered information.

Mixture Map is a quick way to tune but be careful with the data you are looking at.

I try to start with a flat fuel table EG: start with a fuel table with a flat number of say 40 - 50 in every cell.

That way if you look at the surface graph it is just a flat plane.

Having the fuel table cells setup this way, you will be surprised how quickly and accurately the tune will pull together.

A particular cell in a true centred state may start at flat 50 but actually need 52.3 to reach the target AFR.

By having all tables cell values the same the interpolation values are less than from a map with a lumpy surface graph.

So what I am getting at is when the tuned cell is corrected the surrounding cells need very little change to correct themselves and has minimal effect on all other cells around it, leading to a much more accurate overall fuel tune.

As for transitional state changes like TPS increases etc, it is obviously common sense to hold that point until stable and the AFR reading has stabilised to be accurate.

I could rabble on about this for hours however I need to know more information than what you have posted to zero in on how and why for everything.

I don't often use MIXTURE MAP, I do look at the table colours every now and again, but the data log information is what I am interested in most.

Same as QUICK TUNE, set up correctly, I do not use TUNE AHEAD function at all, I adjust time in centre and accuracy but that is it.

With the flat fuel table you will find that in QUICK TUNE it will tune very quickly.

Tune a row and then copy and paste to the next row, you will notice very little change from cell row to cell row.

There will be an obvious fuel trend increase around the point of maximum VE (TORQUE) where the numbers may increase at a more noticeable value changes but nothing huge.

The SURFACE GRAPH will appear a lot flatter than what you probably have seen, but you should not have big troughs and peaks in your graph.

My preferred method of tuning is QUICK TUNE on the dyno and then MATHS/ MANUAL  KEY TUNING from the data logs.

Anyway, that will do for now, I have rabbled on enough.

Regards

Dave.

 

Edited December 10, 2015 by Dave Kriedeman
added information