Feature idea: MAF calibration tool

[subwoofer]

Along the lines of VEAL, but for MAFs.

Rationale: I plan on using a MAF to replace the SD fuel metering, but finding MAF calibration curves is more or less impossible. Apart from the 0-flow point, data sets on MAF flow-to-output transfer functions seems scarce. Only a few points on the curve are given for each model and the plot is not large enough to accurately produce more data (Bosch). I believe the function to be a second or third order polynomial, at least for Bosch MAFs (from just looking at the curves), the job for M(AF)AL would be to determine the coefficients of the different terms for a best-fit calibration curve. It may be useful, at least for a first version, to be able to try different orders for the polynomial to see what creates the best fit. Assuming that full open flow of the injectors is properly set, the generated curve should then be quite accurate and could be basis for a community collection of calibration curves.

The tool could be run in-line, off-line or on-line, depending on how close the starting point is thought to be. If nothing is known, attempting to run MAF fuel mode will of course be futile, so an on-line or off-line calibration while running SD could be used instead. Once in the ballpark, the calibration could be moved in-line like VEAL is today when "update controller" is selected.

A side effect of a statistical analysis like this is that the dead time of the injectors may be calculated, since the same air mass flow will occur at different RPMs depending on throttle position.

I am not expecting this to happen overnight, but I am hoping for opinions.

[subwoofer]

A lot of people have read this, but none have expressed any opinions. What do you think, Phil?

[LT401Vette]

King's equation usually creates a good representation of curve foo a MAF sensor. It will then need correction similar to what describe.
I have several things of this nature on the todo list.

[subwoofer]

Bringing up this again, I think I have an idea of how to adapt a few bits and bobs you have already written to make automagic(TM) happen with the MAF curve. The following assumptions are made:

- Engine up to temperature
- Injector sizing correct
- Dead time more or less in the ball park
- MAF curve close enough that the engine runs, just barely will do, voltage points spaced out to fill expected voltage (or frequency) range
- Battery voltage stable once running

The basis would be a scatter plot. Hopefully the formula -(AFR / (EGO * AFRtarget))*100 should give the required percentage change of flow value for any voltage point. Plotted against MAF voltage (or frequency), it should produce a curve representing the percentage the flow value must be changed to match actual flow for that particular voltage.

2013-04-17 18_50_06-Greenshot.png

Since the mass air flow is the same along the red lines (I have not attempted to place them accurately, it was just to illustrate the point), the dead time can also be analyzed and corrected. Any error will show as scatter in the low-flow region. Depending on the characteristics of the spread, it should be easy to find out if dead time is too high or too low.

This carries a lot of similarities with the WUE analyzer, just the input data are a bit different. What do you think, Phil?

[subwoofer]

I made a thread in the Tuning section with what I have found so far in terms of algorithm. The idea should be easy to automate, and could work in tandem with the an adaptation of the regular VEAL.

http://www.msextra.com/forums/viewtopic ... 34&t=49504

[subwoofer]

Resurrecting this old thread.

In MS3 Development, this thread http://www.msextra.com/forums/viewtopic ... 25&t=50605 seems to be moving the MAF implementation towards a true integrating flow calculation. That has the potential to make MAF metering very accurate over the entire operating range of the engine - IF the transfer function is correct.

I never was very good with statistics, but I believe it would be possible to use a combination of the MAF logger built into MS3 and the ordinary data stream to make a version of VEAL that would work its magic on both the trim table and the transfer function simultaneously.

The practical result would be that you could make a very rough guess at the bottom part of the transfer function and then let TS work it into a perfect tune in a very short time, making the benefits of MAFs easily accessible for everyone.

I'd gladly pay for the upgrade.

[subwoofer]

Resurrecting once again, also requested by another user in http://msextra.com/forums/viewtopic.php ... 0&p=384214

I believe the interest in MAFs is on the increase after the fully integrating MAF code was introduced. MAFs are used by all OEMs, and people who get them to work are not going back to SD anytime soon.

The method I presented in http://msextra.com/forums/viewtopic.php ... 4&p=360825 works very well, but is a manual process. At least the Bosch HFM5s are accurate enough that there hardly is any need for corrections beyond the curve itself. With your (Phil/EFIanalytics) statistics chops, it could be possible to automate the process. VEAL made SD easy (apart from a few hard-to-reach corners of the map), the same type of tool would be a revolution to MAF users. It could even be a lot faster than doing the fuel map for SD and (hybrid-)Alpha-N since one MAF value covers a range of operating points.

Pretty please? Sugar on top? At least an opinion on whether it can be done or not, and where it may be on the roadmap?

[winstonusmc]

We just need VEAL to work with the MAF curve on a tab like VE1, VE2. I can imagine it would be fast.

[cmonref]

We just need VEAL to work with the MAF curve on a tab like VE1, VE2. I can imagine it would be fast.

If you select MAF and "Use VE1 as trim table" that is EXACTLY what MS3 v1.3 already does. I's been using it for two days now in my first attempt to run with MAF.

EDIT: Well, I think I misread your intent. VEAL used the VE1 table as a correction table, building it as you drive around. But then you have to do the manipulation of the MAF Flow Curve manually per SubWoofer's other post. It does NOT do that for you -- and I think that is what you were seeking. Sorry for the incomplete answer.

Brian

[winstonusmc]

Yes, VE1 is a good tool, but its a "work around". If VEAL adjusted the MAF curve directly in a tab, then life would be more simple for someone that doesn't want to mess with scatter plots and figuring out correction factors and such. We all know MAFs are easier, but just imagine if Autotune worked too.

[subwoofer]

We just need VEAL to work with the MAF curve on a tab like VE1, VE2. I can imagine it would be fast.

Me too. I would guess less than 15 minutes of actual driving should give a more or less perfect curve. That would be insane...

[cmonref]

I would guess less than 15 minutes of actual driving should give a more or less perfect curve. That would be insane.

I wouldn't know what to do with the other 7 hours I spent on this today ...

Brian

[mr_minnis]

I agree, I don't think it would be too hard to set something up. I only found PMAS willing to supply a 60 point curve. I am sure the defualt 60 point curve would be good enough to set started with on MFCAL (maf flow curve analyzer). It would make things much easier.

[LT401Vette]

Don't worry this thread has not fallen on deaf ears. I will see about setting something up for the next beta cycle.

[jb007]

Fantastic! Thats what makes your software so great, your ability to listen to the users. Thanks Phil!

[weeblebiker]

sounds like maf will finally be a viable option on ms2 since alot of the ms3 code has been back ported in the recent beta versions. may have to get up to speed and revisit maf
I'd pay for TS again for a better maf tuning capability, (ehm, especially if it was android compatible, ehm)

[subwoofer]

Really looking forward to trying it out!

[weeblebiker]

ballpark date the beta release that includes this feature?
week? months?

getting ready to fire up my latest incarnation and may keep the initial alpha-n/speed density tune for the time being before diving into maf,if the auto tune tool will be available soon.

[5.0Thunder]

Wondering what was done to satisfy these folks. I've been using a maf on an MS2 controller for a while now by coming up with a ballpark curve, then using VE1 as a correction table with VEAL. I never bothered to modify the curve based on correction. Instead, I just drive around with VE1 as the correction. Are other parameters negatively impacted by VE1? Do I need to go modify the curve now? It'd be cool if the curve modification was automated similar to how VE1 is being modified.

Didn't know if anything was ever implemented in to the software.

[5.0Thunder]

Bump