MAF Scaling - Open/Closed Loop Fueling info

[Kodename47]

@solidONE @Wayno the way the OL/CL fuel tables work, if the transitional delays are ignored, are:

The OL fuel table is referenced, if you are above the Minimum OL Fueling value then the CL fueling targets are used. Once you're below this value then the OL fueling table is used.

The standard Min OL Fueling value is set to 14.0 so any time the OL fueling table leaner than this then the ECU is in closed loop and is using only the CL fueling tables and corrections.

[Wayno]

Why you tagging me? I know how it works and my interest in disabling the delays is zero.

He wasn't asking how the computer determines the fuel mode. He asked how the table values worked within closed loop which he can see that he's in plain and clearly in his log.

[Tor]

Sorry if it's annoying a newbie asks questions in every single thread here. I apologize for that. I read through this thread to page 23 and the urge to get started overwhelmed me, I hope I didn't miss any import information. My car runs great and it's probably not worth messing with anything from a performance point of view, but I take as an opportunity to learn.

I experimented with vgi's MAF scaling tool.

I made 3 CL logs (2 autobahn and 1 mixed driving):
http://datazap.me/u/tor/maf-autobahn...-5-19-27-34-38
http://datazap.me/u/tor/maf-0?log=0&...-5-19-27-34-38
http://datazap.me/u/tor/maf-1?log=0&data=5-19-34-38

And 1 log with 4 WOT pulls:
http://datazap.me/u/tor/maf-ol-pulls...=5-16-19-34-37


I loaded my MAF scale out of Romraider from my current tune and my WOT log into "Open Loop" with the following filters: Min MAF V = 3.0 and AFR Error value = 8.0, rest default. Wideband AFR column, I assume is just "AFR" from my Tactrix log.

I then copied the new MAF scale to "Closed Loop". Copied POL from Romraider. Opened my 3 CL logs with the following filters: OL/CL = 2, MAF V max = 3, IAT max = 30 deg C (probably unnecessary as none of the logs have high IATs), rest default.

Now for the questions:
- First off, is the above procedure correct and are the filters sensible?
- I get this "AFR cell hit count". Is that sufficient data to do the scaling?
- These are the suggested changes to my OL and CL scalings respectively. Do they look sensible? Is it okay that they are a bit up and down and the curve is not completely smooth?







Thanks for bearing with my questions once again.

[Wayno]

Be aware when you change CL scale, particularly at the top over 40g/s, it will change the LTFT in open loop.

People worry far too much about fuel trims at idle and other irrelevant low load situations where AFR is a steady 14.7.

You can scale OL for 0 LTFT with tools, but then you have to know how to tune the middle of the scale to generate 0 LTFT reliably when actually driving the car. That's the hard part.

[Tor]

Originally Posted by Wayno

Be aware when you change CL scale, particularly at the top over 40g/s, it will change the LTFT in open loop.

So basically I could apply the corrections to 2.5 volts, leave the middle part and apply the OL corrections, to avoid messing up my pretty good OL fuel trims?

People worry far too much about fuel trims at idle and other irrelevant low load situations where AFR is a steady 14.7.

I was more worried about the -8.44 when cruising at about 3000 rpm would affect my OL LTFT?

You can scale OL for 0 LTFT with tools, but then you have to know how to tune the middle of the scale to generate 0 LTFT reliably when actually driving the car. That's the hard part.

What makes the middle of the scale particularly hard? Is another process required that using vgi's app?

[steve99]

Originally Posted by Tor

So basically I could apply the corrections to 2.5 volts, leave the middle part and apply the OL corrections, to avoid messing up my pretty good OL fuel trims?



I was more worried about the -8.44 when cruising at about 3000 rpm would affect my OL LTFT?


What makes the middle of the scale particularly hard? Is another process required that using vgi's app?

The middle part arround 3v is a bit tricky as its the transition between open on closed loop , so sometimes you have to compromise a bit to get ol and cl to work well in that area. Also as wayne said the last flow band at 60 g\s area sets the ltft for the ol scale so its important to get this area correct. Also irts best to try to keep the whole curve smooth for best results so sometimes you need to overide the utility sscaling a bit

[Tor]

Originally Posted by steve99

The middle part arround 3v is a bit tricky as its the transition between open on closed loop , so sometimes you have to compromise a bit to get ol and cl to work well in that area. Also as wayne said the last flow band at 60 g\s area sets the ltft for the ol scale so its important to get this area correct. Also irts best to try to keep the whole curve smooth for best results so sometimes you need to overide the utility sscaling a bit

Okay, I think I get it. The app is giving a rought guideline, and the rest is feel and common sense to make the final adjustments?

So use the app to get an idea in which direction the curve should be pulled. Then filter out points that doesn't fit the sorrounding points and smooth it out.
Like:

-4.3, -3,9, +0.4, -1.5

change to e.g.

-4.3, -3.5, -2.5, -1.5

Or visually like this?






Change from:



Changed to:

[steve99]

yes thats the idea to keep it smooth and smooth out any bumps in the transition area

[Tor]

So after 2 tries, I ended up with this below. I think it looks pretty good but would greatly appreciate if someone would take a look at it:

Closeup:
http://datazap.me/u/tor/tor-104-pull...5&zoom=554-682
With LTFT:
http://datazap.me/u/tor/tor-104-pull...5&zoom=550-688

My car has the shifts in fuel trims at 5000 rpm with every tune I flashed (including stock http://datazap.me/u/tor/stock-a00g-m...7&zoom=517-651 ).

Is it worth (and possible) getting rid of? I suppose that has to do with the injectors?

[solidONE]

Originally Posted by Tor

So after 2 tries, I ended up with this below. I think it looks pretty good but would greatly appreciate if someone would take a look at it:

Closeup:
http://datazap.me/u/tor/tor-104-pull...5&zoom=554-682
With LTFT:
http://datazap.me/u/tor/tor-104-pull...5&zoom=550-688

My car has the shifts in fuel trims at 5000 rpm with every tune I flashed (including stock http://datazap.me/u/tor/stock-a00g-m...7&zoom=517-651 ).

Is it worth (and possible) getting rid of? I suppose that has to do with the injectors?

Injector scaling and the PI/DI ratio will have an effect in conjunction with your maf scale. Temperatures, fuel quality, and perhaps elevation will also have an effect where your fuel trims end up. Since the latter 3 is continuously variable, I think its best to just to dial your MAF scale under "average" conditions (eg intake air/engine operating temps, elevation etc.) for the least amount of LTFT and call it a day.

Other things you could do is try to match the PI scale to DI as close as you can, dial the MAF IAT compensation table, and change the CL switchover point to where the fuel trims are the most stable on your particular tune. Though, that is a shit ton of work for very little improvement in performance.

Again, it's probably best/safest to dial your maf scale and OL fuel table to where your car runs on target to slightly richer than target with least amount of fuel trims (+LTFT's) on average and call it a day. You much rather have it running too rich than too lean.

[Tor]

Can you help me understand the fuel trims conceptually (or point me to a place where it is explained)? If the car runs slightly lean below 5000 rpm, why am I seeing negative fuel trim? Wouldn't it make more sense it was adding fuel if it runs lean?

I did the MAF scaling logs when it was pretty cold here, only 13-16 deg C. The log above was done at 24 deg C. Was it a bad idea to scale at those fairly low temperatures? The fuel trims here are probably not so accurate since it didn't have enough learn.

Thanks again!

[Kodename47]

Originally Posted by Tor

Can you help me understand the fuel trims conceptually (or point me to a place where it is explained)? If the car runs slightly lean below 5000 rpm, why am I seeing negative fuel trim? Wouldn't it make more sense it was adding fuel if it runs lean?!

STFT is only active in closed loop, LTFT are set by STFT. When the fueling is in open loop it's not comparing the "commanded" fueling to the O2 sensor input.

[Tor]

Doh yes of course it would be OL. What a brain fart. Thanks.

So rather its rich in CL, hence the negative trim. Which is then applied in OL causing it to be lean? Would that be the right way to think of it?

[solidONE]

Originally Posted by Tor

Can you help me understand the fuel trims conceptually (or point me to a place where it is explained)? If the car runs slightly lean below 5000 rpm, why am I seeing negative fuel trim? Wouldn't it make more sense it was adding fuel if it runs lean?

I did the MAF scaling logs when it was pretty cold here, only 13-16 deg C. The log above was done at 24 deg C. Was it a bad idea to scale at those fairly low temperatures? The fuel trims here are probably not so accurate since it didn't have enough learn.

Thanks again!

Hmm.. I'm pretty sure OP of this thread explains how fuel trims and MAF scale works..

The way the stock IAT compensation is set up it tends to run richer in cooler weather but leaner when IAT get warmer for a given value on the MAF scale. So, if you calibrated your maf scale using information taken with low IAT values the scale will tend to be "lean," meaning the MAF scale values will tend to be lower. When this lower valued scale is used in warmer conditions your computer will add even more fuel by increasing fuel trims to compensate for the lean condition in CL.

http://datazap.me/u/tor/stock-a00g-m...2813&mark=2671

Looking at this pull in your log, your commanded AFR and AFR reading also look pretty strange. Towards redline the Commanded AFR goes as rich as 10.8:1 while the AFRs only goes up to 12.13:1 even with +5% LTFT. This tells me your maf scale is quite a bit off. Are you running FI or NA? What are you using to flash your rom and for data logging?