Semi-Auto MAF Scaling with excel! copy and paste!
 

[B]EDIT: VERSION 4 is out, refined formula to determine the gear. It's WAY more accurate, it would write off rows as 0 before because it wasn't in the tolerances but this way redefined the tolerances to be more exact.

Ok so for this to work you must have an Access Port and some basic excel skills (*copy paste*)

A couple of notes to start:
1. This is for NA only... if you blow up your FI car it's your own fault
2. If you blow up your NA car it's also your fault (*YOU'VE BEEN WARNED*)
3. Try to look through my workbook and understand it, if you don't care about your car enough to see what your using to tune it thats your problem.
4. Read the comments and opinions of the forum members, I'm sure their feedback will be usefull.

Set your AP to log the following:
-Equivalent Ratio
-RPM
-Calculated Load
-Long Term Fuel Trim
-Mass Airflow
-MAF Voltage
-Short Term Fuel Trim
-Throttle Position
-Vehicle Speed

STEP ONE: PASTE YOUR LOG IN SHEET ONE

STEP TWO: Go to sheet 2 and BAM you should have a lovely adjusted MAF scale..

Enjoy, I know it's been a while but it's summer so I felt motivated.. I'm also finishing up College right now and juuuuust happened to have a database management course in there. So yeah, that helped a TONNE! If you think of anything I missed in this lemme know and I'll see about fixing it, I wanna get back to adjusting fuel charts now that this damn this is way easier to work with.

OH! Don't be alarmed when the gear column reads weird numbers like 9 and stuff, I made it account for open and closed loop by adding 6 to each gear when it closed loop. Also it will read zero for engine braking and idle conditions.. ;) It's actually a poor column name, if you worked with my old stuff you would see there were like 6 columns to the right of the logs that had to be calculated. Now they're all in that one, and it only runs the parts that are needed now instead of calculating everything.

Please only post constructive criticism, I will correct errors as they come up since this is beneficial to me as well...

This file accounts for as many tables as I could in ATR that deal with AFR, it takes into account which tables to use based on whether its in closed or open loop using the tables in ATR to discern.. It also weeds out useless rows of data from decelerating while in gear, the initial bit when your clutch is still in the process of enganging, and other weird readings. It can tell what gear your and pick tables accordingly, lots of awesome built in features they didn't bother giving us in the AP

also the only table I altered in the stock map that is in there, and you can change it back to normal on the second page is the closed loop exit throttle table

Edit: I'm amazed, this went all night without any comment O.o Still a little anxious to see what peoples reactions will be

This is another pic with about 190000 rows of data and still only using the initial tune from stock, you can see the % difference went down ALOT.

So 2 days worth of tuning and it's gone from 10-20% error across the board down to well under 10% for the majority of the scale..

I will have to try this out on a NA RX-8. Will your calculations be off since your temp data, etc. is metric?

:O I never even thought of that... well, at the moment the only tables involved in my calculations can be found on the second page, if anyone can tell me what other tables from ATR are used with maybe a brief description I will add them.

The system I used works pretty well though, if your fuel trims tell you to take off 12% fuel and the difference between actual and target (*from the tables I'm using*) is only 6% then it will remove 9%. so next time it should be around 3-4% off for fuel trim, and the difference in AFR should be lower as well, so it will pick a mid point again. I knew it was a flawed system since we don't have access to all of the tables so I used that to avoid overshooting and undershooting, and it works. Because everyone hates bouncing back and forth... ^_^

Of course I'm looking to improve this and implement parts of my formulas in other tuning sheets, like scaling injectors and what not

This is a solo project, so don't expect me to rewrite formulas EVERY time something is mentioned, I will wait for reasonable amounts of feedback before each new go

Also, as of yet there are no tables with temperatures that affect my workbook... :S I only see temp tables that affect calc. load, if theres something that uses temp let me know or direct me to a thread perhaps?

This looks really cool
Thanks!

-- edit ---
Not so sure I would be comfortable having a Maf curve that isn't a smooth transition/curve as it progresses.
(mmm I think this is this same kind of method Kane was developing/uses, I think)
Should be interesting to see how it works.

-- edit edit ---
Maybe I missed it but you don't seem to provide any logging method.
How do you do your data logging?
Cruise logs at 500 rpm intervals?

--- edit edit edit ---
Should also caution first timers not to change the first Maf G/S value on the Maf scale.
It's been known to cause weird issues like a check engine light.

--- edit edit edit edit ---
I just quickly reviewed your data log.
I see you've just basically drove around collecting data.
Personally I would be a bit more anal about collecting data.
In order to have reproducible results I would recommend collecting cruise logs in a more static procedure.
For example do cruise logs using the cruise control at 2k rpm all the way up to 8k rpm if desired.

Have you tried it with some WOT data?

same pic with a new line, it shows the stock maps MAF Scale in blue, this shows my averaging system at work, cuts the % error by a bit more then half each time.

also, yes, I guess you would have to use metric since it does use speed to determine gear which pulls info from the fuel tables.. :S I will look into making an Imperial friendly version I guess

Awesome work. I just did something similar myself, but only for closed loop, to use with Mazdaedit on an S2.

One thing I did do was play around with different binning schemes for each MAF cal table voltage. The formula I settled on was to average all valid readings between the previous voltage and the next voltage. You have done the average of corrections >= current voltage and < next voltage.

For example, to get the correction % needed at 0.98V, I averaged all logged corrections between 0.86V and 1.02V (>0.86, <1.02). You have (>=0.98, <1.02).

Any reason you did it that way? I must admit, I just guessed that I should take a range from both sides of the voltage in the table entry.

I found this one out the hard way! I actually though it was a bug in Mazdaedit on the S2 until a kind person told me this applies to all RX8 tuning solutions.

My question is why you believe the complexity of automating this is required when all anyone has to do is log LTFT and adjust accordingly?

Further, having tuned dozens of RX8s using the Cobb system, in OL the numerical values of AFR in the fuel table have no correlation to actual AFR. What it would take for them to have correlation is not worth the effort IMO.

Not trying to be negative. I just don't understand why you believe it's necessary when the factory has already provided a feedback system to accomplish MAF tuning?

I would also be cautious about blindly using the open loop results. It might be better to create two graphs, one for open loop, and one for closed loop. It is also useful to plot the number of sample points alongside the difference % so you can use that to determine if you should ignore certain results (or if you need to get some more logs). Along the same lines, I would also plot the difference from current, and the difference from stock on a different graph to G/s so the y axis can be scaled to see the differences more clearly.

More thoughts:

1. Are you using the AFR reading to try and calculate the correction in closed loop? I would just use the fuel trims in closed loop since I don't think we know all the tables there.

2. In Mazdaedit you can log "Fuel status". This indicates open or closed loop operation. I've only just discovered this. So far I have been using STFT==0 to determine this.

:rofl: yes, please make it as unnecessarily complicated as possible

I did this in a previous attempt and was actually thinking about it earlier today

I like complicated! although I can't complain with my results so far, I chose not to use just fuel trims or just difference in AFR because if you use the average of both, it just verifies that the result isn't far off, assuming they are relatively close and if they aren't then you know somethings up, but you should see that pretty quickly

So, yes it's overly complicated, but I get bored with easy, and it will take the guess work out of it for lazy people if I do it right.

iteration 2!
-I will include an Imperial version at some point (*seperate*)
-a row showing how many valid points there are per voltage
-I will test a version where open loop doesn't use the average of both but just the fuel trims (*TeamRX8*) however the formulas are written and I don't think it will slow down the progress or speed it up significantly either
-I will add a secondary y-axis to the graph and maybe a few other graphs, pretty it up and all that, use the second axis as a 0-100 for percent difference so the lower it gets the closer you are
-I will make a graph with the original maf scale vs. suggested so you can marvel at it... (*Blu3dragon*) although making a graph is as simple as selecting the rows and inserting the graph


I chose to use the >= and < system because I wanted to keep the values in as tight a group as possible, less datapoints per voltage value, but more accurate
If I went with > and < system then that means the voltage value before and after it reuse the same data, I could see that making a smoother curve, however I like it when I see random bumps ESPECIALLY when I know I have enough data for that value, it tells me something else is affecting it, and in the lovely 6-7ish K range where people seem to see weird things happen, it helps me pinpoint my problem areas

and having just got back into tuning this is the first i've heard of mazdaedit, and sorry to say but I won't be buying it, I'm happy with my AP and ATR, does the trick for me!

Have to agree . Although I usually try to get actual vs table values to be close I don't get to anal about it . I also try and be logical about which parameter to adjust to achieve that .
MAF scale , Injector Scale and latency , VE , Calc load and fuel table all influence the final result.

It also pays to be aware of what mechanical issues can influence all this as well . Not a lot of point tuning it to perfection only to find your dirty maf , air cleaner , leaking intake etc. was causing the high LTFT you just tried to tune around. It's rare that a high LTFT exists unless there is either an issue or a discrepancy with the maf tube.

eh, as you can see I have some modifications and it altered my MAF scale a bit, so I did this anyways, and if it turns out I have to tune other tables then this only makes it easier to recalibrate my maf after.. i was doing this more for myself, and figured i would share it

besides, it's a pretty simple sheet to work with, and if it requires no effort on the forum members part and works, then you got something useful for free!

I like options, I'm sure most people do, I don't care if it's a silly endeavor, I have all the time in the world and I like this stuff!
except for the whole work thing... but I would argue I still have time for it there too, haha army...

Was not suggesting you should, only that you might want to see if the AP logs a similar "fuel status" flag so you don't need to try and replicate the calculation the ECU is doing.

BTW, using my own spreadsheet (which is very similar to yours) I seem to get very good results for everything under about 3V or 100g/s. I haven't figured things out above that yet, it almost seems like changes to the MAF table above that have no effect...

It doesn't have a fuel status log unfortunately, and I seem to be having success with everything up to about 180-190 G/sec, I think anything above that is just due to lack of data. My sheet has gotten my error % down to under 10% in every area and under 5% in the majority of the scale, I suspect I will just have to do more logs and filter everything except the trouble areas. I'm happy with my results so far and will be moving onto to other tables soon, and polishing this sheet up for its next use

Version 2!
 

V2!

ADDED:
-Imperial support!! (*Can someone please test this for me, I'm not wasting my valuable logging time/Gas to switch to imperial and test this*)
-I made a row that counts the valid data points and if there are over 100 valid points for that Maf Voltage it will read as a 15 (*I chose this number since it was easily visible on the chart and didn't extend it's borders in anyway or interfere with anything... I'm finding that most values with more then 50 points seem to be spot on
-created a second axis for percent difference so it's more visible

Edit: Wish I thought of this before I posted but I just added value markers to the difference line of the charts so you can see the % difference in numbers as well as a line