Ok - I am finally going to sit down and try to spell this one out as best I can.

With our MAF based car tuning you have two "running numbers"; the fuel flow and the airflow. Once you start to play around with these (or even from the factory) they start to vary from thier intended Volts to Value tables.

Before we address the Fuel Scaling - the first step in tuning a MAF car is to Scale the MAF. This involves correcting the MAF table from the g/sec or lb/min values and the voltage readings.

http://i229.photobucket.com/albums/ee222/kanetomlin/MAFScaling1.jpg

Here is a screenshot of a sample MAF Scale.

Step One: Cruising.
- On your Base Fueling MAP; set the Lamda Values in the cruising range to 1 (AFR 14.7). Be sure to use the LOW LOAD areas in order to avoid engine damage. Additionally; you will need to ensure you are in Open Loop.
- For Closed Loop areas; set the lamda to 1 and compare your Fuel Trims to the % changes you are making in open loop. See more info on this below.
- Go find a flat level surface or dyno and get your car into a stable cruising speed and RPM. Hold it long enough for your AFR's to become stable. Document the Volts, g/sec, measured Lamda like in the picture above.
- You can also log these runs and go over the average (like Baseline Does).
- Adjust your g/sec based on the % deviation from your target. Using a Target Lamda of 1 makes the math easier.
- Hit as many voltage cells as you can in your Low Load cruising areas and look for trends. If your 1-3 Voltage range is all trending low by 10%; go ahead and scale the ENTIRE MAF by that percentage in order to get closer to your targets when you do your WOT MAF Scaling.
- Once you get all your crusing areas within 1-2 Percent then you should be pretty good.

Step Two: WOT / Boost
- The process is basically the same but your obviously NOT going to set your Lambda to 1.... set your target Lambda to a safe number (this AFTER you may have already scaled the entire MAF by the percentage trend in the Cruising Area.)
- Zero the Throttle Enrichment Maps out - you just want to see the Base Map Fueling.
- Be careful about stabbing the GO Pedal; since your gonna get lean spikes without Enrichment.
- Get to your higher Load Areas - Boost and WOT; but don't Punch the pedal, slowly get to WOT and hold it as long as you can.
- Use the same Math Process as Cruising but with a richer Lamda.

Step Three: Connect the Dots.
- Now that you have maybe 40-60% of your MAF scale done; connect the dots by filling in the blanks using the overall shape of your MAF curve.
- Try to make a nice smooth line. You can go back and re-tune the blanks later if you need to.
- Ideally you want to test each voltage in your table; but some interpollation is not going to be a big deal.

Step Four: Idle
- With radical cams / porting idle can be an issue.... the best way to go here is to get a g/sec trend for your car. Like the RX8; all 10-20 stock and boosted ones I have seen, all idle really close to 5 g/sec... so if your car idle g/sec is 7-8 then you can adjust it down to a known approximate value of (in my case 5). If your MAF numbers never get stable; you need to address the turbulence. If your engine is modifyed to the point you can never get it stable; then just do your best and shoot ffor best idle quality.
- You always do idle last because it can be the most challenging; but again if your % changes were all pretty much the same across the board then you can start with that % change at idle as well.

* Remember you don't have to have the same numbers or trendline; if your cruising is scaled at 3-5% and your WOT areas end up being scaled at 15%; that is fine.
* Closed loop actual lambda readings can't be used since the car will hit the target Lamda using Fuel Trims. So in closed loop cruising; look at your logged and stable LTFT instead.

Tuning ANY FUEL MAPS without first knowing your airflow readings are good is a waste - get your sensors right FIRST; then start in on your actual tuning.


Scaling the Fuel Injectors is a bit-o-pain since we have some that only come on later on. As fo the primary injector; use your "known" MAF g/sec value and look at your fuel trims... if your MAF idle is 5 g/sec (good based on known similar cars) and your LTFT is +5%; there is a good chance that is the Fuel Injector scale.... go back and scale the P1 injector a bit and see if it cures it.


I am attaching a sample Excel Spreadsheet you can use to get started.

this is nice kane!

I hope it makes sense.... I talked a few guys through it over the phone - but I was trying to consolodate it into some easier steps.

The big question is when the Air Scaling turn to Fuel Scaling... and short of flow testing your injectors - some educated guessing has to go down.

I'm gonna try a few techniques and get back to you on the Fuel Stuff.

Great info!

dumb question how do you know what is open loop and closed loop?

Nice. Lets automate that!

Cheers,
Hymee.

wow kane I like that a lot, specially since im beginning to maybe want to tune my car with the street tuner and one of the local tuners, so having this info is nice. Thanks!

sticky!

In order to correctly calibrate the ECU for the new induction kit, the relationship between air flow and output voltage of the air flow sensor must be measured. Initial thoughts would suggest using a flow bench to do this, though very few tuners will have access to a calibrated flow bench capable of flowing 300 grams per second of air.

In actual fact, all tuners do in fact have this flow bench in their workshops - the vehicle itself ! By plumbing the the two induction kits in series (the standard system and the new system), it is possible to draw the same mass of air through each induction kit. With a mass air flow sensor in each induction kit, it is possible to measure the sensor output from each kit. By taking the vehicle through a wide air flow range on the dyno, it is possible to measure the relative outputs of the two sensors when mounted in the two induction kits. A tool such as DelatDash External Sensor Logging may then be used to record the data for later analysis.

The result of recording this data is a graph showing the relative outputs of the two induction kits for a range of air flows.

Nice. Lets automate that!

Cheers,
Hymee.

No; that will NEVER work!

You do your part and I'll do mine!

In order to correctly calibrate the ECU for the new induction kit, the relationship between air flow and output voltage of the air flow sensor must be measured. Initial thoughts would suggest using a flow bench to do this, though very few tuners will have access to a calibrated flow bench capable of flowing 300 grams per second of air.

In actual fact, all tuners do in fact have this flow bench in their workshops - the vehicle itself ! By plumbing the the two induction kits in series (the standard system and the new system), it is possible to draw the same mass of air through each induction kit. With a mass air flow sensor in each induction kit, it is possible to measure the sensor output from each kit. By taking the vehicle through a wide air flow range on the dyno, it is possible to measure the relative outputs of the two sensors when mounted in the two induction kits. A tool such as DelatDash External Sensor Logging may then be used to record the data for later analysis.

The result of recording this data is a graph showing the relative outputs of the two induction kits for a range of air flows.

Huh? Parallel logging seems kind of a waste of time in most cases as we have a feedback loop already.

BTW - can an admin fix my crappy spelling on the title... I fail.

Great info!

dumb question how do you know what is open loop and closed loop?

WOT and above 4000 RPMS or so is open loop.

There is a load threshold too - but I don't know it....

Great info!

dumb question how do you know what is open loop and closed loop?

What Kane said, plus a couple of other things...

You can see it with a sCAN tool... (??) :) as one of the params monitored is "fuel system status".

You can also generally see it in logging - when the lambda is around 1.0 give or take a small amount, then it is closed loop. When she drops out of closed loop, lambda will generally be nowhere near 1.0 (consistently).

Cheers,
Hymee.

I'm just going to blurt out my opinion on maf tuning here...

If your maf is reading around 5-6 g/s at idle, has a steady voltage output at idle and light cruise, and if the max flow (normally aspirated) is around 210-220ish g/s then there is no use in touching the maf. If you have a stock intake there is probably no need to touch it and if you have your maf sitting in a different size pipe that should be pretty easy to accommodate with using some math.

I tried to tune only tweaking the maf last summer and all you do is end up chasing your tail.

I'm just going to blurt out my opinion on maf tuning here...

If your maf is reading around 5-6 g/s at idle, has a steady voltage output at idle and light cruise, and if the max flow (normally aspirated) is around 210-220ish g/s then there is no use in touching the maf. If you have a stock intake there is probably no need to touch it and if you have your maf sitting in a different size pipe that should be pretty easy to accommodate with using some math.

I tried to tune only tweaking the maf last summer and all you do is end up chasing your tail.

If running an all stock intake - then there isn't any reason to scale the MAF....

Listen folks - this isn't tuning - this is getting accurate information to the PCM... you can't "tune" with messing with the MAF sensor - you get it right, then get your fuel injectors right, then go tune the fuel and timing tables.


Anything else is a band-aid

No; that will NEVER work!


Huh? Parallel logging seems kind of a waste of time in most cases as we have a feedback loop already.

The procedure I described was to recalibrate the MAF sensor when used in a new tube of a different diameter or new induction kit.
I don't see any reason for messing with MAF when the intake is a stock setup and LTFTs are perfect.

Ah I see ahlan.

Yes that would work; my process is doing the same thing just with the new induction kit already on the car.

If running an all stock intake - then there isn't any reason to scale the MAF....

Listen folks - this isn't tuning - this is getting accurate information to the PCM... you can't "tune" with messing with the MAF sensor - you get it right, then get your fuel injectors right, then go tune the fuel and timing tables.


Anything else is a band-aid

Well said!

Cheers,
Hymee.

WOT and above 4000 RPMS or so is open loop.

There is a load threshold too - but I don't know it....

I started to have a look at my logs and realized most of the cruising speeds where there is a lambda of 1 seemed to be closed loop areas.

What Kane said, plus a couple of other things...

You can see it with a sCAN tool... (??) :) as one of the params monitored is "fuel system status".

You can also generally see it in logging - when the lambda is around 1.0 give or take a small amount, then it is closed loop. When she drops out of closed loop, lambda will generally be nowhere near 1.0 (consistently).

Cheers,
Hymee.

I'd better start sCANing :spank:

I also just read if the STFT is a fixed number - you're in open loop.

SO that is another way to tell.

The way I did it was kind ot a reverse version of Kanes , but I had no instructions so had to start somewhere .

*Made sure my LTFT had settled down
*adjusted entire maf up to 200g/s (typical N/A range) by the LTFT value . IE if LTFT was -10 i took 10% from all values up to 200g/s.
*Did some WOT logs (always monitoring AFRs real time as well) and adjusted maf to approx. get the AFRs i wanted in the ranges above 200
*fine tune with fuel maps

not saying this is any better - just another way to to it .....

BTW - I just checked my idle and it reads 5g/s :)

I have a question about Maf scaling? I installed a AEM intake and I expected my Maf to need recalibrated but my LTFT's are consitently at 0. I've put about a hundred miles on the car with several drive cycles and the LTFT's are still 0. With the stock intake I saw on average +7 for LTFT. Does this sound normal? My AFR's are a little off, so I'm assuming it has to be a Maf issue.

I have a question about Maf scaling? I installed a AEM intake and I expected my Maf to need recalibrated but my LTFT's are consitently at 0. I've put about a hundred miles on the car with several drive cycles and the LTFT's are still 0. With the stock intake I saw on average +7 for LTFT. Does this sound normal? My AFR's are a little off, so I'm assuming it has to be a Maf issue.

AEM is a good stock type intake MAF wise....if you AFR's are off though - retune for it to be sure it is accurate.


Could also be the fuel injectors; or a combo.

AEM is a good stock type intake MAF wise....if you AFR's are off though - retune for it to be sure it is accurate.


Could also be the fuel injectors; or a combo.

Thanks for the advise. I will take a look at the injectors.

A couple of quick points:

1) The MAF sensor's response is not linear. Its not even an exponential curve. Its response not really even a curve. Its a sorta lumpy Bézier. The response curve that is described in the PCM MAF calibration table is an "optimized" average of the expected MAF output.
2) The injector output is also not a perfect curve and changing its scaling changes the slope of that fuel delivery rate curve.
3) The fuel delivery calculation is based on load, not the MAF output, so chasing a lambda value based on the MAF reading wont work most of the time.
4) [TRADE SECRET AHEAD] Open loop is adjusted by several interacting tables based on RPM, MAF, accelerator position, vehicle speed (gear) and load. However, a quick way to force open loop is to disconnect the front O2 sensor.

Bare with me here.

First question. How can the function that describes the output of the MAF sensor, NOT be a 1:1, invertable function? I am not asking about the physics (which, at a first approximation, would suggest that such a descriptive model WOULD be invertable), but the simple fact that the pcm needs to convert voltage back to measured flux. If the function was grossly lumpy then a voltage would be potentially ambiguous, solving for two or more flows.

Second: I am trying to determine if i should proceed with MAF calibration with my current intake configuration. I have the greddy tube, however, I have pretty smooth AFR traces. HOW smooth is smooth. I dont have a problem changing intakes; however, I have more than a foot of straight pipe infront of the MAF housing prior to a ninty degree bend. I think it is pretty good. Furthermore, to smooth out the flow a little, I was considering putting in a screen Could I simply take some flexable screen and just seal it onto the maf housing with the coupler and a t-bolt clamp?

Third: what would be the best base map to start modifing? Would any functional map serve as a base?

Finally, in order to perform the MAF calibration, I need to scale my injectors as a first approximation. I know the scaling function, but how do i know these will provide a sufficiently safely rich mix in boost?

Actual final question. Seriously, how many times can I flash my PCM. I figure, I will be flashing it several hundered times in the next 6 months (the time i figure it will take me to tune.)

The instantaneous MAF value isn't useful.
Its trended.
That is why its non-linear out put is useable.

You can calibrate any intake you want. Just be prepared to fight with it and still come out with an unsatisfactory output.

The "Stock Style" MAP is the latest Mazda flash, unaltered.

The injector scaling is voodoo. Because latency affects it and delivery is a variable, not knowing the MAF output is going to make it impossible to know the proper scaling.
So, just pull it out of your butt and be prepared to address it again later.

I have about 400 flashes on my PCM and counting.

The only thing that works for me is MAF scaling , the fuel maps do not seem to have any effect :dunno:

The instantaneous MAF value isn't useful.
Its trended.
That is why its non-linear out put is useable.

That is why it is important to hold your stable engine condition until the AFR's stabilize when scaling.

That is why it is important to hold your stable engine condition until the AFR's stabilize when scaling.

Not physically possible with the MAF.
Even at steady-state, the MAF is trended.

This stuff isn't theoretical.

The only thing that works for me is MAF scaling , the fuel maps do not seem to have any effect

Huh?

Under what conditions?

Mucho thanks.


I can imagine how the output of a moving average or autocorrelation function could function as a sort of vector, but you would still need some locally invertable MAF scale at startup. Also, if, as you say the MAF scale is non-invertable over a significanly large section of the curve, then if the voltage drops yet the flow increases, you would have to know WHICH local max you just passed. IF that is true, that is a pretty remarkable computer.

But WAIT! Imagine that you are the PCM. You know where you are in the local area of the curve because you have been following voltage from some unambiguous startup value. Now you are at the peak of some local maximum. No matter what happens next, the voltage is going to drop. Lets say that the flow increases and the voltage drops to the right of the local maximum. How do you know, as the PCM, that you fell off the local max to the right. you could have very well dropped off to the left and needed less fuel. Ah! but as the PCM, you also know TP and a host of other outputs. I am begining to see how the interactions become complecated....

...just because I didn't see this in the thread...

If your changing the size of your MAF tube and therefore need to recalibrate the maf curve the starting point should be to multiply the old curve the the ratio of the new and old maf tube areas. In other works your starting scale factor would be...

scale % = ((new tube radius^2/old tube radius^2)-1)*100

This is based on assumptions you have fully laminar flow, the maf remains in the same relative position in the tube, and your not changing air properties or creating turbulence somewhere....

Not physically possible with the MAF.
Even at steady-state, the MAF is trended.

This stuff isn't theoretical.



Huh?

Under what conditions?

only areas i've tried adjusting are load points in the 110-200% range . I've put my target lambda down to as low as 0.5 over the whole 3000 - 5500 /110-200% range and not seen any change in the afrs whatsoever .
Overall my AFRs are pretty damn good with just maf scaling done but i would like to richen up a few peaks for those times the LTFT wanders around a bit .

if the maf isn't just a smooth curve it sounds like Kane's software would work well finding the average? even if it isn't a smooth curve all the time it can obviously be approximated by one.

I reckon the MAF would do stuff all on cranking. there would be cold start parameters in there instead of calculated loads. once it's running then it would use the MAF

Huh?

Under what conditions?

figured it out today - my pcm has only one fueling map for all 6 gears . The other 2 fueling maps do nothing at all .....

figured it out today - my pcm has only one fueling map for all 6 gears . The other 2 fueling maps do nothing at all .....

WHAT?

Which PCM are you using?
I've got just about every ROM available to me and ALL of the maps are active and tied to their appropriate gear.
You've got something wrong there.

WHAT?

Which PCM are you using?
I've got just about every ROM available to me and ALL of the maps are active and tied to their appropriate gear.
You've got something wrong there.

Maybe there is something wrong .
I can tell you i made huge changes on the other two maps and they did absolutely nothing . The map i had thought was gear 1-2 i had not touched much because I was concentrating on getting 3rd gear right initially so was working on what i thought was the 3-4 map mostly .

As soon as i started playing with map 1 it affected AFRs in ALL the gears .

PCM is N3YMEC which is from the first year of JDM production ....

Maybe there is something wrong .
I can tell you i made huge changes on the other two maps and they did absolutely nothing . The map i had thought was gear 1-2 i had not touched much because I was concentrating on getting 3rd gear right initially so was working on what i thought was the 3-4 map mostly .

As soon as i started playing with map 1 it affected AFRs in ALL the gears .

PCM is N3YMEC which is from the first year of JDM production ....
Anything more to tell about this???

Just that now that i have that figured out i have finally been able to fine tune without trying to do it via maf calibration ....

Just a simple stupid question. In order to scale the MAF, I guess that I will have to create a MAP that force the ECU to always put 14,70 AFR into the engine ?

Is it safe to do so ? My map is ready with open loop and closed loop set at 14,7. I'm not planning on running at high load and high RPM with that map but I'm still a bit worried about hurting the engine with that MAP.

Did I understood your recomendations right ? I have the REVI intake and currently my LFT is at +11, I guess that I need a MAF calibration for that reason.

Thanks,

I'm a bit worried about the lamda 1 instructions also.

popping the rear o2 sensor? (check which one is correct) is supposed to force open loop.

then for the different lambdas use different gears. the lamda 1 area changes for gears 1-2, 3-4, 5-6 so you can probably avoid changing the maps and still hit a large portion of the maf range.

also as for FI instructions you really just need to reference back to the lambda target your map is set for. it should be hitting what ever lambda it is set for at the load/rpm.

LOW-LOAD; LOW-RPM IS THE KEY HERE GUYS.

If you want though; you can use whatever lamda or AFR you like; it just makes the math harder.

Detailed and Corrected Scaling Procedure for MAF Calibration
I believe that the majority of the discussion with regards to the DETAILS of scaling the MAF are in error at some level. I want to post my methods of scaling my MAF calibration.

I have not yet seen either an error free or full, detailed description of the MAF scaling here. It is not my intention to call these errors out; however, I will point out that the majority of the errors lie in the very fine details of calculating a correction. They may have been written in haste and escaped into the posts this way, or the poster had no fucking idea what they were talking about. Either way, if you attempt to use these methods, the results are garbage.

The importance of scaling the MAF has been discussed many times, but it deserves repeating. If your MAF is not scaled correctly, and you make alterations to any map, you are NOT tuning. This sounds like semantics, and in a way, it is. If you define “tuning” as making logical adjustments to maps to provide optimum, power, response, transition and safety, then you CAN NOT TUNE WITHOUT A PROPERLY SCALED MAF. Without the proper scaling you can make NO logical adjustment. You can only hunt and peck and hope you get somewhere. I did not come up with this conclusion; I only repeat it because I came to the same conclusion after hours of careful consideration of multiple data that all point, inexorably, to this terminus.

When considering the MAF system, there are some defined concepts to separate. First is air flow. There are THREE air flow numbers to conceptualize. First is true airflow. This is the number we wish to know. Then there is measured airflow. This is the number that we arrive at through the workings of the MAF. They are two different numbers even when they are equivalent (which we wish them to be.) Finally, relevant to this discussion is Adjusted airflow, which is measured airflow, to which we have applied a correction factor. What we want to do is determine if our MAF is reporting measured airflow that is very close to the true value, and we will make corrections through adjusted airflow.

When the engine ingests fuel, and a sub-stoichiometric amount of air is concomitantly consumed, excess fuel is expelled. We have a relatively direct measure of the ratio of how much fuel relative to the amount of oxygen that is consumed (λ). If the computer thinks that more air is entering the engine than what actually is, the computer will mistakenly add too much fuel, creating a rich condition and vice versa. We can evaluate the factor by which the computer made a mistake by comparing the output AFR to the expected AFR. This factor is correctly evaluated as (measured AFR-expected AFR)/expected AFR.

This simple equation can be used as the key to iteratively seeking the correct MAF scaling at any achievable flow.

This calculated value is the unitless correction that can be applied to your measured flow to appropriately scale your MAF calibration at the relevant voltage. Because LOAD is a component in this calculation (as you will see), it is possible that a single iteration of this method will only bring you closer to a correct scale. Multiple iterations may be required.

So lets see this calculation in action. My personal example of scaling the MAF started with my stubbornness to remove my Greddy MAF housing. I still believe that this housing can work with the appropriate physical set up, but I have put an AEM CAI on order to play with as well. I believe that my starting map is scaled to the stock intake box. I believe that this box has an inner diameter in the housing of 3 and 3/8”, but I don’t have one handy to measure. The Greddy inner diameter can be approximated to 3”, but the geometry is somewhat complex.

I noticed that while I could reasonably hit expected AFR at low and higher loads, at medium loads I was incredibly rich. So I decided to scale the entire MAF curve down by about 14%. These curves are plotted in the attached graph as open symbols.

When I did this, although I could target some AFR, I got very lean at load. The reason, I believe, is that scaling by a factor can only be correctly done by parameterizing said factor into the model that controls the shape of the curve. Just moving the curve up by a factor does not take into account the physical model at hand.

Nevertheless, empirical adjustment of the MAF curve is much more accurate and much more useful. The details are such:

Log data IN OPEN LOOP. There are ways to ensure you are in OL, but it is verifiable by checking to make sure that in the areas of interest in your logs, that LTFT-STFT=0. Make sure that when you log your data (Load, RPM, Mass air, MAF volts, AFR) that you have allowed the trims to settle and that you are at operating temp. I personally start the car, bring it to temp for 10 minutes, shut down for a few minutes then restart and drive for 10 more minutes.

Now the process of generating a correction is easy; using your open loop fuel table, look up the expected lambda for each data point using RPM and load as variables. This can be done by hand, but it is tedious. I discovered an Add-on for Excel at http://www.xlxtrfun.com/XlXtrFun/XlXtrFun.htm that allows you to use an indexed fuel table (or a matrix, though I have not tried it) to interpolate (pull-out) expected lambda using your measured load and rpm as variables. This is a very slick way to do this, IMO.

Once you have sets of expected and measured lambda at various measured flow, you can calculate your correction factor (as above), add 1, multiply to measured flow to generate corrected or adjusted flow vs MAF voltage (Plotted in RED). I intend to either fit a single exponential to this adjusted flow and use the parameters to generate an interpolated curve or hand-fit to generate my newly scaled MAF curve. Either way will work. Using this method you can populate all physically achievable areas of the curve. This is recommended.

If you look at my adjusted data, you will see that while some of the curve needs adjustment, some areas do not. That is why I was getting weird results; i.e., rich in some spots, lean in others.

In any case, the only real advancement here is the ability to “pull-out” expected AFR from a fuel table instead of doing it by hand using readily available software and add-ons. I also believe that this method is sufficiently detailed to CORRECTLY scale the MAF calibration.

[FONT=Arial][SIZE=3][COLOR=#000000]Let me know if this clears up any issues, if there are further issues or if any files or screenshots would be helpful.

nice write up Carbon .
My approach was different again but kinda got to the same point you did - eventually :)

Nice - would you be willing to add it to my thread?

Yeah, I actually wanted to post this in it, but didnt want to be insulting. Your method is PERFECTLY fine but there are some math errors in it (in the downloadable excel sheet) that I wanted to correct. If you want, I can just post this over there and delete this. No reason to make clutter. Your thread started me off, so it only makes sence.

i tried this method about a year ago with decent results...

it gets complicated though because different gears had different lambda's due to different fuel maps and the resultant afr error %s were different. You also assume in this method that your fuel injectors are scaled properly which may or may not be a good assumption.


BTW - the black text is really hard to read if you have a dark theme on this site. Its better to use the default color.

I forgot about the text thing. I wondered why it looked funny. I just copied and pasted from Word. If Kane wants me to post over on his thread. I will correct it. (or maybe we can get a mod to merge them and save me the effort. :) )

I have attached a screenshot of the excel sheet of the process. It includeds the addon mentioned.

http://www.rx8club.com/attachment.php?attachmentid=136570&d=1239202003

The gearing isnt an issue because you can set up an equation to calculate which gear you are in from speed and rpm. This way, the excel sheet can choose which map to draw the expected lambda. That is easy.

The injector issue is a problem that I have not yet evaluated experimentally. Depending on how the injectors are staged, it is possible that this calibration proceedure eliminates fine tuning the injector scaling. I will get to that later, but the process to determine the exact injector scale should be analogous to this one, only with using injector duty cycles.

If you want my 2 cents, ignore the data from first and second gears if you are anywhere close to WOT as the latency from the O2 sensor becomes rediculous and somewhat random. An easy why of checking it it before doing a WOT run, start logging, get off the gas pedal completely to do a fuel cut, watch for the afr to go to 20, then do the WOT run and at the end sharply get off the gas to go into the fuel cut. Look at the data and if you have the throttle position logged you'll probably find a latency at the beginning of the run of 1-2 data points (~ .2 sec) and by the end 5-6 data points (~.5 sec) which is enough to be off by a good 1,000+ rpms in second gear.

I found doing my own tuning I had a strange blip around 7000 -7250 rpms in every run in second gear that would not respond to changing the fuel tables in that area. I finally figured out it was occuring at 6250 rpms with the APV opening and suddenly I was able to tune the problem away by changing the map down there. Perhaps this weekend I will get some time to put together how I came up with my final tune...

/\ yes - the point the aux. port changes afrs changes in each gear , getting progressively closer to 6250 as you go up ....

/\ yes - the point the aux. port changes afrs changes in each gear , getting progressively closer to 6250 as you go up ....

Yup, the higher the gear the slower your rpms change and the less the latency problem becomes.

I was hell bent at being super scientific when I started my tuning and have every change backed up with number - then i discovered this and realized the changed needed to really be made with an artistic brush :eyetwitch

Even worse for me because I discovered that i only have one fuel map that covers all 6 gears so i have to richen it up over a wide band to keep it going too lean ....

Ideally you should scale the MAF with the stock injectors with stock injector scaling. Once the MAF is scaled, then go install and scale the injectors. Both scalings affect fuel trims making it very hard to tune both variables at once. I have never tuned an RX8 but carbon asked my opinion since I do tune Mitsus and Subies. When I tune a MAF I log fuel trims and airflow under closed loop fueling. Ideally fuel trims should be close to zero (in reality zero is impossible but anything under 3-4% is good). Bear in mind that STFT will get pushed into LTFT. So the STFT requires the most attention. In Evo tuning we use airflow under cruise closed loop conditions and at idle to figure out the % correction. eg, if at an idle of 50 hz airflow, my STFT is consistently +10%, then the ECU is adding 10% additional fuel to account for 10% excessive airflow. So the MAF scaling needs to be decreased 10% at that airflow range. You do the same for different cruise conditions/airflow ranges. Once the MAF is scaled, you repeat the same for injectors. Although a lot of injector latency numbers are already published and can just be copied.

wouldnt it just be easier to log afr target and afr actual along with maf voltage and compute your adjustments from there?

wouldnt it just be easier to log afr target and afr actual along with maf voltage and compute your adjustments from there?

Uh, that is exactly what I said. Problem is, if you search for details on this process, the are in error or not detailed.

Ideally you should scale the MAF with the stock injectors with stock injector scaling. Once the MAF is scaled, then installed and scale the injectors. Both scalings affect fuel trims making it very hard to tune both variables at once. I have never tuned an RX8 but carbon asked my opinion since I do tune Mitsus and Subies. When I tune a MAF I log fuel trims and airflow under closed loop fueling. Ideally fuel trims should be close to zero (in reality zero is impossible but anything under 3-4% is good). Bear in mind that STFT will get pushed into LTFT. So the STFT requires the most attention. In Evo tuning we use airflow under cruise closed loop conditions and at idle to figure out the % correction. eg, if at an idle of 50 hz airflow, my STFT is consistently +10%, then the ECU is adding 10% additional fuel to account for 10% excessive airflow. So the MAF scaling needs to be decreased 10% at that airflow range. You do the same for different cruise conditions/airflow ranges. Once the MAF is scaled, you repeat the same for injectors. Although a lot of injector latency numbers are already published and can just be copied.


Thanks. Scaling with the stock injectors is not possible as I was not aware at the time I installed them that I would have to do this. And, as you well know, I am not taking them off. In any case, the flow numbers on the stockers are as much a shot in the dark as the upgrades, and the scaling (if necessary) can happen after the MAF is nailed down.

I dont know anything about the evo, (and only infinitesimally more about the RX8,) but as for scaling MAF cali in closed loop on the RX8, it seems that would miss all the RPM ranges above about 4000rpm, below which is barely usable low load territory. Scaling above this is the entire range where you can go "pop!" What you learn below 4000rpm cant be used to scale above this rpm.

I now understand what you were talking about over sushi, but I am nearly certain that the methods are not transferable between the cars.

my take on what you posted, is that you log afr actual, maf voltage, load, etc and lookup in the expected afr table in the map. what i was suggesting is to log both afr actual and afr expected at the same time, as well as maf voltage, then just computing from there.

OH! I didn't understand you. I dont have the capability to log expected AFR. That would make this process MUCH easier. It is possible that those that can do live datalogging with the dongle and a computer can do this. Even if this is possible, I, unfortunately, can not get my computer hooked up to talk to the car. I have basically given up and just log with the AP. But, yeah, logging expected AFR would be a nice thing to have for this process.

OH! I didn't understand you. I dont have the capability to log expected AFR. That would make this process MUCH easier. It is possible that those that can do live datalogging with the dongle and a computer can do this. Even if this is possible, I, unfortunately, can not get my computer hooked up to talk to the car. I have basically given up and just log with the AP. But, yeah, logging expected AFR would be a nice thing to have for this process.

Yeah that is my way of doing it - but you have to set your fuel maps up for it prior to - with your excel file though you can just use the AFR/lambda from your fuel base map.

I am def down to have this in my thread - can a mod move it?

I am def down to have this in my thread - can a mod move it?

Seconded.

Is it in the wrong forum?

Combine them - mine and Carbons?

my take on what you posted, is that you log afr actual, maf voltage, load, etc and lookup in the expected afr table in the map. what i was suggesting is to log both afr actual and afr expected at the same time, as well as maf voltage, then just computing from there.

That would be sweet.....if you could do it. The AP can't log Target Lambda tho.

That would be sweet.....if you could do it. The AP can't log Target Lambda tho.

Set them to 1 like I told you on the phone for low load areas - and you got your target.

AH! So I understand a little more now what you were trying to do, Kane. Your step one is fine, mathematically, but your suggestion to scale the entire maf curve by what ever you see at lower load/flow is not useful. WHat happens at low flow does not predict high. Furthermore, your step two needs, as you say, more math; specifically to convert to using the % error equation to determine the adjustment factor.

Download those excel addons and try them out. I can send you my excel sheet that has a way to convert a matrix of fueling to indexed fueling and the set up to extract expected AFR. Using this, you can see that you could completely automate MAF calibration. Even to the point where you take an excel macro, feed it your fuel table, current MAF cali, and a datalog, and it spits out a new, spline-fit, MAF calibration. Lotta work tho.

Thanks. Scaling with the stock injectors is not possible as I was not aware at the time I installed them that I would have to do this. And, as you well know, I am not taking them off. In any case, the flow numbers on the stockers are as much a shot in the dark as the upgrades, and the scaling (if necessary) can happen after the MAF is nailed down.

I dont know anything about the evo, (and only infinitesimally more about the RX8,) but as for scaling MAF cali in closed loop on the RX8, it seems that would miss all the RPM ranges above about 4000rpm, below which is barely usable low load territory. Scaling above this is the entire range where you can go "pop!" What you learn below 4000rpm cant be used to scale above this rpm.

I now understand what you were talking about over sushi, but I am nearly certain that the methods are not transferable between the cars.

I don't bother scaling for high load and/or high rpm ranges. At those ranges, the ECU is in open loop fueling and I can tune for what I need via the OL fuel map. In Evo tuning, you need to adjust MAF scaling to fix closed loop fueling issues for fuel economy and to prevent lean/rich CELs and for a good idle. These conditions only occur off boost when the ECU is striving for stoichiometric fueling (14.7:1).

Since you are scaling the MAF based on lambda and airflow, I personally don't see how you can properly scale the MAF when you are going to be running rich across the board with bigger non-stock injectors. Also I would assume that the RX8 has a fuel injector scaling table and injector voltage latency table. Or at least I would hope that the Cobb Accessport has those functions. At the very least, get the published latency numbers for the injectors and use a rough approximation of the scaling (roughly 85% of the published flow). Then you will at least be in the ball park for injector scaling. What are the stock injector sizes and what are your aftermarket injector sizes?

I don't bother scaling for high load and/or high rpm ranges. At those ranges, the ECU is in open loop fueling and I can tune for what I need via the OL fuel map. In Evo tuning, you need to adjust MAF scaling to fix closed loop fueling issues for fuel economy and to prevent lean/rich CELs and for a good idle. These conditions only occur off boost when the ECU is striving for stoichiometric fueling (14.7:1).
So, specific to tuning with the accessport, and perhaps specific to the RX8 pcm, the OL maps are afr targets that the pcm wants to hit. The computer doesnt read from the wb02 but it does use the maf to determine how much fuel is required. If the maf is miscalibrated, the fuel is wrong. So duty cycles are determined by the computer, not directly from a map. The maf is critical in that respect. Also, just fyi, the maps are load-based, not pressure-based. After you get used to that, it actually makes a lot of sense.

Since you are scaling the MAF based on lambda and airflow, I personally don't see how you can properly scale the MAF when you are going to be running rich across the board with bigger non-stock injectors. Also I would assume that the RX8 has a fuel injector scaling table and injector voltage latency table. Or at least I would hope that the Cobb Accessport has those functions. At the very least, get the published latency numbers for the injectors and use a rough approximation of the scaling (roughly 85% of the published flow). Then you will at least be in the ball park for injector scaling. What are the stock injector sizes and what are your aftermarket injector sizes? There is a separate individual scaling factor that can be used as a first approximation for injector size. I have left the latencies alone until I can get more information on my specific injectors. Unfortunately, I only know the identity of my 780cc/min injectors. My 650s had no identifing marks.

So, specific to tuning with the accessport, and perhaps specific to the RX8 pcm, the OL maps are afr targets that the pcm wants to hit. The computer doesnt read from the wb02 but it does use the maf to determine how much fuel is required. If the maf is miscalibrated, the fuel is wrong. So duty cycles are determined by the computer, not directly from a map. The maf is critical in that respect. Also, just fyi, the maps are load-based, not pressure-based. After you get used to that, it actually makes a lot of sense.

There is more to it then that though. The amount of fuel the pcm delivers in open loop is also corrected based on the VE tables, load limiting tables, and more... its not just the maf being off

There is more to it then that though. The amount of fuel the pcm delivers in open loop is also corrected based on the VE tables, load limiting tables, and more... its not just the maf being off

Noted. Although relevant to this discussion (with a misguided piston driven tuner) I think it is sufficent to describe it in the simplest terms. Measured airflow drives fueling in OL. How that signal is modulated after the airflow is measured is an important, but secondary, consideration.

So, specific to tuning with the accessport, and perhaps specific to the RX8 pcm, the OL maps are afr targets that the pcm wants to hit. The computer doesnt read from the wb02 but it does use the maf to determine how much fuel is required. If the maf is miscalibrated, the fuel is wrong. So duty cycles are determined by the computer, not directly from a map. The maf is critical in that respect. Also, just fyi, the maps are load-based, not pressure-based. After you get used to that, it actually makes a lot of sense.

There is a separate individual scaling factor that can be used as a first approximation for injector size. I have left the latencies alone until I can get more information on my specific injectors. Unfortunately, I only know the identity of my 780cc/min injectors. My 650s had no identifing marks.

Your 1st statement implies the the RX8 has a wideband O2 sensor from the factory. I agree, if the ECU is shooting for an AFR target than yes, the MAF needs to scaled at all rpm/load ranges. In addition, that technically means that you are no longer open loop... The ECU is using O2 sensor feedback to hit a goal lambda for a particular rpm and load. If that is the case, then tuning is simple. Plug in your goal lambda under boost and let the ECU figure the rest.

There are really only 3 brands of injectors regardless of who sells or labels them:
1. JDM injectors are made by Denso and sold under Greddy, Blitz, HKS, OEM labels.
2. USDM injectors are made by Delphi and sold under PTE, Fuel injector clinic, RC engineering, and tons of other labels (BTW I use these in my Evo).
3. German injectors made by Siemens.

Latency numbers are locatable via google. You just need to know which 650cc injector you have for example (Denso or Delphi). Also different sellers rate their injectors differently. Fuel Injector clinic rates their injectors conservatively. What they call a 750cc injector, PTE sells as a 780cc injector. But they are the same injector. There is a LOT of latency numbers available on the evo forums.

One more point: In Evo tuning, our open loop fuel maps have "AFR" numbers in them for particular loads and rpms. These are not true AFRs in a modded Evo. Once you mod, the numbers are meaningless. You have to make adjustments to these numbers to achieve the wideband AFRs numbers that you want. eg, if my AFR at 4000 rpm and 250% load is 10.5:1 I would increase the numbers in those cells by 2% to get my WB AFR above 11:1. The actual number in the map is meaningless (it could say "9.9"). What is important is what change I make to the number. I say all this because I wonder if open loop fueling in RX8's is similar.

Your 1st statement implies the the RX8 has a wideband O2 sensor from the factory. I agree, if the ECU is shooting for an AFR target than yes, the MAF needs to scaled at all rpm/load ranges. In addition, that technically means that you are no longer open loop... The ECU is using O2 sensor feedback to hit a goal lambda for a particular rpm and load. If that is the case, then tuning is simple. Plug in your goal lambda under boost and let the ECU figure the rest.

Yes, there is a wbo2 on the car from the factory. In CL the pcm uses this information to hit targets. Under high load, I do not believe that the wb02 is used, hence, OL. What specific algorithms that are used to determine fuel is complicated as rotor noted, but the general idea is that, yes, put in your target afr and hit the gas.


There are really only 3 brands of injectors regardless of who sells or labels them:
1. JDM injectors are made by Denso and sold under Greddy, Blitz, HKS, OEM labels.
2. USDM injectors are made by Delphi and sold under PTE, Fuel injector clinic, RC engineering, and tons of other labels (BTW I use these in my Evo).
3. German injectors made by Siemens.

Latency numbers are locatable via google. You just need to know which 650cc injector you have for example (Denso or Delphi). Also different sellers rate their injectors differently. Fuel Injector clinic rates their injectors conservatively. What they call a 750cc injector, PTE sells as a 780cc injector. But they are the same injector. There is a LOT of latency numbers available on the evo forums.

I will post a pic of the 650s to help get an idea of what they are.

One more point: In Evo tuning, our open loop fuel maps have "AFR" numbers in them for particular loads and rpms. These are not true AFRs in a modded Evo. Once you mod, the numbers are meaningless. You have to make adjustments to these numbers to achieve the wideband AFRs numbers that you want. eg, if my AFR at 4000 rpm and 250% load is 10.5:1 I would increase the numbers in those cells by 2% to get my WB AFR above 11:1. The actual number in the map is meaningless (it could say "9.9"). What is important is what change I make to the number. I say all this because I wonder if open loop fueling in RX8's is similar.

It is similar, but how I understand the OL fueling to work in the RX8, if the maf is scaled correctly and the injectors are scaled correctly, and no adjustments are made by the pcm by a multitude of other tables, the lambda you put in should be the lambda you get out. Again this is much further down the line in actual tuning than what I was discussing for maf calibration.

It is similar, but how I understand the OL fueling to work in the RX8, if the maf is scaled correctly and the injectors are scaled correctly, and no adjustments are made by the pcm by a multitude of other tables, the lambda you put in should be the lambda you get out. Again this is much further down the line in actual tuning than what I was discussing for maf calibration.
I would only agree with this statement if you are running n/a with stock injectors and stock exhaust, etc. With boost and all your other mods, the VE of the motor is radically different. I believe that the Evo method of tuning would work fine since your "target" lambda under OL fueling is now meaningless with forced induction. If I am correct then you guys can skip high load/rpm MAF scaling headaches. Just compensate in your OL fuel maps.

AH! So I understand a little more now what you were trying to do, Kane. Your step one is fine, mathematically, but your suggestion to scale the entire maf curve by what ever you see at lower load/flow is not useful. WHat happens at low flow does not predict high. Furthermore, your step two needs, as you say, more math; specifically to convert to using the % error equation to determine the adjustment factor.

Download those excel addons and try them out. I can send you my excel sheet that has a way to convert a matrix of fueling to indexed fueling and the set up to extract expected AFR. Using this, you can see that you could completely automate MAF calibration. Even to the point where you take an excel macro, feed it your fuel table, current MAF cali, and a datalog, and it spits out a new, spline-fit, MAF calibration. Lotta work tho.

I'll look at it this weekend and update the first post. But to clarify - YOU ONLY correct the MAF at high loads and high RPM's if you are trending by a certain %. All that does is bring your higher MAF voltages closer to actual to be safer UNTIL you actually load up the car and scale them. This is critical in FI because if your whole MAF needs to be scaled by say 10% - and you don't scale the higher points as well then your going to be dangerously lean when you access those points in the MAF.


Yes, there is a wbo2 on the car from the factory. In CL the pcm uses this information to hit targets. Under high load, I do not believe that the wb02 is used, hence, OL. What specific algorithms that are used to determine fuel is complicated as rotor noted, but the general idea is that, yes, put in your target afr and hit the gas.


I will post a pic of the 650s to help get an idea of what they are.

Correct - the WBO2 does not provide fuel corrections in OL - but the g/sec calculation determines absolute load, which determines fuel. ACCURACY is KEY - get your sensors and tables right - then tune.

I would only agree with this statement if you are running n/a with stock injectors and stock exhaust, etc. With boost and all your other mods, the VE of the motor is radically different. I believe that the Evo method of tuning would work fine since your "target" lambda under OL fueling is now meaningless with forced induction. If I am correct then you guys can skip high load/rpm MAF scaling headaches. Just compensate in your OL fuel maps.

It is not meaningless - you want accurate sensor information first - then set your fuel maps tp the desired Lambda. We don't need to trick stuff any more with a flash tuner - so don't. Take the time and do it right.

You will never get the open loop AFR to match the targets in the table. To do that you need to get all 3 injectors perfect, you need to get the maf perfect, the engine VE perfect, the fuel VE table perfect, and then deal with the flow disturbances caused by the various ports opening and closing.

You are not going to get all that perfect unless you spend wwaaaayyyy too long on an chassis dyno holding the car in a steady state condition in all areas of the maps.

You will never get the open loop AFR to match the targets in the table. To do that you need to get all 3 injectors perfect, you need to get the maf perfect, the engine VE perfect, the fuel VE table perfect, and then deal with the flow disturbances caused by the various ports opening and closing.

You are not going to get all that perfect unless you spend wwaaaayyyy too long on an chassis dyno holding the car in a steady state condition in all areas of the maps.

I agree . I spent all of yesterday trying to get my maf lined up with my lambda in the fuel tables . It did not work . I got things near perfect at higher flow rates IE actual lambda lining up with expected lambda at 150-170% load . But found that at 110-130% load I was way too rich .
I could not make sense of it but that was the reality .

I'll look at it this weekend and update the first post. But to clarify - YOU ONLY correct the MAF at high loads and high RPM's if you are trending by a certain %. All that does is bring your higher MAF voltages closer to actual to be safer UNTIL you actually load up the car and scale them. This is critical in FI because if your whole MAF needs to be scaled by say 10% - and you don't scale the higher points as well then your going to be dangerously lean when you access those points in the MAF.

I agree with this 100%. Scale the calibration curve UP to make the car rich (make it think it is getting more air than it is.) Then collect data, step-by-step, at increasingly higher flows, calibrating as you go.

You will never get the open loop AFR to match the targets in the table.


Not perfect, no. But if you gather much data under various conditions then plot Adjusted Flow vs. Voltage, you will see the a scatter of data, through the middle of which your calibration curve should go. The relative error looks to be about 10%.

It seems like some people are nuking this thing.... perfect is the enemy of good - a few % points off is no big deal. If you can get +/- 3-4% you are golden.

yup same thing happened to me a couple weeks ago... i tried and couldn't even get the whole range to come in on a NA car

You will never get the open loop AFR to match the targets in the table. To do that you need to get all 3 injectors perfect, you need to get the maf perfect, the engine VE perfect, the fuel VE table perfect, and then deal with the flow disturbances caused by the various ports opening and closing.


this brought up another table I'm interested in learning more about:

the VE table. I was guessing that it should be fairly important as load is calculated based on how much air the engine can flow.

it should have almost as much effect on load as the MAF reading?

Funny, I am looking at a set of data and from it I can conclude three things:

1) My MAF is decently calibrated. Not complete, but decent.

2) Contrary to stated opinion on another thread, you CAN fit a curve to the Airflow vs MAF volts data. You just have to have the right model/equation. I will give you one hint: Belehradek.

3)You can get your measured AFR close to your expected AFR in OL. There is error, but it is close. With limited data, but in a well calibrated region of flow, at steady state, the average percent error is less than 10% and not statistically different at a p=0.05. Certainly this may change under dynamic conditions, but it is supposed to.

this brought up another table I'm interested in learning more about:

the VE table. I was guessing that it should be fairly important as load is calculated based on how much air the engine can flow.

it should have almost as much effect on load as the MAF reading?

Make a new thread, and leave this thread for calibration discussions.

Srsly. I would love to read folks opinion/results on this table, but it deserves its own thread in the engine management section.

3)You can get your measured AFR close to your expected AFR in OL. There is error, but it is close. With limited data, but in a well calibrated region of flow, at steady state, the average percent error is less than 10% and not statistically different at a p=0.05. Certainly this may change under dynamic conditions, but it is supposed to.

would like to see what you have to say about this once you are tuned and have done runs at various throttle positions/loads .

Make a new thread, and leave this thread for calibration discussions.

Srsly. I would love to read folks opinion/results on this table, but it deserves its own thread in the engine management section.

yeah it possibly deserves it's own thread however the calibration procedures described here aren't strictly MAF calibration they are system calibration procedures. it is load you are fudging via the MAF. if the VE table is involved in the load calculation then it is probably part of the reason MAF calibration isn't always perfect.

hymee's explanation of VE and load_abs are a good reference. (http://www.rx8club.com/showpost.php?p=2901148&postcount=5378) I know when the VE tables are expanded for FI they are flat and possibly need tuning for accurate loads?

would like to see what you have to say about this once you are tuned and have done runs at various throttle positions/loads .

Under dynamic conditions (RPM changing with time, throttle changing with time, ) all bets are off. Read exactly what I wrote. It is an hypothesis, currently, and I will try to disprove it as best I can, but I would bet a six-pack of good beer that in OL at steady state (no change in speed, throttle) measured AFR will match expected.

Yes, the VE tables need to be adjusted for boost. I don't know the best way, so I'll just say play with them like me.

There is more to it then that though. The amount of fuel the pcm delivers in open loop is also corrected based on the VE tables, load limiting tables, and more... its not just the maf being off

agreed

Yes, the VE tables need to be adjusted for boost. I don't know the best way, so I'll just say play with them like me.

What exactly did you do with the VE tables?

I would only agree with this statement if you are running n/a with stock injectors and stock exhaust, etc. With boost and all your other mods, the VE of the motor is radically different. I believe that the Evo method of tuning would work fine since your "target" lambda under OL fueling is now meaningless with forced induction. If I am correct then you guys can skip high load/rpm MAF scaling headaches. Just compensate in your OL fuel maps.

This is also correct, IMHO. I have found that playing with the VE for the boosted ranges are more productive than scaling the MAF.

However, since the our stock scaling only goes to 365 g/s, anyone running over 9lbs. of boost has to scale theirs to account for the extra air.

-Yambo

What exactly did you do with the VE tables?

I changed the numbers at the bottom of the y-axis(using ATR) to reflect my new maximum loads under boost. Then I did a few-hundred data logs, slightly changing the VE in boost across the entire rpm range untill I started seeing my target AFRs. I actually came pretty damn close, and I never touched the MAF scaling.

This is not to say it will not work, just that I did this because when I tried only scaling the MAF, the only way I could get my target AFRs was to create unrealistic numbers(400+ g/s w/only 8lbs of boost) on the upper end of the scale.

I still plan on going back to the MAF now that I am comfortable with the VE table.

-Yambo

This is also correct, IMHO. I have found that playing with the VE for the boosted ranges are more productive than scaling the MAF.

However, since the our stock scaling only goes to 365 g/s, anyone running over 9lbs. of boost has to scale theirs to account for the extra air.

-Yambo

Please dont take this the wrong way, but when you just state a fact with no presented evidence to back it up that is completely contrary to the presented data posted, I cant do anything with your post than ignore it.

Post up exactly what you did so that you can at least put some value to your words. Again, dont take this the wrong way. I am looking at data right now, and will post it within the next few days that completely contridicts stated "fact" in a lot of posts, not just in this thread. Specifically, that expected AFR predicts measured AFR with a correlation of 0.9 or better at steady state, that you CAN use the proper equation (an equation built on first principles) to fit your adjusted flow to maf volts and use the parameters to build your new maf calibration (a fact that leads me to state that MAF calibration can be automated in Excel,) and finally, that the GReddy MAF housing CAN be with the proper physical set up (though I agree that it is not the best thing to use.)

No worries. I am not trying to convince anyone that I am right. Just throwing my "limited" experience out there. I have created well over 50 maps for my FI RX8 at this point using the ATR software. My posts are only meant to assist those using said software, since I do not have any experience using other tuning devices/means.

I guess I should have been a little more precise, knowing members of this forum, but I'll try to correct that now. When I stated that using VE seemed to be more productive, what I meant was tuning with VE was the ONLY way I could get consistent AFRs when boosting across then entire rpm range. I don't have any maps, charts, data logs, etc to show this. So sorry.

None of this was meant to say scaling the MAF isn't necessary. I freely admit that I tried several times and never got the results I wanted. I am reading this thread to hopefully learn something about it.

Good luck with your MAF scaling spreadsheet. If you want someone to try it to prove it actually works, I'll be more than willing to.

I love you guys.... but good grief your making this a lot harder that it needs to be.

You guys are jumping the gun - stop messing with stuff. Go one step at a time.... if you are getting turbulence then that is an intake issue - but barring that, you can scale the MAF - it is just tedious.

Keep in mind it is slow... so you have to get it at one consistant voltage for 20 seconds or more to get the AFR's to stabilize. This is hard to do at higher loads without a dyno.

Just wanted to update with some research that i have done while my car has been down (changing out injectors.)

I did some looking around to try and derive a function that could describe the voltage needed to maintain a temperature in a wire when air is passing

it. I did enough to understand the problem, but as you can imagine, this has been studied. In fact, there are patented systems that can

automatically calibrate a MAF when another system determines that the current calibration is out of spec. These systems are built around the

parameterized power equation

http://www.rx8club.com/attachment.php?attachmentid=137001&d=1239851539

where y is air flow, x is voltage and a, b and c are non-linear parameters controling the shape of the graph. The fitting could be any non-linear

iterative process. The output of the parameters are then be used to draw a new calibration.

In fact, I used this process to draw my own MAF calibration. I first generated Mass air, MAF volts and measured AFR and calculated the adjusted Mass air vs. MAF volts curve through the process above. A curve fit through this adjusted mass air data is the new calibration. I have attached a graph of my data showing my previous calibration curve in green, which is the stock calibration adjusted downward by 14%, my data that I used for the fit in open black points and my fit of the parameterized equation above in red. I have also included data in red open circles, which are not included in the analysis. THese points are in the graph representing higher flows but were collected when the AFRs were below 11.1, the limit of the stock wbo2. They are slightly higher than the calibration as I used a higher measured AFR than the actual AFR (which was in the mid 10s)

http://www.rx8club.com/attachment.php?attachmentid=137002&d=1239851539

I only got a limited set of data with this calibration before I figured out I had installed injectors that could potentially damage my pcm. I will likely install with the AEM intake so I will not get full data on the greddy intake. Nevertheless, I did find that my expected AFR predicted my measured AFR with an R-value of 0.82 (essentialy saying that 82% of the variability of the measured AFR can be attributed to changing expected AFR.)

I passed out.... LOL

Just wanted to update with some research that i have done while my car has been down (changing out injectors.)

I did some looking around to try and derive a function that could describe the voltage needed to maintain a temperature in a wire when air is passing

it. I did enough to understand the problem, but as you can imagine, this has been studied. In fact, there are patented systems that can

automatically calibrate a MAF when another system determines that the current calibration is out of spec. These systems are built around the

parameterized power equation

http://www.rx8club.com/attachment.php?attachmentid=137001&d=1239851539

where y is air flow, x is voltage and a, b and c are non-linear parameters controling the shape of the graph. The fitting could be any non-linear

iterative process. The output of the parameters are then be used to draw a new calibration.

In fact, I used this process to draw my own MAF calibration. I first generated Mass air, MAF volts and measured AFR and calculated the adjusted Mass air vs. MAF volts curve through the process above. A curve fit through this adjusted mass air data is the new calibration. I have attached a graph of my data showing my previous calibration curve in green, which is the stock calibration adjusted downward by 14%, my data that I used for the fit in open black points and my fit of the parameterized equation above in red. I have also included data in red open circles, which are not included in the analysis. THese points are in the graph representing higher flows but were collected when the AFRs were below 11.1, the limit of the stock wbo2. They are slightly higher than the calibration as I used a higher measured AFR than the actual AFR (which was in the mid 10s)

http://www.rx8club.com/attachment.php?attachmentid=137002&d=1239851539

I only got a limited set of data with this calibration before I figured out I had installed injectors that could potentially damage my pcm. I will likely install with the AEM intake so I will not get full data on the greddy intake. Nevertheless, I did find that my expected AFR predicted my measured AFR with an R-value of 0.82 (essentialy saying that 82% of the variability of the measured AFR can be attributed to changing expected AFR.)
Spoken like a true scientist. Great analysis carbonrx8!!!

Holy crap. That was a tough read. Care to break that down barney style? The part I am most interested in is the device/software that can calibrate your MAF for you.

-Yambo

I only got a limited set of data with this calibration before I figured out I had installed injectors that could potentially damage my pcm. I will likely install with the AEM intake so I will not get full data on the greddy intake. Nevertheless, I did find that my expected AFR predicted my measured AFR with an R-value of 0.82 (essentialy saying that 82% of the variability of the measured AFR can be attributed to changing expected AFR.)

great work!

there is no denying that you can (can't :banghead: what was I thinking) scale the MAF based on AFR however it is interesting where the relationship between AFR and MAF flow breaks down. I will hazard a guess that the calibration at high flows falls over due to a map that has been extended for high loads hasn't been tuned. possibly the VE map? these maps aren't tuned passed 125%-130% load in most cases.

http://members.iinet.net.au/~landslide/Fitted%20Curves.jpg

Keep in mind it is slow... so you have to get it at one consistant voltage for 20 seconds or more to get the AFR's to stabilize. This is hard to do at higher loads without a dyno.

This point can not be stressed enough as my research found a varying amount of lag in the AFR reading :eyetwitch

there is no denying that you can't [?] scale the MAF based on AFR however it is interesting where the relationship between AFR and MAF flow breaks down. I will hazard a guess that the calibration at high flows falls over due to a map that has been extended for high loads hasn't been tuned. possibly the VE map? these maps aren't tuned passed 125%-130% load in most cases.

Not sure what you mean because of the double negative in your first sentence. In my case, I didnt use the higher flows because I had the AFRs set to very low AFR and recorded very low AFR; lower than what he stock wbo2 can record. Were I to have used them, I would be setting myself up for failure. I dont have my innovate logging to the ATR, which would have solved the issue. I agree with you if you are saying that the calibration fails at high flow because there is no calibration. It is difficult to calibrate it at high load (+140%) as it is hard to get there in steady-state (at semi-legal speeds). However, if you do attempt to approximate steady-state at very high loads, I think that the calibration method will provide good results. This would either take a variable-load dyno or VERY high speeds, somewhere in the neighborhood of 120mph.

What I think you stated is my guess as well. That there are tables that increasingly interact with the OL tables at higher loads. Somehow, these tables do not interact at low acceleration, low tip-in, or low load. The problem is that in the ATR there is no explanation as to how these tables interact with each other. It is reasonable to wonder if these are availible in the pro version as we know that there are some programing restrictions in ATR. I am not complaining as the ATR is free. I just wonder what we are missing. These interactions can be figured out tho. It will just take many tanks of gas at jail-time speeds.


This point can not be stressed enough as my research found a varying amount of lag in the AFR reading :eyetwitch Exactly. WOT runs prolly should not be used to scale the MAF. (Tuning is a different question) Very slow, steady increases in rpm are ok, something less than 10rpm/sec. Anything faster and the wbo2 cant keep up sufficiently for accurate data. Plus the Throttle fuel gear table and, I believe, the Fuel VE table begin to add fuel over the OL maps. All this will fuck with your relationship between your measured and expected AFR and screw up the cal.

Holy crap. That was a tough read. Care to break that down barney style? The part I am most interested in is the device/software that can calibrate your MAF for you.

-Yambo

Glad to help if I can. Just break it down. Start with one part and go from there. I am sure that if you didn't get some aspect it was my failure to explain it clearly. I am also sure that if you are wondering about something, others are as well.

As for the software, I am assuming that you want to do this yourself? Do you have MS excel? That is what I and everyone else based most of this on. I use OriginPro for all my other higher-end analysis but it is kinda expencive just for playing around. You can download addons for Excel to do most stuff and Excel can be macroed to run Levenburg-Marquart nl fitting. The only problem is that you cant really trust Excel to spit out "true" numbers. It has to ALWAYS be checked as there are old, inherent errors in the excel algorithms that MS has famously ignored for more than 13 years.

I should note that you dont HAVE to do this. It was just fun to do. You can make your calibration by hand as easy or easier than curve fitting. You just need to calculate "adjusted flow." Did you understand that part of the above?

I think Morks was hinting I need to get off my butt and build a piece of software to add to Baseline.

This point can not be stressed enough as my research found a varying amount of lag in the AFR reading :eyetwitch

That lag could be OBDII port reading rates. Too bad you can't interface the Innovate LC-1 into the AP (or vice versa).

You can. But I cant get my ATR to talk to my car through the dongle. It might be a computer issue and not an ATR issue. BUt in the ATR there are communication options for what looks to be popular wbo2.

Cool... so I got my logs taken in open loop (~4250RPM) with cobb's stock map. I took a few MAF voltage/flow readings the with good sets of data for averaging. I am using an accessport to change the calibration here, so I wanted to double check this is OK before flashing.

My lambda value for 2.11V is 0.07% off. I did another one for 2.40V, and that was also 0.07%. Next around 7800RPM, 0.09% at 3.07V. Can I safely scale the entire MAF by 0.07%? or am I close enough as is and shouldn't bother with it. Is this premature without more data?

You can. But I cant get my ATR to talk to my car through the dongle. It might be a computer issue and not an ATR issue. BUt in the ATR there are communication options for what looks to be popular wbo2.

So the Accessport has a 0-5V analog input for aftermarket wbo2?

So the Accessport has a 0-5V analog input for aftermarket wbo2?

i believe if you run the ATR software on a laptop and plug an WBO2 sensor into your laptop, Cobb has some drivers available so you can access the WBO2 data from the ATR software :eyetwitch

Cool... so I got my logs taken in open loop (~4250RPM) with cobb's stock map. I took a few MAF voltage/flow readings the with good sets of data for averaging. I am using an accessport to change the calibration here, so I wanted to double check this is OK before flashing.

My lambda value for 2.11V is 0.07% off. I did another one for 2.40V, and that was also 0.07%. Next around 7800RPM, 0.09% at 3.07V. Can I safely scale the entire MAF by 0.07%? or am I close enough as is and shouldn't bother with it. Is this premature without more data?

Honestly at .07% I wouldn't even worry about it - that means you multiply your voltage by .0007... not a very big change - BUT if you want to do it just to see if you can get closer - have at it!

Honestly at .07% I wouldn't even worry about it - that means you multiply your voltage by .0007... not a very big change - BUT if you want to do it just to see if you can get closer - have at it!

Thanks Kane. I decided I'm not going to touch it...yet. Started the procedure for MM's tuning today, so going to follow through with that. But once that's through I'll be having at it :D and perhaps by then Baseline with MAF will be out. :) It will be interesting to compare. This is all new to me...

Honestly at .07% I wouldn't even worry about it - that means you multiply your voltage by .0007... not a very big change - BUT if you want to do it just to see if you can get closer - have at it!


I am wondering if he means .07% or 7%. There is no way that you would fretting about 7 parts in ten thousand. Seven in one hundred is a different issue.

PLUS if you are 7% too high is a completely different animal than if you are 7% too low. One of them could put a hurt on you where the other would "only" make you run too rich.

I am wondering if he means .07% or 7%. There is no way that you would fretting about 7 parts in ten thousand. Seven in one hundred is a different issue.

PLUS if you are 7% too high is a completely different animal than if you are 7% too low. One of them could put a hurt on you where the other would "only" make you run too rich.

Well, it seemed odd to me it was that close to the curve, 0.07% off...
Strikes me as odd because I have an aftermarket intake. :)

Well, it seemed odd to me it was that close to the curve, 0.07% off...
Strikes me as odd because I have an aftermarket intake. :)

Not really - small engineering tolerances could account for it.

This is serious info here.. thanks Kane...

Anytime man.

Ok. I finally got my car running again. I took the downtime as opportunity to do some more editing. So the new ride has a "custom" front crash-bar, K40 and Blinder (using the unused fog light switch as the on/off switch; way cool), new a-pillar gauge pods, LMA3, fuel pressure gauge, Mazdaspeed intake modified to fit the greddy tubing and 650cc/min in bank two.

I also rescaled my injectors in the AP (a guess) and edited the injector latency (based on injector-rehab info). Then I started collecting data to scale my maf.

The issue is that, for some UNKNOWN REASON, while in 3/4 gear, at around a load of 45, at RPMs around 4500, the AFRs will either run at the appropriate AFR or they will drop to super-rich.

Let me restate this. While at steady state in OL at load of 45 and RPM of 4500 where the expected AFRs are just south of stoich, the measured AFRs can be one of two values, either close to expected or WAY too rich.

For the life of me, I cannot figure what tables are interacting to cause this and I need to figure it out prior to scaling the maf completely. I have only two hypotheses. One that it is an injector issue, that one pair of my bank two's are flowing more than the others and they are randomly turning on and adding too much fuel. The other is related to the observation that it seems that if the RPMS are trending upward, then I see the richness. Are there hidden tables that measure rate of RPM change? My accelerator pump tables are set to 1 in the range that I am having the issue.

THoughtful responces are appretiated.

Interesting... did you try to scale your injector bank 2&3 latency and see how that affects it?

It seems like the latency may be screwing with you - - as soon as it hits some magic load area you have more latency calculated than you need.

I wouldnt think so. The lateny table has baro and voltage as independents and microseconds as dependents. At least in that table, latency is not a function of load or RPM.

Here is my best 2d display of the data. I will note again that this is OL (LTFT=STFT=0). I have also filtered this data farther to insure that I was in 3rd or 4th gear (using speed/rpm), to insure I was above 3000rpm and to limit loads seen on the graph. The x-axis is a very narrow representation of availible load. This bifurication is the ONLY "split" that can be seen from low load to well over 1. I am not having any current issues where my measured AFRs are significantly different from commanded (tho I will admit that something might turn up.)

http://www.rx8club.com/attachment.php?attachmentid=141169&d=1246064257

You can see two populations of AFR at nearly identical load and RPM. It shouldnt be this way unless there is another interaction. I ran several tests for interaction across many variables, including rpm delta and saw very little to suggest an issue.

My current hypothesis, as I said, is that one pair of my bank two (P2 and Sec) injectors are not scaled appropriately and that right around this load the injectors occasionally kick-in. But beside this, I am at a loss. Will need to do some more logging.

In the interests of full disclosure, the graph does not represent the dependence on time. I will say that, over time, if the AFRs are high, they will remain high and if they are low, they will remain low. The data does NOT jump between those populations randomly.

I love that analysis. That is really good stuff.

Cheers,
Hymee.

i'm guessing it's something at the quantum level.

get it? since your values are both rich and lean?

yuk yuk.

but seriously, that's about when your vfad is supposed to open, isn't it? maybe something is weird with that. or maybe the coolant temp has something to do with it. I really have no idea though.

i'm guessing it's something at the quantum level.

get it? since your values are both rich and lean?

Actually, you might be more correct than you think. The effects observed at very small scales such that things can appear to move between states without having any values between is caused by the probabilities that follow wavelike or harmonic behavours.

Given that I just put in a new intake (Mazdaspeed) that is rigged very differently than it was designed did suggest to me that perhaps a standing wave is being setup somewhere ahead of the MAF sensor and that this wave randomly "breaks" at different flows. The MAF might then see appropriately bifuricated flows in these ranges of load or some other behavour that could be a observed. I have not really looked into this as it is unlikely given that while the intake is setup quite different than it was intended, the flow is very straight for nearly two feet ahead of the MAF sensor and there is a screen in the flow. If I cant pin something else down, I might start sifting through the data for this.

As for a valve opening, it is possible, but still more likely that it is "user error" in the ATR and can be "tuned out."

Thanks for the thoughts.

you said LTFT = 0 and STFT = 0 as well . OR did you mean LTFT + STFT = 0 ?

you said LTFT = 0 and STFT = 0 as well . OR did you mean LTFT + STFT = 0 ?

Well, I used to just use LTFT-STFT=0 as my filter, but have started using LTFT=0=STFT as this weeds out some deceleration and other datapoints.

I should note that these data were collected by direct connection to the computer. When you datalog via AP, the zero value is 0.16 or something like that. LTFT=0=STFT is never true when logging via AP, IIRC.

Carbon, do you know when the load transition is between the 2 injector banks? I suspect you are correct about the AFR change resulting from an injector issue. I really doubt the intake will cause this.

Yeah me too - I am voting injector issue.

Interestingly, this issue did NOT turn out to be related to the injector scaling; however, I have learned somewhat about the effect on AFR as a function of the injector scaling and believe my bank two injectors are not scaled perfectly.

Honestly, I am not completely sure EXACTLY what caused the above issue, but an inappropriately scaled MAF calibration and a poorly designed fuel map were contributers.

Just to sum up: I got my car running with new injectors and a MS intake. I recalibrated the MAF to stock assuming that because the MS intake is the same size as stock (or close) and as it is installed by the dealer without a reflash, that a new calibration is unnecessary.

I did that and the car was fast! The issue was that I was not hitting my target AFRs and wanted to start a little rich. I logged data, analyzed as above and recalibrated (keeping the stock calibration from 2-20g/sec airflow.) This got me much richer, but with the above result and lots of bog.

I concommitantly had some idle issues with LTFT in the +15 range. I shrank my primaries, and resolved that issue. This had essentally no effect on the issue at cruise.

I then played with the scaling of my secondaries. While I could get richer or leaner with this alteration, the AFRs were still out of line. Some where along here I noticed that my OL fuel maps were very ugly and jumpy. I dont know where this came from. Perhaps an old map? I re-adjusted a map that I believe is a work of art. ;)

At this point I realized that my MAF calibration was just silly. I dont know how I came to this realization, but I reverted to stock MAF calibration AGAIN with results that were promising but not great. I datalogged again, re calibrated by fitting then datalogged AGAIN, recalibrated with fitting and with fine tuning by hand and will datalog again today to see the results.

Again, I am not absolutely sure of what I changed besides the primary scaling (secondaries are back to my first estimate) and some fine tuning of the MAF calibration.

The interesting thing is that with my greddy intake, I obviously had greater turbulance as I observed much greater "noise" in my datalogged air flows, but using a non-linear fit to produce the maf curve was sufficient for a good tune, but now, with the MS intake, the noise is much less and I observe much finer "ripples" in the maf calibration curve.

I am guessing, and will experiment later, that these ripples are a result of mis-scaled bank two injectors. I say this because the ripples occur right around where I have observed the changes in AFR with secondary injectors scaling. The ripples look like what I would think would happen as the different injectors come on line.

The good thing is that if this is true, this will provide the final "equation" so that there is a way to scale your secondary and P2 injectors correctly separate from the MAF calibration (basically, if you have 3 varibles to solve, you need 3 equations. If you have 3 varibles and 2 equations, you can only solve to a function of at least 2 of the varibles.) Essentally, if the MAF calibration is smooth at the injector transitions, then your injectors are scaled correctly. Again, this is my hypothesis.

Be that as it may, it might not matter. It seems that a wobbly maf curve makes up for deficiencies in the injector scaling. Maybe.

P.S. All this above also means that if you change your injector scaling, you MUST recalibrate the MAF immediately following. Period.

Scaling your injectors requires two flashes. One to scale, (datalog) one to recali the MAF.

Any changes to the intake will throw off MAF scaling. Maybe not enough for Mazda to require a reflash for the MS intake on a stock RX8. But with forced induction and greater air flows, the error will be amplified. We discussed this before but ideally you want to tune the MAF with stock injectors, then add the bigger injectors and tune them. Changing injector scaling and MAF scaling concurrently will cause you to chase your tail since both variables affect AFRs. Can you calibrate the MAF with airflow data alone (no AFR)?

Yeah, your point also came up in Jeff's tuning webinars. While I could have swapped back in the stock yellow injectors, I was too lazy. But you are right, doing both maf and injectors on the same flash is a complete mistake, unless you are just throwing a guess out there to start off.

The MAF could be calibrated on a flow bench of sorts, but I dont have an independent airflow meter, nor can think of a way to make one.

You can tune using both MAF calibration and injector scaling but it's knowing when to use them and when not to. The MAF can provide a fine tuning aspect and allows you to change air/fuel within certain airflow regions.

The injectors are a much broader correction. If you have a situation where you're lean across a wide area you can scale your injectors to add more fuel.

They key is knowing which tool in your belt to employ otherwise, as you said, you'll be running in circles.

Good points all - that is one of the reason I do 2D and 3D tuning.

If your 2D is showing consistant errors in a certain AF range - then it is safe to make MAF changes - if your tune just has some small holes in it; injectors might do it; or the base fuel map.

The issue that I posted a while back concerning two "levels" of AFR at one load is resolved. Again I am not completely sure what specific adjustment fixed it, but this is my new AFR vs Load graph:

http://www.rx8club.com/attachment.php?attachmentid=141669&d=1246730580

You can see that with increasing load, I have decreasing AFR, all in reasonable amounts. This plot helped me visualize not only if I was hitting my target AFRs (which Kane's software analyzes) but also helps to make sure that there are no non-normal (non-gaussian) distributions at any particular load.

You can also see that with increasing RPM I have steeper increases in AFR. This was intended.


Now the current issue that I am having is with the scale of my injectors. I installed a new bank of injectors (Siemans Deka; 650 cc/min) and just guessed (with the percent difference function) what the new scalar should be in the ATR software. Having played with this number, I am confident that increasing this number will lead to leaner AFRs and that decreasing this number gives richer AFRs. (I.e., if the computer thinks that the injector is smaller than it really is, the computer will increase the dc and add too much fuel and vice versa.)

Jeff (MM) has posted a rough outline of how the injectors come on line.

http://www.mazdamaniac.com/images/rx8/electronics/injector%20staging.jpg

I dont really know what the axes are but the general idea is obvious.

The question that has arisen several times, both here and during Jeff's webinar is given the influence of both the MAF calibration and the injector scaling on AFRs, how do you scale either independent of the other.

The answer has been that you can not. One should scale the bigger injectors on a stock intake then install and calibrate the aftermarket intake (or, if the stock injectors are perfectly scaled, vice versa.)

Well, I am just not going to go backward and do that. I reasoned that any mis-scaling of the injectors would show up in the MAF calibration curve as "bumps" or systematic excursions from a theoretical calibration curve. For instance, if the calibration curve is smoothly increasing then abruptly alters its shape, what this really means is not that there is a sudden change in airflow with a small increase in volts, but that at this particular TRUE airflow one of the injectors is mis-scaled and that the computer must be LIED TO telling it that more or less airflow is availible so that it increases or decreases the fuel availible to compensate for the error in the injector scaling.

Using the method that I decribed above, I iteratively scaled my MAF calibration until my measured AFRs were close to my target AFRs. What I noticed is that as the voltage increased, the air flows abruptly jumped at two distinct points. Below is my plot of adjusted airflows vs MAF volts with the MAF calibration curve. Note that I adjusted the MAF calibration curve (red points) several times prior to this graph. This graph is not showing ANY adjustment to the MAF curve but is showing how close my adjusted curve is to the curve that is in the ATR software.

http://www.rx8club.com/attachment.php?attachmentid=141670&d=1246730580

I hypothesize that my injectors are scaled too large. Once the bank two injectors start to kick in the pcm has to be lied to telling it that there is MORE air coming in than there actually is. The computer responds by increasing the dc of the injectors to meet the need of the increased airflow. What I think that I need to do is decrease the scalar of both the sec and p2s and then recalibrate the MAF. The question is by how much?

I should mention the method that I used to scale the primaries. I simply kept the stock MAF calibration from 2-20 g/s (idle airflow is 5 g/s) and scaled the primaries until my LTFT at idle was 0.

Look from the 2D maf to a 3D Target vs Actual AFR.... with enough data you should be able to visualized the load and RPM range of your tuned holes - and scale the injectors appropriately.

I've been round in circles a few time on all this - playing with maf scale , AFR table and injector scaling . I came to the conclusion that there is another table that we need to consider as well because no combination of things gave me the right numbers for all load situations .
MM has indicated as much but I don't know which table to alter .
I currently have WOT 2-4% off the actual vs target in one direction and the part throttle target vs actual 2-4% off in the other direction . perfect Maf scaling or perfect injector scaling is not going to fix it from what i've seen .
For now I'm pretty happy as i have tuned the load table to "trick" the pcm into giving me the results i want .

Don't quit man.

The combination of VE%, MAF and Injectors should get you the target AFRs you seek.

An example of this is Chase - his total % deviation is just over 1% across his whole tune.... now he has factory injectors and is not FI-ed.... but if he can do it; then you all should too.

Just need more data!!!!

What does the ve% table look like ?

Trying to load a pic now...

That looks like the one Hymee calls "ETC forward loop gain" ?

my numbers are way different to that map though - do you think i could try loading those numbers in ?

Yep they are the same.


I set all mine to 1 (no modification to fuel table) - scaled my MAF; looked for holes due to injector issues; scaled my new injectors; then looked for any additional holes in my tune - and used the VE% table / forward loop gain.

yeah - Hymee has most of that map above 110% load set at 1 . Could be the issue i'm having ?

Next question
how do you adjust the VE table ? :)

yeah - Hymee has most of that map above 110% load set at 1 . Could be the issue i'm having ?

Next question
how do you adjust the VE table ? :)

What? I just told you dude...

Eliminate all variables - tune one at a time; start with 2D tables - MAF and Injectors... then work the 3D tables - VE first and base fuel last - if you just cannot get one area to line up nice.

But you cannot pull one log and assume you can make changes - you need to get as much data as you can.

What I meant was :

if you change a figure on the ve table that was 1 to say 0.98 what are you actually doing -
Reducing fuel in that box by 2% I presume ?

Yeah - it is a modifier on the Load calcs due to air velocity / resonance I think....

also gives a base fuel modifier before the MAF can read airflow changes.

I should mention the method that I used to scale the primaries. I simply kept the stock MAF calibration from 2-20 g/s (idle airflow is 5 g/s) and scaled the primaries until my LTFT at idle was 0.

that does not sound like a good method as it does not calibrate your maf tube ..... Then again if you changed your primaries at the same time - i see your dilema .

That part of the MAF calibration will have to wait, but will eventually get done when I get around to working in OL in that range. For now it works as I dont get flows while moving below about 20g/s.

Kane, when you say fixing "holes" in the tune, what specifically are you describing? Are you agreeing that the "bumps" that I am observing in this MAF calibration are due to improperly scaled injectors?

As for the Fuel VE table, I cant imagine how that table is helpful at all. That is not to say that it isnt helpful, just that how having one table that describes fueling (OL fuel) and another that alters the first table, is helpful. Perhaps I am not understanding the concept of the VE table?

Also, I searched and did not see any post describing "ETC forward loop gain." Can you elaborate or point me in the right direction?

Can you log airflow? if so, maybe someone with a stock intake can log their airflow (or go to the Mazda dealer and "test drive" a new RX8). Plot the airflow against the stock MAF calibration. Adjust your calibration according. That would be the right way to datalog the MAF.

Can you log airflow? if so, maybe someone with a stock intake can log their airflow (or go to the Mazda dealer and "test drive" a new RX8). Plot the airflow against the stock MAF calibration. Adjust your calibration according. That would be the right way to datalog the MAF.

Sure you can log your airflow. But you are just logging what the MAF calibration curve says. The ECU only sees the voltage of the MAF. From that, it looks at the MAF calibration and deduces what the airflow is. The only way you can do it is to use a know calibrated mass flow meter and compare to the sample. You could run 2 x MAFS, one connected to the ECU, and one just logging voltage independantly, but it would all be relative to the accuracy of the first MAF.

Cheers,
Hymee.

Kane, when you say fixing "holes" in the tune, what specifically are you describing? Are you agreeing that the "bumps" that I am observing in this MAF calibration are due to improperly scaled injectors?

I would not venture to guess what the bumps are at this time... I have seen some interesting MAF curves - it really is not linear. However you have to look at these things in both a 2D and a 3D kinda way - the "holes" in your MAF scale will always occur at the same airflow regardless of load and RPM (a 2D problem); however injector issues will manifest as "holes" that group around a small load and RPM area (3D) I find the the most helpful way to see what I should be tuning. Again; the key to find them is mass data collection - I usually use tens of thousands of data points in a single MAF scale.

As for the Fuel VE table, I cant imagine how that table is helpful at all. That is not to say that it isnt helpful, just that how having one table that describes fueling (OL fuel) and another that alters the first table, is helpful. Perhaps I am not understanding the concept of the VE table?

The VE% table is the instant look up for the PCM - gives a base fuel amount modifier to help get the fueling just right. Look at the OEM settings and you will see this is Fine Fine Fine tuning area. So if you wanna eliminate the variable - set em all to 1.

Also, I searched and did not see any post describing "ETC forward loop gain." Can you elaborate or point me in the right direction?
That is the protuner version of the VE% table - they are one and the same.


I am a wee bit tired and dehydrated... so I may have typed something wrong... but from memory this is the best I got..... LOLZ

Doesn't make sense for the maf curve to not be linear . I have resisted the temptation for making it into spaghetti so far but I just may have to .

i'm not so sure it doesn't make sense, brettus. could be that the air flow levels off between certain loads and then jumps back up at higher loads, so the air flow increases relative to load could be slowing in places and picking back up in others.

I would only worry if air flow in your scale decreases with load or if it's really rollercoaster-like. I think either of those would mean you need more data before scaling your maf.

but at the same time, i've floated the idea elsewhere that possibly a power (or maybe polynomial) trend line of such a curvy line may be the right thing to do to smooth it out.

Here's a scale I made with baseline using a pretty small amount of data (with some junk data in there). THe red linear solid curve is the original scale. and the blue curvy one is the new scale. The two dashed linear curves are different trend lines, one polynomial and one power. Very interesting, I think

http://i27.photobucket.com/albums/c157/myriads/newscaletrends.jpg

I am still hoping that someone will carefully look at my second graph. Is my rationale at least reasonable? I am holding off doing the experiment as I am hoping to see some potential variables that I am not taking into account. I want to highlight the fact that currently I have NO major issues with hitting my target AFRs. There are some things to clean up, but I think I am seeing data that is in the neighborhood of "good." (Average percent error is -1.48%; R=0.78)

http://www.rx8club.com/attachment.php?attachmentid=141732&d=1246847387

Brettus, from a modeling perspective (I have spent a great deal of time in this area) you are absolutely right if you are speaking of a theoretical model. A description of the variation in volts in a heated wire subjected to a moving fluid (with it's own attendant parameters) has been essentially completed and validated experimentally. So if we had a "clean" system, we too should observe a smoothly varying MAF calibration curve. The fact that we have systematic deviations in the physical behavior that is not described by the model means that there is something different in our system than the "clean" system.

So, really, from that perspective, it makes complete sense that we are seeing a wavy MAF calibration. It means that there are problems somewhere. But I think that the objective should be to minimize the wavyness particularly if the wavyness is due to errors in fueling.

From what Kane is describing and from what I have read on his website, I think his approach is exactly this. And I think his insistance on LOTS of data is great, tho, I dont know if his analysis is correct as I am not sure if his stats take into consideration non-normal distributions. But I am sure it is quality.

Kane, What I mean about non-normal is that to use things like standard deviation, you assume that your data is normal. However, when I plot a histogram of the percent error of my measured AFRs from my target, I see a distinctly non normal population. What I see is a histogram with large "wings."

http://www.rx8club.com/attachment.php?attachmentid=141733&d=1246847387

You can see that this histogram fits well to a lorenzian curve as opposed to a gaussian. Lorenz studied behavors that showed a great propencity for outliers, so much so that they could not really be called outliers but were inherent to the system. How this affects your analysis, I am not sure. In fact, as long as you are not seeing bimodal populations, I think running variance based analyses is fine, but maybe this should be thought through.

Brettus, as for your issue with being on one side of your target AFR in WOT and the other side at cruise, are you too rich in boost and too lean at cruise? Have you considered a small vac/boost leak?


As an aside, some folks insist that tuning is more art than science, I think that statements like this come from folks that aren't very scientific. I haven't experienced anything so far that is anything like art and dont expect to in the future.

Brettus, as for your issue with being on one side of your target AFR in WOT and the other side at cruise, are you too rich in boost and too lean at cruise? Have you considered a small vac/boost leak?
.


Just fixed most of this since Kane told me about the Ve chart . It is close enough to spot on now for me to be a happy chappy . :evil_laug

My maf curve looks fn sweet as well now as I've finally cracked all the Greddy wastegate/boost control issues . Not to mention that I'm certainly pushing well over 320whp now :)

Just fixed most of this since Kane told me about the Ve chart . It is close enough to spot on now for me to be a happy chappy . :evil_laug

My maf curve looks fn sweet as well now as I've finally cracked all the Greddy wastegate/boost control issues . Not to mention that I'm certainly pushing well over 320whp now :)

Could you elablorate? Not on your Super Secret Squirrel wastegate control, but on how the MAF curve changed from before to after you implemented the Secret WG control? And as for the VE table, you are saying that this moved your measured AFRs in either direction depending if you were in boost or not???:icon_tup: Cool, but I dont understand how it does this.

carbon, i doubt anyone has thought as hard about tuning as you. i appreciate your efforts.

interesting stuff especially with respect to the lorenz curve.

as to your second graph, I'll think more about it tonight, but best I can tell you're making sense. i would go ahead and run the experiment since I'm not sure what else there is to account for.

carbon, i doubt anyone has thought as hard about tuning as you.

That is very kind but absolutely untrue. I am a rank amateur at best. Thank you, though. Gave me warm fuzzys.

I dont know what you meant about "interesting" with regards to the lorenz curve. If you are following the tacit implications of what I was suggesting, we should talk.

I will try the modification to the injector scalars later this week and see what happens. Thanks for the encouragement!!

Carbon ,
What i've done is basically utilise "boost creep" by restricting how far the wastegate can open - it gives me progressively increasing boost , unlike the stock Greddy which is the exact opposite . The g/s curve is now a thing of beauty and it drives so much better at part throttle compared to when i was using a boost controller. I still use the secret squirell controller BTW except now it operates the boost controller from 6500 onwards so i can dial it up to the max in that range .
Boost controllers suck at controlling this turbo

The Ve table i had was probably very different to what comes with the AP (looking at what Kane posted) so it may not be an issue for you . The issue was always WOT vs part throttle - cruise was never an issue.


I basically adjusted the ve table to correct target vs actual in the areas i have been fighting with

Carbon,

I love your posts.

Cheers,
Hymee.

Sure you can log your airflow. But you are just logging what the MAF calibration curve says. The ECU only sees the voltage of the MAF. From that, it looks at the MAF calibration and deduces what the airflow is. The only way you can do it is to use a know calibrated mass flow meter and compare to the sample. You could run 2 x MAFS, one connected to the ECU, and one just logging voltage independantly, but it would all be relative to the accuracy of the first MAF.

Cheers,
Hymee.

You are correct. You would need to log the raw voltage data and compare it to a known calibrated MAF (ie stock). But if this could be done, then you would eliminate a tunign variable and could focus on other parameters.

One other thought. Carbon, as far as the injectors, are you only playing with injector scaling, latency, or both? I would assume that published latencies for the injectors are pretty much set in stone. You could use the same scaling:injector size ratio as stock for the bigger injectors. Again, I'm just trying to eliminate one of your tuning variables (injectors, MAF, VE). Injector scaling should cause global AFR changes. But it sounds like you are dealing with "bumps" in your tuning curves. So MAF and VE changes would be more appropriate.

I tune Evos and WRX/STis so please excuse me if I am making incorrect RX8 tuning assumptions. I'm friends with Carbon and I am just trying to help out.

Yeah, the latency tables are done. I based them on Injector-rehab.com data. To get any more info on those, I would have to send them out to someone.

The scaling is based on the ratio of the ATR number for the stock injector size to the larger injector size. That assumes that the scaling factor is linear and that all the flows are known accurately. My guess is that the expected flows vs the actual flows are different, and I get to guess what the actual flows actually are.

My bumpy maf curve does a good job at making the best of things and the car is running well, but I figure as I am going to start leaning things up and uping the boost soon, I should get the base tune as good as it is going to get.

BTW, I tested my vac at idle and at 2krpm and got around 18"Hg and 20"Hg, respectively. I will try to get my compression numbers this week so that I can watch them as I go to higher loads. Hoping to get 10-20k miles more out of this motor. I appretiate the input.

Hey Carbon;

Yes I had reviewed other than normal distributions.... however; as we approach N then we will almost always approach a normal distribution with almost any data set (law of large numbers). So I exclude outliers from the AFR logged areas not the MAF scale itself.

PS - Are you accounting for your sensor latency in your calculations? It makes quite the difference in transitional areas of the map.

Again MAF scale issues show up as deviation around a specific voltage - injector scale issues should be shoing up as a specific variance at load and RPM; regardless of MAF voltage. And anything else that cannot be resolved would go over to the VE table - IMO.


Your graphs are cool - much sexier than mine - LOLZ

Hey Carbon;

Yes I had reviewed other than normal distributions.... however; as we approach N then we will almost always approach a normal distribution with almost any data set (law of large numbers).

Not that I know (much) about what I am talking about, I have to disagree with this statement from a fundamental point. First, there are several distributions that bias to one side, including Poisson and power distributions. Both of these will not be normal at any population size. IIRC, both of these distributions can be normalized through various transformations and can be analyzed parametrically after a transformation.

The Lorentz or Cauchey distribution is another thing altogether. This population has as a fundamental characteristic of a propensity for members of the population that posess large differences from a median or mode. This distribution has very large wings, or bias to both sides of a central value. The deeper result of this is that a mean or average is not defined. THERE IS NO MEAN! You can calculate it, but it doesn't mean (no pun intended) what mean means (ok, I intended it that time.)

Now, how this affects us, if at all, may be minor. It also may be that what I found is anomalous or just plain poorly analyzed. In any case, I would think that a bimodal distribution would confuse your software far more than a Lorentz, but what concerns me is that there is a small possibility that an analysis that uses variance to reach a conclusion from such a data set may reach an erroneous one. A simple visual test would catch anything, of course.

Nevertheless, I find this very interesting and am looking into what it means, if anything.

Once again you are correct - however....

Given the noise in the sensors is random (MAF, O2, Injector PW etc); otherwise you would always hit the target AFR's right?

So the law of large numbers does apply to random values....

Either way - I think we are talking semantics unless you want to change my software distribution curves and test.... cause I am inside 1% on most cars... I'll live with it.

Don't get me wrong, I am not dinging you work in any way, shape or form. I think what you have created is an amazing bit of software. I am also not trying to change how you or anyone tunes. I am not even strongly suggesting that I even have anything useful to contribute, particularly with respect to this whole distribution/statistics thing.

I am just curious to understand what is contributing to the observed distribution, if it exists. So as you suggested, one possible sourse of error is the latency, or just goddam random error. All important is certainly in existance.

However, what I suspect is that the observed distribution, with a rich smattering of "outliers" is a result of interaction that betrays the existance of feedback or feedforward algorithms in openloop that we have yet to understand. And if it is not this, then some other physical interaction that might be interesting, again if the observation holds. If so, there might be something to learn and better statistical methods to apply.

I dont expect anyone to be interested or expect that any of this is critical. Hell, most people tune on a wing and a prayer. If it works, fine. I would suppose that such folk would suggest that I am putting too much thought into this whole thing. I suppose I am. Whatever.

So yeah. What you are doing works. I dont disagree a bit.

:Eyecrazy:

I wasn't knocking you at all Carbon - just going with my theories and findings.

Question for you - any chance you have enough data (+100K) records on the same tune that I could put into baseline and see if it is random or if there is a set trend one way or the other? I am down to try anything.

Most of the time i gather about 10k filtered data points per tune (over, perhaps, 3 drive cycles; I might start with 100k of CL and OL points but it usually gets whittled down after selecting OL only, above 3000rpm and load>0.2) before I deside to change something. So, not yet. After this next iteration, we can get together and check something out. I would be happy to collaborate.

Right now, I am supposed to be writing a grant application, so I am going to be a little slow.

Brettus, I appretiate the sentiment.

Let me know bud; I am down.

all very interesting and useful info. i have played around some with accesstuner but eventually when i go boosted ill drop the money on proper AP tuning

O2B1S1w - O2 Bank 1 Sensor 1 Lambda (wide)
EQ_RAT - Commanded Equivalence Ratio

Can I use them to scale my MAF? What about the O2 sensor latency?
Here is a sample log:

http://www.rx8club.com/attachment.php?attachmentid=162351&stc=1&d=1284696911

Lambda Can be used easily.

Latency that I have seen with flashers is somewhere areoung 500ms...but you should test it to be sure.

Commanded EQ is your target lambda and the O2B is your actual. I am not sure how much to trust the commanded eq to be acurate cause I've never used it.