NefMoto

Time to knock another perceivably insurmountable problem with out ECU down. One that says our cars cannot run on anything over Lambda=1.00 or AFR=14.7.

It is truth that ME 7.1 is coded in a way that it is impossible to go over 1.00 lambda. Several hard coded (no maps!) condition checks inside ECU program substitute higher than 1.0 Lambda with a value of "1", effectively making "1" the highest lambda ECU can run at.

The question is though: how does the ECU know what lambda it is running at? It cerainly doesn't just guess or anything like that so some sensor must be telling it the exhaust gas condition. The O2 sensor at reach bank tells it through its voltage output what is the current Lambda of exhaust gas is.

There are numerous aftermarket solution that will allow anyone to run their cars at any lambda they want without the ECU having a clue that it is being lied to. They include stand alone "between the ECU and O2 sensor" boxes, Zeitronix O2 wideband can also supply ECU with narrowband look-a-like signal at any real AFR you set it to stick to.

The result is that these boxes lie to ECU about the actual exhaust gas condition by substituting real O2 signals with their own generated ones. You can set these to report Lambda = 1 (or 0.455 volts) at any AFR you want, commonly being 15.4 which is considered the best economy AFR.

It is all nice but how about doing some shenanigans inside of ME 7.1 to accomplish similar effect? Read on.

Narrowband O2 sensor like S4/allroads/A4s (some) are using, is good only at measuring Air to Fuel ratio at Lambda = 1. They quickly lose resolution once stepped few hundreds either side of Lambda = 1. The following picture illustrates it very clearly:

(http://www.buckeyetriumphs.org/technical/Carbs/AFMonitor/FIG1.jpg)


Some measured voltages with delta (percentage change) from Lambda = 1:

(http://www.btinternet.com/~j_holland/images/o2_chart.gif)


The challenge is to find how the ECU is exactly figuring out what AFR it is running at. The way ECU works in closed loop is very simple. It reads the voltage, determines how much it is off from a sweet point towards lean or rich, adjusts injection time slightly, measures again, adjusts injection time, and  so on, and so on, it is a never ending cycle. The result is that ECU never really sits at precisely Lambda "1", it cruises around trying to eternally reach the goal.

It works since Lambda o2 sensor (narrowband) has great resolution around Lambda = 1. The perfect Lambda of "1" is universally considered to be reached when the O2 sensor voltage reads around 0.450 volts.

I spent some time today analyzing how our car figures out "the sweet spot" and sure enough: module "RL"  translated to "Lambda closed loop control" is responsible for taking the O2 voltage and through a lot of calculations and adjustments, output AFR/Lambda as well as apply corrections to injection time it considers needed at the time to move towards L=1.

Most importantly, there is just one variable that defines what the sweet spot is:

USR - "Control threshold for regulation before Kat"

Address (M-box):  0x11ADE
Data Type: Byte
Conversion formula: (1/192) * x - 0.200
Default:  7B or 0.4405


How do we use this?

Lowering or raising this value will force the ECU to adjust fueling to stick around this voltage value, Looking at the graph above, you want to lower this voltage to a level that will directly correspond to AFR you would like your closed loop to work at.

Due to completely non-linear nature of o2 output and very low resolution of USR, we don't have much room to maneuver if our goal is to reach perfect economy AFR of 15.4, 1.05 Lambda but it is not hopeless.

Here is your graph showing best power and best economy AFRs:

(http://www.buckeyetriumphs.org/technical/Carbs/AFMonitor/FIG2A.jpg)


In my case setting this to hex 34 (0.0708 volts) gave me AFR of ~15.0 in closed loop and setting to hex 33 (0.0656 volts) gave me  AFR around 15.2.

There is a catch though. Since the ECU is set to regulate around 0.45 volts with a lot of room voltage wise around that value with very little AFR/Lambda change, running 0.0656 volts yields a car that like to shoot into very lean territory when you are at very light load/idle. This is really no big deal since you don't really over heat engine at such load or get any knock or misfires.

Massaging PID controller for lambda correction will no doubt make it much smoother but I just wanted to warn any pioneers here that want to use it. The problem doesn't show in normal cruise.

ENjoy!

Just to verify, does this method alter what ECU sees is 14.7, or like you said in your post, does it re-define what ECU sees as sweet spot?

I think a better way to word this would be do you have to alter all fueling maps  to compensate, or just this 1 map?

I looked in the FR, and I'm thinking it just re-defines the "sweet spot" as you say.

It re-defines what ECU hovers around but still reporting 14.7:1 mix.

There are implications when using this. If you use this method to raise your real AFRs, you will no doubt encounter pulled fuel on your LTFTs as your current fueling will use too much fuel.

This will no doubt alter high range frao_w / frao2_w which will directly alter WOT fueling.

The simple solution is to just pull some amount % wise from your KRKTE as your idle trims show or even better just use wideband and see what your car runs at.That value divided by 14.7 will give you an idea how much less fuel you need on KRKTE.

Update.

Run my car around the town with USR = 0.0604 for about 20 miles. This translates to ~15.5 on my AFR gage.

Before changing my USR my fuel trims were pretty close to zero on both banks. Now I am getting 0.90 fuel corrections for closed loop low-mid range operation. WOT Stays at 1.0 which is good.

Since the car requests 10% less fuel to operate than before, it would appear that it works. I don't expect 10% improvement in MPG as I lose just a bit of torque due to leaner mixture, but there should be at least 5% gain here which is 1MPG in our cars. Nothing to sneeze on!

Nice, this is exactly what I was going to start looking into for my sister's allroad, after I got my own car and normal ME7 knowledge up to snuff.  This looks like a much simpler approach than I was expecting to have to take. I'll have a wideband on it as well, so I'll be able to contribute some more data points of voltage to AFR, to see how consistent the O2 sensors read from one car to another.

Nice topic.

I am also looking for fuel save alternates in close loop environment.
The 0x11ADE looks like this? If i am correct mine should be at 0x11ADA with 7B value. Can you upload a screenshot of the hexdump?

If i understood correct this doenst affect WOT conditions?

The only way to figure out if that's it is to compare this region in m-box (or other cars with OLS available for them) to your car and see if stuff matches. RS4 Allroad and other OLS files are posted on this forum - find USR , view in hex and try to find this region in your tune.

ME7.5 wideband. I don't think USR exists (don't have 7.5 FR) but we do have USRJ.

I like the logic of going leaner than stoich, but I don't like this hack, and I will explain why.

1. The mixture is regulated to 15.4 everywhere, but you don't want 15.4 at idle, as it will increase misfires.
2. ME7 stays closed loop and lambda=1 very often. Doing it this way will increase knock at higher loads, but lower RPM's.
3. There will be a very harsh jump between 15.4 and when it goes open loop.

For this to be good, we need at minimum a min rpm/max load threshold value for when this hack is enabled.
Better would be to read a map based on actual load and interpolate the value from there.

Some ASM coding will be required, but it's pretty easy.
Basically I mean dynamically changing USR based on certain conditions, so that we have 14.7 or even a bit richer on idle to help with misfires, 15.4 on light and part load cruise, and normal behavior whenever we do anything but gently cruise along.

1. Maybe. That depends if you already have misfires at idle or not. If you do... time to either tune the fueling properly or get better injectors :)

2. There is really no knock present on my car until hitting heavy at 20psi. At least I never logged any corrections at load below 180ish and at this point the car is running on open loop for a long while already getting much richer mixture than closed loop calls for.

3. There wouldn't be since all tables are geared towards gradual adjustment. The system doesn't drop from 15.4 to 11 in one step as you'd think. The values are interpolated so it would do it gradually. But that's a very extreme example which will not be representative of real life tuning situation as no table should ever be set up like that.

I am not advocating or persuading anybody to use it. I just merely present a simple way to do it others missed to date.

Nvm.....

This will not work for wideband ECU's. Sensor voltage is not used, actual lambda is compared to target lambda.

Not this specific variable. To see how this could be done on WB car, one must think about it a bit. I have a suspicion that there must be a linearization table to wideband sensor that tells the ecu that so many volts equals this lambda.

Since I have no personal interest in pursuing this for a car I don't own (wb one), anybody interested in this will have to hunt on their own. With all the OLS out there this is really not that hard people.

It could work. wideband is just two narrowband sensors working together. 1st sensor measure AFR of a 'sample' of the exhaust in a small camber. 2nd sensor acts as an O2 pump pumping O2 into that sample until 1st sensor reads 14.7. AFR is determined by HOW HARD the 2nd pump has to work (current vs voltage) to make the 1st sensor read 14.7.

So, if the 14.7 voltage calibration of 1st sensor is modified like you are suggesting, it should work the same. Only you offset all AFR readings. So, if you make 15.4 read 14.7 you also make 13 read 12.3. Car thinks it's running 12.3 when it's really running 13. And it might not be linear making things complicated to know your real AFR.

LALIUS : Lambda linearization

input:
USVKK_W : LSU voltage before Kat corrected

elray: i only have LALIUSH, which appears to be behind cat (secondary)... looking for primary now.

Also, you will have to find the code that causes lambda regulation to go open loop when req lambda !=1

I was speaking about using the same method. LALIUS is also available on narrowband cars, but it has  nothing to do with target lambda. You would be recalibrating sensor voltage to equal an incorrect lambda.

Fuel trims are calculated differently though. There isn't a target sensor voltage. On wideband you would need a lean target AFR

While we are at the topic of economy, looking at the economy vs AFR and how steep it is towards the richer side more savings can be made by little leaning out at areas richer than stoic if you have an aggressive driving style. My stock car goes richer than 0.75 at high loads and the wideband voltage is like very close to zero, dont know how rich that is because LALIUS starts only from about 0.3v. That is like more than 40% loss of economy.  I dont want to throw 40% of my fuel in cooling my engine , What a waste.  :'(.  Looks like KFFDLBTS is doing it due to bad quality of fuel I only have access to.

This is a bad way of doing it.  I looked into it a while ago, when I was looking into ways to run leaner.  There are too many drawbacks such as not being able to alter AFR agains load and RPM, and the effect on fuel trims.

There is no need at all to use closed loop at anything  other than idle, as this is where tight control is needed due to very small amounts of air and fuel being consumed.  By tuning KFLF, you should easily be able to obtain any AFR you wish at any load and RPM combination, while still maintaining closed loop at low RPM - I typically use 1800rpm as a threshold.

Rick

Rick how can you dissable closed loop control?

Also Rick is it possible to go openloop in wideband also. Can you please give the details I am intersted.

Why would you ever want to do that?

Because I want to go leaner than 1 AFR by just modifying krkte and use it in my multimap tune using cruse control, so one of the tunes will be lean burn tune. 

Closed loop is enabled only if load exceeds the threshold in RLLRUN for the specified RPM point. This allows you to disable closed loop up to whatever load/RPM you designate by raising the threshold.


This should also apply for wideband ECU's.

Then do it closed loop, that's the beauty of wide band!

As far as everything that's been written on here, that ^^^ is not possible. ME7 hard limits fueling at lambda= 1. So running over 14.7 to be economic is not possible in closed loop unless you redefine sensor calibration via Julex method, or you can disable closed loop and fuel via Ricks method; map RLLRUN and KFLF. Julex method seems only good for narrowband, unless I missed something somewhere... ;)

Oh ok then.......

The only way I see it possible on wideband is to change sensor voltage to AFR calibration with LALIUS. USR does not exist on wideband ECU's as fuel trims a based on lamsoni_w (actual AFR) vs. lamsbg_w (target AFR) and not usvk (sensor voltage) vs. USR.

Spot on. There should be LALIUS map equivalent for front sensors in WB ME unless LALIUUS is used for both sets there which would neatly solve the problem. You find it, I still have enough stuff to search for on my platform :)

Proposed change to WB LALIUS, just adjusting the area around 14.7 - 15.4

voltage x-axis, AFR y-axis

AFR = stock
AFR2 = modified for lean

(http://creativeion.com/rey/vw/help/lalius_lean.gif)

close-up on modifed area
(http://creativeion.com/rey/vw/help/lalius_lean_closeup.gif)

Absolutely incorrect. Depending on cam setups, you won't even have misfire free idle leaner than 13 AFR on more exotic setups.
Of course this is extreme, but your assumption that the injectors are bad is just wrong.
FYI my only car with ME7 is a bone stock RS4 and that would probably be fine idling at 15.4, but I don't only tune my own cars.
If you've never logged any corrections below 180 load, then your timing map is not set up right. That is all I will say.
Please don't tell me what I think. You don't know what I think. You also don't know that I probably have more experience with tuning and ECU's than you, but this is not relevant to this discussion. I just found your statement rude.
With that out of the way, you missed what I told you. You are tricking the ECU to think that Lambda=1 is actually something else. Yet you are not tricking it to think that for example lambda=0.8 is something else, so you will get less gradual adjustment, and not really what is needed in a number of cases.
You will need to recalibrate loads of other stuff to make it operate correctly or update the code a bit.

This with combination of #2 means trouble. If you want to do stuff right.
That's why the *correct* solution would be to have dynamic USR. I would go as far as saying, the USR should depend on LOAD.

And I will write this eventually when I get there, lacking time right now.

look for identical value (search hex). Are your LALIUSH and LALIUSH2 identical? Let's hope there are 4 identical maps. 2 for rear and 2 for front.

We work with what we have and your points are valid... If you read my original opening post and follow ups you'll notice that I point out to possible issues with this approach, namely fueling changes that would have to follow if you want to do it right... or just let the ECU establish long term trims and that's it.

On a target platform for this (ME 7.1 / ME 7.1.1) which uses narrowband, there is no other point of reference but Lambda = 1 = USR switching point when in closed loop and fueling table when in open loop. The only pitfall I see here is when/if  frao and frao2 corrections kick in and it will affect WOT fueling. I guess one would want to massage few other things for this to work, namely fueling adaptation thresholds, PID controller for o2 corrections, etc.

Now, admittedly there also is LALIUS map for rear o2 sensor which appears to be linearizing voltage to Lambda, but it only seems to be used in cat functionality check. At least that's my impression since I experimented with changing all the rear sensors linearization tables with no effect on actual lambda.

I'm not sure if you're in the US or not, but I just thought I'd make a quick comment that I think you might have misinterpreted what he was saying, regarding that statement being personally about you.  He said, "like you'd think," with 'you'd' being a contraction for 'you would.' In that context, 'you' is an indefinite pronoun, and could be replaced with 'one', being that it is a general statement meaning 'as one might think.'  Anywho, I don't think he was trying to be condescending  :-X

Thanks Jibber. I understand that people have bad days and I filed his reply as such. I am still willing to have a meaningful conversation about original topic :)

I might have overreacted a little, or misinterpreted your post, came over a bit "you don't really know what you are doing" way...

I don't think you will get this just right by calibrations alone.
I think the way to hack it in, is make USR dependent on load, so this hack is only active when really *cruising*. Maybe specify some lockout RPM range as well.
Then make tweaks to KFLF to get the trims to zero in the given area.

That would be nice. It is what it is though.

My personal belief is that altering USR is ok for idle and part throttle but not desirable in higher load situation. I believe we share the same point of view here.

This being the case, the ECU already works that way. At higher loads the closed loop is disabled (where lambda < 1), USR is no longer active and fuel trims have no effect on fueling, leaving the car in open loop where it is getting all the fuel it needs.

Geez, so many complainers. Its a hack with pros and cons, which you fully disclosed.

Thanks for making the effort and sharing your findings.

I am running 1.1 lambda now, with lamda .95 at 100% throttle using lamfa, Also decreased torque reserve at idle so my idle consumtion has come down from 1.5 to 1.8 l/hr to 0.9/1.2 l/hr. Mileage has gone up from 8km/l to 11km/l on  daily travel. Had got error code o2 sensor post cat out of limits, so instead of coding it out I have filled the post cat o2 sensor laliush map with all stoic values (will that be a problem). Overall very happy with the setup as the car drives well only problem I think is my catcon is dying as I get a burning smell from the exhaust. Max lambda is limited to 0.75.

Great!

Are you using USR method or redefining RLLRUN/KFLF?

Sorry wrong thread , there was a similar thread for wideband ecu's. ;D I have shifted lalius, reduced krkte by 10%, and increased kfzwop 1/2 did some other mods too dont remember offhand on torque reserve .

would you mind walking us through those changes? did you just mutliply KFMRES by say .80?

I dont have the laptop with me now on which I do all the work , will post soon.

I don't think it is possible to save so much fuel. The possible gain is at about 5% level so your values are not accurate, I am afraid. One thing you might have missed and is affecting your perceived economy.  If you're working off your cluster readout for it though and not calculating gas usage with kms/miles drive vs amount of fuel pumped in since last fill up. Once you change KRKTE, you also need to change KVB so that cluster calculates fuel consumption correctly.

Yes I have been taking these readings from the cluster, Can you explain why kvb has to be changed if we change krkte.

http://s4wiki.com/wiki/Fuel_injectors#cite_note-3 :)

KVB is a constant that is used to define the pulse width sent to the cluster for consumption readout. It is linked to flow rate.

I did an actual mileage check and it matches with the cluster. Did 98 kms to 7.38Ltrs of petrol , cluster reading was 7.7ltrs/100 km for the same. Kvb was left on stock.

What were you getting previously (according to calculation, I don't care about the cluster numbers)?

I cannot say what my mileage would have been as this run was on a totally new route . If my cluster readings are to be believed then there is very substantial difference in mileage when running lean with torque reserve reduced  as I have posted earlier. Honestly I find these figures hard to believe- but as you see better figures in the cluster you tend to try and better it and drive carefully, that may be one of the reasons. I wanted to check whether my cluster was off as I had changed my KRKTE by 10%, looks accurate to me.
My brother used to drive every day to office on stock tune and his figures were like 11l/100km, now its like 8l/100km  ???.

OK. I've spent the last week on this issue (theoretically) for wideband cars there's a point in LAMKO-LAMDSK that sets lamdiag_w=1 when no diagnostic functions are active. The value makes it to the end if there are no other <1 values from the other enrichment functions (LAMFA,LAMBT, etc). In the end lamdiag_w doesn't get MIN with 1, so it makes it as is to lamsons_w as long as it's within lamlgm and lamlgfmn and outside of LAMSOSUF and LAMSOSOF. Those 4 limits/thresolds are fine stock.

So, all that to say this: by changing the one mov line from
mov RAM_lamdiag_w, #1000h
to
mov RAM_lamdiag_w, #10C2h (15.4AFR)

We might be able to idle/cruise lean while maintaining a calibrated LALIUS

BTW, how lean can one run? LAMLGMKT/LAMLGMTM = 1.1 in my AMB ori, but I can change that.

i am running 1.14 @ultra low loads <40% >1.10<55%> and 1.00 > 75% then i use lamfa for enriching for loads above 120% so 1.14 to 1.12 is fine for low load part throttle cruising not good for iddle though

^^What car?

and how are you achieving this? and why is a lean idle not good?

if using LAMFA then are you still closed loop? are you using an adjusted wideband to just alter the input to the ECU? How else would you get LAMBDA above one.

you can dissable closed loop to acheive lambda >1.00 and then lamfa in closed loop to enrich as needed. my fault i just saw the topic was for me 7.1. i have me7.5 wideband. sorry.

There are me7.5 discussion is hear as well. just need to identify which one applies to your comments.

So, how are you requesting 1.14-1.10 during open-loop?

i am using kflf to adjust the mixture.

I implemented this today on a 032HS box. changed it from 1.00 to 1.086. It gave me the lean Target AFR I was looking for. However, there were two issues.

lambas was off making my O2 regulation based off something else... i.e. I expected O2reg to be the same before and after change, but O2reg was -17% which means its cal was not changed with the edit. The edit only changes lamsons_w and doesn't address lambas which I bet O2reg uses.

Also, I didn't see any activity from LAMFA that I made sure to add back in.

So all I saw was 1.086 target execpt for inj cut off/overrun. then I saw target AFR = LASOAB.

Still working on it.


Also, I'm not happy with LAMFA only being a 8bit map. The resolution is not that great. I can't request just a hair under 14.7. The highest I can request that under 14.7 is 14.585. I can't just request 14.69 to get back to near 14.7...

edit... as I type that last statement I realize DLAMFAW is my issue with LAMFA. DLAMFAW = 0.0313 and my LAMFA delta from 1.0 is 0.0078125. Without decreasing DLAMFAW, LAMFA will not activate until LAMFA < 14.24

Ive accumulated the "LAMBDIAG" address for a few different files, ill list them at the end. the only other things ive found needed is to set LAMLGMKT/LAMLGMTM high enough. BTS fueling should be able to request lambda>1 so you can request whatever lambda, wherever you want as long as its lower than lambdiag, but lamfa would be nice if someone could figure that out.

8e0909518ak 368072:  0x72456  offset- X*0.0002441406
8e0909518al 386802: 0x72F1A
4b0906018ch 360101: 0x6654A
4b0906018dp: 0x5F8B0
b5 s4 m-box: 0x3bfb2

Didn't find "LATRO" on my file, but are you saying for set these maps for desired target for narrowband that already had USR tricked?

Idk what youre trying to ask exactly but ive found the only thing that needs done to enable controlled lambda >1 is change your "lambdiag" value to whatever lambda you want to cruise at and raise lamlgmtm. No need to trick the ecu or sensor in any sorta way, although you might just want to disable open loop for low loads with RLLRUN if your ecu uses a narrowband sensor but maybe itd still be helpful to have narrowband lean adaptations, idk try it if you have a wideband gauge.

Maybe your car uses a Wideband, so you can target any lambda higher than one.

With narrowband you can just modify the voltage of the switch point.
I don't remember what the ID is, but I used it to run 1.15-1.2 lambda on LPG.

I made it switchable though based on whether LPG is on or off, otherwise it is not possible to pass emission sniffer test on gasoline and also it runs like shit at 1.2 lambda on gasoline, actually using more fuel.

Trying this on MED 17.5.2 Wideband, but did not found "LAMBDIAG" map name, only similar map i've found was "LAMHAP", are this equivalent?

Lamdiag is not a map, it's limitation in code