ECU Fuel and Ignition Learning Systems

The engine control system is a somewhat complicated system, but it can be broken into multiple parts to be understood easier. This systems primary function is to control the fuel injectors pulse width (fuel amount) and ignition timing (spark plug firing). It may also control engine boost levels, throttle angle, cam advance, or other systems like emissions. It receives information from several sensors including the mass air flow sensor, oxygen or AFR sensor, knock sensor, pressure sensors, temperature sensors, and user input from a throttle or accelerator position sensors.

The engine control unit (ECU) is capable of short and long term learning. The short term is essentially its immediate ability to trim outgoing fuel or ignition signals. When the ECU senses a rich condition via its AFR sensor the ecu can respond by reducing the injector pulse width. In a similar manner when the ECU receives knock feedback from the knock sensor it can reduce the ignition timing. The ECU can then begin to watch for a trend of positive or negative prolonged correction to which it will then learn a value.

So for a prolonged fuel correction or repeated knock event within a specific load and rpm range the ECU will then learn a value. The ECU is typically allowed a range of plus or minus twenty-five percent, where as the fuel learning can correct a range of plus or minus fifteen percent. The fuel correction and learning systems both have limit ranges in the load and rpm range where they can be used. The air-to-fuel sensor also has a limited range feedback and reads effectively down to around 12.2:1 AFR. Unfortunately, this sensor does not read low enough to provide any useful feedback most pump gas setups in full throttle and high load conditions. However, the AFR sensor is quite useful for reading and correcting near stoichiometric conditions (Ideal combustion for emissions 14.7:1 AFR). The stock ECU active fuel correction is limited to a specific range of rpm, temps, throttle, and load. The fuel correction also features a delay timer which can cause some engines to achieve lean conditions which in high boost or load conditions. This being just one reason why custom tuned maps more attractive in terms of reliability. The system is provoked by related sensors, post start timers, and the primary fuel table. The learned fueling is divided into four sections and is limited in its viewing range. So the system can inherently cause incorrect fueling by learning in a range of roughly 2.5 to 3.5 MAF volts and carrying that value up the full range of the MAF. Average cars frequent in the 4.5 volt range with potentially a value it learned from around 3.0 volts. Thus the cars require periodic monitoring to maintain ideal running conditions. Furthermore, the learning system can be used to ones advantage when diagnosing a problem. Excessive learned correction can indicate possible leaks or fueling issues.

The ignition control system is a bit different then the fuel system and includes a ignition advance system which aids to the optimizing of power and fuel economy. This system attempts to add a prescribed amount of timing based on rpm and load to the primary ignition. It can also make immediate real time corrections to reduce ignition advance in the event of engine noise or knock. Most systems also have multi-staged functionality, that will reduce ignition advance in increasing amounts in response to excessive or continuing noise events. The knock sensor provides feedback on engine noise while the ECU compares the current noise with anticipated noise. Depending on how much noise overage the ECU reads, it may reduce ignition advance. The system also can learn a negative value from prolonged or high noise levels and can associate it a specific rpm and load range. When the ECU returns to that load and rpm it will run the reduced learned value,

The ignition system utilizes two tactics to optimize timing advance including a dynamic advance and its multiplier. Dynamic advance refers to the amount of additional timing the ECU may add to the base timing. While the multiplier is a variable that scales the range of the dynamic advance. The factory tunes typically set the multiplier to 50% so that if the car experiences little engine noise, it will increase timing. When the ECU begins to see prolonged learned values or high noise levels it can reduce the multiplier and scale the ignition advance down. This tactic is more effective in a car with unknown fuel and unknown conditions. A car using known fuel in known conditions can run a more optimal ignition curve.

Many drivers have never thought about the functionality of their engine control, or discovered its multifaceted learning system. Many also believe that the factory tune is the safer and more reliable than a customized map which is also mostly false. In fact delayed closed loop to open loop fueling transitions can cause significant knock in forced induction engines. Additionally some factory tunes run additional timing in specific cylinders (usually dog legs cylinders in Subaru’s') which has shown to increase addition engine noise and imbalance. The lack of wide band oxygen sensing is another blunder, that renders the fueling system virtually blind during high boost and load conditions. Overall these system are quite effective at running turbo charged engines, yet these systems can certainly be reprogrammed to not only increase power but more importantly increase reliability.

Www.tuningalliance.com

Thanks for compiling this.

The AFR sensor actually reads down to just about 11, and I believe people have confirmed its accuracy with a WBO2, but I can't say for certain.

As far as the MAF voltages go:

The learned fueling is divided into four sections and is limited in its viewing range. So the system can inherently cause incorrect fueling by learning in a range of roughly 2.5 to 3.5 MAF

The scaling for stock 05 LGT is 0->5.6g/s, 5-6->10g/s, 10-40g/s, 40+g/s

which on the stock MAF scaling is around 0->1.43V, 1.43->1.71V, 1.71V->2.6V, and then everything above 2.6V.

edit: For a 2005 stock ROM at least. The new AF ranges are different I believe.

Thanks for compiling this.

 

The AFR sensor actually reads down to just about 11, and I believe people have confirmed its accuracy with a WBO2, but I can't say for certain.

 

As far as the MAF voltages go:

 

The scaling for stock 05 LGT is 0->5.6g/s, 5-6->10g/s, 10-40g/s, 40+g/s

which on the stock MAF scaling is around 0->1.43V, 1.43->1.71V, 1.71V->2.6V, and then everything above 2.6V.

 

edit: For a 2005 stock ROM at least. The new AF ranges are different I believe.

The stock AFR reads to 11 AFR, but is not accurate at all under about 12.2. Too even effectively read to 12.2 you need to rescale the sensor a bit. Here is an example of a car with stock scaling. As you can see the wideband AEM is reading 12 AFR and the OEM sensor is reading 11.14. If the stock sensor read down to 11:1 AFR we wouldn't need widebands to tune them.

Regarding the MAF, although the learning says 40+ g/s they are limited to 80 to 120 g/s.

The stock AFR reads to 11 AFR, but is not accurate at all under about 12.2. Too even effectively read to 12.2 you need to rescale the sensor a bit. Here is an example of a car with stock scaling. As you can see the wideband AEM is reading 12 AFR and the OEM sensor is reading 11.14. If the stock sensor read down to 11:1 AFR we wouldn't need widebands to tune them.

Regarding the MAF, although the learning says 40+ g/s they are limited to 80 to 120 g/s.

From my own observation, Trim D has always indicated 40+ g/s on my car and during logging, it shows the trim being applied at 40+ all the way to the max I've seen my MAF read (250+). The CL learning itself that takes place for trim D I have seen occur between about 40-60g/s. Above about 55-60g/s (depending on varying conditions) it switches over to OL and the correction is being blindly applied and no learning takes place at all. I've never seen learning take place in the range of 80-120g/s.

Interesting. I stand corrected then. I could've sworn I saw someone do a comparison with another WBO2 and it read very close to the stock AFR sensor even down to 11-11.5, but I guess I'm mistaken.

 

I'm not sure what you mean by that exactly.

 

From my own observation, Trim D has always indicated 40+ g/s on my car and during logging, it shows the trim being applied at 40+ all the way to the max I've seen my MAF read (250+). The CL learning itself that takes place for trim D I have seen occur between about 40-60g/s. Above about 55-60g/s (depending on varying conditions) it switches over to OL and the correction is being blindly applied and no learning takes place at all. I've never seen learning take place in the range of 80-120g/s.

The learning for trim D occurs from 40 to 120g/s, regardless of CL or OL conditions (some cars its 40 to 80). The value is then applied for the entire range (40+). CL mode is only indicating realtime correction, the car can still in fact alter the learned values out of CL mode. As active correction and learning are somewhat seperate functions. So it is important that the ecu does not learn a large negative value in that range as it will lean out the entire high end, which ECU can not truely monitor.

The learning for trim D occurs from 40 to 120g/s, regardless of CL or OL conditions (some cars its 40 to 80). The value is then applied for the entire range (40+). CL mode is only indicating realtime correction, the car can still in fact alter the learned values out of CL mode. As active correction and learning are somewhat seperate functions. So it is important that the ecu does not learn a large negative value in that range as it will lean out the entire high end, which ECU can not truely monitor.

In ATR, in the A/F Learning section, there's the A/F Learning #1 Modify Airflow Limit (Max) table. The Cobb Stage 2 OTS maps have this capped at 80.00, so in that case I presume the D range learning is from 40-80.

In my tune, this is also capped at 80, but the A/F Learning #1 table has the top value set to 90, so the D range correction is always 0 and applies no correction in OL at higher airflow. Seems to be OK for my car. The B and C corrections both sit around -0.5 to -0.8 so looks like things are pretty well sorted in the lower ranges.

What year is your car?

It's possible the RR definitions for the 05 LGT don't show those tables... I'm going to take a look again when I get home, but I don't recall seeing those before. I'm on A2WC522N. Or I might be looking at the wrong area.

I'm not trying to be argumentative or anything with anyone. Just want to get my own facts straight. Thanks.

^ Not sure if that was directed to me, but assuming so, I have an '05 OBXT. My tune is AP-based, but I do have RR and use the RR logger for its (much) faster datalogging polling speed.

I opened up a '06 LGT VF52/stock fueling map posted by mickeyd2005 a few yrs ago, and do not see the tables I mentioned in my prev post. I only see A/F Learning #1 Limits, and A/F Learning #1 Airflow Ranges. I believe I have the latest ECU defs.

Interesting. It would appear that merchgod (the original developer of RR who now works for Cobb) or someone else found new tables and defined them... Very interesting indeed. Thanks for looking into that! This somewhat changes my perspective on OL/CL and AF trim learning.

edit: Jackpot. http://www.romraider.com/forum/viewtopic.php?f=25&t=8121

They say imitation is the best form of flattery

 

I was looking through the COBB SD implementation, and there are a couple of useful tables in there in relation to AF learning and AF3 (rear O2) correction. I'm still busy scratching around for the rear O2 sensor tables, but I've found the AF learning minimum and maximum threshold figures.

 

In a nutshell, if you set the maximum threshold to lower than your D range in the A - D learning - no more learning carried through into open loop. Quite useful

 

Quite cooincidentally, fastblueufo PM'd me to find it for the '04 USDM sti (synchronicity!) so:

For A2ZJ710J:

 

Code:

<table name="Af learning Max threshhold" address="5bb10" type="1D" level="1" scaling="MassAirflow(g/s)"/>

<table name="Af learning Min threshhold" address="5BB00" type="1D" level="1" scaling="MassAirflow(g/s)"/>

A2WC522N

 

Min. C7080

Max. C7090

Interesting. It would appear that merchgod (the original developer of RR who now works for Cobb) or someone else found new tables and defined them... Very interesting indeed. Thanks for looking into that! This somewhat changes my perspective on OL/CL and AF trim learning.

 

edit: Jackpot. http://www.romraider.com/forum/viewtopic.php?f=25&t=8121

Now I see where the disparity between your understanding about Boost Target IAT Compensations, and mine, came from... When tuning, I have always had both RR and ATR up at the same time.

Cobb's ATR maps are much more developed now that Bill works for them, than they used to be. To limit one's tuning to only OS or ATR exclusively limits one's tuning ability, as very often each has something to offer the other does not.

Regarding the A/F Learning ranges, because those ranges are configurable in ATR, I changed them all to suit my MAF Cal and tuning... capping my Max Airflow Limit at 50 g/s instead of 80, for example. Because I had an extremely accurate MAF Cal (AEM CAI) my OL fueling was right on and I had no need for learning above 50 g/s. I did this because of my CAI installation, which caused positive pressure at high cruising speeds and, depending on my driving, skewed my A/F Learning D badly when it went out to 80 g/s. By capping the max learning at 50 (after rescaling A, B, and C, of course) I no longer had to worry about weird values of D learning messing with O/L.

These and other similar things are relatively unimportant on normal everyday tunes. But when tunes are pushed to the edges of the equipment's envelope they become very important. In my case, with OL AFRs of 11.5-11.7, there was no tolerance for having a "few percent" removed by D range learning. Similarly, with IDC's of over 100% in colder weather, ensuring an accurate IAT comp table was also very important. More power due to cold air is very nice, indeed... but higher boost in the wrong place is not.

Edit:

While I was typing this your last post reminded me to add another reason why I had no D range influence on OL... precisely what td-d's post points out. I set my A/F Learning #1 Modify Airflow Limit (Max) to 49.99, which was lower than my D range of 50. It doesn't have to be any more than that to ensure your OL fueling is left alone.

Now I see where the disparity between your understanding about Boost Target IAT Compensations, and mine, came from... When tuning, I have always had both RR and ATR up at the same time.

 

Cobb's ATR maps are much more developed now that Bill works for them, than they used to be. To limit one's tuning to only OS or ATR exclusively limits one's tuning ability, as very often each has something to offer the other does not.

 

I'll have to keep an eye out for that Boost Target IAT compensation, but again, I don't believe there's any sort of compensation like that built in, not just because I don't see it in RR, but because I've never actually seen (while logging) Boost Target affected by temperatures in either extreme.... Only for elevation.

As far as OL fueling being affected, I still have mixed feelings about that. My previous/current position is that the Trims are as much of a safety feature as they are for fuel economy. I've always believed that if a condition occurs that affects fueling that it needs to be compensated for, even in OL. By disabling it, there's no safety compensation available for OL at all.

I still feel that's the case, but I'm becoming a bit more open to the folowing thoughts:

1) it's not worth the trouble... any sort of safety gained from this is next to nothing since any sort of problem that occurs that would affect fueling that severely would manifest itself in some other way anyway

2) the reasons you already mentioned about running lean

That being said, I'm going to give this some good hard consideration the next few weeks.

A/F Learning, by definition, is fueling. It is of relatively little concern for a stock ECU tune. However, once past the threshold of either tuning modification, or equipment AND tuning modification, knowledge of fueling becomes mandatory.

Any owner with a car that is modified by equipment and tuning who does not keep in constant touch with a short-list of important parameters is betting their engine on hope and guesses... akin to running with eyes closed in the dark. Boost and fueling (AFRs) as well as knock are at the top of that list. Logging can keep track of the big picture, that's why I logged 100% of my mileage. But an ongoing awareness of AFRs, Boost, and Knock are part of the tuning equation, without which the owner is betting on something life has precious little of... a sure thing, and unchanging reality. That is why I was always monitoring knock, boost, and AFRs, among many other things, by equipment in front of me in the driver's seat.

During your next few weeks of search and introspection, if you can find someone with ATR to peruse, I'm pretty certain your eyes will be opened... and that you will find IAT Boost Compensations for the LGT.

SeeeeeYa: I presume the ATR table name you're referring to is: "Boost Targets Compensation (Intake Temp)"?

I see the following in Cobb's 6/2013 USDM Table Descriptions document (https://static.cobbtuning.com/support/docs/tuningguide/Subaru%20Docs/Table_List.pdf)

-----

Boost Targets Compensation (Intake Temp) [09-13 FXT, 07-09 LGT MT/AT, 10-12 LGT, 07-09 LGT spec.B, 07-

13 STI, 02-05 WRX, 08 WRX MT/AT, 09-10 2.5GT, 09-13 WRX] -> This is the compensation to the boost targets

based on the current intake temperature. The compensation is applied to the boost targets in absolute pressure

("Boost Targets" table value + 1 atmosphere of pressure).

-----

Based on the above, it looks like the only LGT models that this table applies to, are the 07-12 ones. I don't see this table in my ATR. My ATR instance is for the 05 Legacy GT (my car is an '05 OBXT). I presume iNVAR would end up w/ the same ATR instance if he were to obtain an applicable copy of it, since his car is listed as being an 05 LGT.

SeeeeeYa: I presume the ATR table name you're referring to is: "Boost Targets Compensation (Intake Temp)"?

 

I see the following in Cobb's 6/2013 USDM Table Descriptions document (https://static.cobbtuning.com/support/docs/tuningguide/Subaru%20Docs/Table_List.pdf)

 

-----

Boost Targets Compensation (Intake Temp) [09-13 FXT, 07-09 LGT MT/AT, 10-12 LGT, 07-09 LGT spec.B, 07-

13 STI, 02-05 WRX, 08 WRX MT/AT, 09-10 2.5GT, 09-13 WRX] -> This is the compensation to the boost targets

based on the current intake temperature. The compensation is applied to the boost targets in absolute pressure

("Boost Targets" table value + 1 atmosphere of pressure).

-----

 

Based on the above, it looks like the only LGT models that this table applies to, are the 07-12 ones. I don't see this table in my ATR. My ATR instance is for the 05 Legacy GT (my car is an '05 OBXT). I presume iNVAR would end up w/ the same ATR instance if he were to obtain an applicable copy of it, since his car is listed as being an 05 LGT.

I agree, now that you have posted those descriptions, it is not a part of the '05 LGT ATR. Thank you for that. Gotta love facts.

That said, experience says it is not "there" because Cobb has not added that table, rather than it not being in the ECU. There were a variety of tables added by ATR revisions in the years I was tuning my WRX and STi... mainly because of RR influence and tuner outcry.

Regardless, I'll have to admit to iNVAR my erroneous presumption. I'll also add that that table is one of those that gets modified very seldom, as it is pretty good as it comes, and only at the extremes on well-modified cars was any change warranted and even then very little. Knowing the compensations, however, was very important in understanding of logs across the seasons.

No problem, that realization came about when I started looking for the table in RR and in my ATR.

Wouldn't surprise me to learn, someday, that the table -is- in the ECU for the 05 LGT, just not yet identified/added/etc yet. I wouldn't mind having Boost Target IAT Comp, as my car seems inclined to knock in cooler temps despite the boost curve being nearly if not exactly identical across the seasons. My protune map worked fantastically in 80F temps, but once the temps dipped into the 60s and below I needed to make somewhat substantial changes to WGDC IAT Comp to keep the boost curve consistent. Despite that, I still needed to pull ~2 deg from the torque peak to slightly beyond peak torque, or I'd see FBKC show up. Once the temps go above 80F, I can run the original map no problem.

I realize we're kind of off the OP's topic though, so guess I'll shush now

No problem, that realization came about when I started looking for the table in RR and in my ATR.

 

Wouldn't surprise me to learn, someday, that the table -is- in the ECU for the 05 LGT, just not yet identified/added/etc yet. I wouldn't mind having Boost Target IAT Comp, as my car seems inclined to knock in cooler temps despite the boost curve being nearly if not exactly identical across the seasons. My protune map worked fantastically in 80F temps, but once the temps dipped into the 60s and below I needed to make somewhat substantial changes to WGDC IAT Comp to keep the boost curve consistent. Despite that, I still needed to pull ~2 deg from the torque peak to slightly beyond peak torque, or I'd see FBKC show up. Once the temps go above 80F, I can run the original map no problem.

 

I realize we're kind of off the OP's topic though, so guess I'll shush now

Not entirely, ... there are IAT Ignition Timing Compensations to consider, which you do have.

Most times, however, issues like you mention is due to boost tuning, and boost control itself. If you have the OEM system you'll find its limits under the conditions you have issues. A well-tuned EBCS will usually rectify the OEM deficiencies, leaving the remainder of refinement of boost control to other measures.

Not entirely, ... there are IAT Ignition Timing Compensations to consider, which you do have.

 

Most times, however, issues like you mention is due to boost tuning, and boost control itself. If you have the OEM system you'll find its limits under the conditions you have issues. A well-tuned EBCS will usually rectify the OEM deficiencies, leaving the remainder of refinement of boost control to other measures.

Yeah, but the IAT Ignition Timing compensation table is usually set up to add timing in colder weather, which is not what he wants. Not sure why he would be knocking in colder weather with the same boost conditions. I'm sure he's flowing more air, but the cooler temperatures should allow for that, in general.

A/F Learning, by definition, is fueling. It is of relatively little concern for a stock ECU tune. However, once past the threshold of either tuning modification, or equipment AND tuning modification, knowledge of fueling becomes mandatory.

 

Any owner with a car that is modified by equipment and tuning who does not keep in constant touch with a short-list of important parameters is betting their engine on hope and guesses... akin to running with eyes closed in the dark. Boost and fueling (AFRs) as well as knock are at the top of that list. Logging can keep track of the big picture, that's why I logged 100% of my mileage. But an ongoing awareness of AFRs, Boost, and Knock are part of the tuning equation, without which the owner is betting on something life has precious little of... a sure thing, and unchanging reality. That is why I was always monitoring knock, boost, and AFRs, among many other things, by equipment in front of me in the driver's seat.

Not entirely, ... there are IAT Ignition Timing Compensations to consider, which you do have.

 

Most times, however, issues like you mention is due to boost tuning, and boost control itself. If you have the OEM system you'll find its limits under the conditions you have issues. A well-tuned EBCS will usually rectify the OEM deficiencies, leaving the remainder of refinement of boost control to other measures.

True, and where I was originally seeing knock was right near brief boost spikes (2 psi overshoot originally) shortly after mashing the throttle. I'm running a VF52 w/ a Grimmspeed EBCS. Needed to pull a couple % off the orig tune's WGDC Initial in a couple of RPM areas and needed to further reduce (make more negative) the negative turbo dynamics values. That helped a fair bit w/ keeping overboost to <= 1 psi (and only very briefly at that) and avoided spool/overboost knock. Still would like control a bit tighter, but after the initial ~1 psi overshoot and one smaller magnitude undershoot cycle, the boost tracks target pretty closely. Now I'm dealing with shift knock. My tuner is pretty sure it's drivetrain noise. I've used the AP to pull 3-4 deg across the board and the shift knock persists (usually pulls 2-4, occasionally pulls 6+), so perhaps it -is- something mechanical. Topic for a different discussion though

Sounds like you're dealing with the spikes well. Those spikes would likely be much easier to control if we had the newer LGTs that have per gear tuning in it.

True that. It was per-gear boost and timing that finally made my STi's tune polished. ECUs without the necessary tables to tune per-gear will always have loose ends that cause a dedicated tuner to chase boost/timing issues in circles.