The question was very simple, "Do our stock ECUs learn?"

Three times from the same person:

"They do not learn at all."

It doesnt LEARN AT ALL."

"It doesnt learn, it doesnt adapt to different motors..."

How much damn clearer do you need it? The question only required a one-word answer: NO! If the answer was yes then it would need a lengthy explanation, but it isn't!

 

did you know most standalones dont learn either?



Man, we need to put this one to rest...

 

When you've seen the same questions 50 times they get kind of old. If somebody came to where you work and asked you the same question everyday for 2 years, wouldn't you get kind of annoyed at it?

Not to mention the question was answered 3 (?) times in the thread and still repeated.

I like to think that to, but the facts are higher post people tend to be more knowledgeable (I don't count). Look at the highest post counters in this forum. You, Icemark, Aaron, Hailers, J-Rat, Karack, Rotary Resurrection and others. All of them are the most knowledgeable people in the 2nd Gen section. Granted there's many with low post counts that do know a lot, it tends to be a direct correlation between knowledge and post count.

Even take me for example. I haven't done 1/1000th the stuff to a 7 (or cars in general) that the previously listed people have, but at the same time, I give a lot of answers (most of which I like to think of as correct). Most of that is because I've seen the questions asked many times and seen the answers given by the people that DO know what they're talking about. I spew out the answers given by those pepole. I've also learned how to use the search feature (hint hint new people).

For these reasons I've created the following jpeg.

 

http://en.wikipedia.org/wiki/PID_controller


Let me get to a more technical explanation (I'm killing time before a boring English class). A modern ECU has several layers of correction and learning applied over time that our ECUs do not have. Most of this auto learning stuff is for emissions. Emission regs are way tighter now than they were in '86, and computers are much more sophisticated. You also have more oxygen sensors required on a car. Also remember that a lot of the learning aspects of the ECU don't apply in exactly the same way under WOT because cars don't have factory widebands, so how exactly is it supposed to know just how lean or rich the engine is running?

So if you reset the ECU on say my mom's 01 Corolla, it will run on completely factory programmed fuel maps. Then you drive it around for a while and the ECU will have to make adjustments to keep the AFR's, knock count, etc within the ideal range to minimize emissions. This is called closed-loop oscillation, whereby the ECU is continuously responding to input from all the various sensors. An open loop process would be where the ECU ignores the sensors, such as under WOT or when the car is cold and the oxygen sensor has not heated up to proper temperature.

This mostly consists of EGO (exhaust gas oxygen) correction from the O2 sensors (a modern car has at least two, one before the cat and one after). So if the AFR is not 14.7:1 for emissions purposes, the ECU will continuously vary the fuel injector pulsewidths to try and get it to 14.7 . The goal is to get the tightest oscillation around that ideal ratio as possible--so get it to range from say 14.6 - 14.8 , rather than from 14 to 15:1, like both of my FCs will do under closed loop according to my widebands. Leaner than 14.7 = more NOx emissions, richer = more hydrocarbon emissions. 14.7 just has the minimum overall emissions, not necessarily the minimum of one or the other of those two.

EGO adjustments might be relatively large at first, compared to the stock maps. So over time the ECU will develop what's called a short-term fuel trim. This is a basic correction factor to reduce the amount of error. Error here is defined as a deviation from the programmed ideal value, in most cases 14.7 : 1.

So the ECU will often find the integral of all these little changes made from oxygen center output and that will help create what's called the short-term fuel trim. In that way, it applies the short-term fuel trim first, then makes smaller EGO corrections. Over time, it will monitor all these corrections it has made to build a long-term fuel trim. This is a correction map that changes slowly in order to reflect larger trends in engine operation resulting from factors like ambient temperatures. If you ever look at an AEM EMS you will notice that it has the ability to program a map of integral gain for closed-loop control.

So ultimately, under closed loop the ECU will look at stuff in this order: base map --> modified by long term fuel trim --> modified by short-term fuel trim --> fine tuned by EGO correction and other factors such as coolant temperature, manifold pressure, knock count, airflow meter voltage, etc.

How does this compare to our ECUs ? We defiantely don't have a short or long-term fuel trim. We only have one oxygen sensor, because this was all that was required at the time. We have a knock sensor, coolant temp sensor, pressure sensors, and then a bunch of rotary specific bullshit (such as OMP for an s5).

I'm not knowledgeable enough to be 100% sure about this, but I think even on most modern ECU's the engine cannot really 'learn' under WOT, not with narrowband O2 sensors. All it can do is detect knock and do preprogrammed ignition retard.

Different manufacturers do things differently to an extent, but on a modern ECU you have two factors called short-term trim and long-term fuel trim. These are correction factors applied to the base map within the ECU. As people have mentioned, AFM inputs do not indicate that an ECU can 'learn.' In a sense though it has an advantage over say, the stock FD ECU, because the FD does not have an AFM and cannot use that sensor as a way to move to a different point on the fuel map to accomodate modifications.

 

yeah i guess so. hell i knew after the first time he said it didnt learn that it didnt learn but i wanted to know something else but couldnt figure out how to word it so it wouldnt sound like the same question. Years back I was into hondas and some of the ECU's I dealt with would sense changes to your setup and addapt or change parameters. Im not very knowledgable on rx7 ecu's so i was tryin to ask if they were like that. But yes, this is beaten to death so Im not posting here anymore

 

Well, at the risk of being flamethrowered I understand that it doesn't learn or adapt. But it does make corrections for airflow, right? I'm not trying to say it learns new mods, I guess the question is this: if you do (whatever mods) to your engine, and it flows more air - the AFM senses airflow and corrects accordingly, using it's set parameters, right? Otherwise, what's the point of the AFM?

What I'm thinking the problem is, is that it doesn't know that the engine is flowing that much air at a lower rpm than it's program says it should be. Because, as you say, it isn't a learning computer, so it has no way to figure this out.

And if it does work like that, as I think, than it will be running rich, right? At least down low.

I dunno, it's late and I may be all fuxored up in my thinking. I guess what I'm saying is that it will sense that something has been done, according to the airflow, but it can't interpret it properly.

 

bascially the stock ecu works just like most standalones, although its running more stuff and its not flashable.

so basically it looks at the rpm and afm, and makes a basic calculation, to get a "base" injector time, then it runs thru its corrections, temp, knock, 02, which can affect injector time and timing time, and then somewhere in there is a correction table (this is usually what gets changed when the ecu is tuned), where the base calcs+corrections are corrected. and then the ecu outputs that as timing and fuel injector pulse widths.

so its something like rpm+load+correction table + temp/knock/o2 corrections = output.

as previously said theres no learning, mazda doesnt have that in a rotary until the rx8.

 

most learning is done by auto transmission control units. engine control units have adaptives that can change and learn modes for various solenoids and sensors like the IAC for example as it controls idle.

 

You're right, but this is not learning, adapting or correcting in any way. The ECU simply responds to the various sensor inputs as it's been programmed to. This one of the advantages we have over FD's.