4AGE 16v. Does ECU self adjust?

[Ako]

Isn't it the altezza which will de-tune itself back to factory spec if you use a piggyback like an s-afc?

Sounds like self-learning to me!

[slighty_sykotic]

altezza is just a few years younger than the 16v 4age though.....


--Sykotic

[Mr Revhead]

correct back to factory settings....
is that learning? or maintaining what it thinks is correct? which is what its told when programmed at the factory
to me if it was self learning it wouldnt need the safc in the first place.....

[Ako]

Sykotic - I was just referring to the fact that toyotas apparently "dont learn". The self adjusting does mean that the ECU is paying attention to the AFR and knock count over time.

Revhead - If they can reset themselves to factory settings, that does imply learning in a way, or at least correction over time. The ECU has an AFR its' meant to be aiming for ideally, then it finds it.

I'm curious as to if say - you ran a tezza on 110 octane for a while, it would adjust itself slightly to suit the extra advance it could run? With subbys/mitsis/nissans etc, they do it slowly over time, bout 100km worth of driving or so, but an ECU reset just jump starts the whole process. I've even seen VR4 ecu's cope comfortably with having base pressure raised or slightly larger injectors put in - takes them a while, but they eventually get fuel trims close to ideal all on their own.

So to summarise - Toyota ECU's suck.

[Mr Revhead]

If they can reset themselves to factory settings, that does imply learning in a way, or at least correction over time.

id go more for the correction, it does suggest the tezza ecu is a bit smarter than some others.
my argument is "learning" is the wrong term. if they could learn the nyou wouldnt need safcs etc to run cams etc

[RedMist]

Semantics Warrick.

I suggest that a corrective action taken not to make the same mistakes IS learning.
IE the computer alters timing or fuel to stop detontion or incorrect AFR's. Just because its learning ability is bound (in order to stop fugged up maps generated by faulty sensors) doesnt mean it doesnt learn in the first place.

[Mr Revhead]

yes... but the definition ppl here are working on isnt as described above...

my whole argument is "resetting" the ecu is a myth.
there is no evidence other than ppls butt dyno to suggest other wise.
iv even reset an ecu on a dyno with no effect

[sergei]

Semantics Warrick.

I suggest that a corrective action taken not to make the same mistakes IS learning.
IE the computer alters timing or fuel to stop detontion or incorrect AFR's. Just because its learning ability is bound (in order to stop fugged up maps generated by faulty sensors) doesnt mean it doesnt learn in the first place.

Agreed. In other words learning is: when actions are based on previous experience. In this case if ECU retards ignition due to a knock - it learns, or changes the fuel mixture based on oxygene sensor reading - it also learns. So in other words any ECU which utilises random access memory, sensory input (feedback), and appropriate algorithms (actions based on previous memory), is in fact learning. It even does not matter if the adjustment is only valid between key on and off and reset every time the ignition is turned off. Technically by defention it will still learn during the trip. From my previous experience with SAFC and a 20v ecu, they do learn and will not allow any adjustments over ~20%, plus the ECU will compensate over the time for my adjustment and will basically revert all changes (even at full throttle due to long term fuel trim) in a few trips.
It is possible to tell if ECU has long term FT by looking at the oxygene sensor, if it has got heater the ECU most likely will have long term fuel trim function....

[sergei]

As for reseting ECU on dyno, did you do half throttle runs and compare them together before and after reset? Also the learning with the engine warmed up will take only a few seconds, so capturing that on dyno is not really possible.
Best way to test is to take and replace a variable on the tested car (if it has AFM - untighten the spring) and take it for few runs and see if the mixture will return to same level before adjustemnt...

[RedMist]

As for reseting ECU on dyno, did you do half throttle runs and compare them together before and after reset?

Its a good point... closed loop isn't used during WOT, when acceleration fuel is pending, or a few minutes after start. As such you'll never see it on a typical dyno run.

[RomanV]

The reason the Altezza guys wanted to run the SAFC, was to lean out the factory fuel at high RPM...
As they run a bit rich, as a safety margin.

They did this, and gained some HP.
However after a week, it was back to normal.
I think there was dyno plots to prove it, but I cant remember.

And regarding the WOT comments not using the sensors....
I think that they do adjust the WOT map, but indirectly.
At partial throttle conditions, the ECU richens/leans it out on the fly.
However if it can establish a trend over a period of time (Eg. constantly having to richen it at 3000rpm)
Then it bumps up the fuel map, so it doesnt have to go through the lean/rich cycle anymore.
If it is capable of doing this, I'd imagine it has the capability to alter the WOT map as well, based on the revised long term maps which are altered while at partial throttle conditions.

Just because it's running WOT and from a map, instead of the sensor info, doesnt mean that it cant use the sensors to establish trends, and alter the map to suit.

I dont think anyone can say for sure though, either way.

Personally, I think it does learn, because of:
The autoshop101 documents saying that this is how it works
The altezza guys having their AFCs tuned out
And some things that I noticed with my redtop engine.

The pipe that my AFM is in, is bigger than the standard cross sectional area of the pipe.
When I first started it, it ran rough as guts, and would hesitate and splutter when the throttle was opened.
However after running for 1/2 an hour or so, it came right.
After this I turned the car off.
Next time I started it, it fired up and idled and revved fine from the start.
After this I disconnected the battery to work on something, and when I started it up again it was running rough again.
I took it for a drive, and it was hesitating slightly again. (only drove for a few mins)
The next time after that, it was running great when I drove it.
Perhaps there is some other reason for this, but the facts seem to point towards an ECU that adjusts itself.

However this is a relatively modern engine, AFAIK one of the first to come with DFI, VVTI, etc.
It wouldnt surprise me if it got an IQ increase to suit the rest of the motor.

I also wouldnt be surprised if earlier ECUs didnt have these capabilities, especially something made when commonplace fuel injection was in its infancy.
(And at a time, where emissions regulations werent so strict.... Why spend 50x longer making an ECU that performs 3% better?)

I also wouldnt be surprised if the increased intelligence of ECUs is not only thanks to better hardware, but also increased technology, eg. Toyota being able to make complex simulators of engines, to help design and 'debug' their ECUs. (Which perhaps becomes more necessary, as they become more complicated?)

However all of that is mere speculation, dont quote me on it.

[RomanV]

BTW....

Adamal, didnt you say that there's a knob or dial or something on your car, which leans or richens the mixture?

If this is the case, then isnt this whole thread rather redundant?
As obviously turning the knob richens or leans it.

Interesting discussion none the less however.

[sergei]

knob exsist only on NZ/Australian New vehicles which come without oxygen sensors.
The only other screw which is responsible for mixture adjustment on EFI cars, is on the AFM, and that only adjusts idle mixture.

And that incremental adjustment at WOT is due to ability of ECU to compensate for sensors which eventually go out of calibration. It is long term fuel trim.

[RedMist]

RomanV, I doubt very much that a consistant enrichment of a zone in the 3k mark will alter master fuel. But hey this argument could go around and around for a very long time as I dont know of anyone to have cracked the ECU code.

The way standard ECU's work is they generate an overlay map which is then added to the original map which is of course added to a single master fuel value. These overlay values must be within certain bounds, so as to prevent sensor failure screwing up a map. So total injector time is calculated from a host of factors including both the master fuel, baseline map and an overlay. A modification to one zones can only alter (through interpolation) up to the sorrounding 8 zones. As such if we manage to hold 7.5k rpm, just off WOT for enough time, the zoning for that zone, and possibly up to 8k (I dont know the column spacing) will be altered. The reason we ignore data at WOT, accl pending fuel and cold, is the narrow band O2 sensor is just too inaccurate, or we dont wish to achieve stroch in those circumstances.

[pc]

Yup... if you can't get the ECU code/program then you would have to put it on a bench and connect it to your PC... with the PC creating sensor inputs for the all of the sensors and logging all the outputs the ECU generates.
With some good test software, logging, and analysis you could determine the logic that the ECU is programed with, and therefore what it will do given a set of circumstances. If the ECU 'learns' over a long time then even this sort of testing could be difficult.
Without something definitive like this, we just have idle speculation... pointless.

[sergei]

The only problem that the rig which will emulate sensors will cost as much as motec...