ECU learning issues

[BSOD2600]

Ever since I swapped my engine/performance stuff from my '94 touring wagon to the '94 sport sedan, the ECU just hasn't been running at 100%. For swap, I transplanted the TW ecu into the SS, which people assure me is ok. I'm running the Revtronix stage 2 v2 chip, which might attribute to some of these issues too.

Several weeks ago, I noticed that the ECU just wasn't adapting to the fuel trims during WOT and in fact, was getting richer over the weeks (ended up in the 9.5 AFR range). This isn't right, since I was in the 11.5 AFR range in the TW. Finally I got a CEL, code 22 - knock sensor and it was replaced. I thought I noticed a little better performance, but it really didn't change much.

Few days ago, I decided to switch from using my MBC and go back to the BCS (never actually used it in the SS). I would watch the boost gauge needle oscillate around 13-15 PSI and hear the air hissing pitch change from the engine bay, so fairly sure it was working. After a few miles of driving and WOT runs, the BCS disabled itself. Boost wouldn't go above ~8 PSI, which I assume is TD05 wastegate pressure. According to Revtronix, this is because the ECU is in fail safe mode. Looking in the FSM / BBS ECU Codes document, nothing in fail safe mode would actually do this specific thing -- unless it's tied to fuel cut (which I'm not experiencing) that the revtronix chip is ignoring (but I doubt that since I have gotten fuel cut with their chip).

I digress.

Finally I got my hands on a laptop so I could do some data logging. Just as Revtronix suggested, the 'spark learn factor' was well below 25% (which means fail safe mode) -- it was at 0! Other key sensors like the TPS, temperature, MAF, and O2 all appear to be within normal operating ranges.

I attempted to reset the ECU (test mode and read memory connectors connected, ignition on, throttle full/half way depressed, drive for 1 min, bla bla ba). Except, the whole time I was driving the CEL was flashing solid with no breaks. I also tried entering just D-check (green connector only, throttle full/half way, then engine on). With the ignition on, CEL is solid (good according to FSM), engine on, CEL is off (good again), yet soon as I drive for 2 seconds, the CEL is back to a solid flashing. Reading further into the FSM, this appears to be normal and means a lack of any codes. Even though I didn't see a change in the flashing CEL while driving in clear mode, I think it was cleared, since the BCS was enabled again (aka getting boost above 10 PSI). Yet, a few more miles of spirited driving on boost, the BCS got disabled yet again.

Many months ago, I recall that when I had the test/read connectors connected and drove, the CEL wasn't on at all -- until the ECU was cleared and then it was just an occasional flash. But now with the ecu flashing solid while in d-check or clear modes, how will I know when the ECU has actually been cleared?

Basically, I'm trying to figure out why the ECU is in some sort of fail safe mode, even though there are no CELs'?

[ericem]

the car you bought is it original 5spd or was it a auto converted ? Wonder what tells the ecu its in auto mode and how it will effect the stock ecu.only other suggestion is swap ecu.

[log1call]

The workshop manual is talking about the original subaru ecu, you have a revtronics, you need to refer to their recommendations surely.

[klin7757]

if your ecu is reset the cel should be flashing, if not, there is still an error in the system. the bcs is disabled due to excessive knock. if you've replaced the knock sensor with a totally new one i would check to see if there is an issue with the sensor wire downstream.

[ScottyS]

The workshop manual is talking about the original subaru ecu, you have a revtronics, you need to refer to their recommendations surely.

Clarification: the Revtronix chip plugs into the original ECU board to provide new mappings, it does not change how the ECU works.


BSOD - it's not clear to me if you have the correct ECU setup for your drivetrain - are you still using an auto tranny? You should be able to fully reset the ECU if all the wiring is OK, and go from there.

[log1call]

Well if the knock correction isn't doing anything and you have a knock trouble code.... it would seem you have a faulty knock sensor circuit.
If the circuit was connected and there was knocking the knock correction would be occuring and you would get a negative figure on the knock correction.

The long term fuel correction and the boost pressure may come right by them selves if you fix the knock circuit.

[wtdash]

How many miles are you allowing for 'learning' before going WOT?

FWIW, my experience w/my Rev chip is it takes 50+ miles. I put about 25 miles on the chip, the 'spark learn' was @ 80%, I went WOT, and it dropped to 50%, or so. And it ran poorly. So, I had to start over - staying off boost, and just easy cruising for another 50 miles.

[log1call]

Gee you guys, now you have got me confused. I don't know much about these revtronics but you tell me they just alter the fuel maps but the ecu still behaves like a standard subaru one. A standard subaru doesn't have a long term spark correction. It doesn't in the JDM ones we mainly have here in New Zealand, nor in Toyotas, nissans or anything else I am familiar with. Spark correction is instant and temporary. Likewise, the standard subaru ecus learn fuel correction to compensate for wear in the motor and dirty filters etc, but they don't apply it to WOT do they? Not as far as I know. The long term fuel correction only applies to low and mid loads,at wide open throttle the fuel maps are set to run safely rich. If the revtronics do it some other way then I'd say they are not the same as he original subaru ecus.

As I understand it, if the knock sensor trouble code comes on it either means you have a defective circuit or the knock correction has corrected all it can and there is still knocking going on... too many planes of the head, too much boost, too hotter motor, bad fuel or something like that. If there is no knock correction showing in the laptop and the code is set then the circuit is crook. The knock code will limit the boost. The fuel will probably come right when the first problem is cured.

If your cars over there do have some other system of engine control I'd be keen to hear what it is and how it works.

[ScottyS]

log1call, Revtronics does not replace the ECU, it simply is a memory chip that plugs into the existing processor. As far as I know, the Subaru ECU that we have will take a long time to raise the spark correction once Safe Mode has been entered through knock or some other factor.

[log1call]

Hi, yes I know the revtronics is a chip, but that chip has the entire operating instruction set and all the maps in it as well. If it only has different maps, and all the operating instructions are the same as the original subaru ecus then there is no long term timing correction. The timing is corrected for temperature, load, throttle etc as the motor runs, and the knock correction is instant and alters back to the map straight away as soon as the knock stops.

If the motor enters safe mode it will reset if the key is tured off and back on, unless the fault is still there in which case it will go to safe again after a few instances of the fault manifesting. Depending what the reason was for entering safe mode, it will likely reset after a few seconds/minutes if the fault clears anyway.

[ScottyS]

Hmm, if you want the technical specifics, as it seems you do, the best place to ask would be the company themselves. All we can do is tell you what they've told us, implied, or that we've observed.

[log1call]

What has been observed is that the trouble code is on for the knock sensor and that there is no knock correction occuring according to a laptop. Those two circumstances together tell me that there is a fault in the knock sensor circuit.

I was trying to ascertain whether the revtronics handles faults differently than a subaru to see whether it is worth trying to fix the other symptoms first. From the sound of it, it doesn't, therefore those other symptoms will probably resolve themselves after the knock sensor circuit is repaired.

[BSOD2600]

ScottyS:
Running the stock SS driveline, 3.90 5MT.

wtdash:
Well I've allowed the ECU to learn a good ~500 mi since the whole engine swap ordeal. Then another 300+ mi since the knock sensor replacement (no ecu reset then). Since the recent ECU resets of a few days ago, I haven't driven 50+ mi yet.

log1call:
I had a CEL for the knock sensor on 8/11 and have since replaced the sensor with a new one. I have not had any CELs since that date. Yet, obviously something is unhappy and putting the ECU info fail safe mode.


Last night, on the way home from work and data logging, I put the ECU into clear memory mode. Drive a good 2mi home, all above 35 mph. Entire time, ecu was flashing the steady blinks. Idling in the driveway, I did some data logging of everything. RevScan SparkLearn was still at 0. Shouldn't it be at least at 50 or 75% right after a reset?

Mike @ Revtronix still thinks it's something with the knock sensor which is causing this all to occur and wants more data logging...

[ScottyS]

Why not use the ECU that came in the car instead of the one that came in the auto? I would start there, do a reset of that ECU, and see what you get.

The CEL blinks steady once clear memory mode has been entered. After that, drive home, shut it off, and disconnect the test plugs.

[log1call]

The knock sensor may not be faulty... the wire going to the ecu may be though. Have you checked the wires all the way to the ecu?

As I have said, I'm not really familiar with the revtronics chip, but I am familiar with how these cars work, how these ecus work.

Revtronics say that their chips have all the standard subaru features and systems, therefore as far as I know the knock/spark timing should behave as a normal subaru except for the rate at which it pull spark timing, but not when it pulls it, which means that the spark at idle should be about twenty odd degrees, it should advance with revs and, when the cars comes under a load which causes knock it should show a knock correction figure.

The knock code comes on, so either it's got detonation which retarding the timing can't correct or the wiring is faulty. The code comes on but the knock correction isn't showing as happening so I'd assume there is a wiring fault. If it was knocking which the retarding cant compensate enough for,(bad fuel say, high tempretures, over boost or too higher compression), it should be showing full correction which in a standard subaru is about thirty-five degrees.

[ScottyS]

It seems obvious because it's been covered elsewhere, but the wiring for the knock sensor at the white connector on the firewall side of the connection has a tendency on these vehicles to fray and even separate over time. Check this connector and take it a part and re-solder if it looks at all suspicious.

[log1call]

The knock correction figures should only show a figure while the knock is happening, it should only show a figure at all for a few seconds because the knock correction process should stop the knock. If the motor is in good nick you should only get knock under heavy load, high IPW%, low revs. If you do get a knock correction figure that persists then you have a problem like bad fuel etc.

Can you post your data log and I will see if I can see anything suspicious.

If I get a chance later I will log some knock correction and show you what it normaly looks like.

[asc_up]

It seems obvious because it's been covered elsewhere, but the wiring for the knock sensor at the white connector on the firewall side of the connection has a tendency on these vehicles to fray and even separate over time. Check this connector and take it a part and re-solder if it looks at all suspicious.

I was also having this problem, so I completely re-pinned the connector at the ECU with a new sensor wire and ran it to the knock sensor; it fixed the problem completely. If you do this, be sure to use a shielded wire (such as coax cable) and run the knock sensor ground pin to the shield wiring.

[BSOD2600]

ScottyS: Yea, I'm strongly thinking about putting the stock SS ECU back into the car and installing the revtronix chip in that. The white KS connector doesn't really seem that frayed. Going to see what Mike @ Rev says about these logs first.

log1call: From the data logging tonight, I believe the KS is working correctly, since the ECU only sees knock under heavy load. Would suggest to me that it would eliminate a frayed wire, jiggling connections, etc, right?

ECU was reset ~10 mi ago. I spent the evening doing data logging with RevScan. I think I either broke something now or the ECU is in a serious bad state of learning.

Datalog-2009-09-07.xls is the datalog.

Narration from each of the various data pulls:

• coast1: Coasting down a long hill and driving along flat road, right after cold start
• pass 1: Pass #1 up steep hill. 1st - 3rd gear. 1-3 degrees of knock retardation logged during WOT.
• coast 2: Coasting back down the steep hill
• pass 2: Pass #2, same hill, 1st-3rd. This time, BCS disabled itself (or it did that end of pass 1 -- I forget). No KS logged this time.
• coast 3: Enabled all logging, coasted down a different / long hill. At the bottom in 3rd, went WOT and BCS re-enabled itself again.
• pass 3: Pass #3, same steep hill as before, 1st-3rd. No KS logged, but various boost DC learned values are changing.
• pass 4: Pass #4, same hill. This time on the shift from 2nd to 3rd, there was a sudden power loss the remainder of the hill. Knock retardation as high as 7 was logged.
• coast 4: After the power loss from the last pull, re-enabled all logging and drove/coasted down another long hill. Once again, at the bottom, I went WOT in 3rd and there was jerking/power loss.
• coast 5: Turned car off and back on again. Back to original logging. Coasted down a hill.
• pass 5: Pass #5. Attempted to WOT up another hill, but lots of power loss, hesitation, bucking. Had to downshift back to 2nd. No knock logged.

After that last bad pull, I left it idle in the driveway and popped the hood. Noticed the ticking was louder than normal. Got a long screw driver and 2 of the 4 injectors seemed noisier. These are the 440cc injectors I purchased from the japan ebay dude, which claimed they were already serviced. Otherwise, all seemed ok since the engine has always had somewhat of a tick.

Thoughts?

[log1call]

Good stuff. I just had a quick look and it will take a while to come to terms with it all because your nameing convention is different to what I am used to. I did notice though that it seems to go really rich where you say it missed, might be the result of something or it might be a cause. I'll have a look later tonight and see what I can figure though.

I got out and did a quick run up a dirt road and have posted my results here... http://cid-4ca3c3459aaa7f7f.skydrive.li ... ublic?uc=1

I dont know how to uplaod them to here


None of your knock correction figures look bad enough to cause a trouble code... look at mine! I was thrashing/abusing it up the road in third gear though. Your ignition timings in general actually look rather extreme. If those advance figures were true, your motor would be dead by now!

Oh, my throttle position sensor goes the opposite way to yours by the way. Another cause for confusion.

[BSOD2600]

I would be logging my LC-1 if revscan could do that. I also have evoscan, but since revscan wanted me to do data logging for them, figured it would be better to use their software .

As for the rich O2 and extreme timings, yea that's something I've noticed with the revtronix chip. I assume that's where their changed fuel maps / other come into play over the stock maps.

Your Throttle position sensor (TPS) goes the other way? I thought all of our cars the TPS reads ~4.7V when closed and ~0.9V when fully open -- via the ECU. If you're talking about my TPS (Scaled) column... that's something Revtronix threw in there which I assume calculates it off the acceptable range of the min/max for the TPS voltage. They stay quiet on how all their data is calculated, which bugs me. If you want to dig, revscan.dat contains the location they pull things from the ECU and I assume the calculations... I hate reading binary data though.

[log1call]

I've just had another look at those figures.

I presume that that "load-G/rev", is airflow?

I notice that where you say it ran bad in pass 4, the airflow was consistant, well fairly much as I'd expect it to be for the revs, and the knock correction seems to be doing it's job. It would have been good to have the O2 and the base and ignition timing reading right then.

In coast three when you floored it things seemed to behave correctly, ie timing retarded with knock. The IPW went up and the O2 readings with it, but then the O2 dropped off before the IPW did, I thought that was suspicious. It's always a bit hard to be sure because the runs are short, there are anomilies, lags in recordings etc, you know how it is, but... could you have a dirty fuel filter by any chance?

I'd try recording revs, speed, throttle position, MAF, IPW, O2, knock, base and ignition timing, then when it started to miss, I'd try to keep it missing, for long as I could, I'd drive it to a standstill if I could, so I got a long record.

As you will have noticed, there are nearly always lags between changes of throttle and mixture changes being recorded, a long steady missfire will show up better after a few seconds of labouring.

The knock does seem to be working, well it's recording figures and the timing is retarding.

I'd suspect fuel problems myself so I'd try to devise a test drive and record sensors that might detect that.

ANyway, it's been a long day so i'm off to bed. I'll have another think about it there.

I would be interested to see which addresses the revscan logs from but I couldm't figure where that revscan.dat file would be. If you could direct me to it, it would be appreciated.

[BSOD2600]

...could you have a dirty fuel filter by any chance?

Possibly. I haven't replaced it on this car...

I would be interested to see which addresses the revscan logs from but I couldm't figure where that revscan.dat file would be. If you could direct me to it, it would be appreciated.

You have to open it up in a binary text viewer; I used Visual Studio. Then once the file is open, you'll see the bits of parameter text you recognize. I forget how many bytes after the text was the address they're using... it's a pattern. Most of them are the same which I put into my EvoScan file (and posted it in that thread), but several are not there. Not sure if they're parameters which b10scan missed or something they're calculating instead.

[log1call]

Hi, yeah I can read the dat file, but where do I get a copy of it?
I looked in the revtronics site and did a google... nothing. I thought I'd had revscan once as well, it sounds familiar, but I can't find it in my software now so I must not have.

When interperating these logs I have always found that the behaviour and relationships between figures is far more telling than the figures themselves. So apart from what the figure/number is, how it reacts to some other figure tells a lot. It's also important to realise that the times between figures varies and to take that into account when interperating them. In this case, even though it's only one figure, in one columm, that is out of place, it's significant because of the relatively large time factor. It means it probably isn't a glitch in readings or a delay in the O2 sensor reading.

Looking at "coast three" logs 56 to 60... See there is a delay of two point three seconds between each reading(approx), and at fifty-nine the throttle is still down, the IPW has gone up higher than ever, but the O2 reading has fallen right off.... So after only seven seconds of full throttle it seems like you are running short of fuel.

If we assume it is a fuel problem, and ignore the ignition/timing for a bit, you could log only TPS, IPW and O2, and go push it till it starts to die out, then just keep your foot there and see what happens to the car and the logs.

You are lucky you have clear roads there by the look of it. In some situations it is hard to get the pedal down for more than a few seconds, or to find a hill that takes as long as you have there to get up.



Now, off topic but... I like the way your spreadsheet has the discriptions at the top locked so when we scroll they stay visible. I do that to mine every time I open them using the "freeze" feature, but I have never yet figured out how to save that so it will do next time I open the spreadsheet. Can you explain how you do that?

I'd also like to know how you get those links to your spreadsheet into your posts. I have read and read and tried all sorts of things here and in other forums and I can never get it to work!

I repair computers for a sideline to me mechanicing, I can connect laptops to all sorts of cars, I can usually figure things out but I can not figure those two things out even after several years of trying. Crazy huh? So... If you can explain in simple terms how to save the freeze feature and how to post spreadsheets I would be ever so grateful.

[BSOD2600]

You have to email revtronix for a copy of it... but I've got a copy mirrored here.

As for the "coast 3" run, I'm logging so much data that it takes a good ~2+ seconds for it to get all the parameters. By that time, the data has surely changed a bit. Simply not enough resolution to any sort of granular troubleshooting. Initially I logged everything to just get another baseline of driving, not to capture a problem. I looked at the fuel filter today and the canister seems newish, but would still be a good idea to replace it. Although, last week before I was having these problems, messing around with the BCS, etc WOT runs on the highway, the wideband was seeing down to 9.5 AFR, so I'd think that means the fuel is flowing quite well .

Per Mike @ Rev, I pulled all my sparkplugs today. All looked the same. Can post pics if wanted. As for the O2 sensor drop, I've noticed that more and more in the logs that it reads 0, instead of 0.01 - 0.1 mV when in a lean condition -- think that's a sign it's old/time for a replacement? Otherwise, it seems to react and be in spec for rich/lean conditions. Mike also noticed that the Spark Learn Factor was still at 0, instead of 75%, which means something is preventing the ECU from getting reset properly...

------------------------------------
In Excel, I used the Window -> Freeze Panes feature in each worksheet. Then saved it. Nothing magical. Maybe it's a version issue? I used office 2003.

As for making the links in my posts, it's called BBCode.

I'm actually a system admin / programmer by profession so all this computer stuff comes easy. It's the mechanical side of cars which some times irks me