2005 GT Rough Idle

[lagwagon]

Hi Guys,

 

I'm having a rough idle issue that I need some help with. Here is the run down:

 

2005 GT manual, Accessport V2 on stage 0 OTS map.

 

In the summer, AF Learning C value was really negative so I had the mechanic smoke test it and it was leaking from a few places. I had them replace all the seals from the turbo inlet to the TGV seals and the boost leak went away but it seemed like it now had a slight vacuum leak.

 

When it is cold out, the car idles fine on cold start until it hits some point (I assume it is cross over from open to closed loop). Then it idles like crap until it is fully warmed up (coolant, and general engine bay temp needs to get up as well).

 

AF Learning A was pretty high and was swinging from 0 to 18 over the course of a few drives in a never ending cycle.

 

My mechanic did a smoke test, pressure test, fuel pressure test, propane test and full diagnostics. The coolant temp sensor was out of range so they changed that out but there was no vacuum or boost leak. They did say the car stuttered when they propane tested near the front O2 but that makes sense to me since they pull calibration air through the wire harness, right?

 

They also did a fuel system pressure test and service starting at the pump housing (not just a pour in cleaner).

 

Now the car is tending to run really negative AF learning values: often correction will be at +8, Learning A will be at -8 to -10 (currently about -3.5, so I think it's mid cycle again), and the AFR will be in the 13s at idle, especially when I come to a stop, AFR will drop and hold at high 12s low 13 for 5-10 seconds. Learning B is at -14.8 right now, C and D are fine.

 

It's got me a bit confused as to why AF correction is so high when the car is running super rich at idle.

 

I did some data logging but was having problems with losing connection so I reset the log list and wasn't able to log all four AF learning values or AVCS.

 

Anyways, with the data available, can anybody see anything obvious going on here?

 

Log 7 is just general driving, log 8 is a third gear pull from 2500-6000 rpm.

datalog7.csv

datalog8.csv

[Pleides]

Have you replaced or cleaned your mass airflow sensor?

 

I replaced mine shortly after I bought the car and it solved a few issues. I've been told my a reputable Subaru tech I work with that he replaces the MAF sensors on old turbo Subarus every time the timing belt is replaced.

[lagwagon]

Have you replaced or cleaned your mass airflow sensor?

 

I replaced mine shortly after I bought the car and it solved a few issues. I've been told my a reputable Subaru tech I work with that he replaces the MAF sensors on old turbo Subarus every time the timing belt is replaced.

 

I put in a known good one from my other car after confirming no difference in operation with a brand new one. I did clean it as well. Swapping the new one in doesn’t change anything. I haven’t ruled out a wiring issue with the maf but wiggling the wires doesn’t seem to change anything.

[Pleides]

OK, good to know. I've replaced my front O2 sensor and that helped with some issues with what felt like misfiring in my car that left no check engine light. If your shop (which sounds like a pretty good shop with the types of tests they've been running) found something awry there, that might be a good place to look.

 

Is this a dedicated tuning shop or a shop that otherwise can boost leak test above 10 PSI? Some mom and pop shops do tests on our turbo cars at 3 to 5 PSI and find nothing without knowing you need to load the car up with as close to factory boost as possible to diagnose. I had to use a tester we use for turbodiesel trucks at work to vacuum test my car awhile back because I, like an idiot, kept pressurizing my car up to 5 PSI after a rebuild. It seems like your shop knows what they're doing, but thought I'd ask.

 

Also, what is your DAM reading when shit hits the fan?

[lagwagon]

DAM is 1 all the time, no changes there at any load or rpm. Fine Knock Learn and feedback knock correction are both 0 all the time. There is some Dynamic Advance at high load, but I don't actually know what that means, it gets as high as 5° and spark timing peaks at 41° at high load.

 

Several of these guys used to work at diesel tuning shops so I think they understand the value of higher pressures during leak tests but I haven't actually asked what pressure they take it to. The problem only seems to be at idle anyways so I wouldn't expect a high pressure needed to find anything. I actually thought it was a vacuum leak until they changed the coolant temp sensor and the learned values flipped from positive to negative, but still very high.

 

I do have a new O2 sensor but I was a bit too tired to install it this weekend and I want to try and get a log of the AF learned values and AFR before swapping it out so I have before and after data.

 

Another weird thing is that the AF learning 1 doesn't match AF learning A, B, C or D when I enter logging view (logging or not) but does match when I enter live data. You can see this in the logs as well, where AF learning 1 uses a huge range of values, instead of just the 4 learned values. It seems like entering logging view starts to mess up the computer or something.

[lagwagon]

I replaced the O2 sensor last night. The sensor is contaminated. Either some sort of sealant or possibly coolant. It's very white and flakey, but no green crystals I would expect from coolant.

 

Also found the ground strap near the front O2 is missing and the heat shield has a sheet metal screw between it and the manifold and other tears in the metal and contact points with the manifold. Overall a bit of a mess down there.

 

I think there was actually some mechanical damage to the sensor and/or it wasn't screwed in all the way since the car sounds way different now.

 

I didn't data log last night but AFRs were pretty steady at idle and while driving, very nice to see compared to the old sensor that oscillated like a narrow band sensor. AF Learning B started to come down as well (only 15 minutes of drive time so only dropped 1%.

 

Now I wonder what the EGT, pre-cat, cat and rear O2 sensor look like and what actually caused the contamination...(don't yell at me for still having the EGT and pre-cat at 250,000 kms)

 

EDIT: also, Denso 234-9120 is not an exact replacement sensor. Harness is about 4" longer and the attachment clips are not in the right locations and can't be moved without destroying them.

Edited February 2, 2021 by lagwagon
more info

[Pleides]

https://www.walkerproducts.com/wp-content/uploads/2020/06/O2-Silicon-Contamination.jpg

 

Did it look like this? Supposedly that's silicone contamination.

[lagwagon]

It looks more like the coolant contamination picture on walker’s site. I’m having issues with mobile uploads, I’ll try to post pics tomorrow.

[lagwagon]

Here are pics of the sensor. The front of the sensor is very consistent, white coating and the back is a bit flakey. Comparison to another old sensor that also failed but likely due to age.

 

The fuel trims are back to where they were before changing the sensor, despite the AFRs looking more stable, but still rich. AF Learning A is -7, B is -14.4. C and D are -2.3 and -1.4.

 

Comm. Fuel Final is 14.3 but at hot idle the car is running 14.1-ish. For a while, fuel correction 1 and AF learning 1 were both at 0 at idle, then AF learning 1 started to pull fuel and fuel correction 1 started adding fuel. The sum stayed within 2% of 0 and AFRs are consistent at 14.1.

 

I'm not sure why AF Learning 1 would swing negative if correction 1 is at 0. I'm missing something about the logic here but isn't correction 1 short term and drives changes to AF learning if it trends away from 0? Why would they stay in lock step but opposite signs? Doesn't that just mean that the ECU is correcting for an erroneous correction (learning 1) it made? Especially when AF Learning is changing rapidly without any change in correction 1?

 

Can anybody confirm potential causes of negative Learning A and B? Leaking injector, fuel pressure too high, failing FPR, pin hole vacuum leak on the vaccum line to the FPR (too small to cause an overall decrease in vacuum but close enough to the FPR to raise fuel pressure???), MAF problem (its stock FYI so not a scaling issue per se).

 

I've found very limited search results of similar fueling trends, none on legacygt.com, and none with an actual resolution.

 

Sorry, I don't fully know all the Subaru systems like the TGVs, DBW system/IAC system etc. so not sure if there is something obvious that should be checked.

[Pleides]

It's worth considering that tuning a car (if the car has been tuned) on any sort of bad sensor can mean you'll require a retune with a new sensor.

[lagwagon]

I would agree if it wasn’t a stock tune* that there could be a scaling issue with the MAF, but I think there is something else going on to cause AF Learning values and correction values to oscillate almost 180 degrees out of phase, with the learning values being leading slightly, when it should be the opposite.

 

*I know stage 0 isn’t exactly stock but in this case the fuel algorithm should be identical with slightly tweaked table and target values.

[lagwagon]

Bump and small update.

 

I'm now getting apparent rev hang between shifts, which is new.

 

AF Learning B is now pegged at -15%. AF Correction 1 is still fighting AF Learning A. Car still wants to stall when coming to a quick stop, RPMs dropped down under 500. Brand new oil smells like fuel, but I need to compare it to another car to be sure it is more smelly than normal but it certainly smells far worse than fresh oil in the jug and it only has like 200-km on it at most.

 

I might need to have the injectors pulled and tested.

 

Is there a link to a thread with fuel pressure test pieces/requirements that lists the right fittings? I have various NPT hose barb fittings and -4AN pushlock line to run a gauge on the windshield but I'm guessing I need a 1/4" tee fitting and a small piece of fuel line of some sort, not pushlock.

 

EDIT: I guess I need a 5/16" x 5/16" x 1/4" tee and some 5/16" hose.

Edited February 17, 2021 by lagwagon