Rough Idle - Engine Dies and Won't Idle after 15 minutes of Hot Idle / Traffic

2007 Subaru Legacy Special Edition 5MT 220k Miles

This has been going on for about 2 months now:

If the vehicle sits in traffic moving slow, or a drive-thru or in a hot idle for a while (~15 minutes) it will start to develop a rough idle.

It will begin to feel like it is noticing a miss and catching that and idling up, increasing fuel trims as a result. Then as it tries to get back to a steady state it will miss, RPM's will dip, idle up, fuel trim up, and the cycle continues.

This cycle continues to go at an increasing rate until eventually the computer can't keep up and the engine dies. If you restart it it can't idle on it's own and you have to keep it >1500 rpm to keep it running.

Some investigation I've done while hooked up watching live data just shows the fuel trims going crazy showing the computer trying to keep the engine going. (No codes btw, just get an EVAP code often after I replaced the gas tank and filler pipe a few years ago, I know where the evap leak is from a smoke test, just too lazy to pull it all out and replace it again, thanks Rock Auto).

I have also replaced the coil (Was losing Cylinders 3 and 4 randomly and getting misfire codes for them) for a few months before this issue and that resolved that issue.

I also removed and cleaned the throttle body. Plenty of plugs, wires, etc.

This weekend will probably be another compression test.

Previously it had around 130 psi of compression on my cheap harbor freight compression tester with a weak battery and not all the spark plugs removed at the same time so probably not a high rpm either.

The only thing I have noticed watching the data is that it seems to be about when the intake air temps reach about 140-150°F the issues start to happen. STFT and LTFT otherwise aren't too crazy beforehand, everything seems normal and no codes.

If anyone has any ideas of anything to check, they would be greatly appreciated!

Have you tried cleaning the maf sensor hot wires/resistors that are kind of buried in the sensor?

Have you tried cleaning the maf sensor hot wires/resistors that are kind of buried in the sensor?

I have tried cleaning it with just a general MAF cleaner, but I will try a parts store replacement MAF sensor (I am thankful for generous parts store return policies if it does not fix it).

crankshaft sensor...

But why only after a long hot idle? It's fine all the time otherwise and fine just off idle.

Further investigation has shown me that if it starts to get crazy I can shut the engine down for 5-10 minutes and start it back up and it'll idle fine for a further 5 minutes or so before starting to rough idle and die again.

I know when the engine in my wagon did that type thing, it was the small black gaskets between the TGV's and the heads. The new larger orange seals fixed it.

Not sure your engine has them ?

I know when the engine in my wagon did that type thing, it was the small black gaskets between the TGV's and the heads. The new larger orange seals fixed it.

 

Not sure your engine has them ?

+1 to what Max said. Had the same issue with my old Civic, leak in intake manifold gasket. And few years ago my wagon had somewhat similar but not that severe issues with idling until I had the intake gaskets replaced.

Some pics from my son's 05 OB rebuild, should not be too far from your 07. There are at least 3 places where air can get in the intake, huge paper gasket between throttle body and intake manifold, and then two sides of intake manifold to the heads.

Some pics:

https://drive.google.com/file/d/0B9dWWCx5l4RAbS05Q3lJWm9WTDA/view?usp=sharing&resourcekey=0-hIXj7FFy5LI-1-JU0Mn9dQ

Intake manifold to heads (I believe that is what Max was talking about):

https://drive.google.com/file/d/0B9dWWCx5l4RAc25kSnAwbE03VEU/view?usp=sharing&resourcekey=0-rP2MYQdsQGbOrHWpg0oc2g

I know when the engine in my wagon did that type thing, it was the small black gaskets between the TGV's and the heads. The new larger orange seals fixed it.

 

Not sure your engine has them ?

Yeah it does not, but it still does have that block, and the upper and lower gaskets. I only replaced the lower gaskets (I believe the ones you are talking about) when I did the head gaskets @130K with OEM ones.

+1 to what Max said. Had the same issue with my old Civic, leak in intake manifold gasket. And few years ago my wagon had somewhat similar but not that severe issues with idling until I had the intake gaskets replaced.

It's interesting because I didn't think about a vacuum leak because it felt like a miss and the idle up was the ECU correcting and trying to keep the engine running as a result though. And the fact that that this only happens after a long hot idle.

Some pics from my son's 05 OB rebuild, should not be too far from your 07. There are at least 3 places where air can get in the intake, huge paper gasket between throttle body and intake manifold, and then two sides of intake manifold to the heads.

 

 

Some pics:

 

https://drive.google.com/file/d/0B9dWWCx5l4RAbS05Q3lJWm9WTDA/view?usp=sharing&resourcekey=0-hIXj7FFy5LI-1-JU0Mn9dQ

 

 

 

Intake manifold to heads (I believe that is what Max was talking about):

 

https://drive.google.com/file/d/0B9dWWCx5l4RAc25kSnAwbE03VEU/view?usp=sharing&resourcekey=0-rP2MYQdsQGbOrHWpg0oc2g

 

 

 

 

Yeah so I just replaced the TB gasket when I cleaned it last weekend, the upper and lower gaskets I can replace if the MAF doesn't fix it.

I appreciate everyone's suggestions. If the MAF doesn't work, I'll swap out the upper and lower intake manifold gaskets as well and see what happens. And I'll get the compression test results too when I get the chance to do it as well.

Quick update:

New MAF had no impact.

I watched the MAP sensor readings this time and noted that it lost vacuum for a snap second every time it missed but it returned to normal when the ECU tried to correct and raised the engine speed. To me I would expect the vacuum to go down if the engine rpm went down because of a miss, so that makes sense. But if it were a vacuum leak because of an intake manifold gasket leak, I would expect engine speed to increase and vacuum to decrease, but I saw the opposite.

To me this seems to be more electrical still. But that's just my opinion.

Front O2 sensor. Overheating and failing causing ECM to over drive injectors.

OP seems to be able to monitor live data, if so the O2 sensor can be monitored too. Interesting if that's the case.

To OP: Back in the day I recall replacing spark wires on 05 OB to cure some stumbling issues and that did help. Then later on replaced ignition coil which made no difference. Not sure if in your case coil the culprit but parts store return policy should let you try that too.

Front O2 sensor. Overheating and failing causing ECM to over drive injectors.

I suspect you are correct.... Thanks to more parameters from FreeSSM I was able to see the Injector pulse width go crazy high each "cough" and go higher and higher as it progressed until it died. It also looked like it was giving crazy current values (positive and negative. therefore lean and rich). It looked to be telling the computer it was lean when it was adding pulse width. So I'll be ordering a primary O2 and hopefully that'll be the issue, I'll post an update for sure! Thank you very much for your suggestion!

OP seems to be able to monitor live data, if so the O2 sensor can be monitored too. Interesting if that's the case.

 

 

 

To OP: Back in the day I recall replacing spark wires on 05 OB to cure some stumbling issues and that did help. Then later on replaced ignition coil which made no difference. Not sure if in your case coil the culprit but parts store return policy should let you try that too.

So this was an initial suspicion due to my strange issue with the coil last year when losing cylinders 3 and 4 on the highway at steady state cruising, that was resolved with a new NGK coil.

Fuel pressure? I have seen the O2 do exactly what you describe without setting a CEL, and my first place would have been there. Front O2 can also cause surging, showing STFT and LTFT bouncing all over the place, again, without setting CEL.

I would check fuel pressure and make sure it's consistent.

The other weird thing I've seen is TB getting very hot and stretching and slapping around inside the covers. I've seen it on a couple of Subarus and also on a couple of other cars, so depending on how old your TB is, that might be something to explore.

Fuel pressure? I have seen the O2 do exactly what you describe without setting a CEL, and my first place would have been there. Front O2 can also cause surging, showing STFT and LTFT bouncing all over the place, again, without setting CEL.

 

I would check fuel pressure and make sure it's consistent.

 

The other weird thing I've seen is TB getting very hot and stretching and slapping around inside the covers. I've seen it on a couple of Subarus and also on a couple of other cars, so depending on how old your TB is, that might be something to explore.

What's strange is it was doing exactly as you described, injector pulse width was going up as it was receiving strange values from the primary O2 sensor.

Unfortunately I haven't found a way to record data with a fast refresh rate yet to be able to share it here.I don't see an option with FreeSSM and with the Torque app I only saw an option to export a screenshot of a graph of data but I didn't spend much time investigating that avenue.

Also I would mention that the catalytic converter is dead, not sure how that could be ruining a primary O2 at a long idle though.

Yes fuel pressure is another thing on my agenda to check along with general engine health with a compression test, plus I'll reconfirm with FreeSSM that the injector pulse width and primary O2 sensor readings are going crazy again, even with the new sensor.

The TB was something else I was considering, that's why I initially pulled it off and cleaned it but short of going to a junkyard and pulling one I'm not sure what else I can really do (a new one is like $600).

What O2 sensor did you install? Some aftermarket never work right from the beginning, this is on of the parts that should be OEM.

I installed a Denso sensor, who's probably the OEM supplier (and also my employer, so I do trust it).

Just FWIW I recently replaced bank1 O2 sensor on wife's Tribeca that I replaced 3 years ago with aftermarket when I did the HG on it. It ended up being the cause of elusive valve knocking that never threw any codes but was clearly heard from the cabin. Finally it died and threw a CEL. It was Denso from Amazon. Replaced it with OEM part from Subaru, car seems to be happy.

In your case if the life data is still showing 02 sensor readings being abnormal, its either sensor itself or the wiring to the ECU from said sensor. Check the connectors for any damages like bent pins, or dirt in the connector, or cracked wires and such. But since the issue seems to be temperature related the O2 sensor remain the main suspect.

Dead cat would throw P0420 but other than that should not affect engine operations much less make it stall.

Just FWIW I recently replaced bank1 O2 sensor on wife's Tribeca that I replaced 3 years ago with aftermarket when I did the HG on it. It ended up being the cause of elusive valve knocking that never threw any codes but was clearly heard from the cabin. Finally it died and threw a CEL. It was Denso from Amazon. Replaced it with OEM part from Subaru, car seems to be happy.

 

 

In your case if the life data is still showing 02 sensor readings being abnormal, its either sensor itself or the wiring to the ECU from said sensor. Check the connectors for any damages like bent pins, or dirt in the connector, or cracked wires and such. But since the issue seems to be temperature related the O2 sensor remain the main suspect.

 

 

Dead cat would throw P0420 but other than that should not affect engine operations much less make it stall.

While I agree that the heat and O2 data is still suspicious, I find it strange that changing the O2 sensor made no change, so I can only suspect the wiring, pins and ECU, etc. Not the sensor itself.

I was able to take a log with Torque, just not with Inj pulse width, for anyones' viewing pleasure, formatted into a CSV. RPM is high at first with the AC on (It's very hot in southeast TN at the moment).

Basically everything goes haywire, not just O2 sensor readings, but this is to be expected as the engine is dying and the ECU does whatever it can to try to compensate.

However, as I saw before The O2 sensor (and calculated A/F ratio) goes lean right before RPMs drop, then I'm assuming pulse width goes high to compensate, then RPM goes up eventually, then it's rich as a result, then the cycle continues.

I can confirm pulse width width with FreeSSM again, just can't log with it (That I know of).

Anyway, if anyone has any ideas from this data, let me know.

https://drive.google.com/file/d/1ZaO46GIZBrr8L1AjnaRVLghow0x6aT0e/view?usp=sharing