poor acceleration

[Fairtax4me]

So I cleaned the MAF sensor with the special cleaner, didn't look particularly dirty. No change in performance. Then I tried unplugging the MAF with the ignition off, this did not work. When started, it idled rough and when I tried to give it some gas it died, couldn't be driven.

 

 

This usually means the MAF isn't the problem. With it unplugged you should at least still be able to drive it though it may not run its best.

Tried this just recently on a friends 99 forester and got the CEL and blinking AT light but the car actually ran much better with the MAF unplugged.

 

If it runs worse with the MAF unplugged the problem is likely elsewhere, such as the O2 sensor. You could try unplugging the front O2 sensor.

[alpop]

So I want to drive it a few more times for up to 30 minutes with the knock sensor isolated to see if the observed behavior of - malfunctions 'till its warmed up then drives fine after that - holds.

 

Then I will re-attach the knock sensor and see if symptoms worsen significantly.  If so then its probably a knock sensor/ecu thing, if not then it is unrelated to the knock sensor. 

 

I will then try unplugging the front O2 sensor.  I also still need to take it  on a longer drive/trip with the fuel injector cleaner in the gas to see if an extended drive with the cleaner in the gas will help.

[alpop]

Hmmm, well when I reattached the knock sensor it drove perfectly for 1.5 drives but then acts up intermittently.  Whereas before it always malfunctioned when engine cold, now it can be fine at first when engine's cold, but then act up later.

 

Anyways, things do seem to change when I fool around with the knock sensor

 

Then I looked at the manual (a little late) and saw that you are supposed to torque it to only 14 ft/lbs (this is nothing)  -  oops -  I loosened it and torqued it right but still off and on malfunction. 

 

When I originally put the new one on I specifically remember thinking  - "this thing vibrates around so I better tighten it down good"  heh, I'm sure I way way over torqued it, and from what I've read this can damage the knock sensor.

 

I ordered a new one and will torque it right this time.  Worth a try for $13 anyways-

[alpop]

So put in another new knock sensor and it had no affect on performance.

 

On to the O2 sensor, if outward appearance had any significance this would be it, amazingly dirty and grungy, wire and all.

 

I now have unplugged the upstream O2 sensor and I don't think it changed much of anything - still malfunctions sometimes, maybe a little more than before, but hard to say, of course the CEL came on.

 

Don't know what this means, question is what happens when you disconnect the O2 sensor on a properly performing car?

[alpop]

So after driving it more with the upstream O2 sensor disconnected it malfunctioned (very poor acceleration up to 2500 rpm) on 10 of 12 attempted accelerations on a short in town trip - It is definitely doing the same thing but more of it.

 

Any one want to hazard a guess on what this means?

 

Anyone ever tried disconnecting the O2 sensor on a properly running Subaru and seen what happens?  Guess I could try it on one of my other (non-subie) cars-

[Fairtax4me]

Unplugging the sensor makes the computer default to an "open loop" status, which means it runs on a preset fuel mixture and doesn't adjust the mixture (adjustments to the fuel mixture are called "fuel trim") based on O2 sensor feedback (because it can't when the sensor is unplugged).

This means that if the O2 sensor is causing a problem due to incorrect feedback to the computer, the engine should run better with the sensor unplugged.

The computer doesn't always revert back to the default fuel setting right away, especially if the fuel trim is off by a substantial amount. Sometimes the computer needs to be reset by unhooking the battery for 15 minutes or so to reset the fuel trim to 0.

 

With the problem occurring more often with the O2 sensor unplugged, probably means the O2 sensor is not the problem, but try resetting the ECU and driving with the sensor unplugged again.

[alpop]

ok so disconnected battery, reset computer and as usual it drove beautifully for 20 minutes and also CEL did not come on, then CEL came on and it still drove pretty good.  But tried it this morning and same-old, same-old.

 

I may not have mentioned that when its malfunctioning the shifts are sometimes rough also, when driving beautifully you can hardly tell its shifting.

 

This driving perfectly for 20 minutes after disconnecting the battery should allow us to eliminate certain possibilities?

 

Seems to me it means either 1. the ECU is receiving bad information from one of many information providers (sensors) and reacting correctly to the wrong information or  2) the ECU is receiving good information but reacting incorrectly to it.   So either a sensor is bad or the ECU is bad  - 

 

My tests and replacements have not identified a bad sensor.

 

Are there other possibilities?

 

I guess I really need to have someone in the know hook up a scan tool and drive it to see what is happening when it mal-functions.

[Fairtax4me]

Still a chance this is a knock sensor issue.

 

Go to radioshack and buy some resistors. Wire them up in series to total 560-580k ohms. Plug this get-up in place of the knock sensor and ground the other end to the block. Take it out for a drive and see what happens.

 

Could also be a fuel pressure issue. Have access to a pressure guage you can leave hooked up to it while you drive around some?

[nipper]

OK explain to me how the resistors work, as the knock sensor is a piezo-electric sensor. Basically a peice of quartz that generators an electrical signal that the ECU deciphers as good bad or otherwise. 

 

Not being snarky, but am interested in hearing how this works as opposed to just unplugging it.

[Fairtax4me]

The resistance of the sensor element is about 560-580k ohms. The resistors simply keep the ECU from saying "where the hell is the knock sensor?" and putting the car into limp mode with no ignition advance. Instead you get full advance (based on the timing map of engine load vs RPM).

This is used to rule out a false signal from the knock sensor.

[nipper]

Ok more of a checking it is there then actually trying to send untrue data.

[alpop]

Well I have had 3 different knock sensors on it, the original and two new ones (all read 560-580 k-ohms) so if it is a knock sensor issue it is probably the ECU responding incorrectly to each of them.

 

I did the isolation test (removed knock sensor from engine, left plugged in, grounded it and wrapped in foam and taped to air box) with the first new knock sensor.  I did conclude that things changed when I did this, and then changed again when I put it back.  

 

However this problem is so variable I can't be sure the changes in behavior were really because of my actions.   Today instead of normal acceleration being restored at 3000 RPM, it was closer to 2200 several times.

 

Don't have access to a pressure gauge, why would it do it only at lower RPMs if fuel pressure were too low?

[alpop]

The thing I keep coming back to is that it runs beautifully for 20 minutes when you reset the ECU, then the ECU decides to make things better and screws it all up. 

 

No CEL codes are being thrown legitimately, I only get them when I fool around with a sensor (very consistent - IAC valve, TPS, MAF sensor, etc). 

 

So it appears that the ECU does two simple things it is supposed to:  1. run on default settings for a while when its reset; 2.  Throw a code for a sensor when it is disconnected.  Yet when asked to do the more complex task of monitoring all the sensors and fine tuning timing, fuel/air ratio etc., it fails somewhere.

 

Given the above, is it possible that when resets to the default settings that it could overcome something more basic like a faulty FPR?

 

I will take a look at the FPR and make sure vacuum hose is ok, etc.

 

The fact that (very likely) either the ECU or some sensor is screwing everything up, AND it is very capable of - but doesn't throw any codes for sensors (or anything else) on its own, makes me suspect the ECU more and more.

 

I know - they almost never go bad, however this one may have really been stressed when the PO on purpose burned up the previous engine. 

 

After all, turns out you couldn't trust Hal -

[1 Lucky Texan]

does the ECU send a 5v signal to some of those sensors? maybe that is screwed-up or those sensors have a common (but poor) ground?

[alpop]

I checked on one or two of the sensors and they had the right signal.  They all have ground wires going back into the wiring harness, I assume grounded through the ECU?

 

Not absolutely sure where the ECU is grounded, but I cleaned the likely spot near the firewall in the engine compartment.

 

No codes are being thrown unless I mess with the sensor, wouldn't codes be thrown if ground were bad?  I have shown that the ECU is not hesitant to throw a code when a sensor is disconnected.

[1 Lucky Texan]

yeah, likely to spew codes if a ground were really bad. But, some codes stay 'pending' unless they occur 2 or more times within X drive cycles w'ever.

 

might be nice to see what the fuel trims are if you can get frezeframe or live data. Some smartphones apps can do that with a $20 elm327 BT adapter i think. My old Innova scanner can get FF data.

 

related to other posts - if the car runs better in open-loop or immediately after a hard reset, there's likely bad data from an A:F sensor. or related sensor. O2, MAF, Knock, maybe engine temp, etc.

Edited March 12, 2015 by 1 Lucky Texan

[Fairtax4me]

ECU is grounded to the intake manifold. On 99 the ECU ground should be right on top by the ignition coil.

A loose or corroded ground isn't unheard of, but it usually causes major problems like a No -start or random codes.

 

A bad FPR is usually pretty obvious because you'll get a very rich idle AF mixture which will cause it to run rough and be hard to start when warm.

 

Is the idle rough or too high/too low? Is it constant or does it jump up and down?

[alpop]

Yes, I cleaned those ground points a while back.

 

Occasionally the idle seems "a little" rough but not very often.  It does move around a little, it can vary from ca. 600 - 900 RPM when the engine is warm.

 

When started cold it acts perfectly, starting maybe at 1400 rpm and gradually working its way down over a period of ten minutes or so to ca. 600 rpm.

 

I haven't yet cleaned the IAC valve and figure this needs to be done.

 

Never hard to start warm or cold.

[alpop]

Well I checked out the FPR, vacuum hoses and such look fine.  Then I followed one tiny vacuum hose over to the pass side strut tower and said what are these darn things?

 

The MAP (manifold absolute pressure?) and pressure switching solenoid valve are found here.  The Haynes manual has a procedure for checking both of these.

 

The resistance on the switching solenoid came out fine, and the 5v reference voltage for the MAP sensor was fine.

 

However, when I backprobed the MAP sensor with the engine running, I found 1.37v at idle, and this went to around 2.0v at 1/2  throttle.  1.37 - 2.0

 

Haynes says it should be about 3v at idle, and should decrease to around 2v at WOT, 3.0 - 2.0

 

Not only are my values way off, but they go in the wrong direction with opening the throttle!  Bingo??

 

Anybody know if Haynes is right?  I have found some mistakes in it.

[Fairtax4me]

It's been a while but I think it should be 1.2-1.5v at idle and around 3.5-4v with no vacuum. (Pull the vacuum hose off)

[alpop]

well that sort of popped by bubble, 1.4v at idle and 3.4v with no vacuum.  Why would Haynes be so wrong?

[alpop]

Hmm, only other thing is I had a real hard time both tries with getting the ground probe to work.  Both times initially I got no reading until I grounded to the chassis then after a minute or two of fooling around the ground probe started working.

 

I left it unplugged and it just drove fine for 10 minutes, 'course that means nothing since its done that before.

[Fairtax4me]

The map sensor is very low priority on that engine anyway. It is mostly used for barometric pressure reading to help the ECU determine elevation.

The main air density calculation is made by the MAF sensor.

 

Haynes manuals are often wrong, Take anything you read in them with a grain of salt.

[alpop]

So in my continuing effort to test sensors I looked for the "rear" O2 sensor today, found it easy on the back part of the cat convertor but a long wire left the sensor and instead of taking the easy route of up to the firewall and the wiring harnesses there, it went back and disappeared above the back part of the transmission.  I could find no sign of a plug or where the wire went after that-

 

Anyone know where the rear O2 sensor plugs in?

[Fairtax4me]

It runs along the left side of the trans until about halfway the. Loops across the top in a diagonal fashion towards the right side. It then runs up with some other wires and is clipped to the bracket on the top right corner of the bellhousing. It's a 4 wire plug, either grey or white. Should look the same as the plug at the sensor end of the harness.