95 Legacy EJ22 computer is cutting fuel pulses short while under heavy throttle opening

[White95Legacy]

Problem: 95 Legacy engine computer is cutting fuel pulses short while foot is still heavy on throttle, causing hesitation while power is demanded by the driver. The hesitation lasts a few seconds then it is fine for tens of minutes. Car is not disabled by this problem but it is an occasional annoyance. Sometimes it stalls while coasting in neutral or clutch is pressed, but doesn't stall while idling while stopped. At first we noticed the problem as a hesitation, but have since narrowed it down to ecu not providing long enough fuel pulses for the amount of air the cylinders are breathing.

 

My friends car:

1995 Legacy AWD 5speed with hill holder. EJ22

 

MODS: None

 

MAINTENANCE:

New spark plugs.

Inspected the Mass Air Flow Sensor visually but not electrically.

New front oxygen sensor.

New fuel pump and sump filter when the pump failed recently, new engine compartment fuel filter as a precaution, the stuttering/hesitation was there before and after the replacements.

 

No check engine lights.

 

We hooked a multimeter to the fuel pump to watch how much current it was pulling to let us know if it was shutting off, current stayed steady while hesitation occurred while driving.

 

Hooked a wire to one of the spark plug wires and ran it inside the car, when the hesitation occurred while driving we brought the wire 5mm away from a grounded part of the car and saw the arcs, meaning the coil was firing just fine during the hesitation, ignition is not the cause of the hesitation.

 

Then we hooked an oscilloscope to one of the fuel injector wires and watched the pulses while driving and sure enough we saw the pulses cut down to almost non-existent when the hesitation occurred.

 

Something is telling the engine computer to kill fuel delivery to a very small amount while foot is still heavy on the throttle.

 

Swapped the front oxygen sensor with the rear one and it didn't fix the hesitation. Then he bought a new oxygen sensor for the front and it still didn't fix the problem.

 

Throttle position sensor?

Mass air flow sensor?

Coolant sensor?

Loose wire somewhere?

 

I don't see a MAP sensor anywhere on the engine or in the shop manual, so vacuum leaks seem unlikely.

 

I have a few ideas of what to do:

The MAF seems like it can be probed with an oscilloscope

I read this on Wikipedia about MAFS:

 

Hot wire sensor (MAF)[edit]

A hot wire mass airflow sensor determines the mass of air flowing into the engine’s air intake system. The theory of operation of the hot wire mass airflow sensor is similar to that of the hot wire anemometer (which determines air velocity). This is achieved by heating a wire suspended in the engine’s air stream, like a toaster wire, with either a constant voltage over the wire or a constant current through the wire. The wire's electrical resistance increases as the wire’s temperature increases, which varies the electrical currentflowing through the circuit, according to Ohm's law. When air flows past the wire, the wire cools, decreasing its resistance, which in turn allows more current to flow through the circuit, since the supply voltage is a constant. As more current flows, the wire’s temperature increases until the resistance reaches equilibrium again. The current increase or decrease is proportional to the mass of air flowing past the wire. The integrated electronic circuit converts the proportional measurement into a calibrated signal which is sent to the ECU.

 

Seems like the white wire of the MAF is the signal wire from the MAF integrated circuit since we see it is shielded from EMI/RFI in the schematic.

Anyone have an idea of the frequency range or any advanced knowledge of Subaru MAF? All I have for an oscilloscope is my laptops sound card with 44khz sample rate, "soundcard scope" software and some 1 megaohm resistors to protect my laptop from dangerous current levels.

Edited December 4, 2015 by White95Legacy

[crazyhorse001]

If the ecu is shortening the inj pulsewidth, then for some reason it believes its overfueling.

 

Check the tps, and make SURE you have a smooth transition from low to high ohms. They can develop a dead spot at cruise.

 

The maf MAY also be the culprit. If at all possible swap it with a known good one.

 

Id mention the o2 sensor, but you've already changed that. Double check the wiring harness to the o2 anything that runs near the exhaust can get bad places in the wiring.

 

Maf should be able to be read by a meter. A smooth swing from 5v to i think .5 v from a lo at idle to hi at wot. More air=more volts. Personally its easier to find a good one and swap to see if the hesitation goes away. A maf is basically a heater element, and a thermocouple, and should read as such with your trusty meter.

 

Kudos to your advanced troubleshooting skills tho!

Edited December 4, 2015 by crazyhorse001

[Numbchux]

I'm fairly sure those have a MAP, mounted over on the RH strut tower. I know a '96 does, and while the '95 ECU is a unique animal, the rest of the management is pretty similar. With or without, a vacuum leak can cause problems (letting in air after the MAF = low reading at the MAF = less fuel).

 

A '95 2.2 does have a fairly primitive OBDII, it would be worth hooking up a code reader and looking for stored codes, and watching live data of the TPS, MAF, and O2 while acting up. Probably way easier and more accurate than trying to read it with a scope.

 

Also, a pre-O2 exhaust leak can throw the reading off....make sure the manifold gaskets aren't blown out, and all studs/nuts are present.

[Fairtax4me]

Knock sensor causes hesitation. Common failure.

[White95Legacy]

I'm fairly sure those have a MAP, mounted over on the RH strut tower. I know a '96 does, and while the '95 ECU is a unique animal, the rest of the management is pretty similar. With or without, a vacuum leak can cause problems (letting in air after the MAF = low reading at the MAF = less fuel).

 

A '95 2.2 does have a fairly primitive OBDII, it would be worth hooking up a code reader and looking for stored codes, and watching live data of the TPS, MAF, and O2 while acting up. Probably way easier and more accurate than trying to read it with a scope.

 

Also, a pre-O2 exhaust leak can throw the reading off....make sure the manifold gaskets aren't blown out, and all studs/nuts are present.

Do you mean MAF instead of MAP? Cause the MAF is above the right strut tower, I can't find a MAP anywhere.

 

The fuel pulses cut wayyyy down to like 1% duty cycle during the hesitation, as if the ecu thinks we are coasting downhill with foot completely off the throttle, even if foot has the throttle all the way to the floor. Pulses are 1% when they should be about 10-20-30-40-50-60% duty cycle judging by how far we have our foot on the throttle, then the ecu comes back to life several seconds later and we see pulses the way they should be, 10-20-30-40-50-60% etc.

 

The hesitation comes and goes, happens for about 1-5 seconds at random times, then completely disappears for maybe 5-30 minutes and the engine does it's job perfectly, happens when driving fast and maintaining speed, accelerating from a slow speed, maintaining a slow speed. Feels like a loose wire.

 

Thank you, making a list of what to look for when we feel like playing with his car again.

 

 

 

When I get a chance to look at my friends car again we will try this trick with the connectors under the steering column http://legacycentral.org/library/literature/codes.htm

Edited December 5, 2015 by White95Legacy

[White95Legacy]

If the ecu is shortening the inj pulsewidth, then for some reason it believes its overfueling.

 

Check the tps, and make SURE you have a smooth transition from low to high ohms. They can develop a dead spot at cruise.

 

The maf MAY also be the culprit. If at all possible swap it with a known good one.

 

Id mention the o2 sensor, but you've already changed that. Double check the wiring harness to the o2 anything that runs near the exhaust can get bad places in the wiring.

 

Maf should be able to be read by a meter. A smooth swing from 5v to i think .5 v from a lo at idle to hi at wot. More air=more volts. Personally its easier to find a good one and swap to see if the hesitation goes away. A maf is basically a heater element, and a thermocouple, and should read as such with your trusty meter.

 

Kudos to your advanced troubleshooting skills tho!

I should of described how severely underpulsing the ecu is being in my original post, but read my reply to Numbchux for a better description, it's like 1% duty cycle when it should be like 40% duty cycle. I see tiny pulses, then when the hesitation ends and the engine runs fine the pulses go back to like 40% duty cycle.

 

Pumping the throttle doesn't immediately end the hesitation, so it doesn't feel like a dead spot in the TPS. The hesitation ends several seconds later regardless if the throttle position is changed after hesitation initiates.

 

But i'll try probing the TPS at some point if the MAF checks out fine. Unfortunately my friend doesn't want to fork out the cash for a new MAF, they run between $100-$250, I see one for $35 but who knows about the quality. I'll try probing stuff near the ECU to let us know if there is a loose wire between the MAF and ECU.

 

I'll closely examine the MAF for an integrated circuit, if it looks too simple then the measurement circuit is probably inside the ECU and the MAF connector can be probed with a simple multimeter, if it has an integrated circuit built into the MAF then the measurement circuit is probably inside the MAF and the integrated circuit sends a digital or analog signal to the ECU, like it says in the last sentence of that Wikipedia paragraph I linked. Or the calibrated signal it talks of is voltage based as you say and not a digital or analog pulsed signal.

 

Here is the Wikipedia link: https://en.wikipedia.org/wiki/Mass_flow_sensor

Under section named Hot Wire Sensor (MAF) read paragraph 4, it sounds like what is in this car since the white wire from the MAF to the ECU is shielded.

 

I'll probe all the MAF wires for voltage till I understand, then I'll hook the Oscilloscope up if it doesn't make sense, to see whats hiding in the wire signals.

 

Thank you

[White95Legacy]

Knock sensor causes hesitation. Common failure.

I looked around the internet and it sounds like all a knock sensor does is notify the ECU that detonation is occurring and the ECU retards the ignition timing.

 

Unless the ECU is cutting fuel pulses to 1% duty cycle for several seconds as a test to somehow diagnose a knock sensor reading malfunction.

[White95Legacy]

MAF pinout:

Pin 1 Connects to Batt+ when the main power relay comes on when the ignition is on

Pin 2 Connects pin 1 of the TPS and pin 2 of the Coolant Sensor and connects to pin 25 of the ECU

Pin 3 Connects to engine ground

pin 4 Connects to pin 26 of the ECU, and the shielding jacket around the pin4-pin26 wire connects to pin 54 of the ECU

 

I read the schematic and this appears to be the way the MAF is wired.

 

I'll first check to make sure pin 1 is Batt+ with the multimeter. Then I'll check that pin 3 really is grounded. Pin 2 seems to be either a solid positive or negative charge switched by the ECU so I will investigate that. And pin 4 seems likely to be sending some kind of signal or voltage or current reading.

[Fairtax4me]

Yes.

Pin 1 = 12v batt supply

Pin 2= 5v ECU reference

Pin 3= ground

Pin 4= signal voltage to ECU

 

Pin 4 wire is sheilded to prevent interference. Sheilding is grounded at the ECU.

 

Knock sensor is easy to check and easy to replace, and often causes the symptoms you have. Intermittent hesitation at low rpms. If the housing of the sensor is cracked, the sensor needs to be replaced.

 

If you have some resistors laying around, wire a few together in series to about 580k-ohm, stick one end in the harness connector for the knock sensor and bolt the other end to the block.

This will eliminate false signals from the knock sensor, and keep the ECU from going into limp mode.

[Cougar]

Maybe something inside the ECU is causing the trouble. Replacing it shouldn't be much of an issue to see if that is where the trouble is at. Another thought is, have you verified that the TPS sensor is getting the 5 volt reference power when the trouble is occurring? Maybe there is a intermittent problem with that power source (which as we know, comes from inside the ECU). I assume you have already made sure that the 12 volt power to the injectors is good when the trouble is happening along with the return side of the injector circuit back to the ECU.

[White95Legacy]

Thanks people, I made a checklist from all our posts.

 

1.Verify 12v to injectors is staying on.

2.Verify 12v to TPS, MAF and Coolant sensor is staying on.

3.Check pin 4 of MAF for signal.

4.Disconnect knock sensor and bypass it's wire with 580kohm resistor to ground.

5.Verify pins 2 and 3 of TPS have normal readings.

6.Verify O2 sensor readings are reaching the ECU connector.

7.Verify Coolant Sensor signal is reaching ECU.

8.Use http://legacycentral.org/library/literature/codes.htm green connector trick for engine code tests while driving.

 

All of these tests will be tested before driving, then tested while driving while the hesitation occurs.

 

I came up with a nifty test probe trick, take a sewing needle, put a wire thru the eye, wrap the eye end with gorilla tape to hold the wire on and create a handle, then stab the needle thru a wire you want to test. Just leaves a small hole and allows you to tap into a wire with minimal damage if probing a connector is not an option.

 

We have to take out the engine in a few days to replace the right head gasket cause it's leaking oil, and replace the clutch. Would be the time to replace gaskets and search the intake plenum/PCV/EGR etc for leaks.

Edited December 5, 2015 by White95Legacy

[Fairtax4me]

#8 won't work because 95 is OBD2. Engine codes have to be retrieved with a code scanner. This can be done while driving. If you buy a Bluetooth OBD2 adapter you can use an app like Torque to monitor live data while driving. But data is limited on these older OBD2 models, you will probably not be able to see knock sensor inputs or injector pulse width.

 

Avoid sticking needles through the wire insulation. This allows moisture into the wire and will cause corrosion which will cause problems down the road.

 

Back-probe the connectors for any wires you need to test by sticking the needle between the wire insulation and the weather seal grommet in the back of the connector. Push it in until you feel it hit the metal terminal.

 

http://www.enduringautomotive.com/how-to-back-probe-a-sensor/

These guys are using paper clips. T-pins or needles work well, especially if you apply some silicone spray or di-electric grease to help it slide between the grommet and wire.

 

 

If you need to probe multiple wires, be sure to wrap your needles in electric tape to avoid them touching each other or other metal parts and causing a short circuit.

Edited December 5, 2015 by Fairtax4me

[White95Legacy]

#8 won't work because 95 is OBD2. Engine codes have to be retrieved with a code scanner. This can be done while driving. If you buy a Bluetooth OBD2 adapter you can use an app like Torque to monitor live data while driving. But data is limited on these older OBD2 models, you will probably not be able to see knock sensor inputs or injector pulse width.

 

Avoid sticking needles through the wire insulation. This allows moisture into the wire and will cause corrosion which will cause problems down the road.

 

Back-probe the connectors for any wires you need to test by sticking the needle between the wire insulation and the weather seal grommet in the back of the connector. Push it in until you feel it hit the metal terminal.

 

These guys are using paper clips. T-pins or needles work well, especially if you apply some silicone spray or di-electric grease to help it slide between the grommet and wire.

If you need to probe multiple wires, be sure to wrap your needles in electric tape to avoid them touching each other or other metal parts and causing a short circuit.

Going into the city with friend today in a different car to buy tools for the head gasket and clutch job. We will stop by O'reiley's auto parts and see what they have for scan tools for rent.

 

When I do the needle trick I put tape there so it won't short. When I am done I put a piece of electrical tape around the holes. The connectors have seals around the wires sometimes and I don't want to damage them, some allow you to pull back the seal while leaving the connector connected, but sometimes it's just more practical to poke the needle thru the wire insulation then patch it.

[Fairtax4me]

Meant to post a link in my previous reply and it didn't paste for some reason. Added that in.

 

I've never found a situation where it was necessary to poke a hole in the wire insulation. The grommets around the wire are silicone and stay pliable for a long time. You can slip a needle in them easily without causing any damage.

[White95Legacy]

Clutch is really worn out, Clutch kit and Gasket Set are coming tomorrow, going to pull the engine today or tomorrow. We're going to test drive and troubleshoot the hesitation after.

 

Let you all know what I find about the hesitation in a few days.

[Cougar]

I'm not sure if 12 volts is supplied to the TPS sensor. I thought it was only 5 volts. Fairtax4me, can you help with that?

[Numbchux]

Do you mean MAF instead of MAP? Cause the MAF is above the right strut tower, I can't find a MAP anywhere.

 

The fuel pulses cut wayyyy down to like 1% duty cycle during the hesitation, as if the ecu thinks we are coasting downhill with foot completely off the throttle, even if foot has the throttle all the way to the floor. Pulses are 1% when they should be about 10-20-30-40-50-60% duty cycle judging by how far we have our foot on the throttle, then the ecu comes back to life several seconds later and we see pulses the way they should be, 10-20-30-40-50-60% etc.

 

The hesitation comes and goes, happens for about 1-5 seconds at random times, then completely disappears for maybe 5-30 minutes and the engine does it's job perfectly, happens when driving fast and maintaining speed, accelerating from a slow speed, maintaining a slow speed. Feels like a loose wire.

 

Thank you, making a list of what to look for when we feel like playing with his car again.

 

 

 

When I get a chance to look at my friends car again we will try this trick with the connectors under the steering column http://legacycentral.org/library/literature/codes.htm

 

Nope, they have a MAF and MAP

 

 

Like has been mentioned, that link is for OBD I cars, and yours is OBD II. Get your hands on a scan tool, just about everything on your list (and more) can probably be diagnosed by reading the live data on one.

 

 

Something is telling the ECU to cut fuel. Plug in a code reader, go to the live data page, and look at TPS, MAF, MAP, and O2 signals, and see if they're doing anything strange.

 

 

 And yes, TPS is 5v

[Fairtax4me]

TPS gets 5v from the ECU.

 

95 was a weird year. Some just had a MAF, others had both MAP and MAF.

[White95Legacy]

Nope, they have a MAF and MAP

 

3923890001_large.jpg

 

Like has been mentioned, that link is for OBD I cars, and yours is OBD II. Get your hands on a scan tool, just about everything on your list (and more) can probably be diagnosed by reading the live data on one.

 

 

Something is telling the ECU to cut fuel. Plug in a code reader, go to the live data page, and look at TPS, MAF, MAP, and O2 signals, and see if they're doing anything strange.

 

 

 And yes, TPS is 5v

We took out the engine today, there is no MAP sensor on his strut tower like there is in your pic.

 

When we get it back together we will go rent a scan tool.

 

Thanks.

[White95Legacy]

You might be the winner Fairtax4me, we got the engine out a few hours ago, I found the knock sensor hiding under the intake manifold earlier but didn't think much of it. After reading your older post again, I just went and took a 2nd good look at it with my 500 lumen flashlight and it has a big ole crack in it. He ordered a new sensor. We will try that before the scan tool then.

 

So limp mode causes the ecu to cut fuel off or something?

[Fairtax4me]

It would only be in limp mode if the CEL is on, in this case with a code for the knock sensor.

 

Ayer you replace the sensor re-check your pulse width and see what it does with a good sensor. Maybe the ECU also cuts fuel with input from the knock sensor.

[mikaleda]

The knock sensor is supposed to change ignition timing to compensate for low octane fuel. Its possible it can change fuel pulses as well and being damaged its sending the wrong info to the computer. Its possible.

[White95Legacy]

Got engine back in, head gasket/valve cover oil leaks fixed, clutch is new and grabs perfect now.

 

New knock sensor installed to replace the cracked one.

 

Took car driving today, bad hesitation, worse than before, lasts longer and happens more often. Happens at a variety of speeds, lasts anywhere from .001-60+ seconds. It will not go above 1300 rpm regardless of what you do with the throttle while the hesitation is happening, then it drives fine when the hesitation clears. The hesitation lasts long enough for the car to slow to a crawl sometimes but it avoids totally stalling if you hold throttle some random amount, if you hold-clutch or switch-to-neutral and let go of throttle it's idle bounces between 300-1300 rpm (RPM hunting) and it sometimes stalls cause it danced too close to stall speed, if you press the throttle it stays right at 1300 rpm no matter if you press it 5% or 100%. It's doing something stupid with it's idle air control motor, and obviously the injector pulses cause we saw them with an oscilloscope, delivers spark fine.

 

Idles fine when hesitation is not happening.

 

RPM hunted and stalled while idling when warming up, just like the hesitation while car is moving and driver is trying to get torque from engine, another time it idled fine while warming up.

 

Engine temp makes no difference, random.

 

It's like the ECU is switching to some bullshit idling mode and ignoring the amount of air it is breathing thru the throttle body, only providing fuel for the air it thinks it's getting thru the idle air motor.

 

We ordered http://www.newegg.com/Product/Product.aspx?Item=9SIA2VF2W40245

 

Mini ELM327 V2.1 OBD2 II Bluetooth Diagnostic Car Auto Interface Scanner. Will take a week or 2 to get here.

 

Either the new knock sensor made it worse, or more likely, we disturbed something when we disconnected the engine wire harnesses, disconnected all the engine sensors, and replaced head gaskets, rocker cover gaskets and camshaft gaskets.

 

We disconnected TPS and nothing changed, car drove ok without TPS but hesitation showed itself occasionally just like with TPS connected. No CEL even with disconnected sensor.

 

We disconnected MAF and nothing changed, car drove ok without MAF but hesitation showed itself occasionally just like with MAF connected. No CEL even with disconnected sensor.

 

We will wait till bluetooth scanner comes. Might play with disconnecting the Oxy sensors and temperature sensors tomorrow.

 

Feels like a loose wire shorting something and putting the ECU into idiot mode. We took off the passenger carpet and foot plate and found the ECU, no corrosion or signs of water intrusion, no wires appeared frayed or tampered with.

 

We keep looking everywhere we think important wires are and checking for frays/damage/water/tampering.

 

Seems like more trouble than a 20 year old car is worth, but my friend is enjoying taking apart a car and learning how it works while I walk him through. Electrical problems like this requires input from experienced Subaru techs like you since idk how Subaru programs their ECU's.

 

Anyone got detailed info on ECU tech data, programming logic tables, explanations of open/closed loop modes, missing sensor fallback modes, short circuit/overload fallback modes?

 

OBD scanner is next best bet.

 

This could either turn into replacement of all sensors, harnesses and the ECU, or conversion to carburetor intake.

 

Thank you all

Edited December 15, 2015 by White95Legacy

[White95Legacy]

I'll have to check the 5v line to MAF TPS and Coolant Sensor tomorrow, since it seems to be trying to idle and acting oblivious to driver requests.

 

Weird that it idles with throttle wide open, might have to reconnect oscilloscope to fuel injectors to see pulses. I suspect it is skipping some combustion cycles to regulate 1300 rpm speed but maintaining correct air to fuel ratio by giving a large pulse to accommodate the air charge from a WOT, either by skipping some fuel pulses or skipping ignition pulses or both. I could wire the left mic channel for my laptop with a 1 megaohm resistor to part of the ignitor circuit and the right mic channel to an injectors wires for a cylinder on that part of the ignitor. To get an exact picture of what the ECU is doing.

 

Tried starting it with TPS and MAF disconnected and it turned over but didn't start.

[White95Legacy]

I'ma take the engine harness connectors apart and clean them real good. They looked a little dirty and corroded. Bright flashlight and a paper clip to gently scrape the pins and pick away stuff to make the pins shiny.