Just another misfire/rough idle thread...

[birkhoff]

^^^^ beat me to it! I was trying to find the location -- still not sure. Somewhere on the transmission but the FSM doesn't seem to be clear about this.

 

Could be interesting.

[StkmltS]

Holy schniznit, we might be getting close to a winner. Heiche and Birkhoff (edit), it's funny you just mentioned the neutral position switch. Out of sheer coincidence when I pulled into my driveway just now I happened to put it into neutral and it felt like something changed when I popped it out of gear. When I went from neutral to in-gear (clutch pedal in the whole time, zero MPH) the misfire count stopped increasing. I tried it several times, and sure enough, each time that I went from neutral to in-gear the misses stopped increasing. My beautiful wife and kids won't be home for a couple more hours so I have some free time on my hands to look at the log and explore this a little deeper. A new development! Edited September 29, 2016 by StkmltS

[StkmltS]

Here's what it looks like. Notice how much less fuel correct jumps around when it's in neutral (especially around time = 50000). In this log all other cylinders were behaving so I kept them off the chart.

 

I'm going to get back in the car and play around a bit more and see if keeping it in-gear for any length of time can get the roughness counter to reset back to zero.

BtSsm_20160929_174705_Neutral-vs-In-Gear.csv

[heiche]

I think you're on to something! Do you know how to disconnect the neutral position switch?

 

Also, do you know if you flashed a different version onto your ECU before this started happening a while ago? For example, was it one of the ECU ID's below, then you flashed it to a different one? I ask, because there were problems related to the neutral position detection in some of the versions. I believe the newest one on the bottom is probably the "best" version, and the one you might want to try.

 

A2WC510N

A2WC511N

A2WC522N

[StkmltS]

I think you're on to something! Do you know how to disconnect the neutral position switch?

 

Also, do you know if you flashed a different version onto your ECU before this started happening a while ago? For example, was it one of the ECU ID's below, then you flashed it to a different one? I ask, because there were problems related to the neutral position detection in some of the versions. I believe the newest one on the bottom is probably the "best" version, and the one you might want to try.

 

A2WC510N

A2WC511N

A2WC522N

 

I don't know how to disconnect the neutral position switch but you can bet BtSsm's sweet little tush I'm going to look it up!

 

I've been using A2WC522N the whole time. I started out with CovertRussian's stage 0 safety tune (A2WC522N) and I've just been modifying that map since.

 

The roughness count definitely resets to zero if I keep it in-gear long enough. I did see a few misses in-gear just now, and I'm not really sure what to think of those yet. My wife is on her way home so tonight's car time ends now

BtSsm_20160929_191517_Neutral-vs-In-Gear.csv

[StkmltS]

I just realized I probably could have just added throttle to the chart instead of all of those little text bubbles

 

Sent from inner space.

[birkhoff]

If you look at the CSV sheets, the misfire count resets to zero every 8 seconds, and it does so simultaneously across the cylinders. There must be a counter in the ROM. I don't think it has anything to do with how long you are in gear or not, or how high the count goes (although there is evidence from Bonbon that it cannot go past 99).

 

Every 8 seconds all four cylinders are set to 0 and the sum begins again.

[StkmltS]

If you look at the CSV sheets, the misfire count resets to zero every 8 seconds, and it does so simultaneously across the cylinders. There must be a counter in the ROM. I don't think it has anything to do with how long you are in gear or not, or how high the count goes (although there is evidence from Bonbon that it cannot go past 99).

 

Every 8 seconds all four cylinders are set to 0 and the sum begins again.

 

Look at my first log (post 528). It stays level at 5 for about 38 seconds (time 7315 to time 45281)

[xt2005bonbon]

^ There must be some additional 'if statement' in the ECU code for how long a misfire count stays at a given value. I've also seen certain misfire count stays at a fixed value for a while.

Getting sick of this. I want to see the code governing this roughness business now :mad:

[birkhoff]

Look at my first log (post 528). It stays level at 5 for about 38 seconds (time 7315 to time 45281)

 

Maybe it's another clue

[StkmltS]

Messing with the neutral position sensor didn't yield any changes. I tried unplugged, partially connected, twisted, and kinked. I even drove around for a little while with it disconnected and nothing out of the ordinary happened.

 

http://uploads.tapatalk-cdn.com/20161002/808f30ae7c787058ccc3d4687d0294b3.jpg

[StkmltS]

Applying load with the e-brake engaged changes the misfiring... in a weird way.

The gradual ramp-up in throttle is the ECU doing its own thing as I slowly let out the clutch. The spike in RPM is when I applied (plenty of) throttle to keep the motor from stalling.

 

http://uploads.tapatalk-cdn.com/20161002/115bb156be45c71410f7e1e6ef45cdc8.jpg

[StkmltS]

Starting at a complete stop at the top of a slight incline, with the engine idling and misfiring (both in-gear and in neutral), and using only the weight of the car to start movement (frontwards and backwards), the roughness count stops increasing every time vehicle speed changes from 0mph to >/= 1.2mph (1.2mph appears yo be the min measured by the ECU).

 

Controlling misfires by moving in/out of gear doesn't always yield the same results that I/we saw the other day. This morning in-gear @ 0mph misfired just as much as it did in neutral.

 

http://uploads.tapatalk-cdn.com/20161002/9b4771ed13b03b0deb149764de247851.jpg

[StkmltS]

And yes, I was under the car and experimenting this morning before 5am. Hey man, when something big is on my mind sometimes I can't sleep.

[MTBwrench]

All this is absolutely boggling my mind... why on earth is vehicle speed impacting spark timing/fueling enough to misfire?!

 

If I were you, I'd *try* sourcing an oscilloscope to look at the injector signals as vehicle speed changes.

[fahr_side]

You know there is a separate 2D table for idle timing eh? That is active or not depending on a threshold value for vehicle speed in another 1D table. The stock value is usually 1mph. Have a look at the idle timing value vs. what's in the base timing table.

 

I'd have thought such a light flywheel would exacerbate any tendency toward high misfire count.

[birkhoff]

I think we need to start a project to figure out how the subaru misfire detection system (MDS) works. Unfortunately, we don't have much to go on. I've not found anything definitive in the FSM. From the web, you get mostly shallow articles; the best I could find was this one. Unfortunately, it is a discussion of generic systems with lots of 'some manufacturers do this . . . ' and we don't know specifically what is going on with our cars yet.

 

Let me start trying to collate this information. I'll start with what we 'know' from the article.

 

When/why does the MDS decide to reset the count to zero?

 

From the article:

 

Any misfires that occur are recorded every 200 revolutions of the crankshaft. The tally is stored in 16 memory blocks, so every 3,200 rpm the misfire count starts over again.'

 

Something like this is mentioned in other articles, so let's assume for the moment it is approximately correct. What I previously thought was a timed reset, is in fact a rotation count reset. Since we were mostly monitoring around idle, time was a reasonable surrogate for revolutions. We can backwards engineer the count used by Subaru by integrating the RPM log between two hard resets in a good log. I tried this on some of StklmS's recent logs but got unreliable results. However, we may have some confounding factors in those logs.

 

Is the MDS active at all times?

 

Apparently not. For example

 

The OBD II system is usually programmed to ignore misfires when the engine is cranking, and when a cold engine is first started. It will also ignore misfires during decel when the fuel is momentarily cut off.

 

and

 

One of the drawbacks of using a crank sensor to detect misfires is that it can sometimes be fooled by normal powertrain vibrations. Driving on a rough road, for example, may produce variations in crank speed that seem like misfires but are not. Some OBD II systems monitor inputs from the ABS wheel speed sensors to tell when a vehicle is driving on a rough road, and disable misfire detection until the road smoothes out.

 

So reports that the car only misfires after it warms up could be misleading. Also, if your ABS sensors are messed up, that could interfere with the MDS! Our cars have another system monitoring pitch and yaw. Inputs from that system may be part of the subaru MDS algorithm. Who knows?

 

 

Could a lightweight flywheel be a factor in causing `phantom' misfires reported by the MDS?

 

I posted a link to a discussion about this in earlier posts. This seems probable, as in, while it is not the case that every LWFW will cause problems in every car, it will increase the chance of the MDS observing unsteady crank rotation, especially at lower RPM.

 

<< space for further discussion/comment >>

Edited October 2, 2016 by birkhoff

[birkhoff]

I didn't want to confuse the above post with wild conjecture about what this means for our specific problem here. However, as an attempt to widen the focus, consider the following scenario.

 

Suppose that the misfire free periods are actually just the MDS being suspended by some other glitch. This would include suspending the reset count, leading to long periods with the system holding whatever count it currently shows.

 

Now suppose that you have a bad or marginal crank position sensor; the observation that the problem is mostly showing up in #2 and #3, two cylinders that are in exactly the same position on the crank (but 360 degrees out of timing phase) may not be a coincidence. Suppose the MDS is reporting misfires, like, all the time.

 

Finally, suppose you have an overly active rough-road suppression signal disabling the MDS whenever the car is moving. That would lead to a very different interpretation of your recent logs. Periods of no activity or flat counts would indicate a problem, not a lack of problem.

 

Up to this point we have been thinking exactly the opposite.

 

Question: Do you have the same crank sensor as pre-rebuild? Have you gone through the FSM diagnostics on it?

[StkmltS]

I really like what you're getting at. It could be really useful if we knew more about the MDS.

 

My crank sensor is the same one that's always been on the car. I haven't done anything to verify that it's working properly because...

 

I can feel, see, and hear the misfires happening. I knew they were happening before I had the tools to monitor the engine (Torque, then BtSsm). Sitting in the driver's seat you can feel the shudder, under the hood you can see the motor shake, and in the exhaust you can hear the misses.

 

Sent from inner space.

[Perscitus]

One interesting note to above info on MDS on recent-gen Subarus (say 2005+) is that at least under SSM, MDS is active on cold start as well and appears fairly accurate on a per-cylinder basis.

 

Monitoring 'Misfire (counts) Cylinder #1-6' SSM params over CAN or Kline (as applicable: so 1-4 on H4s, 1-6 on H6s) on Scoobies of varied ignition sys health should help figure out at least part of this MDS logic.

[birkhoff]

I really like what you're getting at. It could be really useful if we knew more about the MDS.

 

My crank sensor is the same one that's always been on the car. I haven't done anything to verify that it's working properly because...

 

I can feel, see, and hear the misfires happening. I knew they were happening before I had the tools to monitor the engine (Torque, then BtSsm). Sitting in the driver's seat you can feel the shudder, under the hood you can see the motor shake, and in the exhaust you can hear the misses.

 

Sent from inner space.

 

I know, I remember you saying that you could feel the misfires. But if the crank angle determination is marginal, it could be messing up the MDS as well as causing general timing issues with the trigger signal to one or more cylinders. A scope (as someone suggested before) would help identify the latter, but we don't have a scope handy. In the old days, a simple timing light would show this kind of thing very clearly. A sloppy or loose distributor or worn timing gear line would be the culprit and unsteady idle, general $hitty running the symptom. A bad crank sensor could be the modern equivalent.

 

Ignition trigger is going to be some combination of crank and cam sensors, RPM etc; cam sensors mainly to determine what stroke the cylinder is on. The lead angle parameter must be crank position because the cam sensor does not seem to be designed (from the look of it) to be nearly as accurate for this. If the crankshaft trigger signal is moving around, that would explain a number of symptoms, but not all.

 

As far as MDS cutting in and out, I think the most curious things about your logs are the long long periods of positive but stationary count. I wasn't actually joking when I commented this could be a clue. To me, it is pretty bombproof evidence that the system is blasting through what should have been a count reset under some conditions (wheels moving as a best guess) and therefore must be inactive at those times. Contrast this with a long string of zeros, which contains no information about active/inactive status for the MDS. What is causing the MDS system to cut out, we don't yet know. But it may be quite unrelated to the actual misfires, if you see what I mean.

[birkhoff]

One interesting note to above info on MDS on recent-gen Subarus (say 2005+) is that at least under SSM, MDS is active on cold start as well and appears fairly accurate on a per-cylinder basis.

 

Monitoring 'Misfire (counts) Cylinder #1-6' SSM params over CAN or Kline (as applicable: so 1-4 on H4s, 1-6 on H6s) on Scoobies of varied ignition sys health should help figure out at least part of this MDS logic.

 

Tell us more!

 

Do you know any conditions for sure that put the MDS to sleep in the 2005 logic? It would be great if you could help to move us from pure speculation to a reasonably accurate reverse engineering of the system!

[StkmltS]

Once my WBO2 is wired up I be able to use it to help narrow down the source of the misfire?

 

To clarify:

If I have a spark problem and the fuel isn't burning, then my AFR would be lower when each misfire occurs.

If I have a fuel problem and there isn't fuel to burn, then my AFR would be higher when each misfire occurs.

An air leak could also cause a high AFR, but I've all but written off that as a possible cause.

 

Will I be able to measure this small of an amount of change in AFR for just one misfire?

[xt2005bonbon]

I am not sure. When I had mine, I had the same thought process as you. So I carefully watched my fuel trims, the AFR, fuel correction at idle. They were still fine . Or at least, I could not tell the diff.

I think the sure way would be to install o2 sensors on the exhaust header itself, one for each side of the engine at least. That would be pretty nice I think.

 

You could also inspect the spark plug after a few hundred miles. It may look different than the other cylinders. Mine was still fine in #2 but did look slightly different than #4.

[rhino6303]

Once my WBO2 is wired up I be able to use it to help narrow down the source of the misfire?

 

To clarify:

If I have a spark problem and the fuel isn't burning, then my AFR would be lower when each misfire occurs.

If I have a fuel problem and there isn't fuel to burn, then my AFR would be higher when each misfire occurs.

An air leak could also cause a high AFR, but I've all but written off that as a possible cause.

 

Will I be able to measure this small of an amount of change in AFR for just one misfire?

I would think you'd only be able to tell a difference if you had 4 wideband sensors plumbed at each exhaust runner. Then you could narrow it down to a specific cylinder. Without an aftermarket header, this would be difficult.