OB99W

I reformatted the data into tables, which makes it somewhat easier to see relationships and trends. Since some values in the trigger frame data seem to be "outliers" (look at the "Calc Load" values in particular), I'm only showing frames -19 thru -1. I added leading zeroes ("0") and other spacers as needed to keep the columns as even as possible. "CL--" = "CL" (closed loop) "OLD" = "OL-Drive" (open loop, due to driving conditions) All below have the form: {frame #}*{Fuel System Status}*{STFT B1}**{LTFT B1}**{Calc Load}**{STFT B1 S2} -19*CL--*+00.7**-1.5**42.7**0.7 -18*CL--*+00.0**-1.5**44.7**0.7 -17*CL--*+01.5**-1.5**47.0**0.7 -16*CL--*-01.5**-1.5**51.7**0.7 -15*CL--*-02.3**-1.5**42.3**0.7 -14*CL--*+02.3**-1.5**36.4**0.7 -13*CL--*-03.1**-1.5**39.6**0.7 -12*CL--*-01.5**-1.5**39.2**0.7 -11*CL--*-00.7**-2.3**43.9**0.7 -10*CL--*-03.1**-2.3**45.0**0.7 -09*CL--*-03.1**-2.3**43.5**0.0 -08*CL--*+03.9**+3.9**28.2**0.7 -07*CL--*+00.7**+3.9**39.2**0.0 -06*OLD*-04.6**+3.9**24.3**0.0 -05*CL--*-25.0**+3.1**17.6**0.0 -04*OLD*-00.0**+3.1**08.6**0.0 -03*OLD*-00.0**-1.5**10.1**0.0 -02*OLD*-00.0**-1.5**12.1**0.0 -01*CL--*-10.9**-1.5**07.4**0.0 -19*CL--*-00.0**-3.1**32.5**0.0 -18*CL--*-01.5**-3.1**24.3**0.0 -17*CL--*+01.5**-3.1**28.6**0.0 -16*OLD*-00.0**-3.1**43.9**0.7 -15*CL--*-00.0**-2.3**46.6**0.7 -14*CL--*-10.9**-3.1**42.7**0.0 -13*CL--*+03.1**-3.1**34.9**0.0 -12*CL--*-16.4**-3.9**15.2**0.0 -11*CL--*-02.3**-3.1**30.1**0.0 -10*CL--*+01.5**-3.1**27.8**0.0 -09*CL--*+02.3**-3.1**33.3**0.0 -08*CL--*+03.1**-3.9**25.0**0.0 -07*CL--*-16.4**-3.9**10.9**0.0 -06*CL--*+04.6**-3.9**24.7**0.0 -05*CL--*+03.9**-3.1**32.9**0.0 -04*CL--*+04.6**-3.9**22.3**0.0 -03*OLD*-07.0**-3.1**08.6**0.0 -02*CL--*+03.1**-3.1**30.9**0.0 -01*CL--*+01.5**-3.9**16.0**0.0 -19*CL--*-00.0**-4.6**07.4**0.0 -18*CL--*-10.1**-4.6**21.9**0.0 -17*CL--*+02.3**-3.9**27.4**0.0 -16*CL--*-02.3**-3.9**26.6**0.0 -15*OLD*-00.0**-3.9**06.6**0.0 -14*CL--*-11.7**-3.9**06.6**0.0 -13*CL--*-23.4**-3.9**06.2*0.0 -12*CL--*-20.3**-3.9**04.7**0.0 -11*CL--*+04.6**-4.6**28.6**-0.7 -10*CL--*-00.0**-3.9**45.4**0.0 -09*CL--*-03.9**-6.2**17.2**-0.7 -08*CL--*-00.0**-4.6**50.1**-0.7 -07*CL--*-00.0**-3.9**39.6**0.0 -06*OLD*+02.3**-3.9**16.4**0.0 -05*CL--*+02.3**-3.1**32.5**0.0 -04*CL--*-00.0**-3.9**09.0**0.0 -03*CL--*-01.5**-3.9**32.1**0.0 -02*OLD*-02.3**-4.6**10.5**0.0 -01*OLD*-07.8**-4.6**07.8**0.0 No details of running circumstances available? -19*CL--*-12.5**-7.0**32.5**0.0 -18*CL--*+09.3**-7.8**24.3**0.7 -17*CL--*+00.0**-5.4**28.6**0.0 -16*CL--*+09.3**-6.2**43.9**0.0 -15*CL--*+02.3**-3.9**46.6**0.0 -14*CL--*+07.0**-4.6**42.7**0.0 -13*CL--*-00.7**-7.8**34.9**0.0 -12*CL--*-10.9**-7.8**15.2**0.0 -11*CL--*-07.8**-9.3**30.1**0.0 -10*CL--*+08.5**-6.2**27.8**0.0 -09*CL--*+04.6**-7.8**33.3**0.0 -08*CL--*-18.7**-7.0**25.0**0.0 -07*CL--*+06.2**-8.5**10.9**0.0 -06*CL--*+03.1**-8.5**24.7**0.0 -05*CL--*+06.2**-8.5**32.9**0.0 -04*CL--*+06.2**-7.8**22.3**0.0 -03*CL--*+03.9**-9.3**08.6**0.0 -02*CL--*-20.3**-8.5**30.9**0.0 -01*CL--*-03.1**-10.1**16.0**0.0 In the above tables, there are several instances of the STFT being in the region of -20 to -25. (That's due to the ECU seeing the need to substantially lean the mixture.) It's typically happening about the time that the load is significantly reduced (probably with foot off throttle), and mostly shortly after the ECU switches from closed-loop to open-loop-drive. In all cases, the system is initially CL, which means the A/F sensor is in play. It apparently transitions to OL-Drive during deceleration when the throttle is closed (expected behavior). So, if either the fuel pressure is out of spec or the A/F sensor is sluggish, the mixture might overshoot enough to cause the problem. (The dealer induction service and replacement of IACV & injectors, and further IACV cleaning that was done, none of which resolved the problem, hopefully allows a rule-out of those as possible culprits. ) I'm back to the suggestion of checking the fuel pressure, doing something about it if it's out of spec, and if that leads nowhere then chancing a new OEM front A/F sensor. (Hmmm, did anyone check for holes in the exhaust system? :-\) Good luck!

No problem. I'm trying to make sure that the fuel pressure measurement at idle is done with the FPR seeing vacuum typical of a well-functioning engine. That's why I'm suggesting taking the pressure reading while there's 20" applied, rather than something possibly several inches lower (that your intake might be at). Yes to 1) and 2). I'm not sure exactly what you're asking. Assuming the FPR's diaphragm is okay, the vacuum will hold at whatever it's been pumped to. If we're talking about engine idle, change should be minimal as compared to (let's say) 17". Once the fuel pressure gauge and vacuum hand pump are attached, the engine should be run. A fuel pressure reading can be done with the FPR initially having no vacuum applied, and then 20". You could even monitor the change in fuel pressure as the vacuum to the FPR is changed. With no vacuum applied to the FPR, fuel pressure should be about 43 psi. With 20" vacuum applied, about 32 psi. Both should preferably be within plus/minus 5% (10% max). Be sure to take care when "T"-ing in the fuel pressure gauge, since there will initially be pressure in the line even with the engine off (unless your injectors are very leaky). The diagnostic info in the FSM for P0172 goes into detail concerning the above. Please read it. Those readings are strange. I suspect two of them are typos. "Calc Load" can't exceed 100% (and I doubt the engine was that stressed ), so it's probably "13.3", not "133". For "MAP, it seems likely you hit one key over, so that should be "16.2". The rest of the readings don't fit the car having just been driven 25 miles. Unless things are so bad that the catalyst is ready to melt, a misfire requires two trips to trigger the CEL. Therefore, the freeze frame data was likely captured at the first trip (when the code was "pending") -- this was likely at a cold start, which would explain being "OL", the ECT of only 78 degrees, the 1429 RPM (cold idle?), and the Veh Speed indicating the car wasn't moving ("0"). This is already getting long, so I'll comment about the captured live data in a following post.

I don't think I've ever before made this much effort to diagnose in order to avoid "trying" parts. Don't get me wrong, I'm not complaining, it's good brain exercise. After posting last night, I awoke this morning and realized that we should look more closely at what I wrote in the second paragraph of my previous post (#158). I'd like you to try the manual trigger live data capture a couple more times, with the engine at least somewhat warm. In order to see if the problem relates to CL - OL-Drive - CL transitions, it would be very helpful to have slightly different data. The values to obtain are the Fuel Sys (OL, CL or OL-Drive), STFT, LTFT, and Calc Load for all 20 frames. If the results are sufficiently clear, this could point more directly to the culprit. If not, the fallback is what I wrote in the last paragraph of post #158.

The resistance test didn't actually check the knock sensor, since by dismounting it, it was disconnected from the circuit. As I tried to explain, one connection of the sensor is made to ground via being mounted by the bolt. The way your meter was used, it was connected to (and measured) the ECU's knock sensor input. (This isn't meant to fault you -- it's more of a realization on my part of the difficulty in conveying certain information.) With no change occuring in the ignition timing when tapping on the knock sensor, it might be that the sensor is okay, even with the single hairline crack you noticed. As you noted, based on the data, the system is going from "CL" to "OL-Drive" at some point preceding the stall. As I explained previously, that would be expected when decelerating for a stop (closed throttle). When "OL" (due to drive conditions or otherwise), the A/F sensor output is ignored by the ECU, and therefore there's no feedback to the ECU at that time about the mixture, meaning it falls back to default values and can't correct for certain possible faults. Once the engine revs drop to near idle speed, the system should go back to "CL", when the ECU uses A/F data again. However, if the engine is already faltering, the feedback might be too late to allow for a recovery, especially if the A/F sensor is sluggish in operation. So, since you have the Adult Auto Class coming up, I have one more idea before you "bite the bullet" and replace the A/F sensor (and possibly the knock sensor). If it can be done at the class, I suggest that you check the fuel pressure that the FPR is providing to the injectors at idle. In order to make sure that low engine vacuum isn't causing the FPR to raise the pressure, you can disconnect the vacuum line coming from the engine, plug it, and connect a hand vacuum pump to the FPR, pumped to about 20 in Hg. If the fuel pressure is too high then, either the FPR is bad or the return line is blocked. If the pressure is correct, I'd take a chance on the A/F sensor. I'm wishing you some good luck at this point!

If all the ABS-related fuses are good, then I'd suggest reading the ABS code(s). In general, a continuity tester could prove useful.

Tap on the sensor itself, both on the flat area near the bolt head, and around the side. Connect wherever it's more convenient for you. The knock sensor wiring may have one or two electrical connections. If there are two, one of them will go to an electrical shield that doesn't actually connect to the sensor element. The element is connected via it's output terminal and the ground connection made by mounting it. The resistance can range from 400 kohms to 700 kohms, and is typically closer to the middle of those values, about 550 kohms. If you ground one lead of your ohmmeter, then placing the other meter lead on the knock sensor output terminal should result in a reading in that resistance range -- a shield connection should be "floating", and have no continuity (infinite resistance) to ground. Assuming a firm connection of the meter leads, the resistance should not vary much, if at all, when the sensor is tapped. Can we assume that you initiated the trigger as soon as possible after the stall?

I meant to say "clock spring/roll connector", not "roll spring" I need more sleep.

John, start by rechecking fuses, specifically #18 in the interior fuse box (10A, center one in bottom row). It's common to the ABS and cruise control modules. The roll spring is common to the cruise and horn. Theory: I wonder if a problem in the steering wheel might have taken out that fuse.

It can happen (as can a bad radiator cap), but "air" in the coolant is more typically due to leaking head gaskets, assuming the cooling system was properly filled to begin with. EDIT: Dave beat me to it!

Very possibly. Even a slight contamination can cause a problem. The '97 has a hot-wire type MAF. MAF wires are pretty fragile -- as you found out, compressed air of any significant velocity damages (breaks) them. The replacement, if a good used unit, will probably resolve the problem.

For checking whether the sensor is responding to "real" knock, I agree that tapping with metal is the way to go. However, when a knock sensor is failing (usually due to cracks), it sometimes develops a sensitivity that causes it to produce output when it shouldn't. Since there's some (apparently erratic) spark retardation happening, tapping with something non-metallic would tend to show up that sort of failure. So would a resistance check, if the reading is out of spec or changes with impact. I'm just trying to determine if the later than expected spark is a result of the ECU thinking there's too much load, or because the knock sensor is misleading the ECU. (Chicken or egg, again. )

It's probably covered in the manual on Josh's server: http://www.main.experiencetherave.com/subaru_manual_scans/2000+_Legacy_FSM/

I think that the "End" trigger point would provide the most useful data. By the time the button is pressed, the "event" (stall) has already occured -- being able to look back and see what change(s) the ECU saw preceding it should be helpful. Data after a stall would likely be unrevealing, since most of the sensors will have seen dramatic shifts in values at that time. Having all 20 frames would make it more likely to spot a trend. I can't detect a pattern based on the data you've provided so far. Let's see if some more, taken using the "End" trigger point, will help. If you're finding that the stalls occur often enough with the engine warm, then not all the tests have to be run with a cold engine. "OL-Drive" is open-loop operation, but related to driving conditions. The ECU can't use the A/F and other sensors to regulate things well under certain circumstances, such as heavy acceleration (tendency to go too rich) or closed-throttle deceleration (tendency to go too lean), so it uses default values. By the way, there's also "OL-Fault" and "CL-Fault", which the ECU can use if it detects certain sensor failures. Subaru engines from the early 200x's are all very similar, so the 2003 knock sensor info should be applicable to your 2000. With the engine warm and idling, you could check "Spark Adv" with the 3130 while tapping on the knock sensor, as Fairtax4me suggested (although I'd use a wooden dowel or similar material, not something metal), and see if the timing changes much. I've found that checking the sensor's resistance (electrically disconnected from the harness, engine off, measuring between the sensor output lead and engine ground) while tapping can be a better indicator of cracking, etc. If you do either test, let us know what you find.

That is a higher than normal idle for a warm engine -- it should be closer to 700 rpm.

When you replaced the knock sensor, did you follow the torque spec when tightening the mounting bolt? The spec is for relatively low torque -- exceeding the spec significantly can damage the sensor, but that may not be immediately obvious. Also, the rotational angle the sensor is mounted at can affect operation. If you don't have the info, it's in: http://www.subaru-car.ru/leg3/leg3_sec1_1-1.pdf I'm sure that was somewhat disappointing. The old switch was apparently acting up, but not to the degree that would likely cause what you're experiencing. However, eventually it would probably have gotten bad enough to create problems, so it was prudent to change it anyway. The mass of the clutch is apparently acting as an additional flywheel, helping the engine get past cylinder misfires, bogging, etc. Sometimes people who install a lightweight crank pulley and clutch/flywheel (with the expectation that reducing rotational mass will improve free-revving) run into stalling problems.

Good question.

Using the PCV connection isn't practical, hence the ";)". As we've discussed, the best place to measure vacuum is at a central location on the intake manifold. Sometimes that isn't a choice. Since the FPR needs to know how "all cylinders" are doing, even if the point isn't ideal, it's adequate for your needs. The main difference the vacuum take-off point on the intake manifold will make in the measurement is that a central point tends to average it better -- that is, there's less pulsation. If you see more needle vibration than you'd like, you can dampen the movement by restricting the flow. That can be done with an orifice in the gauge's vacuum line, or something as simple as pinching the hose. Just be sure the needle movement isn't indicating an actual problem before you damp it out.

Are you saying that the high idle speed began occuring right after you replaced the knock sensor? No, there is no "screw" adjustment for idle speed -- the ECU controls it, based on data from the engine sensors. If the change in idle speed occured after replacing the knock sensor, then it might be that the ECU will need a few drive cycles to adjust to that. Or, you could try resetting the ECU. Another possibility is that in the process of replacing the knock sensor, you did something to create a small vacuum leak, which sometimes can cause what you're experiencing.

I thought we were discussing the connection of a vacuum gauge. If there's more that this is related to, you'll need to be specific.

Sorry, I finally had an opportunity to mull things over and get back to this. The biggest difficulty in pinning down the cause of the stalling is that of the old "which came first, the chicken or the egg?". The ECU is programmed to deal with changing conditions, as it determines from sensor data, and "corrects" things. Nothing in the OBD-II data stands out as obviously wrong, and we've been able to verify that data from several of the sensors is apparently correct. The parts that were already replaced and inspections that were done have ruled out other things. Although we've been trying to avoid "throwing" parts at the problem, I'm inclined to give replacing the front A/F sensor consideration again. But first I'd run one more diagnostic test, if it's practical for you and can be done safely (it might be best to have a passenger along to help with this). Connect the 3130 scan tool, put it in Live Data mode, and take a short drive with the engine cold. As soon as the engine stalls or bogs, make note if the system is "CL". If this test can be run with the engine cold over a few days, and the system is consistently "CL" when the stalling/bogging occurs, then the A/F sensor becomes a better candidate. The 3130 has a very handy function related to the Live Data Mode. It can record 20 frames of live data. One way of telling it when to do that is described in the manual in the RECORDING (CAPTURING) LIVE DATA section "Record by Manual Trigger". If you have the patience to read that part and can practice how to use it a few times before hitting the road, the test I mentioned in the previous paragraph can be easier to do and even more revealing, and you might not need a passenger/helper. If/when you run the test, let us know the result. Only the system state ("OL" or "CL") is needed for this, unless you see something else that seems worth noting. IMPORTANT EDIT: I almost forgot this -- given the erratic ignition timing (Spark Adv) readings, the knock sensor might be failing. They can be troublesome long before being bad enough to trigger a code. Cracking can lead to generating a signal when engine noise isn't high enough to nomally cause retardation. Sufficient retardation can cause inefficiency leading to low intake vacuum, leading to the FPR increasing the fuel pressure, leading to the ECU trying to control that via fuel trim, ... (chicken or egg? ) It would be prudent at a minimum to visually inspect the knock sensor, or consider replacing it even in the absence of a code, perhaps even before further investigating the A/F sensor.

There's the PCV, too. Looks like you'll have to "T" into the FPR vacuum line, since I doubt you'd want to drill and tap the manifold.

Can you post a pic of the intake manifold, near the TB?

You obviously first connected to a source of ported vacuum. What you want is a connection on the intake manifold, to the front of the TB.

One possibility is a defective engine coolant temperature sensor. If you have access to an OBD-II scan tool, you could check the live (real-time) data with the engine cold and see what temp is indicated. If it's not near to the ambient temp, the temp sensor might be bad. However, check first for corrosion at the connector before considering replacement.

Auto or manual trans? If auto, first make sure there isn't a fuse installed in the "FWD" holder/switch.