Xantia V6 broken exhaust (update: and rough low rpm running)

[stevieb]

Are you sure you're not confusing the fuel pressure regulator and the pulsation dampener Steve ??!

The fuel pressure regulator definitely does not stare anyone in the face when you open the bonnet. Looking into the engine bay its on the front (closest to the radiator) rail on the left hand (cam belt) side where one of the fuel hoses comes in. You can't even see it without removing the upper manifold much less get to it.

After the manifold is out the front rail has to be unbolted and lifted up, and the regulator comes out downwards towards the bottom manifold. As its usually very tight its very difficult but not impossible to get it out without completely removing the rail, but the risk of damage to breather hoses is high trying to do it in place.

You might be thinking of the pulsation dampener which is near the throttle body on the end of the rear rail - it looks just like a pressure regulator but without a vacuum hose. I could easily see that being removable without taking the manifold off, but that's not the one we're after...

Nope. it's definitely the regulator I'm thinking of. I'm sure it's accessible once the plastic engine cover is off, albeit a bit fiddly to get to... Mind you, I haven't been under the bonnet for a while now, so my memory might be letting me down. I know the front injector rail looks like it can be undone and rotated on its hoses without removing the manifold, which may prove handy as one of my front injectors is far from silent, although not as noisy as the one I took out.

[stevieb]

If it's any use, a lot of the sensor ranges are listed in the document here - http://www.coupe406.com/documents/Tutor ... nOu6vX4Bwh" onclick="window.open(this.href);return false;

Of course I've been assuming you've seen that document before. I'm sure I linked to it in my Coupé Club Invisible Caravan thread. I should apologise if it's new to you Simon, I should've brought it up before

I never paid any attention to them before (as they're no use to me) but there are scope readings for most of the devices. Hopefully they add to the raw data that's quoted there and make life a bit easier.

[Mandrake]

If it's any use, a lot of the sensor ranges are listed in the document here - http://www.coupe406.com/documents/Tutor ... nOu6vX4Bwh" onclick="window.open(this.href);return false;

Of course I've been assuming you've seen that document before. I'm sure I linked to it in my Coupé Club Invisible Caravan thread. I should apologise if it's new to you Simon, I should've brought it up before

I never paid any attention to them before (as they're no use to me) but there are scope readings for most of the devices. Hopefully they add to the raw data that's quoted there and make life a bit easier.

Nope, I've never seen that document before! Thank you! I don't mean to sound ungrateful, but is there an English version anywhere ? French is not something I learnt at school...

By the way have a look at the knock sensor page, in the text it says "avec une décrémentation maximale de 12 degrees". I don't have to read French to know what that is saying! Google translate confirms - "with a maximum decrease of 12 degrees". Here's an image capture of the page in question:



Given that its talking about the knock sensor that seems to nicely confirm my suspicion that the timing is being retarded by the MAXIMUM possible amount at low rpm when the performance is poor - I actually measured 1ms / 12 degrees difference at 2000 rpm between immediately before and after a battery off reset, which exactly agrees with the maximum figure in that text.

Anyone with a bit of French care to translate the pertinent details from that text ? Addo ? There could be some key information in that text that helps us better understand the functioning of the knock sensor and the specific values.

Really making some headway here. (or at least confirming what I already thought)

[Mandrake]

Well I manually typed it into google translate (complete with diacritical marks, ugh) and this is what it came up with:

The knock sensor , piezoelectric type is mounted on the engine block. On V6 engines, the knock sensor is at number two .

These sensors can detect knocking, vibration phenomenon due to an explosive ignition of the mixture in the combustion chamber.

This repeated phenomenon can cause destruction of the mechanical parts by abnormal rise in temperature of the walls . This sensor provides a voltage corresponding to engine vibrations. After receiving this information , the computer performs a reduction of the advance to the ignition cylinder or 2 degrees concerned with a maximum decrease of 12 degrees .

The réincrémentation will gradually ( 0.5 degrees every 120 PMH vicinity)

Parallel to the withdrawal of advance , the computer applies a richer air / fuel mixture to prevent excessive temperature rise of the exhaust gas which could lead to the destruction of the catalyst.

Rapid acceleration , the voltage delivered by the knock sensor should be between 0.1 volts and 1 volt : this voltage is alternative.

So it looks like it says it takes 2 degrees off the timing each time it detects knock until it stops knocking, to a maximum of 12 degrees, then with no knocking detected advances it at 0.5 degrees per whatever "120 PMH" is !! I don't think that bit translated very well, but I'm going to guess that its 120 revolutions of the engine. If that's the case it would take 2880 revolutions of the engine with no knocking to adjust the timing from fully retarded 12 degrees back to normal - at 2000 rpm that's 86.4 seconds. That also assumes that you stay at the exact same rpm and engine load for that time as every cell learns it own value and would have to be individually re-learnt!

So yes, it takes a long time for the timing to fully recover from being heavily retarded across a wide range, when you consider there will be at least a dozen load/rpm cells.

[stevieb]

I saw the 12° bit and wondered if you'd notice (not that I had any doubt ).

You might be right about PMH, as the crank position sensor data states something about "two teeth being deleted for the recognition of PMH" (and no, I don't speak any French - I've been using Google Translate too). It crops up under some of the other sensors too. Something makes me wonder if the H stands for "heures" though. Could it be some kind of a refresh cycle in the ECU? Or worse, maybe PMH means startup...!

EDIT - Oooh, ooh!!! Just Google Translated the Le Capteur Regime Moteur Actif" and PMH comes up as TDC.

Are we to assume the ECU reads TDC on every revolution of the engine, or is it something that calibrates at startup based on the crank position sensor's data? I'd assume the former, otherwise how would it know where the engine is in the cycle, but in the case of "PMH", maybe it does mean when it calibrates TDC at startup.

One other question arises. Do you honestly think there's enough knocking going on in the engine, caused by what is comparatively mild lean-running, to force retardation all the way back to the maximum 12°? Especially if you can't even hear the knocking...? I'd be heading back to the MAP sensor as a possible culprit - is it understating the vacuum readings and causing the ECU to under-fuel?

[Mandrake]

EDIT - Oooh, ooh!!! Just Google Translated the Le Capteur Regime Moteur Actif" and PMH comes up as TDC.

Ah, nice find. So it is engine revolutions then.

Are we to assume the ECU reads TDC on every revolution of the engine, or is it something that calibrates at startup based on the crank position sensor's data? I'd assume the former, otherwise how would it know where the engine is in the cycle, but in the case of "PMH", maybe it does mean when it calibrates TDC at startup.

No I don't think so - the way I interpret it is the reference to TDC just means it's counting engine revolutions - every 120 revolutions it will try to sneak the timing forward half a degree if it's currently retarded, then after another 120 revolutions it tries another 0.5 degrees and so on until it reaches maximum/optimal advance or it starts knocking. When/if it knocks it then immediately drops it back by 2 degrees and the process starts over. So it's very quick to retard the timing (probably the very next spark for the next combustion stroke) but only slowly and carefully "probes" to see if it safe to bring the timing forward.

One other question arises. Do you honestly think there's enough knocking going on in the engine, caused by what is comparatively mild lean-running, to force retardation all the way back to the maximum 12°? Especially if you can't even hear the knocking...?

Well the 12 degrees of timing retard below 2500 rpm is definitely there - I've measured it. As for whether there is enough knocking to trigger it, yes I do believe so.

Knocking isn't really audible on modern engines with knock sensors, at most barely audible if its severe. Remember that the moment it detects a knock or "ping" in a cylinder (which is a momentary high pitched sound about 6.3Khz for this engine bore) it yanks the timing back almost instantly so the knocking stops. If it pulls out 2 degrees per knock event it would only take 6 knock events in a row to pull out all 12 degrees.

Even at a low 2000 rpm that is 33 1/3 revolutions per second, on a V6 there are 1.5 power strokes per revolution so 50 power strokes per second. 6 into 50 is 0.12 seconds - so it could knock 6 times in a row and pull out the maximum 12 degrees of timing in just over 1/10th of a second - could you really hear knocking that was a high pitched noise that only lasted 1/10th of second or less amongst all the other noise ? Doubtful.

I have never "heard" knocking on this engine audibly, except when I've driven it with the knock sensor unplugged. Then I've definitely heard it, but even then it's not loud or obvious, it's a relatively subtle noise and it's only because I was listening carefully for it in just the right driving conditions. It was audible at WOT at low rpm in the 1500-2500 range. (Surprise surprise)

As soon as I plugged the knock sensor back in I couldn't hear it any more - because the knock sensor was doing its job and causing the timing to get retarded to the point where it was no longer knocking. If you could "hear" the knocking it would mean the knock sensor wasn't doing its job and allowing the engine to continue knocking.

Another thought I have is whether a lean misfire could also cause a noise that "sounds like" knock to the ECU - any erratic combustion event where the pressure oscillates up and down very rapidly due to the mixture not burning smoothly, instead of a smooth increase and decrease in pressure might "sound like" knock. (see the example pressure diagram on that knock sensor page)

Either way it doesn't matter what the precise mechanism is if the root cause is the lean mixture - it still has to be fixed. I think its just too much of a coincidence that the lean O2 reading and the retarded timing occur at exactly the same rpm range with the timing retard cutting off at the same rpm (2500) that the mixture is going lean.

[Mandrake]

That is EXACTLY what was happening to mine this time last year. The car would run perfectly fine to work in the morning, then be sat in the sun all day and 100 yards out of the car park on my way home performance would drop like a bomb. I'd actually feel the power drop off, like the handbrake had been pulled on or the invisible caravan had been hitched-up. I'd always claimed the ignition was retarding a huge amount, and you've probably just confirmed it. It sounds like you're getting the point I was making earlier about the IAT sensor (though I made the point very badly).

Yes, sounds like the same thing. With warm coolant from having been run earlier in the day (even 8 hours later your coolant is probably still 30+ degrees) plus sun warming the car the cold start enrichment period would be very short - maybe a minute at most before it switched back to a "normal" driving mixture, normal in this case being lean because of the fault.

If that IAT has been heat soaked in the engine bay all day in the sun it could be reading 30 degrees or more at first even though the air might only be 17. The cooler denser air would immediately be entering the engine but the engine would be calculating injection pulse for less dense 30 degree air so the already too lean mixture would go even leaner, possibly to the point where you get an actual lean misfire - that's sure how it felt to me yesterday, a kind of "tugging" back and forth when trying to accelerate quickly.

Today the weather was very different - overcast all day, rainy, a maximum of 9 degrees and the symptoms from yesterday were not present at all.

Again - I don't think there is anything wrong with the IAT - if the mixture was correct in the first place the small error in mixture due to the IAT temporarily reading too high due to heat soak would have an insignificant effect. But if the mixture is already very lean and on the verge of producing a lean misfire or inducing knocking, it could be the straw that broke the camels back.

Like so many other things that have a minor effect on the mixture - when its so marginal to begin with the effect is blown out of all proportion.

If I get another really hot day (Friday and Saturday are looking promising) I'll try substituting a resistor that will trick it into thinking the air temperature is near zero and see whether the same symptoms appear as yesterday - if they don't then its likely the heat soaked IAT theory has some merit.

I know you've said in the past that there was a difference in performance between setting off quickly after a start-up and setting off after a short delay (though you put that down to the oxygen sensor), but I'm sure you've also said that the IAT only influences the ECU for a set period after start-up or within a certain rpm-range.

I don't recall saying that. The IAT is used all the time in open loop mode - its a fairly important input along with the MAP sensor for calculating the weight of air entering the engine. In closed loop mode the O2 sensor has the final say, but closed loop mode is only used during constant partial loads.

[stevieb]

Well the 12 degrees of timing retard below 2500 rpm is definitely there - I've measured it. As for whether there is enough knocking to trigger it, yes I do believe so.

Knocking isn't really audible on modern engines with knock sensors, at most barely audible if its severe. Remember that the moment it detects a knock or "ping" in a cylinder (which is a momentary high pitched sound about 6.3Khz for this engine bore) it yanks the timing back almost instantly so the knocking stops. If it pulls out 2 degrees per knock event it would only take 6 knock events in a row to pull out all 12 degrees.

Even at a low 2000 rpm that is 33 1/3 revolutions per second, on a V6 there are 1.5 power strokes per revolution so 50 power strokes per second. 6 into 50 is 0.12 seconds - so it could knock 6 times in a row and pull out the maximum 12 degrees of timing in just over 1/10th of a second - could you really hear knocking that was a high pitched noise that only lasted 1/10th of second or less amongst all the other noise ? Doubtful.

I have never "heard" knocking on this engine audibly, except when I've driven it with the knock sensor unplugged. Then I've definitely heard it, but even then it's not loud or obvious, it's a relatively subtle noise and it's only because I was listening carefully for it in just the right driving conditions. It was audible at WOT at low rpm in the 1500-2500 range. (Surprise surprise)

As soon as I plugged the knock sensor back in I couldn't hear it any more - because the knock sensor was doing its job and causing the timing to get retarded to the point where it was no longer knocking. If you could "hear" the knocking it would mean the knock sensor wasn't doing its job and allowing the engine to continue knocking.

Another thought I have is whether a lean misfire could also cause a noise that "sounds like" knock to the ECU - any erratic combustion event where the pressure oscillates up and down very rapidly due to the mixture not burning smoothly, instead of a smooth increase and decrease in pressure might "sound like" knock. (see the example pressure diagram on that knock sensor page)

Either way it doesn't matter what the precise mechanism is if the root cause is the lean mixture - it still has to be fixed. I think its just too much of a coincidence that the lean O2 reading and the retarded timing occur at exactly the same rpm range with the timing retard cutting off at the same rpm (2500) that the mixture is going lean.

My thought is that the ECU can only retard to 12°, but that's no guarantee that it's the end of the knocking. Regardless of cause, if there's sufficient knocking to cause the retardation and it goes on long enough (six knocks and it's out - but obviously not all in quick succession in your case), then unless you're very lucky and 12° of retard is precisely the right amount required to stop the knocking, then surely given the choice the ECU would choose to retard even further if it could. Assuming the ECU doesn't re-advance the timing by any noticeable margin it would suggest the knocking is still present even at 12° retardation (am I right in thinking you've seen the knocking on the scope at 12° retardation?)

I would've thought in this case either the knocking would be pretty severe to start with - to the point that full retardation would be almost instant - which I think was happening with mine on the warm afternoon journeys home from work last year, or if it's a gradually increasing "minor" knocking that gently pushes the tables back one at a time until the ECU hits its limit (which yours sounds like) then I would expect it would continue to show itself until it'd be audible (working on the theory that the first knock pushes the timing 2° back, things settle, then it reoccurs and timing goes back to 4° etc - if the knocking continues beyond 12° then surely it would eventually start to become apparent). But if this is the point at which the ECU changes octane map then shouldn't the knocking disappear (at least temporarily), because by now the engine would be running 95RON on a much lower (91RON) map and be able to re-advance again.

When mine was knocking I found it very easy to instigate, and it was very audible. There's a road local to home with a steep climb and a set of lights right where it levels-off at the top. At off-peak times the lights are approach-controlled. If you ease off at the right time you can trigger the lights to change without having to touch the brakes or come to a stop, but you have very little kinetic energy left in the vehicle, so easing away from a crawl in second or third labours the engine a little if you're only just above tickover. I could guarantee mine would knock at that point, every time. Based on your findings, my ignition would've been at full retard and still knocking. As I said though, I believe the cure was the new MAP sensor.

[stevieb]

I don't recall saying that. The IAT is used all the time in open loop mode - its a fairly important input along with the MAP sensor for calculating the weight of air entering the engine. In closed loop mode the O2 sensor has the final say, but closed loop mode is only used during constant partial loads.

I could've sworn it was in this thread - maybe it was in my Coupé Club thread. I know it was suggested by somebody that the ECU uses readings from the IAT at startup alongside readings from the MAP sensor to allow the ECU to calculate/calibrate the air density, but when the cold start enrichment drops off the ES9J4 ECU then only polls the IAT on a set cycle, and not constantly. The light-hearted criticism that was raised was that in the UK where weather conditions can change in minutes, the ECU might be fuelling for 15°C atmospheric temperature, but then ten miles down the road you hit a hailstorm where the temperature is nearer 5°C, but the ECU is still fuelling for 15° because it isn't constantly monitoring... I guess if this were the case then some wayward readings from other sensors might force the ECU to poll the IAT sooner, but I'm guessing you're going to tell me I've been told wrong...?

On the off-chance you don't though, let's just assume that once the enrichment has stopped the ECU ignores the IAT for a period of five minutes. Taking the engine bay heat-soak idea into account, the IAT could read artificially high at startup, and if there's little or no enrichment required then the IAT won't have much chance to cool in the airflow before the ECU starts ignoring it - so those initial IAT readings may stick for the first five minutes until the ECU re-reads the IAT and realises the change. However, in this time the ECU will have been under-fuelling very slightly, and possibly enough to cause the lean ignition knocking (in those very specific rpm/load circumstances) that's pushing your ignition tables back one at a time.

I think you could definitely do worse than trying the resistor trick. If you can coax the ECU to fuel that little bit more, then it'd be interesting to see if the leanness seen at the O2 sensor disappears, and if so whether the knocking disappears with it. If the leanness drops but the knocking still occurs then I'd concede that you've probably got a duff injector that can't just keep up and is causing one cylinder to run very lean and knock. But if a tiny bit of enrichment via the IAT eliminates both problems, then I'd be more suspicious of the sensors (my money's still on the MAP sensor ).

On the subject of MAP sensors, there was a part revision somewhere along the production run. PSA ref:1920 1K, Bosch ref: 0 261 230 034 was replaced with PAS ref: 1920 AN , Bosch ref: 0 261 230 057.

[stevieb]

Another thought.

The weird reset/re-calibration thing with the ignition key helped once before, so it must've changed something. Have you been able to establish with your readings what, if anything, changes after this little trick? It makes you wonder if it's a shortcut for the dealer grease-monkeys to reset the octane maps or the knock tables without losing the customers' radio presets

Or maybe something else...

[Mandrake]

Another thought.

The weird reset/re-calibration thing with the ignition key helped once before, so it must've changed something. Have you been able to establish with your readings what, if anything, changes after this little trick? It makes you wonder if it's a shortcut for the dealer grease-monkeys to reset the octane maps or the knock tables without losing the customers' radio presets

Or maybe something else...

That key on/off three times for 10 seconds sequence reset hasn't worked reliably for me unfortunately. I don't think it does anything that a 10 minute battery off reset doesn't do, except its a lot less reliable. (probably because we don't know the exact timings that its looking for) Because of that I haven't checked any of my measurements after attempting it - I'm happy that the battery off reset told me what I need to know.

[Mandrake]

I could've sworn it was in this thread - maybe it was in my Coupé Club thread. I know it was suggested by somebody that the ECU uses readings from the IAT at startup alongside readings from the MAP sensor to allow the ECU to calculate/calibrate the air density, but when the cold start enrichment drops off the ES9J4 ECU then only polls the IAT on a set cycle, and not constantly. The light-hearted criticism that was raised was that in the UK where weather conditions can change in minutes, the ECU might be fuelling for 15°C atmospheric temperature, but then ten miles down the road you hit a hailstorm where the temperature is nearer 5°C, but the ECU is still fuelling for 15° because it isn't constantly monitoring... I guess if this were the case then some wayward readings from other sensors might force the ECU to poll the IAT sooner, but I'm guessing you're going to tell me I've been told wrong...?

Nope, the IAT is monitored all the time and in open loop mode the IAT is always a factor in calculating the mixture. It's the MAP sensor barometric pressure calibration that only occurs at key on engine off and WOT at low rpm.

I think you could definitely do worse than trying the resistor trick. If you can coax the ECU to fuel that little bit more, then it'd be interesting to see if the leanness seen at the O2 sensor disappears, and if so whether the knocking disappears with it. If the leanness drops but the knocking still occurs then I'd concede that you've probably got a duff injector that can't just keep up and is causing one cylinder to run very lean and knock. But if a tiny bit of enrichment via the IAT eliminates both problems, then I'd be more suspicious of the sensors (my money's still on the MAP sensor ).

I will try the resistor trick - I have a box full of small resistors and with the Lexia to tell me what temperature it thinks it reads I should be able to find one that reads somewhere around 0-5 degrees. If it permanently reads the same even after the engine warms up the ECU might get a bit upset and stop believing the sensor, (rationality testing) but we'll see. If I'm right and the ECU is fooled by the low air intake temperature reading it should run significantly better (but probably still not normally) especially on a warm day. (In fact the warmer the weather gets the richer it will run with a fixed temperature reading going to the ECU) If it makes enough difference I might even leave the resistor connected until I've fitted the cleaned injectors, as running lean under load is not good for the engine even with the knock sensor trying to protect it.

I've sent my injectors off for cleaning today, so there's no turning back! I've literally put my money where my mouth is with my lean running restricted injectors diagnosis...

Here they are in the process of being individually wrapped and packaged like small Christmas gifts.



I'm sending a spare fuel pressure regulator too so he can verify their 3.5 bar type is physically compatible before sending it - if it is I'll fit it, if not he's going to refund me the cost of the regulator and I'll fit the best one I have of the ones I already have - which measures 2.9 bars.

By the way, some interesting reading on dirty injectors here:

http://injector-rehab.com/shop/tsclean.html" onclick="window.open(this.href);return false;

[stevieb]

I tried the resistor trick for a couple of niggles last year. The main one was the CTS. At first I suspected the reading wasn't steady, so obviously I took meter readings at all temps and they consistently came back accurate. Then I began the suspect the plug, because on mine the wiring into it is strained slightly by the cable guide and the rubber boot is past its best. With a warm engine and a matching resistor in place I could wiggle the plug and wiring to my heart's content to see if I heard any obvious changes in engine note - hardly scientific, but good enough for me to conclude that the wiring was fine.

I wish you luck with the injector cleaning, but I still wonder how much improvement you'll get. The peace of mind of knowing they've been done will definitely be worth it though. It's for that reason I'd be getting mine done if I had long-term plans for the car, but seeing as I'm heading back to the realms of 20,000-30,000 miles per year and an ageing V6 petrol isn't exactly the most practical solution any more, it just doesn't seem worth the hassle, not to mention "if it ain't broke don't fix it".

I know I suffered the same problems in-extremis last summer, but that was with an injector that was ticking loudly and clearly giving trouble. Without any extreme or obvious signs like mine had I can't imagine your problem is being caused by the injectors, maybe just exaggerated by them, so it'll be interesting to see the outcome.

For the benefit of anybody chancing across this who suspects their MAP sensor, I found some meter readings I took from the three MAP sensors I had. The figures are based on having the MAP sensor opening flat down on the table with the three terminals facing you, numbered 1 to 3 left to right. Readings were as follows:

Original Bosch sensor:
pins 1-2 - 4k ohm
pins 2-3 - 6k ohm
pins 1-3 - 5k ohm

Intermotor sensor:
pins 1-2 - 420k ohms
pins 2-3 - 11k ohms
pins 1-3 - 430k ohms

New Bosch sensor:
Pins 1-2 = 4.4k ohm
Pins 2-3 = 5.7k ohm
Pins 1-3 = 5.4k ohm

I think both the Bosch sensors were electrically sound, but I couldn't vouch for the mechanical element. The Intermotor sensor was clearly shot. Hope it helps somebody

[Mandrake]

I wish you luck with the injector cleaning, but I still wonder how much improvement you'll get. The peace of mind of knowing they've been done will definitely be worth it though.

[...]

I know I suffered the same problems in-extremis last summer, but that was with an injector that was ticking loudly and clearly giving trouble. Without any extreme or obvious signs like mine had I can't imagine your problem is being caused by the injectors, maybe just exaggerated by them, so it'll be interesting to see the outcome.

I'm surprised you're so sceptical, given the change in your car when you swapped the injectors. Or are you just sceptical that the cleaning wouldn't restore normal operation if the injectors were blocked ?

For the benefit of anybody chancing across this who suspects their MAP sensor, I found some meter readings I took from the three MAP sensors I had. The figures are based on having the MAP sensor opening flat down on the table with the three terminals facing you, numbered 1 to 3 left to right. Readings were as follows:

I hate to be a party pooper but the MAP sensor has an integrated circuit inside it - so taking resistance measurements of the terminals is not a valid test of whether it's working ok or not.

Anyway, I have some excellent news! I've just tried the resistor in place of the IAT trick tonight not really expecting it to do much, or make a very small change if anything, and I'm absolutely blown away by the difference! For such a simple test I wish I'd tried it months ago!!

The Lexia reported the IAT temperature was 17 degrees (after a short drive) and an ohm meter showed it was 2.5k ohm at that temperature, I wanted to trick it into thinking the air temperature was just above freezing, a little bit of experimentation and I found that 5.6k ohms registers as +2 degrees C.

I left the Lexia on the car during a test drive to make sure it didn't detect a fault and substitute a "replacement value" - for example if you unplug the sensor (completely open circuit) it will initially read -48 degrees C then after a few seconds it detects a fault and uses a replacement value of +39 degrees. Anyway, it seems to be happy to believe the intake air is 2 degrees all the time, so the ruse worked. I didn't disconnect the battery during the change or attempt any kind of reset.

As soon as I started the engine I noticed a difference - on a warmed up engine the oxygen sensor was previously showing that the mixture was lean during the first 30 seconds of idling before it went into closed loop mode - with the change it was now showing as rich before entering closed loop mode (as it should be) and the idle was smoother too. So straight away proof that it's running significantly richer.

The next thing I tried is flooring the accelerator at idle which was previously causing a severe stumble. There was still a slight hesitation but nothing like before, with the engine accelerating from idle MUCH faster and smoother than before. Another good sign.

So off I went for a drive - immediately I noticed it had a lot more pull and eagerness below 2500 rpm. Because I hadn't disconnected the battery the memorised retarded timing would still have been there to begin with but in theory in the absence of lean induced knock it would gradually start to correct by itself as you put the engine through different load/rpm combinations for long enough.

Sure enough performance got better and better in leaps and bounds the more I drove, after 10-15 minutes it was massively better. The flat spots in the torque curve that were causing power fluctuations during acceleration were slowly disappearing as if someone was ironing out the kinks.

The scope on the O2 sensor was clearly showing rich under WOT with almost no sign of the lean running I've been seeing for months.

I decided to hasten the learning process by resetting the ECU, another big jump in performance and responsiveness, and still showing rich under WOT at all rpm. After the ECU reset the power "fluctuations" during acceleration that have been so pervasive for months are all but gone with only the tiniest hint still remaining. Acceleration is strong and smooth right through the rpm range.

The extra performance seemed to be confusing the gearbox which was now changing gears somewhat roughly, I ended up doing an auto-adaptive reset on the gearbox and after that and 10 minutes of driving the gear changes are now lovely and smooth again for the first time in quite a while. (The ECU controlled gearbox is very susceptible to being confused and giving poor gear change quality if the engine performance is erratic and unpredictable - it relies on smooth, accurate power delivery to achieve smooth gear changes and smooth torque converter lockup transitions)

I did maybe another 30 minutes of mixed urban and motorway driving and it just got better and better the whole time, rather than getting worse as it usually does when it heats up. It was absolutely flying up the hills, in fact I'd say that performance seemed fairly close to normal. I also noticed the tendency for the rpm to drop below normal idle speed then surge when pulling to a stop seems to have gone.

Now clearly this is not a proper fix, I've just tampered with the IAT signal to fool the ECU into providing a wider than normal injector pulse width to partially compensate for a lean condition - but how much richer ? 5% ? 10% ? 15% ?

I looked up the density of air at 15 degrees and 0 degrees on Wikipedia:

http://en.wikipedia.org/wiki/Density_of_air" onclick="window.open(this.href);return false;

This should be near enough to 17 degrees and 2 degrees. Dividing the two density figures shows that 0 degree air is only 5.4% denser than 15 degree air - so if we assume the injection quantity increases proportionally to the air density that means my tweak to the IAT signal which has caused such a massive change in performance has occurred (at the prevailing ambient temperature) with a mixture only 5.4% richer! (The higher the ambient temperature in the intake the richer it will run with a fixed IAT signal)

This is much less change in mixture than I had expected for such a big result, although I did read something yesterday that said that an injector flow rate reduction of only 10% from normal is enough to actually cause a lean misfire, so it does seem reasonable in light of that.

Even with this tweak in place I still don't think it's quite as rich as it should be, but I think this is fantastic proof that the lean mixture is the problem. It helps eliminate all sorts of other possibilities like a blocked exhaust, (although I'd long ago eliminated that) mechanical fault in the engine (compressions, valve gear etc) or ignition misfire - the car was absolutely flying tonight and running far far better than it has in a long time and showed no sign of "sagging" after being driven for a while, it just continued to improve.

From what I've seen tonight I'm completely satisfied that the lean mixture is the entire cause of the power loss issue, and that the timing retard is indeed resulting from the effects of a lean mixture. I still don't know for certain that it's injector restriction causing the lean mixture, but we will soon know.

In the mean time it's running VERY well with a little 5p resistor poked into the end of the connector and taped up...I now also understand why it runs so poorly with the IAT completely unplugged - after a few seconds it inserts a "replacement" value of 39 degrees - which is going to make the mixture MUCH leaner than it already was, probably to the point where it lean misfires under load.

[RichardW]

Density close enough Simon - it varies by the ratio of the absolute temperature (low temp / pressure pretty close to ideal gas) so (2+273)/(17+273) = 0.948 = 5.2% difference.

I too am surprised it makes that much difference, but you can't deny the outcome! Didn't this effectively start all at once though, so something must have broken or gone wrong to cause it - unless you just got a cruddy tank of petrol that blocked the injectors - and the ones you are running now are little better....!

Edit - my only confusion is why the O2 sensor doesn't overule the ECU, by saying "too lean" - it appears that the ECU is taking the IAT as gospel over the oxygen sensor? Why would it do that? Or is it because it is running in open loop at WOT? In which case why doesn't it advance the timing back once you are off WOT? Or maybe that should just be ignored, and revel in the fact that it WILL run OK

Now all we have to do is make it stop!