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

[Chris570]

I've personally seen three or four engines where the worn timing chain has chewed through the timing case or rocker cover.

If Simon's problem is knock sensor related, I'll eat one of my 39 hats.
Sent from my GT-I9300 using Tapatalk 2

Will you video it?

As a side note my drivers door doesnt lock on the remote but locks on the key, do Wynns do something I can pour into the oil to make it right? I know there will be a broken wire in the door shut but I'd rather go this route than spend the money on 6 inches of wire and a bit of solder

[Ben82]

As a side note my drivers door doesnt lock on the remote but locks on the key, do Wynns do something I can pour into the oil to make it right? I know there will be a broken wire in the door shut but I'd rather go this route than spend the money on 6 inches of wire and a bit of solder

Think the one at 0:21 is what you need

I'm skeptical it's the additive that has fixed the issue too, but will keep an open mind. I guess it could be a combination of the knock sensor being faulty and the extra noise from the tappets sending the ecu into a tiz. But I very much doubt it's the tappets alone.

[Mandrake]

I'm not saying the additive has done nothing, but its addition also corresponded - did it not - with a measurable rise in ambient temperature?

Nope.

The weather is much the same before and after it was added, and there hasn't been any recent correlation between ambient temperature or weather and performance, at least that I've noticed.

Many months ago when there was an intermittent ignition related misfire due to a weak spark the car ran a lot better in very dry low humidity conditions and really struggled in wet high humidity conditions, to the point of spluttering, but since the ignition was fixed that correlation has gone.

The only things that I can positively correlate with changes in performance in recent times are:

1) Runs a lot better immediately after resetting the ECU, but then deteriorates with driving, sometimes minutes, sometimes hours of driving time.

2) Runs a lot better if its left to idle for a while before driving, especially if the ECU has just been reset. Runs much worse if driven away quickly after a cold start, and continues to run poorly for the rest of the trip.

3) Performance only ever gets worse the more you drive it in a given driving session, never better. If it starts off bad it stays bad. If it starts off good it typically gets worse. The only variation is how quickly it gets worse.

Given that the ECU reset restores normal performance immediately, suggests to me that the adaptive behaviour of the ECU is learning from "bad" data.

Yesterdays driving session post additive was the first one I've seen for a long long time where the performance started off pretty good but consistently got better and better the longer it was driven and was running the best at the end of the journey instead of the beginning as was previously the case.

People are free to form their own conclusion from the above, I'm only presenting the data.

[Mandrake]

I'm skeptical it's the additive that has fixed the issue too, but will keep an open mind. I guess it could be a combination of the knock sensor being faulty and the extra noise from the tappets sending the ecu into a tiz. But I very much doubt it's the tappets alone.

Ben I agree that it may not necessarily be noisy tappets causing the problem, but mechanical noise in the engine in general - I said as much in my posting where I said that 2-3 tappets were a bit noisy again at the end of the trip, however the performance of the car was still excellent and had not deteriorated at all. It could be noise deeper in the engine that isn't obvious when listening under the bonnet. However from the drivers seat the engine is markedly quieter when revving and under load now, so something has got quieter...

[addo]

That gives you an apparently anomalous situation once, against a sample of many dissimilar outcomes. Perhaps part of what's happened is simply a change to oil viscosity via the additive bottle.

What if the notch filtering ability of your ECU was at fault, and it misread knock sensor input?

[Mandrake]

That gives you an apparently anomalous situation once, against a sample of many dissimilar outcomes.

Agreed, no matter how dramatic the change, it needs more time to see if its a fluke or a permanent change.

Perhaps part of what's happened is simply a change to oil viscosity via the additive bottle.

There are other additives that do thicken the oil up to try to quiet down the engine (I believe Lucas is one) but Lubegard specifically claim that they don't modify the viscosity of the base oil, so I don't think viscosity is the difference.

What if the notch filtering ability of your ECU was at fault, and it misread knock sensor input?

Different ECU's vary greatly on just how smart they are at filtering out the sound of knocking from other mechanical noise. It's not nearly as easy as it seems, and other experiences with this ECU give me the impression that as ECU's go its a bit on the "dumb" and simplistic side, (it can't even detect misfires for example) and not nearly as sophisticated as the ME 7.4.6 on the ES9J4S for example.

[addo]

It is a very basic ECU; my understanding though is that they are dynamic always rather than taking a while to settle into a constant state during operation. So if a detrimental aspect changes during engine running time, after a few samples have reaffirmed the change in sensor reportage the ECU will modify its behaviour to take in the new situation.

I wouldn't credit the knock sensing with any great degree of sensibility either (eg, experimenting with retarding one coil vs another based on amplitude or attack/decay characteristics) and feel you may be overrating it. To my way of reading about them, the sensing circuitry constantly samples, saying "Background noise? Check. Spike above threshold?" and then reacts or doesn't react. 8000RPM is only 66 firing events per second, so it's not a demanding situation.

Why is it when someone says "eat", my mind goes out the window?

[addo]

New product I found on the shelf.

[Mandrake]

New product I found on the shelf.

[Mandrake]

It is a very basic ECU; my understanding though is that they are dynamic always rather than taking a while to settle into a constant state during operation. So if a detrimental aspect changes during engine running time, after a few samples have reaffirmed the change in sensor reportage the ECU will modify its behaviour to take in the new situation.

Yes, they're always dynamically learning the running characteristics of the engine, when the ECU is reset the long term fuel trim is reset to zero which means it may run a little rough if the engine is well out of spec (vacuum leaks, low fuel pressure etc) until the oxygen sensor kicks in and it starts correcting the long term fuel trim.

The knock retard table I'm not sure whether it starts with a fully advanced (high octane) map or whether it starts with a middle of the road one and creeps it forward in the absence of detected knock. Either way, retarding enough to switch to the low octane mapping is a long lasting change, the ECU is quick to knock the timing back but very slow to creep it forward again... so it wouldn't take too much intermittent false knock signal to retard the timing and keep it there semi-permanently if it happens frequently enough.

I wouldn't credit the knock sensing with any great degree of sensibility either (eg, experimenting with retarding one coil vs another based on amplitude or attack/decay characteristics) and feel you may be overrating it. To my way of reading about them, the sensing circuitry constantly samples, saying "Background noise? Check. Spike above threshold?" and then reacts or doesn't react. 8000RPM is only 66 firing events per second, so it's not a demanding situation.

I've already done quite a bit of research into how ECU knock sensing works, there are a number of different levels of complexity it can follow depending on how modern and fancy the ECU is, and at least a couple of different sensor types as well, resonant sensors or broadband.

Really old systems work more or less how you say - they measure the background noise as a sanity check and watch out for spikes that are well above that background noise, any large spikes are assumed to be knock. The problem with that is that knock isn't that much louder than the background noise so newer systems filter the frequency range that it's listening to down to a narrow range to filter out as much extraneous noise as possible.

The resonant type (which is what I suspect the ES9J4 has but don't know for sure) is a piezo that's mechanically resonant at the frequency pinging occurs at forming a narrow mechanical bandpass filter with a Q of about 3 to 5. The frequency of knock is said to be about 573/bore in mm, so for an 87mm bore that's about 6.6Khz.

Basically the shock wave from the knock hits the inside of the cylinder causing a bell like resonance at 6.6Khz, as if you'd tapped the inside of the cylinder with a small hammer, and the cylinder block rings at its natural bell resonance frequency which is largely determined by the bore but also a little bit by the bore material.

The ECU may apply additional filtering or the band pass filtering may be done entirely in the ECU instead of the knock sensor. Additional cleverness that can be added other than just looking for that one frequency is to only listen to the knock sensor at the part of the combustion cycle where cylinders may be knocking, so random noise appearing outside that time window will be ignored (I don't know if the MP7.0 does this or not) and even cylinder individual retard based on determining WHICH cylinder knocked. I believe Subaru Imprezza's etc have per cylinder timing control but I doubt the MP7.0 does that.

The ECU also has a pre-programmed table of how much noise to expect at different engine RPM and loads, so allows for more noise at higher RPM for example, by raising the acceptable threshold.

The problem with bandpass filtering is that impulsive noise will still excite a bandpass filter if its loud enough. Any mechanical tapping or clicking near the bores or anything "hitting" the bores (piston slap for example) will cause the block to ring and produce a small knock signal...if you tap the block with a small hammer it rings like a bell and will set off the knock sensor.

You might find these interesting - actual recordings of the output of a knock sensor played back as audio of an engine at different RPM both with and without knocking. (Best heard with headphones)

http://theknockbox.com.au/the-knock-box ... ecordings/" onclick="window.open(this.href);return false;

The knocking sounds a lot louder than the still clearly audible valve gear noise but remember it has already been bandpass filtered to accentuate the frequencies around knocking and filter out the other noises as much as possible, and that the human brain is remarkably good at separating different unrelated sounds, far more so than a computer. See how much harder it is to tell them apart at high rpm, which is why knock sensitivity is usually reduced high rpm as the engine noise itself is so loud.

Anyone who doesn't believe a knock sensor can hear valve gear needs to hear those recordings, especially 5000 rpm Of course the ECU should be smart enough to tell them apart, but its not necessarily foolproof if the engine is unusually noisy.

[CitroJim]

Simon I'll catch up on all thjis later - far too much for my feeble mind to even try digesting right now...

I love that 'Start ya Bastard' video

At the end of it a random video of a Class 37 locomotive attempting a cold start on a very cold day appeared..

Think it needs new glowplugs...



It gives an excellent demonstration of why train buffs call them 'Tractors'...

[CitroJim]

Simon, I've just changed my oil in the XM after about a thousand miles. I did a quick change as I'm not sure how often it was changed before I had the car...

I can say that mine was still quite clean and nowhere near as black as yours or really even black - brown but not really black. Yours is not normal... I'd get a couple of rapid changes done and see how it goes..

You may have found an issue with blow-by there and this may well account for a lot of your woes. Did you ever do a compression test?

Mine took exactly four quarts with a filter change...

[addo]

Excess blow-by and blackening of the oil happens with a constricted exhaust.

[Mandrake]

Compression test was good, 210 psi +/- 3 psi. No exhaust back pressure measured.

One thought I did have though is that it was running for a few months with an under load ignition misfire on one possibly two cylinders, could there have been unburnt fuel causing cylinder wash past the rings on that cylinder causing fuel contamination in the oil ?

[Mandrake]

Some more observations and thoughts in the ongoing quest to pinpoint the intermittent engine performance issue...

Tonight I was installing Diagbox on a netbook from work and while testing it on the car I noticed a fault logged on the gearbox ECU that I have seen appear sporadically before:



The gearbox ECU receives the engine speed signal from the engine ECU (which in turn gets it from the crank sensor) and uses it as a critical piece of information for regulating the torque converter clutch slip etc. If I understand it correctly the above fault is indicating that the speed sensor signal from the engine ECU was temporarily missing or incoherent.

From memory this fault has only ever been logged in neutral, and I have a theory on why that might be - if we imagine that a problem with the speed sensor signal causes it to receive less (or no) pulses per second, how would the gearbox recognize that ? My guess is that it compares the engine speed signal to the torque converter output speed sensor in the gearbox and checks the plausibility of the two speeds.

If you're in neutral then the engine speed signal should always be equal to or faster than the torque converter output speed signal - if its less for any significant length of time the gearbox ECU can assume there is a fault with the signal and flag a fault code. If it can detect the output of the torque converter is turning at 640 rpm but the input according to the engine is only turning at "300" rpm, (due to missing pulses) its safe to say there is a fault.

However if you're driving in gear then its normal for the engine speed to be lower than the torque converter output if you're on the overrun with the throttle closed, so the gearbox can't check the coherency of the two speed signals in gear, which I believe is why the fault is only ever logged in neutral even though the fault itself is almost certainly occurring while driving in gear as well.

Anyway, I let the engine idle for several minutes while testing the Lexia, I noticed at first that there was a bit of an intermittent roughness in the idle, with what seems like maybe one miss/stumble every second or so, which gradually went away over 3-4 minutes. This symptom has been there for at least a few months now, and I haven't been able to find any cause for the rough idle when cold.

I then went for a drive and surprise surprise, the engine is going like a rocket again.... Tons of power, very eager to accelerate, not quite 100% at low rpm, but anything over about 2000 it seemed perfectly normal. If it ran this well all the time I'd be very happy indeed.

Only this morning when I drove to the train station I let the car idle for only 20 seconds before driving (hey I was running late! ) and when I tried to accelerate up a steep hill in 1st it was really struggling, it felt like it was missing badly, struggling to even reach 3000 rpm in 1st with 3/4 throttle, as soon as it hit 3500 rpm the "missing" stopped and it suddenly roared away like an invisible turbo boost button had been pressed...

This symptom of poor performance below 3500 rpm for the first minute or two after starting, particularly noticeable if trying to climb a steep hill has been going on intermittently for quite a while now, and has been really obvious since we moved to a house which is on the side of a hill.

So what the heck is going on ? Putting together the clues of the engine performance and the speed sensor fault logged by the gearbox (not for the first time) I'm now thinking about a poor quality crank sensor signal - something I've speculated on before in this thread.

This reminds me of the intermittent misfire the car in this video had, see around 12 minutes in:



In that case the problem was a crank sensor with an excessive gap. The crank sensor signal is a sine wave which increases in both amplitude and frequency with engine rotation speed which is then converted to a pulse waveform inside the ECU. If the gap is too big the amplitude is too low which means at LOW rpm there will be insufficient amplitude for the Schmidt trigger (or whatever the ECU uses) to produce a clean reliable pulse waveform.

At high rpm the pulse waveform will be fine because there is amplitude to spare, but at low RPM the pulse waveform will start to become intermittent causing missing injector and/or ignition pulses and intermittent errors in timing and dwell times. (See the video at 12 minutes)

What I'm thinking may also happen under these circumstances apart from power loss through lost or mistimed ignition pulses is that if the timing of ignition pulses is not steady and reliable, when the fault is symptomatic it may sometimes be causing pre-ignition / knocking due to the spark being intermittently too far advanced, thus triggering the knock sensor and retarding the timing map. So I really need to get my scope onto the crank sensor signal.

I know I've asked this before, but can anyone tell me where the crank sensor connector is, and how I can access the crank sensor itself to check/adjust the gap ? I would look up the wiring for it on Sedre unfortunately Sedre is on my dead laptop...