Yes I do have my original MAP sensor - in fact I think its on the car right now. I say think, because the replacement one I have is also a genuine Bosch sensor and they're absolutely indistinguishable despite 15 years of age difference between them!
I've swapped them over a few times and honestly there is no difference between them, so I'm confident my issue is not the MAP sensor. I've also monitored the output of the MAP sensor on my scope on a few test drives and can't see anything wrong with the output at all.
More importantly - at the wide throttle openings where I'm seeing lean running reported by the O2 sensor the MAP sensor is already reporting maximum voltage by then (maximum pressure / zero vacuum) so the ECU is already being told the engine is at maximum load.
The TPS signal is also working perfectly, tested on the scope, and the air intake and coolant temperature sensors have been tested with the Lexia and are giving sensible readings.
I've monitored the crankshaft sensor signal on the scope and it too is perfect - about 5 volts peak-to-peak at 2000 rpm which is more than enough from a variable reluctance sensor. So ALL input signals that contribute towards calculation of injector pulse width in open loop mode have been tested and are working normally.
That means if its still running lean at WOT its not due to the ECU calculating too small an injector pulse width due to a sensor problem. The ECU is doing what it should be.
It has to be due to something the ECU can't detect - low fuel volume delivery, low fuel pressure, low injector flow rate, or excessive airflow due to exhaust/inlet changes.
Although a change in the exhaust/inlet flow rates could cause it to run lean, I don't think that's the issue - and the main reason is that I started to notice poor performance a few weeks before the exhaust broke, also at the time when I changed the old exhaust system to the new one I didn't notice any change in performance, likewise when I changed the air filter it made no apparent difference to the performance. So for those reasons I don't think it can be that.
Low fuel volume delivery - the new fuel pump has a measured 30% increase in flow rate over the old pump but it hasn't made any real difference to the problem so its not that. Fuel pressure - the regulator in the car at the moment delivers 2.8 bars (its the one that came with the 406 rail) and the original delivered 2.7 bars - both are a bit low but bear in mind that the car was originally running perfectly on the regulator that only produces 2.7 bars, so I don't think that alone is enough to explain the symptoms.
Last weekend I went a bit further with my scope to try to analyse what is happening. The first thing I did is put it across the knock sensor output. (At the ECU terminals)
Quite interesting. The first thing I noticed is that the engine is rather noisy at high rpm, at least according to the knock sensor! Below about 1500 rpm there is little activity on the scope, as the rpm goes up the "noise" on the scope goes up dramatically, by the time you get to about 3000 rpm there is about 500mV peak-peak of semi-random noise, (from the mechanical noise of the engine transmitted through the block) it's no wonder that the knock sensor input thresholds are higher at higher rpm to minimise false triggering. Of course without comparing it with another car I have no idea whether it's noisier than average of course.
Knocking looks different to random noise though, it should look like a short "sine burst" at about 6.3Khz, repeating once per power stroke, and its hard to pick that out visually on a scope amongst the other noise (a spectrum analyser would make it MUCH easier) but I'm fairly sure I did see intermittent examples of knocking at 1500-2500 rpm under load.
Not always reproducible but it happened often enough in a single test drive that I'm reasonably satisfied that if the ECU is retarding the timing due to the knock sensor the knock sensor IS genuinely detecting actual knock, not just getting overwhelmed by the mechanical noise of a worn, noisy engine.
The next thing I wanted to check was to try to watch the ignition timing in real time - and I came up with a very clever solution, if I do say so myself.
(Jim will love this if he's reading
)
Working on the assumption that the injector timing is fixed in relation to the crank position (no VVT on this engine) I displayed both injector voltage waveform and coil primary voltage on two different traces and triggered on the injector pulse. Here is an example:
The cyan trace is the injector waveform, the yellow is the coil primary, both of Cylinders 1/5 in this case.
If you follow the cyan trace from the left it starts at 12 volts while the injector is off, where it drops to the bottom is near zero volts - this is the injector on period, where it spikes right up near the top is the moment the injector is turned off and the voltage rises to about 80 volts from the collapse of the magnetic field.
The yellow trace for the coil is similar - at the left the coil is off, when it drops down the coil is charging (the dwell period) where it spikes up high the coil is firing the spark, the wriggly line after it drops down a bit is the "spark line" which shows that the spark is occurring in the combustion cylinder.
By trial and error I found the point where the injector turns OFF is fixed relative to crank timing - when the injector pulse width increases with more throttle it opens sooner rather than closing later, so I set the trigger to trigger on the point where the injector turns off.
There is a LOT of information in this capture! Especially watching it change under driving conditions. You can see the injector pulse width, the coil dwell time, the ignition spark line (shows you that the spark occurred in the cylinder and not externally) and the ignition timing - if the timing is retarded the whole yellow waveform moves to the right relative to the blue one.
I checked all three pairs of injectors/coils, and this is what I learnt:
1) All three injector waveforms are perfect, and identical to each other during the test drives. The large voltage spike when they close shows that the magnetic field collapse is strong and indirectly shows that the injector is drawing a normal amount of current. If any injector had a high resistance connection (wiring etc) that spike would be lower than the rest but they were all identical. So I'm satisfied that there is no electrical drive problem to the injectors that would cause a lean condition.
2) All three coil waveforms were identical - most importantly the "spark line" is present and very close to identical on all three even under wide throttle which pretty much rules out any sort of ignition problem. If the spark was jumping outside the cylinder due to ignition lead breakdown or not jumping at all there would be no spark line for example. The presence of a healthy spark line on all three confirms the ignition is working and firing properly.
3) Ignition timing is indeed a lot more retarded below 2500 rpm than it is above, especially when low rpm torque is lacking. Looking at timing on a time scale instead of a degrees scale requires a little bit of interpretation - part of a normal advance curve is rpm related - the higher the engine rpm the more advance in degrees you need - but this is to give a constant advance in milliseconds. So if you need 10 degrees at 2000 rpm but 20 degrees at 4000 rpm, that is actually the same advance in milliseconds. (Combustion takes the same time regardless of engine rpm...)
Looking at it on the scope you're seeing the advance in milliseconds, not degrees, so the timing advance with rpm is automatically factored out, leaving the corrections for engine load and knock feedback. What I saw was that when you stamp on the pedal quickly the timing is instantly retarded a LOT for a fraction of a second (to minimise knocking during a momentary lean spot) then returns to normal, (as expected) and that between 2500-6000 rpm the advance in milliseconds is pretty much constant for constant throttle over a wide throttle range.
Below about 2500 rpm the timing drops back a lot, very quickly (it fully transitions between about 2400-2600) and between 1500-2500 it is considerably retarded compared to above 2500, even with a light throttle. I then tried a battery off reset for 10 minutes and went for another test drive - the timing above 2500 rpm was the same as before, but as well as the performance improving the timing between 1500-2500 rpm was quite obviously more advanced than it had been just before the reset - not as advanced as at higher rpm, but maybe halfway towards that figure, and although I didn't try to calculate it exactly it must have been many degrees.
So I have my first conclusive, objective proof that the timing is indeed getting retarded excessively below 2500 rpm when the engine is performing poorly, and that the battery off reset is temporarily restoring at least some of the lost timing - until now it has been a theory with no hard evidence, not any longer.
So we have:
1) Definite, somewhat intermittent lean running at WOT from 1500 to 2500 rpm. (proven by monitoring the oxygen sensor on the scope)
2) Fairly convincing signs of knocking being detected by the knock sensor from 1500 to 2500 rpm. (proven by monitoring the knock sensor on the scope)
3) Definite proof of the timing being unusually retarded below 2500 rpm. (proven by monitoring the injection/coil pulse timing, then observing a big difference when re-testing after a battery off reset)
So we're back to intermittent lean running as the root cause of trouble with retarded timing being a "knock on" effect (ha ha) from the lean running, and I believe that most of the loss of torque and throttle responsiveness actually comes from the timing retard that gets applied rather than from the lean condition itself, even though being lean seems to be the root cause of trouble.
So I think I'm down to the point where it can only be low and/or intermittent flow rate from the injectors causing this lean condition. To be honest it has actually got a bit worse since I swapped the injectors over for the second hand ones - despite all the testing and cleaning I gave them before I fitted them they may actually be worse than the originals!
The stumble flooring the accelerator at idle has actually got worse with the S/H injectors as well.
So I think what I'm going to have to do is send my original set of injectors away to be professionally tested and ultrasonic cleaned - at least I won't be without the car while they're away!
I really can't see anything else that could be causing it to be so lean under WOT and stumbling when flicking the throttle when so much else has been ruled out...
I don't know if its the same kind of joint with an inner and outer skin on the 406 but on the Xantia the outer skin rusts away until it breaks free from the weld and then there is nothing holding it together even though the inner tube is fine.
So, if an ECU disconnect reset temporarily restores a bit of normality it shows it's learning from whatever the problem is and subsequently acting upon it.