Leo ECU frying Lambda sensors

[LairdScooby]

Here's the set-up :

1995 Rover 827 V6 (same as Honda 2.7 V6) running single point, Lambda controlled Leo system with 42mm venturi mixer (changed from 33mm mixer today) with just the purple wire from the Leo connected to the rear bank Lambda sensor as per advice from Gilbertd about not using the grey wire. The vap is an OMVL R90/E but i don't think that's any part of the problem.

This is the problem - every 6 weeks or so, the rear bank Lambda dies (or is killed) and as far as i can see, the only common part is the Leo. If it was overfuelling (due to the previously restrictive mixer) then it would surely fry the front Lambda as well, not just the rear one that i have connected into for the Leo. I run the same wiring set-up on my Jeep with no problems.

The Lambda sensor gives 0.48V output regardless of cold, hot, idle, constant high revs, flooring the loud pedal to intentionally make it rich or on over-run cut-off which would obviously be very lean. There is no cycling up/down round the centre point as it should on petrol and as the Jeep does and also as it did when the Lambda was first renewed.
I tried another Leo unit and that had similar results for 5 minutes then switched back to petrol and wouldn't go back to gas - refitting the original Leo unit restored what i had beforehand.
The only thing i haven't yet tried is disconnecting the Lambda to see if it still shows 0.48V but i have a sneaking suspicion it will.

I have a Millenium ECU that i plan to fit on the Jeep and put the near-new Leo from the Jeep into my Rover with (yet another!) new Lambda

Hopefully this will sort me out! However, i'd like to know if this is a known problem with older, pre-abused Leo ECUs or if it seems i have an oddball fault here?

[LPGC]

Why would a completely different ECU, if assumed to be (another) good unit, switch back to petrol and refuse to switch back to gas? You know about the 'errors' screen?

Usually takes far longer to kill a probe due to an overly rich mixture on gas than on petrol, less soot and other contaminants. A probe on only one bank could be killed if that bank is burning enough oil for whatever reason (piston rings, breather pipes feeding that bank engine oil, etc). Head gasket failure - coolant kills probes pronto.. Losing any coolant?

They can be killed be applying a voltage to the signal wire but wouldn't expect .48v to be enough to do that - Some older petrol closed loop systems are designed to hold the lambda signal line at approx .5v during engine warm-up and/or when running open loop due to detecting a fault.. i.e. where newer ECU's would just switch to an open loop mode, some systems just hold the signal wire at stochiometric voltage so fuel trims are not affected while the fault or open loop condition exists. The Rover ECU may hold the lambda signal wire at .48v when the engine is cold or if, for example, fuel trims have drifted to one extreme. In such situation, a probe you think is broken might actually be OK and connecting the grey wire might be a cure to the 175 reading the constant .48v (because connecting the grey wire means you cut the direct connection between the petrol ECU and lambda probe - don't forget to do that if you connect the grey!). Like you said, disconnecting the probe and checking probe signal voltage would confirm. Reading between the lines, the front probe doesn't seem to be suffering any problems - could swap the front probe to the rear to see if it lasts in the rear position.

There's usually voltage (maybe .5v) on a disconnected lambda signal pickup wire from ECU's (Petrol ECU's and AEB175s) but this voltage should easily be pulled up/down by the probe. This means that should the purple wire become disconnected from the probe, the 175 will maybe show a constant voltage of around .48v - worth checking for such bad connection... Could be the case that disturbing the loom when you re-connected the original 175 re-made the purple wire connection, the second ECU may have switched back to petrol after a while due to settings in the errors screen.

I have known AEB175's to fail in a way which sees them have higher voltage on the purple wire and instead of this voltage being pulled up/down by the probe it remains more constant - means the AEB175 is unable to read lambda voltage and has the potential for damaging the probe. Most I've known that have failed in this way also exhibit other signs of failure, such as losing setup information and 'crashing' (can sometimes work properly again, at least for a while, by resetting the ECU and re-entering all setup info). In some cases the 175 will show it's reading the approx .5v while holding the purple wire (and probe signal wire) at a higher voltage, in which case the reset etc won't work.

Simon

[LairdScooby]

Thanks Simon, i've just double-checked the fault code on a hunch. The hunch being i may have misread the flashes due to reflections and the flashing of the yellow LED next to the red one. With just the ignition on, the yellow LED doesn't come into play as it is only to indicate out of spec idle speed on the OEM ECU - the red LED flashes the fault codes.
The code for the front bank Lambda is one flash and the rear bank Lambda two flashes - it's only giving one flash.

This makes a lot more sense all round and explains why i thought the rear Lambda (where i've connected the purple wire) was FUBAR again. The front Lambda was renewed at the beginning of April with the rear one with a pair of ebay cheapies (£10 each) from Northern Ireland. When the rear one died mid-May, i renewed it with a dearer one from mainland UK and have ordered another one of those to replace what i thought was the rear one again.

This means that with the front one dead, the OE ECU has gone into "Limp Mode" and as you rightly point out, put a stoichiometric nominal value on the Lambda inputs so that the ECU doesn't see a discrepancy between the two Lambda values and doesn't try to alter the mixture accordingly - it just runs on a preset "best guess" map until the fault is cleared and new component fitted.

Your last paragraph sums up exactly what i was thinking had happened. The other Leo ECU was not a known good one, just assumed good so will need some investigation once i get things fixed.
No coolant loss, burns a bit of oil when i blip the throttle (suspect PCV valve in front rocker cover) but isn't constantly burning oil. I did check the errors screen - all good there. I assume the OEM ECU saw the dead Lambda and went into limp mode which would have not shown a fault on the Leo - in any case it turns out to be the front Lambda so the Leo wouldn't see that in any case.

Serves me right for buying cheap Lambdas but they were the quickest delivery at the time! Also had Lambdas from the same seller before and no problems so i blame the Polski Postie for playing football with them!

That said a few weeks after buying those Lambdas the listing disappeared from ebay, perhaps it was a bad batch. Who knows now?

I'll see what happens when the new Lambda lands tomorrow (hopefully!) and i have the pleasure of fitting it! You'd think the front one would be easier but it isn't, too many things like cooling fans, radiator, air-con hoses, compressor etc in the way of making it easy at all!

Hopefully this is just a glitch down to a pair of dodgy Lambdas - might still fit the Millenium to the Jeep and the Jeeps near-new Leo to the coupé though, especially as i'm thinking of selling the Jeep because i've just won another estate car on ebay that doesn't have a torpedo tank straddling the load space between the suspension turrets!

Meanwhile my 827 Sterling will be getting the Zavoli SGi treatment sometime soon..........

[Gilbertd]

If the Leo sees a constant voltage with no switching for a 'long time' (although nobody seems able to tell me how long a long time is), it will flag a Lambda not working error. That will cause the yellow LED on the switch to give the slow flashing while running on gas.

Your comment that the Leo would not switch to gas does suggest it's reached it's sell by date. I checked one for somebody that wouldn't change over even though everything suggested that it should. Did a full reset and then programmed it with all the settings and it worked fine but only for a couple of days before it went back to not switching again.

[LPGC]

Tch Dave not reading the flashes correctly and 'sun in your eyes' excuse! Bit off subject here but a bit of ribbing seems in order and a loose dark analogy comes to mind.. Remembering the plane that crashed on the motorway 20+ years ago due to the co pilot reading instruments wrong and shutting down the wrong engine (for vibration) when instruments were showing it was the other engine with the vibration... Calling for a bit more thrust on approach after shutting down the good engine (doh), the vibrating engine broke.. Led to a redesign of cockpit instruments locations because in the pilots (highly questionable) defence instrument location was maybe a bit confusing (2 columns/banks of instruments, 1 for each engine, except vibration indicators were both under the same instrument column). i.e. He didn't just have to count number of flashes on an LED, often best to pay close attention and confirm stuff during diagnostics I've converted vehicles for a few airline pilots and been underwhelmed by some of their car driving skills, a bit clunky, not too smooth.. Even more underwhelming when you learn a few months down the line that they've crashed the car and had a few other at fault crashes.

So there might not be anything wrong at all really.. With the 175 connected to the rear bank the front probe might flick similarly to the rear probe or might not. Even if the front probe is broken, since you'll be running on LPG most of the time (which only uses the rear bank probe) you might think it's not worthwhile changing the front one, particularly if the petrol ECU ignores the front probe on petrol after running on LPG.

Simon

[LairdScooby]

Sadly it isn't the first time i've miscounted the red LED flashes as a result of the yellow one flashing because of an "idle speed out of spec" condition being flashed by the yellow LED and i daresay it won't be the last either - hence the hunch! As it happened the sun was in my eyes the first time as well which obviously didn't help.

Anyway, if the OE ECU didn't put a stoichiometric value on both Lambdas for Limp Mode, i could run on gas and ignore the front one. The new sensor landed a short while ago and i've just changed the plug on it to suit my car and the ECU memory fuse has been out so the ECU should show no faults at all.

As for the other Leo not switching to gas - i had already tried the reset/reprogram stage so is strongly suspect that one is dead as the proverbial, however i will double check once i have everything running as it should be again. I also have a Millenium behaving strangely when i connected it to the Jeep so for now i'm going to put everything back as it was and try the coupé with the new Lambda.

Will report back once i have something more to report.

As for that plane - are you on about the one that landed on the M27 Simon?

[LPGC]

I didn't quote which motorway because I couldn't remember and didn't want to bother to check... I used to know which motorway, not the M27, more like M5, M42 or M1, I drove past the site en-route from here to Cornwall and thought the plane was on approach to East Midlands airport.

Only kidding with the stick I've given you lately, for the most part anyway!

Simon

[Gilbertd]

The plane was on approach to East Midlands and landed a bit short so it dropped onto the M1. The approach glide path if coming in from the East means they are, or should be, at about 1,000 feet when they pass over the Motorway.

[LairdScooby]

Thinking about it, the M27 incident was a bit longer ago, near enough 30 years no. A plane over-shot the runway at Eastleigh Airport and dropped onto the M27 between the airport and the Transit plant (as was) one morning.

As for the Leo/Lambda problem - the plot thickens!

New front Lambda fitted and working but rear Lambda still showing between 0.46 and 0.5V, even after changing the engine management ECU to eliminate that as a possible cause, fitting another Leo from the Jeep (that didn't work because of some wiring anomaly it seems), a spare Leo that worked and the Millenium that wouldn't work on the Jeep but works fine on the coupé except they all show the same voltage on the Lambda sensor.
Having already proved it isn't the cars ECU putting the voltage on it, i disconnected the rear Lambda at the OE connection and measured the voltage - varied between just under 0.2 - 0.378V so it seems the Lambda is at least giving an output that varies with mixture.
Reconnecting the Lambda saw the voltage shoot up to the 0.46 - 0.5V and there it stayed, even with the engine off. In fact, the only way to get rid of that voltage on any of the LPG ECUs was to remove the two fuses from the supply and refit them and it stayed at 0V until i switched the ignition on.

The purple wire is loomed into a 4-core loom from the Leo (mounted centrally on the bulkhead) which runs to the drivers side suspension turret where i've picked up the injector cut, TPS and Lambda signals.
The TPS signal is ok and the injector cut (yellow wires) works fine so in (not quite) desperation, i cut the purple wire about halfway along and checked the voltages on each end. The car end/Lambda was showing 0.2 - 0.378V and the Leo end was showing.............
.......... you've guessed it! 0.46 - 0.5V.

My next step is going to be to check that wiring, it shouldn't have got hot enough to cause a problem on the insulation but maybe that's the problem. However, with the ignition off, the voltage wouldn't be present even if the injector feed had melted into the Lambda feed and in any case, that would be battery voltage.

I don't believe three separate Leos would exhibit the same fault, i've checked the plug to make sure the grey wire isn't shorting to the purple one and causing this although there's a possibility (albeit slight) that the grey wire has melted into the purple one inside the loom.
All that being said, there's no sign of heat damage on the insulation tape i used to loom it all up.

I'll get to the bottom of it one way or t'other and maybe even the wiring anomaly on the Jeeps Leo install, whatever it may be! Meanwhile the Jeep works fine on its own Leo but that won't work in the coupé and those that work in the coupé won't work in the Jeep. Strange but true!

[LPGC]

When measuring voltage at each end of the cut purple wire, did you use the same earth point?
Otherwise voltage could be due to purple wire connected (melted) to TPS? If connected to grey, programming the Leo's grey wire output signal should see the voltage change according to your settings.
Why not just run a separate purple wire directly between lambda signal and Leo?

Take it you set the Leo's yellow wire utilisation to 'disconnect injectors', not 'reset ECU'... Not a biggie but personally I don't like the idea of using the yellow wires to disconnect injectors directly - no injector emulation for the petrol ECU, the further possibility of if something goes majorly wrong with the Leo buggering the petrol ECU. Ideally would use a Pitagora, or if unconcerned about emulation use a DPDT relay connected to gas solenoid wire so the default off position connects petrol injector positive feeds, one relay channel per cylinder bank. Would at least isolate the petrol system using the yellows via a relay, seen quite a few burned Leo/175 boards with various shorts, unknown how a short to earth on pet injector + would effect chances of driving home (on petrol), considering that in this situation driving home on gas would first involve some re-wiring of the LPG system and removal of the stepper, then hoping the pet ECU isn't fried and will still run the ignition. Relays in Leos might handle 5 amps, could be close to that with 6 petrol injectors at full chat, current through the board tracks will raise the board temp a bit, only narrow tracks.

I just finished installation/re-installation on a 91 Defender that's just had a new chassis fitted, a really nice example looks like new. In it's previous life it had only 2 sill tanks, a Tartarini reducer, a different switch. Came to me with 3 tanks fitted, no piping, no LPG wiring, no switch fitted. I did all the piping, wiring, fitted an R90, etc. Twin carb V8. Wired both the main switch feed and switch LPG solenoid outputs through relays to prevent problems... 3 Tank solenoids + 2 reducer solenoids will draw about 5 amps pushing the rating of the switch relay but not the 30amper.. The other relay is there to take the ignition on feed from the coil, a few advantages including easier to follow wiring on the install, all switch wiring is under the bonnet.

Simon

[LairdScooby]

Yes, used the same earth point as the 0V reference point Simon - a convenient earth point on the manifold that has two OE earths running to it. Just to be on the safe side, i rechecked the voltages using several other earth points including the earth for the Leo (which used a spare way in a factory MEP), the earth on the front side cover plate, alternator body and one or two others. All voltages read the same within 10mV.

I haven't used the grey wire but it did cross my mind that maybe a stray strand of wire from either the purple had got into the grey or vice versa and the output from the grey was corrupting the signal. Bear in mind i had to rewire the whole loom as some clever article had removed it prior to sale on ebay by amputating the wires about 2" from the multiplug.

Running a separate purple wire direct to the Lambda might indeed be an option, maybe even run it to the front bank Lambda just for the sake of eliminating any potential problems with the rear one.

The plan for tomorrow is to deloom the wiring and inspect/test with the wires separated. Didn't get anything done on it today as i was out picking up another candidate for LPG conversion but that's another story!

The yellow wires are indeed set to injector cut, i used a relay to switch the BRC version of a Pittagora on the Jeep when i converted it to Leo operation as it needed the opposite of the output i had from the Leo - can't remember exactly how/why now but it works which is the main thing. However the Rover/Honda ECU doesn't need emulation and although i'm not fully au fait with the achitecture of the PGM-Fi ECU, i'm 99% certain that the +ve output from it for the injector common feed is taken straight from the +ve input so it will be fused on the main EFi supply. That said, semiconductors can fry many times over in the time it takes even a fast blow fuse to rupture - it may well be something to look at when i get things sorted.

Meanwhile once i've de-loomed the wring tomorrow i might just run all 4 wires (2 yellow, TPS and Lambda) separately while i check things out. I'll also check the voltage on the grey wire as well, something somewhere isn't right obviously and i'm wondering if maybe i should wire it "as per spec" (grey wire to OE ECU and purple to Lambda) just to be sure. I just hope this fault hasn't caused a problem with the rear Lambda, if it has then i think i'll have to "piggy-back" the front and rear Lambda inputs and connect them to the grey wire and the purple wire to the front Lambda to get running once i've found and removed the ffault. Of course the reaer Lambda might still be ok, i seem to recal at one point it was quite common (Bosch comes to mind) to feed Lambdas with a small voltage to help them warm up quicker and also to help burn off excess carbon but that was only for a short while.

Fingers croossed for the moorning!

[LairdScooby]

Finally got to the bottom of this one! THANKFULLY!!!

After double checking the pin outs on the Jeep Leo multiplugs and confirming they were the same, i fitted the Jeeps Leo onto the coupé and still got about 0.5V on the purple wire!

I disconnected the rear bank Lambda and still had the same so left the multimeter probe connected and grounded it out causing the voltage to drop to zero. This made the mixture go rich and i could just hear and feel the movement of the stepper-motor to prove this.

On a hunch, i made a lash-up to piggy-back into the front bank Lambda and connected it to the purple wire and the voltage was flicking up and down and eventually settled about 0.7V but would alter dramatically when revved or the throttle was closed from high revs. I also doubled up the connection from the front Lambda into the rear input to the OE ECU to prevent the possibility of it throwing up an error code.

Once i knew the Jeeps Leo was working on the car, i refitted the original Leo and hey presto, it works again! That means the other Leo and Milly i have are probably ok as well. I'll have to test them out soon.

After various other interruptions today it was getting near feeding time so i called it a day and will tune the vap tomorrow. Been several red herrings along this particular fault, it seems the rear Lambda has gone weak for some unknown reason. I have a pair of Lambdas from a later KV6 engine, when i work out the heater circuit for them i'll fit them in place of the single wire jobbies. I'll also fit a SPDT switch into the purple wire/Lambda pick up circuit so i can switch between front and rear Lambdas so if a future fault occurs, i can simply flick a switch and see if there's a difference.
I recently bought one of these "Carlos Fandango" air-fuel ratio gauges that simply measures the Lambda voltage, when i fit that i'll hook it in to the switch so i can see what the Lambdas are doing at any time as well. If nothing else it will give me an indication if one or other of the Lambdas is getting lazy.

Maybe once i've tuned the vap i can get on with what i was intending to do this week and get the car cleaned/polished ready for the show on sunday!

[LairdScooby]

Bit of progress on this one! After finding the new, 42mm venturi mixer was too big (went scarily lean above 3000rpm) i bored out the original 32mm mixer (that was going way too rich above 3000rpm) to a smidge under 36mm and refitted it.

After a bit of tweaking and twiddling, the default position on the stepper motor is about 110-115 (can't remember exactly now) and the mixture now stays the same throughout the rev range, even when hoofing it with only minor adjustments of the stepper to correct itself.
The performance is certainly acceptable (if anything it goes better now than it ever has - threw me back in the seat at 70 when i floored it to overtake) and the economy seems much better, ball-park maths suggest it is nearer 30mpg than 20mpg as i was getting!

Time will tell, as will the next fill up!