Mysterious Occasional Misfire - VVC 160

[revilla]

Towards the end of last year I had an occasional misfire which I thought I had cured by swapping the coil packs and HT leads, but after standing for the winter, now I've got it back in the road it's back.

What happens is:

Either at idle, or at very low throttle setting such as coasting down to a junction with just a light touch of throttle, is feels like it drops one cylinder for a few seconds. Just a touch of throttle normally clears it pretty much immediately and it doesn't then return when you release the throttle.

It most commonly happens maybe half a mile after setting off from cold. It starts up fine, idles fine, drives fine for a little while and then when it drops to idle or approaching the a junction, very occasionally it seems to drop a cylinder. It has happened occasionally with the engine properly warm too. I'm always relatively gentle on the engine until the oil is up to temperature so at the time when it occurs I've not usually give it any heavy throttle since start up.

It's never hesitated under load and in general drives perfectly. It's been on the rolling road since this first happened with very strong results and no indication of a problem, other than the mixture being maybe a little leaner than ideal (but still quite acceptable according to Troy at Northampton Motorsport). The initial lean mixture was traced to a weak fuel pressure regulator which was changed and this brought it back within acceptable limits although still slightly leaner than he's used to seeing on a Kmaps mapped ECU.

OBDII scanner has occasionally reported a misfire code after this has happened, but the ECU doesn't seem to be able to determine which cylinder is misfiring - when it does report a cylinder it varies and can report any one of the four, more often it reports the random multi cylinder misfire code. The misfire does however feel like a regular miss with one cylinder not firing. Unfortunately when it happens, it lasts for such a short time that I don't stand a chance of doing any diagnostics while it's misfiring.

Since this first started it's had brand new coil packs, HT leads, plugs (swapped from Rover NLP100290 to NGK BCPR7ES gapped to 0.9mm) and the ECU has been changed. Plugs removed looked OK and all similar. Engine has recently been rebuilt, compressions and leakdown test results are excellent.

It doesn't feel like an electrical connection type issue in the way that it clears as soon as you touch the throttle.

I'm currently thinking I may have an intermittenly faulty injector. Either sticking and sluggish (this would probably show more at low duty cycle where it would have a bigger percentage effect - this might fit with the slightly lean mixture) or possibly leaking (which would probably show more at low throttle where a little extra fuel would enrich the mixture most - this might also fit with it happening more when cold when the ECU will be enriching at idle already).

I can try to get an oscilloscope onto the injectors in a morning to see if I can see any differences in the responses, particularly when it's cold.

The only other thing I can think of is that several years ago I had a light aircraft that would drop one cylinder when cold and this was caused by a hot (and therefore expanded) exhaust valve sticking in a relatively cold (and therefore contracted) head. That would only happen with high throttle settings though (most disconcertingly, just after the wheels left the ground on the first takeoff of the day!) as these were needed to heat the exhaust valves significantly and differentially. When I rebuilt then engine, valve stem clearances certainly weren't unduly tight.

Is there any known history of occasional sticking valves in the K Series?

Any other ideas worth looking at?Thanks,Andrew

[nickh7]

Lamba sensor ? Slightly different to yours but Mine was stalling in traffic as I came off the throttle revs dropped off and then it coughed and died or ran rough. 

I was told to disconnect the Lamba sensor to see if that made a difference. Yep no issues since

 

[rj]

Either lambda or TPS.  If not, then valve. I have heard of it on a K before, but I think only once.

[revilla]

OK yes, lambda is one obvious thing I haven't tried for some reason. Will try that next.

[revilla]

Hmmm thinking about it though, I'm sure it's something that is cylinder-specific. It doesn't just run rough and stumble, it's clearly one cylinder that just drops out and then drops back in again. It's like a switch, the fault is there or not, no half way house.

[rj]

Are you able to check EGT when it happens? This would give you an ide of which cylinder

[revilla]

I don't get a chance. I'm driving it at the time. It probably only lasts for 2-3 seconds then it's gone.

[rj]

Analogue gauge is you  friend :-p (or datalogging)

[revilla]

Just scanned it after the most recent episode and I had a "Cylinder 2 Misfire" code and a "Misfire with Low Fuel". Fuel has never run low and fuel pressure is spot on (I have a permanent inline gauge now). Suggests the lambda was reading lean when it misfired. So again points back to lambda or injector.

[rj]

I trust you have a few spare injectors to test with?

[revilla]

Yes. I think I'll keep monitoring the DTCs over a few episodes and see if there is a consistent pattern to what it logs. I'll probably get the injectors ASNU'd - the report may actually show the problem as well as getting it fixed. When I'm driving on my own I'll probably leave the OBDII scanner plugged in on the passenger seat. If I'm quick I might be able to see what the lambda is reporting when it does it.

[rj]

I think I have a spare Innovate LM1 you can borrow. Do you want me to look?

[revilla]

Regin,

Thanks for the kind offer. You have everything!

I may well take you up on it in the future but I'll see what I can work out with I've got for now. If something is throwing the mixture enough to make it misfire I'm hoping I can find it. My reasoning is that if the ECU is consistently logging it as a lean mixture with a misfire then my current narrow band lambda is either picking it up or giving a false lean, which is suppose could lead the ECU to enrich it to the point if a rich cut.

I've just had to take one of Chia's friends back to Leicester so we took the Caterham and I made about an hour's run of it, during which it didn't misbehave. I'd cleared the DTCs before setting off and nothing was set when I got back. I was slightly concerned that the pop and crackle that Kmaps had mapped in could have been tripping misfire codes as it does make it a little ragged on the overrun, but the fact that it logged nothing tonight gives me some confidence that the misfire events it is logging are genuine misfires and probably correspond to the events I'm noticing.

I'll keep clearing the codes before a run and checking afterwards for a while and see what pattern emerges; whether I can tie the codes to these misfire events for sure and whether there is any consistent pattern to what is recorded. I've read that the misfire cylinder identification on the MEMS is a little bit approximate and can offset be a cylinder out in firing order, so for example if it keep logging 2 with some 4 and some 1 it could quite easily be 2 consistently in reality.

Although I don't have proper data logging I can probably rig a digital scope and laptop up to the lambda sensor and put it on slow record to see if I see anything odd.

I'll also try to get a scope onto all four injectors and compare the traces. You can see the current and voltage humps corresponding to the pintle moving within the injector so if one is sluggish to open or close I should be able to pick it up as a difference compared to the others.

I've got plenty of spare lambda sensors in my box so I can swap that out too to see what happens. I've also got a set of fresh recently ASNU'd injectors so I could swap them in without leaving the car off the road while they were off for processing; I could then get the ones from the car done to see if the initial report picked up anything out of the ordinary.

[StevehS3]

I had a very similar sounding problem. It happened when it was warm (but not hot). It turned out to be a faulty heater in the lambda sensor which is used to heat the probe when the engine isn’t up to full temperature. In my case it was reported by the ODB reader. If you have a spare lambda it might be worth swapping it.

[rj]

Andrew, him who has the most gadgets, when he dies, wins. I'm not a looser. :-p

[revilla]

Regin, I thought it was him who has the most spare parts!

[rj]

Andrew,

Maybe you have the most spare parts, but tools, gadgets and cars... I have way too much and am trying to scale down. Yet I'm about buying the fire engine I used to drive in the '90s to convert it to transport my Caterhams to trackdays.... The big question is whether I can register it or not because it in reallity has been marked "scrapped"

(help...)

[revilla]

Taking your Caterham to a track day on a fire engine? Are you SURE you're not secretly planning to try out Evans Waterless Coolant?

Regin, between us we have well and truly hacked my thread! Trying to get things back on track (no pun intended)...

What are the possible causes of a sticking valve? I can only think of two:

1. Insufficient valve stem to guide clearance causing the valves to stick when they expand, especially when the head hasn't yet reached temperature.
2. Carbon or other deposit build up on the valve stems.

Both of these sound very unlikely in my case. The head was very recently rebuilt. When I checked the valve guides my initial concern was excessive clearance although Dave Andrews declared them to be fine; they certainly weren't tight though. 

Following inspection, all of the guides were honed out with a 6mm flex hone to remove oxides (by Dave Andrews) and the head was built up with brand new valves and stem seals (Dave Andrews specials). The valves were the later "carbon break" types. My earlier oil consumption problems are long gone, cured during this rebuild. I have no reason to imagine that the valves could gum up and stick in about 3000miles or so from this level of rebuild.

I did think about a lifter (which were not new and so could potentially have some gumming up) pumping up, but I can't imagine that acting like a sharp on-off switch; I can't imagine a pumped up lifter bleeding down instantaneously with a blip of the throttle.

Maybe a bit of injector clearer and lifter treatment through the system for a short while wouldn't be a bad idea?

[rj]

Always remember only to change one parameter at a time. Otherwise you won't know what cured the issue.
 

[revilla]

[revilla]

Got a scope onto the injectors tonight. Nothing obviously wrong there, but then again it wasn't misfiring at the time, but I thought something might show up.

Below are current and voltage traces for one of the injectors at idle; all four were pretty much identical - I was mainly looking for differences. All of the injectors saturated at between 0.84A and 0.87A. This led to a rise in the switched ground drive voltage of around 25mV, which is close enough to zero to be negligible (it equates to a total impedance through the ECU driver and ECU ground wiring of about 0.03 ohms).

No signs of problems with the injector drivers or the mechanical mechanisms of the injectors there, but nothing there shows me anything about whether the injectors are flowing properly when open and closing properly. So nothing found, but nothing really ruled out either.

[revilla]

PS: I know I could have measured the rise in the switched ground voltage much more accurately by measuring it relative to ground, where here I was measuring the voltage across the injector so referencing it to the injector supply voltage, but it was clear without swapping the wiring round that there wasn't a problem there so I didn't bother.

[Wrightpayne]

Where is the ‘over my head’ emoji?

Top investigation work Andrew!

Ian

[revilla]

OK done a bit more digging with the oscilloscope. Investigating possible lambda sensor failure, but didn't want to just swap it and see as the fault is so intermittent I could kid myself I'd fixed for ages before seeing it again, so I wanted to see if I could actually find something wrong.

Some of the traces look a bit iffy to me, but I'm only comparing with "textbook" clean examples on the internet rather than real life experience, so I'm not sure how unusual they are. I think it's worth me swapping the lambda at this stage and repeating these tests.

So here goes ...

This is the lambda sensor output with the engine hot and running a fast idle at 3000rpm. Ignore the sharp noise spikes, there's something in my garage coupling noise into the breakout leads I'm using, the spikes are there even with the ignition turned off. And annoyingly the scope software hasn't captured the voltage scale cursor I was using when I saved the image, which means that one of the key observations is not easy to see: The peaks in the waveform are only reaching 0.70-0.72V whereas from what I understand they should be 0.85V minimum, preferably closer to 0.90V. The cycle time appears good at around 0.8s (looking for 1s or less).

This is what happens when, after holding the RPM at a steady 3000 for a minute (feeler gauge under the idle stop screw) I chop the throttle and allow it to return to idle. The lambda sensor reads it going rich for a good 6 seconds (2s per division) before starting to cycle again. This only starts once the engine has settled back to idle speed which happens quite rapidly.

And this is the output after it settles back a hot idle. A rather slow and erratic cycle, again only peaking at 0.70-0.72V with cycle time highly variable but generally a lot longer than 1s. From what I read, I would expect it to continue to cycle around once a second at a hot idle, although this depends on the ECU behaviour too I guess.

This doesn't look entirely healthy to me, bit I guess the next thing to try is to put a new lambda sensor on it and see how it compares, as I'm not really 100% sure how this engine should look and some of this may just be the way it is, although the peak lambda voltage of 0.70-0.72V does sound like a faulty sensor. Worst case I'll just end up with a spare.

My lambda sensor is not well placed at the moment, it's in No 4 primary rather than the boss down on the catalytic converter, so it's only really reading one cylinder and if that one is out of balance with the others it won't be seeing that. The reason it's installed there is because the boss on my cat is positioned almost right at the bottom and the lambda sensor installed there becomes about the lowers point on the car and prone to being left hanging on speed bumps. However I'm booked in to get a decat pipe made at the end of May and I will get the boss put in a better place when that is done.

I don't know at this point whether the 6 second rich period is actually the ECU running the engine rich, or whether it is a false read from the sensor, in which case it would cause the ECU to run the mixture lean to compensate. It does seem to correspond with the point where my misfire has been occurring though, so could be significant. If it still does that with a new lambda sensor I may look at what happens with a stock ECU as this one is a Kmaps remap with the "pop and crackle" mapped in, and I'm just wondering if that could be running it rich on the return to idle.

That's all I know for now! Any comments gratefully accepted as always. Will come back when I've had a chance to swap the sensor and / or ECU map and report what I find.

(Edited to correct for the fact that a low lambda sensor voltage means a lean mixture and a high lambda sensor voltage means a rich mixture - I'd originally got it the wrong way around.)

[revilla]

Oh well ... Nice theory!

Swapped in a brand new lambda sensor. Absolutely everything was the same. Same voltages, same cycle rates, same rich up after returning to idle.

Swapped in a stock ECU and couldn't really get anywhere, the adaptive fuelling needed time to adapt to my engine I think as it just ran rich all the time without any cycling of the lambda sensor output.