How could I detect an intermittent misfire on a spark

  • Thread starter brewnog
  • Start date
  • Tags
    Spark
In summary: If you mean a failure to spark, it could be detected/isolated by using an inductive pickup on each wire (like a timing light). I assume here that there wouldn't be a magnetic field in the wire unless the spark actually jumps to ground (unless there's an unintentional ground downstream of the pickup but before the plug). The pickups could feed into a computer that's programmed to notice a missing signal. As for finding the cause, I dunno...Do you mean system monitor or fix a problem?
  • #1

brewnog

Science Advisor
Gold Member
2,750
7
Given simple apparatus, how could I detect an intermittent misfire on a spark ignited internal combustion engine?

Novel ideas are welcome!
 
Last edited:
Engineering news on Phys.org
  • #2
I think current knock sensor design is basically a microphone that listens for pre-ignition in the engine, and retards the engine's timing to get rid of it (and/or turns down boost in a forced induction engine).

Alternative designs might be able to watch pressure in the cylinder, and any spikes in pressure before the spark are a giveaway for pre-ignition. This would also be a helpful way to measure engine health and cylinder compression.

EDIT: Are you interested in spark knock or misfire due to lack of fuel/spark?
 
  • #3
If you mean a failure to spark, it could be detected/isolated by using an inductive pickup on each wire (like a timing light). I assume here that there wouldn't be a magnetic field in the wire unless the spark actually jumps to ground (unless there's an unintentional ground downstream of the pickup but before the plug). The pickups could feed into a computer that's programmed to notice a missing signal.
As for finding the cause, I dunno...
 
  • #4
Do you mean system monitor or fix a problem?

Mech_Engineer has a good idea for monitor.
I think light sensors, current monitoring and instantaneous crank velocity changes are also used.

If your trying to fix something then.

Examine spark plugs for fouling.

Open the hood at night with the engine running.
Look for sparks.
Picking a day with high humidity helps.

Since things are often obscured, examine the wires (cap and rotor if they exist) for burn marks.

Just replace wires, cap, rotor and plugs.

Ideas for complete shutdown would be different than misfire.
 
  • #5
Thanks guys, let me be more specific!

I'm not trying to fix a problem here, just trying to monitor a suspected one.

I want to monitor a suspected problem with suspected intermittent misfire. Not knock, and not necessarily failure to spark. Just failure of the mixture to ignite, whether it's due to spark strength, timing, mixture or compression.

I could monitor the HT leads as suggested by Danger (in fact, I have!), but this will only show if the engine is failing to spark. Well, not even that; just whether the ignition system up to the plugs is working.

I could monitor the exhaust temperatures, but this would only show up a continuous lack of firing on that cylinder, not an intermittent misfire.

I've checked compressions too. I can't really get away with an intrusive sensor, so the pressure transducer idea is out too (good idea though!).

Think I might have to butcher a knock sensor, unless we have some more ideas?


Notime, do you have any more info about the instantaneous crank velocity method, ie suppliers of equipment, and whether the resolution is good enough to work on, say, an 8 or 12 cylinder engine?
 
Last edited:
  • #6
Got fuel injectors?
One of those could be sticking.

There are temp monitors that are placed in the exhaust header.
One for each cylinder.
They sell aftermarket systems for aircraft and many are capable of detecting an individual miss.

Edit: No I don't have additional info on the crank velocity but I recall it was in conjuntion with crank position sensor. If you have one that scans flywheel teeth an ADC and software could do the trick.
 
Last edited:
  • #7
Edit Edit. Think it was experimental, not commercial
 
  • #8
Ooh all good stuff! I've got individual thermocouples in the exhaust runners but the response is far too slow to detect an individual miss. Will look into that one though, especially if they'll fit in the existing tappings.

It's carburetted, not fuel injected, good thought though. I have a crank position sensor scanning flywheel teeth, and an ADC, but I fear the natural speed swing of the engine will mask a slight decrease in speed due to a single misfire, particularly since it's a big, heavy engine.

Thanks!
 
  • #9
Do you have any way of monitoring chemicle composition of your exhaust? If you've got one cylinder occassionaly failing to ignite, you should be getting occassional puffs of exhaust with unburnt fuel in it. In fact, if you could test exhaust in the manifold, you could detect the exact cylinder that's not always firing.
 
  • #10
If you already have exhaust sensors, it may be just a matter of increasing the sample rate and/or resolution.
I would think the blast of cool air would raise a small signal even in a fairly masive sensor.
I recall most setups having very low sample rates(well below firing rate) with resolution in 10s of degrees.

Don't know if engine mass was an issue.
I think reflected load changes were a problem.
Along with a small s/n ratio to start.
Lots of play in even a tight engine.
 
  • #11
Thanks chaps,

Lurch, yes we do monitor exhaust emissions but the misfire is so intermittent, and the response of the emissions kit is far too slow to detect one single misfire; over ten seconds or so it would barely show at all. Like the idea though!

NoTime, the sensors are 10mm K-type thermocouples, read off an analogue system, no sampling rate to speak of. The 'blast' of cool air isn't as cool as you might think, - the heat of compression means that exhausted air from a non-firing cylinder is only 10% (ish) cooler than that of a firing cylinder. Not convinced that my thermocouples would detect that over such a small time, do you think otherwise?
 
  • #12
I have no idea if this is even possible, or if it would count as 'intrusive'. Is there any way that you can monitor rod bearing loads? A cylinder that fired will have a crush load on the top shell, whereas a piston that's being dragged down by the crank will have the load on the bottom shell.
 
  • #13
Yep certainly could! Strain gauges on the mains is the most common way of doing it. But much too intrusive for my purposes, I don't want to have to take the engine to bits at all ideally!
 
  • #14
I might think 10% is low.
The air does expand on exhaust and the compresion cycle would actually cool it baring input heat transfer from the cylinder walls.

Since I don't actually know.
My sugestion is to connect a high resolution ADC to the sensor, have the sample rate syncronised to the piston stroke rate or multiples (not the firing rate).
Force some one shot misfires and examine the output data stream.
They sell or perhaps you can borrow PC ADC card that can do this. A second chanel can record the one shots to locate the area of interest.

An analog meter system (or common digital meter) wouldn't have the instantaneous response or resolution necessary.

One posible problem is that aircraft engines tend to be operated at constant power. This is not the case for cars.
So an effective monitor would have to compare several cylinders or perhaps throttle plate position.
 
  • #15
There are wideband AFR logging systems that are capable of simultaneously looking at all of the cylinders (one O2 sensor per exhaust port) and are fast enough to detect a mis-fire, but I don't think they would be very helpful in diagnosing the CAUSE of the problem. They are not cheap, but are commonly used on racing teams, especially drag.

Depending on how bad you need to find out what's going on, this could be for you... and it would do an excellent job of measuring precise AFR out of each cylinder as well.

Here's what I would recomend:
http://www.innovatemotorsports.com/products/st12.php

Also a possible logger:
http://www.innovatemotorsports.com/products/dl_32.php

And then the AFR Sensor Controller, you'd need one per cylinder. This is the sensor I use on my Supercharged 4Runner, but only one:
http://www.innovatemotorsports.com/products/lc1.php

You would need one of these per cylinder too, since you will be very close to the exhaust ports:
http://www.innovatemotorsports.com/xcart/customer/product.php?productid=16148&cat=250&page=1

You also might just consider piecing together your own modular system:
http://www.innovatemotorsports.com/products/MTS.php
 
Last edited:
  • #16
Hmmm... then; just hmmmm...
What about replacing your exhaust runner thermocouples with pressure sensors? Would a fired cylinder have higher exhaust pressure as well as temperature? Or would backpressure from other cylinders mess it up?
I'm just free-wheeling now, guys.
Any chance of X-raying the thing while it's running, or would that be useless?
 
  • #17
Danger said:
Hmmm... then; just hmmmm...
What about replacing your exhaust runner thermocouples with pressure sensors? Would a fired cylinder have higher exhaust pressure as well as temperature? Or would backpressure from other cylinders mess it up?
I'm just free-wheeling now, guys.
Any chance of X-raying the thing while it's running, or would that be useless?

I think his best bet for detecting a mis-fire would be wideband Air/Fuel ratio near each exhaust port on the header. Now determining the cause, that's another story all together; if you see a severe lean condition it's fuel, if you see an extreme rich condition it's spark, but especially for spark it could be difficult to hunt down why it didn't fire, short of replacing the ignition system.
 
  • #18
I'm a bit behind the times, Mech. I didn't notice the posts after Brewski's last one. That sounds like a pretty good approach.
 
  • #19
Look into flame sensors. A flame, or ionized air, is conductive. Each time the exhaust valve opens there will be a certain amount of flame present in the port. You should be able to see something on a scope display from the flame sensor.
 
  • #20
Brewy, i may be missing some thing, but an intermittent miss fire is hardly
ever fuel related, may be a sticking valve or in extreme cases oil blow by,
so if the ignition checks out ok surly it is a head off job.
 
  • #21
Thanks for all the great responses guys.

Mech Engineer; thanks for the links! I had no idea those AFR systems were so responsive, I might have a chat with one of the reps.

Danger, good idea with the pressure transducers too, I'll try and have a play.

Wooly, I can't say too much but with this application the fuel side could indeed be a problem! I'm reasonably certain it's not spark or timing, since this system is well proven. It better not be blowby with the amount of time the engine has run!

Thanks for all the help chaps.
 
  • #22
brewnog said:
Thanks for all the great responses guys.

Mech Engineer; thanks for the links! I had no idea those AFR systems were so responsive, I might have a chat with one of the reps.

Danger, good idea with the pressure transducers too, I'll try and have a play.

Wooly, I can't say too much but with this application the fuel side could indeed be a problem! I'm reasonably certain it's not spark or timing, since this system is well proven. It better not be blowby with the amount of time the engine has run!

Thanks for all the help chaps.

Nods as good as a wink, sort of closed door new thingumyjig, good luck :smile:
 
  • #23
I'll keep thinking on it. Meanwhile, happy hunting. :biggrin:
 
  • #24
Any one remember the colour tune plug? i guess if one of those could be videoed in slow mo.
 
  • #25
Here's an interesting note from the LC-1 Documentation:

Innovate LC-1 Documentation said:
6.5.1 Advanced output programming

The normal state of the analog outputs is to update the outputs every time the LC-1 takes a new measurement. The LC-1 is fast enough to distinguish individual pockets of exhaust gas. For many applications this will be too fast. The advanced programming allows to set the analog out
update speed.
 
  • #26
Couple of thoughts here (just thinking out loud):

I have used the Innovate products. They are 'ok'. I have used the multi-O2 sensor unit on an engine dyno. http://www.gearchatter.com/Test/Innovate_8WB.JPG

I have never tried to use my sensor(s) for identifying knock.

Ion Sensing
I spent a couple of days on-site http://www.woodward.com/engine/smartfire.cfm?" [Broken] to learn about there engine control system. You probably don't need an ignition controller, but it might give you some ideas. They use ion sensing to determine knock and mis-fire. It is expensive and I was not that impressed with the unit particuarly with regard to knock detection. I believe the new BMW M5 uses ion sensing for the same functions.

Cylinder Pressure Analysis
A fairly affordable (and kick-ass) solution would be http://www.tfxengine.com/index.html" [Broken]. Check out their low cost unit. You can get it with spark plug based sensors, so you don't have to tap the cylinder head. I know of a couple professional drag racing teams that have started using this system.
 
Last edited by a moderator:
  • #27
Thanks Tex. I really wanted to use cylinder pressures but can't tap the heads for transducers. I'm not trying to detect knock. I've been trying to find suitable spark plug sensors but still haven't had any success, the link you provided does seem to be what I want but they won't suit my application sadly! Cost isn't an issue if I can get away with not having to tap the heads.

It was starying to look like I'm going to be having to use an intrusive technique, but one idea I'm going to pursue first is logging of knock sensor signals through a DAC.

However, that ion sensing could be just what I need. I'll have a chat with the manufacturer this week, but are you aware of an experimental counterpart to the SmartFire system?
 
  • #28
Ok, I've spent half the day playing with pressure transducers and knock sensors. Getting there, I think!

Thanks everyone.
 
  • #29
Sounds good. I would like to know more about the application and the data you log (if you get any pressure analysis data). If possible. :)

I am not for sure what you mean about an experimental counterpart to SmartFire, so I would have to say 'No'. I believe OEM's will be looking at this technology for both knock and misfire. I will be really interested to see how it works for knock.

The basic premise (IIRC) is to monitor resistance through the plug after ignition. There are 2 distinct events that can be tracked. One is a chemical reaction and the other is a thermal reaction. IIRC, this data is available by way of the Nitrogen-oxides allowing a current to flow during combustion of which the resistance can then be monitored. Knock was/is detected by comparing output to a library of data or algorithms particularly comparing the number of inflection points in the wave.

Misfire was simply like a flat line. If the ignition system has the capacity to fire again and the ECU can determine misfire quickly enough, then a second ignition could be ordered and emissions are likely to be greatly reduced.
 
  • #30
Ok, I've logged cylinder pressures and have got traces for each cylinder, the misfires are easy to spot. I've also been playing with sound recordings of each cylinder using knock sensors on each head, same story there. The misfires are easy to spot by blips in throttle response, but this alone can't attribute it to a single cylinder (the pressure/noise traces obviously can).

I sometimes use a very similar system to SmartFire for knock detection and retard, but this engine shouldn't be anywhere near to misfire anyway so I wouldn't need such a system. But for troubleshooting, a similar system would be useful, especially if I didn't have to fit knock sensors or pressure transducers.

As for how my current system works for knock detection, it just listens to the combustion on each cylinder. If it 'hears' a characteristic knock sound (compared with a recording of the engine's signature knock), it will retard the timing until it no longer knocks. I guess this is how SmartFire will work.
 

Suggested for: How could I detect an intermittent misfire on a spark

Replies
5
Views
1K
Replies
7
Views
1K
Replies
30
Views
2K
Replies
0
Views
228
Replies
1
Views
429
Back
Top