I just took my 98 Mazda 626 for it's inspection and they said the carbon monoxide is slightly high. The test is done using a tail pipe gas analyzer while holding the RPMs between 2500 - 2800 for thirty seconds.

From what I've read, this is normally an indication that something is causing the engine to run rich:

Carbon monoxide (CO), is incompletely burned fuel or to be more precise are hydrocarbon molecules that split apart but don't burn in the combustion cycle. High (CO) is the result of one problem, a rich air/fuel mixture but may have several causes.

My CO reading was 0.6% by volume, with the manufacturer's limit being 0.3%.

The was no problem with NOx emissions.

However, during this test they also take a Lambda reading. My reading was slightly lean at 1.021, with the limit being 1.03. This reading seems reasonable as my fuel trims are close to zero. According what I've read in Engine Management: Advanced Tuning, this slightly lean lambda is a very reasonable target for steady state cruising conditions:

For best economy we swing to the other side of the stoichiometric balance with a target of about Lambda = 1.05 ( about 15.5:1 A/F ). ... To achieve this, we mix just enough fuel that, if burned completely, makes enough power to maintain current conditions with enough excess oxygen to ensure that no fuel molecules go unused.

So based on the objective readings of the gas analyzer I'm running both rich and lean at the same time!

Now, for anyone who's been following the adventures of my 626 project, you'll know that I've got a brand new O2 sensor, and that I had been having problems with hesitation, lack of power and rough idle. These problems were greatly improved, but not eliminated, by replacing my standard spark plugs with fine tip low voltage NGK g-power plugs.

Now I think these things are related. Having both incompletely burnt fuel and excess oxygen says to me that I'm getting poor / weak ignition of the A/F mixture, causing a slow, incomplete burn.

I've used a spark tester at night set to a 30kv gap which produced a weak thin looking orange spark, when based on folk wisdom I should be getting a thick bluish / white spark. I've tested resistance on my wires and they're all in spec.

My theory is that I've got a weak ignition coil. The Mazda WSM calls for three tests of the coil pack. The primary and secondary resistances are in spec, although it takes a second or two for the reading on the primary coil to settle. It's starts high and then slowly settles to the correct value of between 0.45 to 0.55 ohms. I was unable to do the third test, which is the "Insulation Resistance of Case" test calling to measure resistance between the ground terminal and the case with a "500 v Mega tester". I was also unable to do the conductance test suggested by another user.

The I've checked pretty much everything else under the sun, and this is the only thing that's making sense to me, so I ordered one which should be here in about a week.

I'd like to hear what people think about my theory.

Postscript April 8th, 2016

So I went in to do the test again today, and it passed emissions this time. This time my lambda was 1.009 and the CO was 0.1%. I'm not completely sure why, so I'll just list everything that was different.

First off, I had recently installed a fuel pressure gauge and had to remove the air intake hose. I forgot to tighten down the band where the hose meets the throttle body, so I tightened it down. It might be that a tiny bit of air was getting in there.

The first time I did the test I was idling in line for 20 minutes and they read engine temp as 126 C*. This time I went straight in, it was 96 C*, and they had me rev the engine hard three times before starting the test, and also had me rev up to 3000, although their equipment showed 2770.

Post-Post Script March 3rd, 2017

Thought I'd follow up again after this years emissions test.

Well, it failed emissions again when I went for the test last week, but just barely. Idling at 670 rpm according to their equipment, CO was at 0.51% with a maximum allowed of 0.50%, so it just barely failed there. Held at 2770 rpm and measuring 80*C both according to their equipment for 30 seconds the lambda read 1.017 which was fine, but the CO read 0.46% with a max allowed of 0.3%.

After reviewing what @FredWilson wrote here last year, I figured that since 98% of my driving is low speed, low rpm in the city, maybe taking the car for a high rpm drive on the highway for 15 minutes might burn off any carbon or sulfur that was interfering with the cat's performance. So I did that and went for a retest ( a week after the first test ). This time it read 0.04% idling at 670 rpm. At 2770 rpm and 91*C according to their equipment they read a 1.014 lamba and 0.18% CO, so it passed this time.

Some side notes; on the second test I was again very close to 3000 rpm according to my tac, which makes me think there is a problem with their equipment maxing out at 2770 rpm. Also, the first time I did the test the car was idling pretty rough, while, for no apparent reason the idle was much smoother the second time I did the test. Not sure if this was a contributing factor, or just coincidence.


4 Answers 4


The average gas engine exhausts between 1 and 1.5 CO during normal combustion at stoiciometric mixture (lambda 1.00). The catalysts is expected to burn this excess CO. So if there is .6% at the tailpipe then the most likely cause is a catalyst that is either not lit or the mixture is out of the catalysts working mixture range. The farther the mixture gets from lambda 1:000 the harder it gets for the catalyst to clean all the gasses. At 2% lean CO and HC should be easily combusted by a good quality, fully lit, catalyst. NOx would not be reduced very well at all. If the mixture was 2% rich the situation would be reversed. NOx is reduced better but CO and HC would not burn well.

Assuming the readings obtained are accurate the CO is not an indication of a lean mixture. If there are no exhaust leaks and the lambda reading is correct the engine is running lean. Optimal catalyst efficiency is found between lambda .995 and 1.005. Most systems maintain lambda well inside this range in most driving conditions. Again assuming the stated values are correct the catalyst is either not lit or failing and the system is doing a poor job of fuel control.

  • I'd been idling in line about 20 minutes so I would think the cat should be up to temperature, no? I'm not sure if they test for NOx locally, but they didn't say anything about NOx. So could this be a dirty cat which might benefit from running laquer thinner through the tank, or taking the cat off and washing it like Kilmer suggests here? youtu.be/5icTmYItwiE Commented Apr 3, 2016 at 7:53
  • @RobertS.Barnes Catalyst testing should done at 2500 rpm. At idle the catalyst can cool off and become unlit. Also exact lambda 1.000 fuel control is not always maintained at idle. Some systems will command a lean mixture to save a bit of fuel. They can because NOx is less of a concern. I do not recommend any added fuel type catalyst cleaning, this can seriously damage a Cat. You can burn out any embedded sulfur with a long high power uphill run. Commented Apr 4, 2016 at 1:47
  • The problem is solved, could you give your input on my postscript data. Commented Apr 8, 2016 at 12:35
  • @RobertS.Barnes On this test the catalyst is lit due to your efforts to keep the exhaust volume higher. The mixture is still a bit lean but closer to what the catalyst needs. Not enough data to say for sure but this catalyst is likely on the old and weak side. Commented Apr 9, 2016 at 2:29

The data doesn't support it

  • You measured your coil resistances to be within spec
  • You measured your wire resistances to be within spec
  • You've already replaced the plugs with ones that require less voltage to operate

So while the bad-coil theory may explain what's going on, I'm not confident it will.

Possible explanations for the readings obtained

Off the top of my head:

  • Exhaust leak

    A fine crack in the exhaust manifold or compromised head gasket could introduce oxygen to the exhaust stream. This could explain why lambda is slightly lean even though there are higher-than-expected levels of CO present.

  • Bad cats

    An ineffective three-way catalytic converter may not be scrubbing residual CO out of the exhaust. If your emissions report has NOx readings this may help to confirm/rule out this possibility.

  • Systematic measurement uncertainty

    So the emissions analyzer is reporting the CO & lambda values with some systematic error. Let's hope this is not the case.

  • I've visually checked the exhaust manifold for cracks, and I've also looked at the connection and pipe between the manifold and the cat and didn't see anything obvious, although it's possible I might have missed something. They didn't report any NOx problems, although I'm not sure if they check that were I live. See my edit regarding the coil pack. Commented Apr 3, 2016 at 7:32
  • @RobertS.Barnes I'll have to check the coil test update when I get home from work. Another possible explanation did occur to me - a clogged EGR. Do you know if the EGR is working as it should?
    – Zaid
    Commented Apr 3, 2016 at 9:05
  • This car has an electronically controlled EGR, so I tested the resistance on the pins like it says in the WSM, and they were in spec. There's not any other test listed. I've also seafoamed the intake manifold. Commented Apr 3, 2016 at 10:55
  • @RobertS.Barnes the issue may be mechanical, not electrical. If there is blockage it could explain the symptoms.
    – Zaid
    Commented Apr 3, 2016 at 10:58
  • Wouldn't blockage of the EGR result in NOx emissions due to higher combustion temps? Commented Apr 3, 2016 at 11:29

The highway drive at high RPM for 15 minutes is highly effective at incinerating carbon not only from the cat. but also the spark plugs, exhaust valves, combustion chamber and O2 sensor. Ford TBA recommends RPM be at least 3,500 for 4-6 minutes, but Chrysler does not recommend sustained RPM above 4,300 for more than 15 seconds. The run can done with the transmission shifted into first gear (on an automatic, "D1" or "L1"). I have found this type of drive to be highly effective in curing detonation ("ping"), poor acceleration, poor idle, excessive gas mileage in my cars, especially since I, too, only drive slowly in the city.

You can further aid carbon removal with Marvel Mystery Oil (MMO). A one once treat rate per gallon of gas in the gas take is an aggressive treatment; perform the high RPM drive every 1/4 tank of gasoline after driving around stop-n-go. I have seen smoke pour out of my cats and improve performance! Use it in the crankcase with fresh synthetic oil by substituting twenty percent of the motor oil with MMO, drive at the high RPM, then change oil again after 200 miles. Once the engine is clean, use MMO in the gas tank 4 oz. for every ten gallons of gas -- you will see mileage improve 8% and will never have a problem with carbon again. Add 8% to the crankcase during winter months.

You can combine Berryman's B12 with MMO for better cleaning of the entire system, especially the cats and EGR system. It is mostly acetone and some other volatile organic compounds that work synergistically with MMO.

I've been doing these things to one of my cars since new for 23 years and it still runs like brand new. I just had an emissions test done here in New York and the levels were far, far below threshold values.

  • I just use home-brewed seafoam to clean my intake valves and pistons. Commented Mar 4, 2017 at 17:23
  • Sure, using Seafoam through the induction will clean valves & heads, but will do nothing for the the fuel tank, fuel pump, or fuel injectors. Also, it will not help the "dry" side; i.e., the exhaust components of O2 sensors, EGR, and catalytic converters. I have personally seen the difference after experimenting with numerous cars over the past 35 years. I have researched the best method of cleaning for countless hrs. over the Internet, so give it a try & you'll see.
    – Carguy
    Commented Mar 6, 2017 at 8:04
  • If you put oil to your petrol, you will see smoke, that is normal. MMO won't improve mileage. If anything it is known to cause engines to perform badly and even cause plane crashes (according to NTSB : en.wikipedia.org/wiki/Marvel_Mystery_Oil ). I feel that this magical 8% number is just made up. It is impossible since mineral oil has lower energy density. Not to mention it is illegal in some places to add MMO into your fuel due to harmful emissions (since you are burning oil!) it causes. Commented Mar 11, 2017 at 18:55
  • You've never never used MMO, so how can you be so sure it doesn't improve gas mileage? It improves compression, and in turn mileage, by forming a better seal around the top piston rings (much like a two-cycle engine that has oil mixed into the gas). It only caused one plane crash, reasons not clear, after millions of uses. Where is it "known to cause engines to perform badly"? I would also like to know in which locales it is expressly illegal, or any shred of evidence that it contributes to harmful emissions to any measurable degree (it's not at all like burning motor oil).
    – Carguy
    Commented Mar 13, 2017 at 14:31

Another thing not to overlook is a dirty air filter. I was just working on a European OBD1 2000 Honda Civic D16Y8 which had failed emissions. The car had already been to someone else who said it needed a new engine. The first thing I found was an O2 sensor code, and it turned out the pintle was physically broken off the sensor. However, after replacing the O2 sensor and doing a good long high rpm pull up a mountain, it was still failing with a lambda of 1.042 and CO of 0.72% at the tail pipe at 98*C and 2720 rpm.

All the scan data looked normal, with normal fuel trims and the O2 sensor switching normally. After failing to find any exhaust leak and on a hunch I took a look at the air filter. It was extremely dirty, probably hadn't been changed for a few years.

After changing the air filter the only difference I noticed in the scan data was that the IAC was opening between 2-3% instead of 4-5%. However, the lambda went down to 1.020 ( still a little lean but under the 1.030 limit ) and the CO went down to 0.16% with the limit being 0.3%.

So I would say first thing to do if you can't seem to find any problem yet it's still failing CO, change the air filter before running to get a new catalytic converter!

  • Why would a clogged air filter make the system lean?
    – ronenfe
    Commented Sep 9, 2020 at 21:24
  • @RonenFestinger It's likely due to the air flow assumptions that the manufacturer made regarding the air filter. Not everything is calculated real time, they make certain assumptions in their computer models and an extremely dirty air filter might make enough of a change to slightly throw off their calculations. Commented Sep 14, 2020 at 9:31
  • A clogged air filter will make less air in the system so it should be richer.
    – ronenfe
    Commented Sep 15, 2020 at 10:09

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .