O2 Sensor

O2 sensors are used in all cars made for the United States. They are used in cars made for many other countries as well. When the O2 sensor in your car goes bad, your car runs differently. This is a story of my experiences with the O2 sensor in my car. Your mileage may vary.

Update! See the bottom of the page for info on cars with multiple O2 sensors!

Background info:
Gasoline engines run best when the air-fuel mixture is correct. The air-fuel mixture must be controlled to reduce exhaust pollutant emissions. Lowest emissions are achieved with a slightly lean mixture. The catalytic converter, an essential part for emission control, also likes a controlled lean air-fuel mixture. The O2 sensor detects the air-fuel mixture of a gasoline engine by measuring the amount of oxygen in the exhaust gas. The fuel-injection system will trim the mixture richer or leaner based on the signal from the O2 sensor. The typical installation is in the exhaust manifold, where the hot exhaust gases will pass by it. High temperature(>350 C) is required for the sensor to operate. My car has a heated sensor, so the sensor starts to work very quickly after the engine has started. It has three wires. Two are for the heater, and one is the sensor output. The metal case of the sensor is grounded and is the return for the sensor's output. Unheated sensors have one or two wires. I do not know the exact composition of the sensor, but I do know how it is constructed and how it operates. A ceramic-like material is exposed to the exhaust gas on one side, and to the outside air on the other. A voltage is produced as oxygen travels through the material. A rich mixture will produce almost 900 millivolts. A lean mixture produces about 100 millivolts or less. The sensor does not produce in-between voltages with any regularity. I think the material in the sensor is made to match the optimum air-fuel mixture, and the output of the sensor is pretty much on or off(too rich or too lean). The car's fuel-injection system picks a voltage to compare with, and modifies the mixture to try to maintain an on-off ratio close to 50%, so if you averaged the voltage over time it would be about 400 or 450 mV. The on-off transitions occur about one or two times per second in my car when it's running properly. When the sensor is cold, no voltage is produced, and the sensor is an open circuit with almost infinite resistance. When the sensor heats up, the impedance drops and it can produce a little bit of current. If I had to guess, I'd say that mine has an impedance of about 5000 ohms when hot. The car presents a very high impedance load, in the megohms.

My car, and what it did:
The car is a Dodge Caravan, the engine is a Mitsubishi 3.0 liter V-6 with Holley multiport electronic fuel injection. At 100,000 miles(what a coincidence) the car started to run a little badly. A bit of hesitation, and some slight bucking at highway speeds. Full throttle performance was good. Exhaust was a bit smelly, both at idle and when driving. After a day or two of this, the "check engine" light would come on whenever the car had warmed up. I did check, and there was an engine :-) Fuel economy had slipped about 35 percent.

What the car says is happening:
Diagnostic codes can be read on my car by turning the key on-off-on-off-on and watching the "check engine" light for patterned blinking. The car reported a "51" which is "lean condition detected". Table of diagnostic codes. Fortunately, when my car was new I ran an extra wire when I was installing an alarm system and had to run wires from inside the car to the engine compartment, and I now attached this extra wire to the O2 sensor signal wire. Attaching a digital voltmeter inside the car to the wire and to ground showed that the oxygen sensor output never got above 250 mV, showing a lean condition. The car was compensating for this lean signal by running as rich as it possibly could.

Bench test:
Removing the sensor was easy, and I attached an ohmmeter to the sensor and found it to be infinite ohms. At this time I did not know that this was normal, so I went to the auto parts store and asked for a new O2 sensor. I had my ohmmeter with me, and tested the new sensor right there at the counter, under the apprehensive gaze of the counterperson. Many mechanics think that an O2 sensor can be damaged by an ohmmeter, but they are wrong. I found the new sensor also had infinite ohms, but bought it anyway. About $55.

Installation and road test:
Smooth as silk. Sensor output toggles up and down, the car runs better. It takes a couple of days before the car is running optimally. I hypothesize that the fuel-injection system has a group of mixture profiles for various operating conditions that are updated very slowly. All is well for 3 months, and I surf the web finding out information about O2 sensors to satisfy my curiosity.

When nice cars do bad things:
My car blows a head gasket. (See a picture of the motor torn apart. 60K bytes) After a long and expensive repair, I'm back on the road. Very soon I realize that my car is running less than perfectly, gas mileage is down, and it takes a lot more throttle to get it to go down the road. The transmission was shifting harder, and at higher engine speeds than it should. I suspect all sorts of things, mostly related to the major repair that just happened. Feeling like an O2 sensor specialist, I decide to check the output of the sensor to see what I can see. Very unusual results. Sometimes all appears to be well, other times the output is either on or off. The "check engine" light reports no codes. After a week of this I decide to take further action. I removed the O2 sensor and tested it by heating it with a propane torch while attached to a voltmeter. Voltage was produced as I moved the torch around the sensor, but I really couldn't make any judgement of the sensor based on this.

I build a simulator:
I couldn't figure out if the oxygen sensor or the fuel system was malfunctioning, so I decided to build a simulated O2 sensor to send signals to the car. This way I could vary the signals to see if the car responded in a predictable way. I also wanted to monitor the signal from the O2 sensor, to see if it responded in a manner consistent with the way the car was operating at any given time. I only had the one test wire run into the engine compartment, so I had to get clever to do both. I ended up with a setup that gave me a voltmeter to show the signal the car was getting, a switch that would give the car the signal from the O2 sensor or from my simulator, and a knob I could adjust the simulator's output infinitely from full lean to full rich. The simulator produced an on-off oscillating signal with a period of about one second. I made a web page for the simulator if you want to build one yourself. The link is further down the page.

Fooling the car:
The car responded well to the simulator. I could adjust the mixture while driving. I drove my workday commute for two days under the simulator. Since this was a dumb simulator, the mixture would eventually get too lean or too rich, and I'd have to crank the knob one way or the other until the car started running properly before moving the knob back somewhere near the middle. Flipping the switch over to monitor the O2 sensor was not giving me consistent results, but I was too busy with my commute and keeping the vehicle on the road to figure anything out. Finally I tried running while parked, looking at the output from the O2 sensor. It seemed to produce low voltage, but if I revved the engine really hard, the output would start toggling normally for a while. I took the O2 sensor out again and put the torch to it to clean it off. I noticed a rattling noise if I shook it, and upon close examination found that a piece of ceramic material was floating around inside the metal shield. I decided to try to return the sensor in exchange for a new one. Luckily I had saved the box the sensor came in. I also saved the old sensor, and had used its plug and wires as part of my simulator cable. The auto parts store replaced my sensor for free(because of the rattling sound). The engine runs fine again!

Two sensors in three months:
Upon close examination of the original sensor, I could see that the ceramic element was broken and wedged against the slotted metal cover. Both sensors actually had similar failures, but the replacement sensor failed in a way that still allowed it to function sometimes, and kept fault codes from appearing in the diagnostics. The hard shifting continued, however. I tracked that down to the kickdown linkage being assembled incorrectly when my head gaskets were replaced. At light pedal the transmission was sensing more throttle than it should have, so it waited longer and shifted harder.

Conclusions:

You spend a lot of time with your car, so pay attention to it because you must diagnose your car yourself. Cars are so complicated that a mechanic can't spend enough time with your car to properly diagnose it. The exception to this is a specialist or dealership, where a mechanic may have prior experience with your model car, or factory support. The disadvantage to this approach is that dealerships are always out to take as much of your money as possible, and are frequently dishonest. My approach is to find an honest mechanic who will honor my diagnosis and do the repair I suggest, or just do the repair myself.
Replacement parts are usually(but not always) lower quality than original parts. Save your receipts and packaging for the parts, it helps when you return them after they go bad.
Buy a service manual for your car, and learn how to read the fault codes out of the car's computer.
Whenever possible, earn an enormous amount of money so you can buy a new car every few years before trouble strikes. I’m still working on this one :)

Click to go to the O2 sensor simulator web page

Cars with multiple O2 sensors:
Starting around 1994, and definitely by 1996, all cars have multiple O2 sensors. There's one in the normal position between engine and catalytic converter, and another behind the catalytic converter. The engine control module monitors this extra one to verify the efficiency of the catalytic converter. With normal operation this O2 sensor produces almost no voltage because the catalytic converter gobbles up all the good stuff left in the exhaust.
I got a message from a fellow in Venezuela who had to remove his catalytic converters because they burned out, unleaded fuel being unavailable where he lived. The O2 sensors also failed (leaded fuel is bad for them, too). His car noticed, lit up lots of dash lights and ran poorly. He built an O2 sensor simulator, connected it to the front O2 sensors(I didn't say, but cars with dual exhaust have them in both pipes, so he had a total of four) and to the rear sensors. This didn't completely work, of course. The car thought the catalytic converters were bad since the signals were the same "after" as "before". I suggested he hook the O2 simulator only in place of the front sensors, and ground the rear sensors connections. I think that worked. You could also attenuate the O2 simulator signal with a couple of resistors and feed that to the car instead of the rear sensors, the front signal coming directly from the simulator.

Finally:
Using an O2 sensor simulator on a permanent basis isn't a solution, the simulator is not ideal since it needs constant adjusting. I only used it as a diagnostic aid and to limp around a little better when my O2 sensor was bad. It was quite valuable for that, and I learned a lot!