How Oxygen Sensors Keep an Engine on the Right Air-Fuel Mix
Features

Oxygen Sensors Hold Petrol Engines Near 14.7 to 1 From the Exhaust

Photo: User:Mgiardina09  This photo was taken with Samsung Galaxy S4, CC BY-SA 3.0, via Wikimedia Commons · Source · Licence

A probe in the exhaust that never sees the fuel yet steers every injector pulse: what the lambda sensor measures, how the computer reads it, and why wideband and narrowband differ.

FeaturesBy 11 min read
The short answer

An oxygen (lambda) sensor sits in the exhaust and reports whether the burnt mixture was rich or lean. The engine computer uses that signal to keep petrol engines near 14.7:1 air to fuel, which suits the three-way catalyst. Volvo first used the technology in 1976.

The sensor that decides whether your engine is burning its fuel properly never touches the fuel and never sees the air going in. It sits in the exhaust pipe, samples what is left after combustion, and from that single clue the engine computer works out how to adjust the injectors. Bosch developed it during the late 1960s, and Volvo became the first manufacturer to put the technology on the road in 1976.

What does an oxygen sensor do?

An oxygen sensor is a small electronic part that works out how much oxygen a gas contains. In a car it goes by several names: O2 sensor in the workshop, lambda sensor in engineering, and sometimes just the probe. Its job is to give the engine computer a running verdict on the mixture that has just burned. Too much fuel and the mixture is rich. Too much air and it is lean.

The clever part is that it does this from the wrong end of the engine. Since the sensor lives in the exhaust stream, it cannot measure the air or the fuel that went in. It infers. Combine its report with information from other sensors and the engine management system can work out the air-fuel ratio indirectly, which is enough to run modern electronic fuel injection and emissions control.

Earlier fuel systems read a quantity of petrol off a fixed map and hoped for the best. With an oxygen sensor, the injectors vary what they deliver according to real-time data, check the result in the exhaust and correct themselves. Engineers call this closed-loop control, and it also lets the computer judge whether the catalytic converter is performing properly.

Closed-loop controlInjectorsEngineExhaustOxygen sensorEngine computerSignal0.2 V lean0.45 V ideal0.8 V rich
The oxygen sensor in the exhaust reports to the engine computer, which adjusts the injectors; the sensor voltage reads about 0.2 V lean, 0.45 V ideal and 0.8 V rich.
The sensor does not measure oxygen so much as the gap between the exhaust and the outside air, and that gap is enough to steer the injectors.

Why does the engine aim for 14.7 to 1?

Petrol needs air to burn. The ideal quantity, the one that burns every drop of fuel with no oxygen left over, is called the stoichiometric mixture, often shortened to stoich. For petrol it is about 14.7:1 by mass, so every gram of fuel wants 14.7 grams of air. Anything above that figure is lean, anything below it is rich.

That figure moves with the fuel. Pure octane, the tidy laboratory version of petrol, comes out at about 15.1:1. Real pump fuel is a cocktail of hydrocarbons and additives, and oxygenators such as ethanol, methanol or MTBE carry extra oxygen into the combustion event, which pulls the number down. A fuel laden with MTBE can sit as low as 14.1:1.

Because the target shifts with the fuel, engineers prefer to talk about lambda. Lambda is simply the actual air-fuel ratio divided by the stoichiometric one for that fuel. A lambda of 1.0 means stoich, anything under 1.0 is rich, and anything over 1.0 is lean. Multiply a measured lambda by the stoichiometric ratio and you recover the air-fuel ratio again. The Greek letter on the sensor's name is that same lambda.

Why not just run rich for power or lean for economy and be done with it? Because every point on the scale trades something away. A rich mixture wastes some fuel yet can make more power and burns at a lower temperature. A lean mixture squeezes more from the fuel but runs hotter, and those higher temperatures lead to nitrogen oxides. In a naturally aspirated engine burning octane, peak power often arrives between 12.5 and 13.3:1, which is a lambda of 0.850 to 0.901. A ratio of 12:1 is generally treated as the maximum-output mixture, and 16:1 as the maximum-economy mixture.

Stoich itself has an awkward side. A mixture at exactly the stoich ratio burns extremely hot, and under high load that heat makes detonation possible, the uncontrolled burning that drivers know as knocking or pinging and that can push cylinder pressures very high. So stoichiometric mixtures are reserved for light to low-moderate load. For hard acceleration the computer richens the mixture, which produces cooler combustion products and protects the cylinder head.

Why does the catalytic converter care?

A petrol engine's tailpipe emissions come down to three culprits. Hydrocarbons escape when fuel is not burnt completely, as in a misfire or a rich-running engine. Carbon monoxide is the result of running slightly rich. Oxides of nitrogen dominate when the mixture is lean, and they form when combustion chamber temperatures climb past 1300 kelvin because of excess air. Nitrogen oxides add to smog and acid rain.

The three-way catalyst fitted to petrol, propane and natural gas engines is built to deal with all three at once, but only inside a narrow window. The O2 signal lets the computer hold the mixture within a whisker of 14.7:1, the best operating point for that type of catalyst. Drift to either side and one pollutant rises as another falls.

Real combustion never quite reaches the textbook. At 6,000 rpm one crankshaft revolution takes 10 milliseconds, and most of the fuel burns in roughly 2 milliseconds, measured from the spark until 90% of the mixture has burnt. Catalysts work best on exhaust from nearly perfect combustion, so the engine is trying to hit a target it can only approach.

How does a lambda probe measure oxygen?

The classic probe is built around zirconia, a ceramic. The original element was a thimble-shaped piece of zirconia coated on both the exhaust side and the reference side with a thin layer of platinum, and today's sensors keep the same principle. Think of it as a battery that runs on the difference between two atmospheres. One face sees exhaust gas, the other sees ordinary outside air, and the ceramic generates a voltage that depends on how different the two are. Engineers describe it as a solid-state electrochemical fuel cell, known as a Nernst cell.

The sensor therefore does not report a concentration of oxygen in the usual sense. It reports the gap between exhaust and air. A rich mixture leaves little oxygen in the exhaust and creates what amounts to an oxygen demand, so oxygen ions move through the ceramic and the voltage rises. A lean mixture leaves oxygen to spare and the voltage falls.

The numbers are easy to remember. An output of 0.2 V (200 mV) represents a lean mixture, and 0.8 V (800 mV) a rich one. The ideal set point is about 0.45 V (450 mV), which corresponds to a mixture roughly 0.5% lean of the stoichiometric point, where the exhaust carries minimal carbon monoxide. In normal use the probe will produce up to about 0.9 volts when the mixture is rich.

The response is far from a straight line. The voltage is most sensitive close to lambda 1 and much less so when the mixture is very lean or very rich. In practice that makes a standard zirconia probe behave more like a light switch than a dimmer: it flips decisively as the mixture crosses stoich but says little about how far away from stoich it is. That is why it is called a narrow-band sensor: it responds usefully only across a slim span of ratios.

Heat is the other requirement. The ceramic only begins to work effectively at about 316 degrees C (600 degrees F). Old unheated probes had to wait for the exhaust to warm them, which can take several minutes depending on ambient temperature and the layout of the exhaust. That slow start-up is linked to pollution problems, as is a similar delay in the catalytic converter reaching working temperature. Modern probes carry a heating element encased in the ceramic that brings the tip up to temperature quickly.

The planar-style sensor, which reached the market in 1990, took that idea furthest. It cut the mass of the ceramic sensing element and built the heater into the ceramic itself, so the probe warmed up earlier and reacted more quickly. The stainless steel construction resists corrosion in a hot, aggressive exhaust. A typical probe has four wires, two for the signal and two for the heater. Some makers use the metal case as the signal ground and get away with three, while older unheated designs had one or two.

How does the engine computer use the signal?

The engine control unit is the device that takes the signal and does something with it. Among its main duties are fuel injection, ignition, idle speed control and variable valve timing or lift. It also sets spark timing and the width of each injector pulse, so a distorted oxygen reading can ripple into several systems at once.

When the engine is working lightly, such as gentle acceleration or a steady cruise, it runs in closed-loop mode. The computer changes the amount of fuel and then expects to see a matching change in the sensor's response. This feedback loop deliberately pushes the engine a touch lean, then a touch rich, then lean again. No one position is held. Think of steering down a straight road: the wheel is never perfectly still, yet the small corrections keep the car on the line. The mixture averages out near stoichiometric even though it never sits there.

Speed of response matters in any control system, and this one is no exception. Quicker swings mean a higher cross count, as engineers call it, and a higher count means the system reacts sooner. An ageing or fouled sensor tends to react slowly, and that dulls the whole system.

Put the engine under high load, such as wide open throttle, and the arrangement changes. The sensor output is ignored. The computer enriches the mixture automatically to protect the engine, because a misfire under load is far more likely to cause damage. This is open-loop mode, and any change in the sensor's output is disregarded while it lasts. In many cars, apart from some turbocharged models, the air flow meter input is ignored too.

None of this appeared overnight. Ford's Electronic Engine Control system, built around a Toshiba TLCS-12 microprocessor, entered mass production in 1975, and Bosch's first engine management system, Motronic 1.0, arrived on the 1979 BMW 7 Series (E23). Lambda sensors reached the United States from about 1979 and by 1993 many European countries insisted on them for every car model.

Wideband or narrowband?

TypeWhat it reportsHow it worksWhere it is used
Narrowband zirconiaVoltage: about 0.2 V lean, 0.8 V rich, 0.45 V idealNernst cell comparing exhaust with outside airFactory standard from the late 1970s and early 1980s
TitaniaResistance, which changes with oxygen and temperatureTitanium dioxide ceramic that makes no voltage of its ownLess common narrowband type, sometimes paired with a gas-temperature sensor
Wideband zirconia (UEGO)Pump current, which indicates oxygen content directlyPlanar zirconia element plus an electrochemical gas pump and a feedback circuitEngine management since 1992, dyno tuning, driver air-fuel displays, stratified injection, diesels

NTK brought out the wideband sensor in 1992 to meet rising demands for better economy, lower emissions and stronger performance all at once. It is built on a planar zirconia element, but adds an electrochemical gas pump. An electronic circuit with its own feedback loop varies the pump current so the cell's output stays constant, and that current then shows the oxygen content of the exhaust directly. Inside there are three elements: an ion oxygen pump, a narrowband zirconia sensor and a heating element. The wiring is more elaborate too: typically six wires, one of them for a calibration resistor.

The practical gain is that the lean-rich cycling built into narrowband control disappears. With a reading that says how lean or how rich the mixture is, rather than merely which side of stoich it sits, the control unit can alter fuel delivery and ignition timing much more rapidly. The sensor is also called a UEGO, short for universal exhaust-gas oxygen sensor, and it is commonly found in aftermarket dyno tuning and in the air-fuel displays that performance drivers fit to their dashboards.

Titania takes a different route. Its ceramic is titanium dioxide, which does not generate a voltage but changes electrical resistance with oxygen concentration. Since that resistance also depends on temperature, some titania sensors are used alongside a gas-temperature sensor to cancel out the effect.

Aftermarket air-fuel meters simply read the sensor's voltage. Narrowband versions usually have 10 or 20 LEDs in round housings of 52 or 67 mm, to match standard gauges, and wideband sensors are much more accurate, though more expensive.

What happens when a sensor wears out or is tampered with?

Sensors do not last forever. Normal ageing wears them, leaded fuel poisons them, and fuel contaminated with silicones or silicates can foul them. The consequence is rarely confined to the sensor. A failed unit can damage the catalytic converter, and that leads to expensive repairs.

Tampering is a separate hazard. Interfering with the signal the sensor sends to the engine computer is bad for emissions control and can damage the vehicle. Suppose a modification pushes the engine moderately lean. There may be a slight gain in fuel efficiency, sometimes at the cost of more nitrogen oxides, much higher exhaust gas temperatures and a small increase in power. At ultra-lean ratios, that small gain can turn quickly to misfires and a sudden collapse in power, with potential damage to the engine and, through the misfires, the converter.

Push it rich instead and power rises slightly, but only up to a point; beyond it the engine floods on unburned fuel. Economy gets worse, unburned hydrocarbons in the exhaust increase, and the catalytic converter can overheat. Prolonged operation on a rich mixture can cause catastrophic converter failure. Because the computer also controls spark timing and injector pulse width, a mixture that is far from correct can leave the fuel igniting too early or too late in the cycle, which wastes fuel in its own right.

What this means for buying and driving

For anyone shopping for a used car, the era of the car says a lot about how it copes with fuel. A car with oxygen-sensor feedback measures the exhaust and adjusts the injected volume, so it compensates automatically when the fuel's stoichiometric ratio changes. That covers seasonal blends and fuels containing ethanol or other oxygenators. Cars from before the mid-1980s lack this control. They may have difficulty with certain blends, especially winter fuels in some regions, and can need different carburettor jets to restore the right fuelling.

On a modern car, the practical advice follows from the failure modes above. Keep leaded and silicone-contaminated fuel out of the tank. Treat any change to the sensor's signal as a modification to the emissions system, not a harmless tweak. And if you want to see what the mixture is doing on a performance car, an air-fuel ratio meter can read the sensor and monitor the ratio on any vehicle that has one.

Where this goes next

Wideband technology keeps spreading. It is already used in stratified fuel injection systems, and it has also been adapted to diesel engines, helping them meet EURO and ULEV emission limits. The reason is the combination the sensor was created to serve: better economy, lower emissions and better performance together, which a sensor that only flips between rich and lean cannot deliver on its own.

Frequently asked questions

What is the difference between a wideband and a narrowband oxygen sensor?
A narrowband zirconia sensor mainly flips between lean and rich around stoichiometric, giving about 0.2 V to 0.8 V. A wideband sensor adds an electrochemical gas pump whose current shows oxygen content directly, so the computer can react much faster.
Why does an oxygen sensor need to be hot?
The ceramic only works effectively at about 316 degrees C (600 degrees F). Modern probes have a heater built into the ceramic to reach that quickly; unheated ones can take several minutes to warm from the exhaust.
What damages an oxygen sensor?
Normal ageing, leaded fuel and fuel contaminated with silicones or silicates can all cause failure. A failed sensor can in turn damage the catalytic converter, which makes the repair expensive.
What is lambda?
Lambda is the actual air-fuel ratio divided by the stoichiometric ratio for the fuel. A value of 1.0 is stoichiometric, below 1.0 is rich and above 1.0 is lean.

Sources

  1. Oxygen sensor - Wikipedia
  2. Air-fuel ratio - Wikipedia
  3. Engine control unit - Wikipedia
Next story / Engines

Motronic: How Bosch Put Fuel and Spark in One Engine Computer