Tuning Wiki

Air-fuel ratio (AFR) is the mass of air divided by the mass of fuel; lambda (λ) is that ratio divided by the stoichiometric ratio of the fuel in use. Lambda 1.0 is stoichiometric, below 1.0 is rich and above 1.0 is lean. A wideband oxygen sensor measures lambda across the range a tuner needs; a narrowband sensor only indicates which side of 1.0 the mixture is on. This page is the overview. How mixture is read and calibrated, including closed-loop control, transport delay and gas analysers, is on mixture tuning: AFR and lambda; sensor generations, connectors and controllers are on wideband oxygen sensor types.

Lambda rather than AFR

Bosch's training literature uses 14.7:1 by mass as the stoichiometric ratio for the gasoline it describes; ethanol's is about 9.0:1 (Haltech). Because the stoichiometric ratio changes with fuel, an AFR figure means different things on E10, E85 and diesel, while lambda does not. A display fixed to 14.7 misreports any other fuel; the lambda channel does not have that error. Lambda is therefore the unit used for targets on this wiki. Fuel effects are covered on E85, ethanol and flex fuel.

Target values

No single mixture is best everywhere. A three-way catalyst needs stoichiometry; best torque and best economy lie on either side of it. The figures below are starting ranges, graded by source. The right target for an engine is found by measurement on that engine, and full-load targets are further constrained by knock and exhaust temperature.

ConditionLambdaSource and grade
Warm idle and part-load cruise with a three-way catalyst1.00 (catalyst window about 0.99 to 1.00)Bosch training literature (manufacturer)
Maximum power, manifold-injected petrol0.85 to 0.95Bosch training literature (manufacturer)
Best economy, homogeneous petrolAbout 1.1 to 1.2Bosch training literature (manufacturer)
Full load, naturally aspirated petrolAbout 0.80 to 0.90HP Academy (tuning school)
Full load, moderate boostAbout 0.80 to 0.85HP Academy (tuning school)
Full load, high boostAbout 0.75 to 0.80HP Academy (tuning school)
Full load, turbocharged on E85About 0.80 to 0.82, against 0.78 to 0.80 on pump petrolHP Academy tutor, forum answer (single source)
Diesel at the smoke limitLean of 1; patent literature gives about 1.1 to 1.3Patents (indirect); no Bosch EDC figure found

The richer full-load targets on boosted engines are mainly for protection, not torque. On one HP Academy test engine, torque was almost flat between about lambda 0.86 and 1.0, peaking at 0.92. Running richer than best-torque lowers combustion and exhaust temperature and adds knock margin; Bosch ME7 component-protection logic enriches specifically to bring exhaust temperature down. E85 can often run nearer best-torque because its knock resistance does not depend on charge cooling from extra fuel. Why boosted engines run richer, and how a request differs from a delivery limit, is set out on mixture tuning.

Narrowband and wideband sensors

A narrowband zirconia sensor produces a voltage from the oxygen difference between exhaust and reference air. Its output changes steeply around lambda 1 and says little about how rich or lean the mixture is away from it, which suits an ECU holding stoichiometry and does not suit measuring a full-load target.

A wideband sensor such as the Bosch LSU 4.9 is a planar zirconia dual-cell limiting-current sensor with an integrated heater. The controller varies the pump-cell current to hold the Nernst cell at a set point, and that pump current is what correlates with oxygen content; the output voltage is produced by the evaluation circuit, not by the sensor. The LSU 4.9 reads monotonically from lambda 0.65 to free air, on petrol, diesel and E85, with a stated accuracy of ±0.01 at lambda 0.80 and ±0.007 at 1.016. It works only with a matching controller, and its connector carries a trim resistor that defines the individual sensor's characteristic. The LSU ADV has no trim resistor and needs its own evaluation unit.

Installation

RuleReasonSource
Mount at least 10° above horizontal, cable end up; Innovate specifies between 9 and 3 o'clockCondensate must not collect on the hot ceramicBosch LSU 4.2 and 4.9 data; Innovate LC-2; AEM 30-4100
Before any catalyst, in the collected streamA post-catalyst sensor reads the catalyst's oxygen storage, typically leanInnovate LC-2; AEM 30-4100
After the turbine on a turbocharged engine; makers' distances vary (Innovate at least 24 in from port or turbo, Haltech about 1 m)Exhaust pressure exaggerates rich and lean readings; the LSU 4.9 is specified to 2.5 bar with reduced accuracy aboveBosch LSU 4.9; Innovate; Haltech; Tech Edge
Exhaust below 930 °C continuous (1,030 °C briefly); sensor hexagon below 600 °CSensor temperature limitsBosch LSU 4.9
Away from the tailpipe, with no leaks upstreamAir drawn in at a leak or open end reads lean, especially at light loadBosch LSU 4.9; Innovate LC-2
Controller powered whenever the engine runs; heater not run at full power before the exhaust is dryAn unpowered sensor in exhaust is damaged; water on a hot element cracks it. Controllers use a reduced-power condensation phaseBosch LSU 4.2; Innovate LC-2; Haltech
Free-air calibration where the controller requires it, with the sensor out of the exhaustSets the lean referenceInnovate LC-2

Leaded fuel, silicone and oil contamination shorten sensor life; AEM advises mounting at least 36 in from the port when leaded race fuel is used. Bosch requires the reference-air path to stay clear of oil and wax.

What a reading does not prove

One sensor in the collector reports a mixed sample of the cylinders feeding it. An acceptable lambda does not show that every cylinder is the same, that the engine is free of knock, or that exhaust temperature is safe. Readings lag the combustion event by the gas transport time and the sensor response; a full-load figure is judged after throttle and boost have settled, and fuel-cut sections are excluded from averages. Lambda is read with ignition and knock correction, boost, fuel pressure and temperatures: see ignition tuning, engine knock, exhaust gas temperature monitoring and data logging. Carburettor tuning uses the same sensors: carburettor tuning.

Not established

No manufacturer response-time figure for the LSU 4.9 was found; AEM's claim of about 20 ms for its X-Series controller is a vendor test. No Bosch EDC figure for the diesel smoke-limit lambda was found.

Sources

Continue the conversation

If this topic interests you, consider joining the following forums. They host material referenced in this article.