MINI Commanded vs Measured Lambda: Reading O2 Data Properly

Diagnostic Process

Test to a moving reference, not to a snapshot. The whole capture was taken at warm idle and through repeated idle → roughly 2,500 RPM → idle excursions. Every other value in this write-up is read against that RPM trace. A single displayed number tells you what one PID said at one instant; the time relationship between RPM and the PID that is supposed to respond to it is what tells you whether the DME is actually controlling the engine.

Start with the disagreement, stated precisely. Engine RPM 846.5, fuel/air commanded equivalence ratio 1.000, equivalence ratio (lambda) B1-S1 0.923, charging voltage 14.22 V. Those are two different PIDs answering two different questions. Commanded is the DME's target. Measured is what the upstream wideband sensor reports the exhaust contained. Lambda 1.000 is stoichiometric, below is richer, above is leaner, and 0.923 × 14.7 gives approximately 13.6:1 for gasoline. AFR conversion is only there to make the number intuitive — lambda is the better diagnostic language because it is fuel-independent and because the DME's own target is expressed in it.

Eliminate the things that corrupt mixture data before interpreting mixture data. Misfire counters were zero throughout, which matters because unburned oxygen from a dropped cylinder walks straight past the upstream sensor and makes rich look lean. That is only true of the captured windows; it does not rule out an intermittent misfire at some other time. Air mass flow rose with RPM and load and came back down, throttle actual tracked throttle setpoint closely, intake and exhaust VANOS changed with operating condition, spark advance moved with RPM and load, and charging voltage held around 14.2 V. Manifold pressure read 9.5-9.7 psi at idle — absolute, so roughly 5 psi of vacuum below the 14.7 psi atmospheric reference, which is exactly what an idling engine should show and emphatically not 9.7 psi of boost.

Note what those checks do and do not buy you. A MAF that responds believably supports the air-measurement side of the picture, but a responsive MAF can still be biased, which is why it has to be correlated against manifold pressure, throttle and calculated load rather than trusted alone. Dynamic spark, VANOS and throttle data prove the DME is actively managing combustion. None of it is a direct test of the oxygen sensor.

Cross-check the generic PID against the manufacturer's own mixture data. This is the step that changed the picture. Read through the MINI application rather than generic OBD, lambda actual value oscillated between roughly 0.98 and 1.03 around a lambda setpoint pinned at 1.00, repeatedly and across several separate captures — the small, continuous hunting around stoichiometric that closed-loop mixture control is supposed to produce. The oxygen sensor controller value sat at 0.98-1.01 over the same period.

Read the adaptations, because they are the DME's own accumulated opinion. Multiplicative mixture adaptation was 1.00 at high load and 1.02 at low load; additional mixture adaptation was 0 percent, with an additive mixture adaptation factor of 0.14 percent. Note carefully what these are: manufacturer-specific mixture parameters with their own scaling, not generic STFT and LTFT percentages, and they must be read against the service-information definition of each PID. Taken at face value, though, they say the DME had not learned any meaningful correction — and a DME that genuinely believed the engine was running 7.7 percent rich, in closed loop, would be expected to have learned something.

Look downstream for plausibility, not for a mirror image. Pre-catalyst sensor voltage held steady around 1.48-1.49 V. Post-catalyst voltage sat low, about 0.06-0.10 V, during stable idle captures, then rose to roughly 0.65 V and began oscillating once engine speed was raised. B1S2 is primarily a catalyst monitor and should never be expected to track B1S1 exactly, because the catalyst stores and releases oxygen. It is still useful context, and a downstream sensor that starts oscillating when load changes is behaving like a sensor watching a working converter rather than one drowning in unburned fuel.

Frame the conclusion as a question about the disagreement. Two readings from the same module, minutes apart, disagree about the same mixture. Either the engine is genuinely richer than commanded and one channel is failing to show it, or the mixture is under control and the generic equivalence-ratio PID is being derived or scaled in a way that does not represent the exhaust. Everything measured on this car — neutral adaptations, closed-loop lambda hunting on setpoint, plausible downstream behaviour, no misfires, believable airflow — leans towards the second possibility, but leaning is not proving. Neither reading was thrown away, and no part was recommended on the strength of one of them.