Scope and prerequisites
This is a log-review companion for the VAG ME7 examples in the MxT guide. Identify the exact ECU and matched logger definition first: signal names, trim representation, feedback strategy and sensor configuration vary. A believable MAF curve cannot compensate reliably for unknown injector data, fuel pressure, air leaks or exhaust leaks.
Closed-loop data
When the applicable lambda control is active, collect stable idle and part-load samples across the voltage and airflow range the vehicle actually reaches. The MxT workflow groups corrections by MAF voltage and filters rapidly changing samples. Keep the number and spread of observations as well as the average: a bin with sparse data or mixed operating states is weak evidence.
| Example signal | Role in the cited logger references |
|---|---|
uhfm_w, mshfm_w | MAF voltage and converted air mass; verify units and sensor range. |
fr_w, fra_w | Lambda-regulator output and a multiplicative adaptation in the alias list; confirm sign, scaling and the relevant bank or region. |
B_lr | Pre-catalyst lambda-control condition for bank 1 in the alias list. Other banks and software require their own definitions. |
nmot, wdkba, temperatures | Identify the operating state and exclude unsuitable transient or warm-up samples. |
Keep the same hardware, fuel and calibration when combining comparable logs. Retain adaptation state and review each bank. A correction can also reflect injector, fuel-pressure, leak or sensor problems; it is not a direct measurement of MAF error.
RPM/load trim heatmaps. The NefMoto heatmap discussion describes a tool for examining ME7.1 KFKHFM against logged lambda corrections. In the reviewed source revision, it reads RPM, load, fr or frm, and MAF; filters changes over a rolling sample window; and groups suitable samples into RPM/load cells with a minimum sample count. This answers a different question from the voltage grouping above. Check which trim is selected and its scaling. The source uses fixed example RPM/load axes, so compare them with the actual calibration before associating a displayed cell with a map cell.
The reviewed code can average the two bank corrections when both are present, which can conceal a bank-specific difference. Its rolling stability filter does not by itself establish that lambda control was active or that warm-up, overrun or other unsuitable states were excluded. Select appropriate log conditions, inspect each bank and retain sample coverage. A trim pattern may reflect fuel pressure, injector behavior, leaks or sensors as well as the air model; a heatmap is not a direct measurement of MAF error.
Check the tool revision. The README at revision 6212eb2 describes weighting samples by distance from the cell centre. Inspection of that revision's bucketed calculation indicates that the selected samples are already inside the cell and are compared with their own coordinates clamped to its bounds: their computed distance is therefore zero and the default weights are equal for those samples. This is a code-reading inference about that revision, not a claim about every release or a vehicle test. Check the implementation in use and compare its output with the underlying samples before applying any suggested change.
Richer or open-loop operation
For an application whose fitted feedback cannot measure the rich mixture, use a correctly located and calibrated wideband. Compare its measured lambda with the applicable final requested lambda, not merely the driver-request table. If the wideband and ECU are recorded separately, match the same events and allow for time, transport and sensor delays. If using AFR, ensure the display and logger use a consistent stoichiometric reference for the fuel.
MxT’s ME7 procedure uses separate logger files and describes pairing corresponding pulls. Before accepting its proposed curve, inspect event matching, sample coverage and the effect on already-correct regions. A requested-versus-measured difference is an investigation, not proof that the MAF is the only error.
For feedback dropping out during a pull, compare the sensor measurement, final request, regulator condition, immediate correction and learned adaptations as explained in ME7 lambda control state during boost. Identifying injector and MAF errors illustrates why matching lambda at one point does not uniquely establish either calibration.
Sources and attribution
Adapted and rewritten from TracqiTechnology/MxT contributors’ ME7 Calibration Guide, Apache License 2.0, full text, with signal labels checked against the ME7Logger alias list. Additional M-box feedback context follows S4Wiki contributors’ Tuning, revision 5324, Creative Commons Attribution 4.0. For injector calibration, adaptation state and lambda paths see ME7 fueling and MAF calibration and ME7.1 S4 fueling and MAF examples.