B58 / S58 log reading: start with the question, then the graph

Jump to the recorded example · What changes for S58? · Tuner handoff checklist

1. Decide what you are trying to learn

“Is this tune safe?” is too broad for one CSV. A more useful question is: “During this acceleration, did measured lambda follow the logged target?” Or: “Did the pressure change happen before or after the driver lifted?” Write down the engine variant, calibration revision, fuel, relevant hardware, gear and starting temperatures. If those details are unknown, keep them unknown rather than filling gaps with a model badge or a filename.

EcuTek describes the B58 controller as a modeled, torque-based system. Pedal input is a demand signal, not a promise of a particular throttle opening or boost value. That is why several synchronized channels are more useful than an isolated peak. Source: EcuTek B58 Tuning Guide.

2. Check what was actually recorded

Before interpreting a graph, check the exporter, column labels and units. Request time, RPM, pedal, throttle, actual and target pressure, actual and target lambda, cylinder timing corrections, actual and target rail pressure, and intake temperature. Gear and relevant torque-limit or intervention channels help explain transitions. The exact channel list depends on the logger and vehicle software.

MHD lists boost, lambda, temperatures, ignition, fuel pressure and other monitors for B58, while noting software-dependent availability for advanced parameters. A channel missing from your export is not evidence that the corresponding event never occurred. Source: MHD B58 monitor documentation.

3. Follow one real B58 excerpt

Our contributor authorized this anonymized eight-channel excerpt: 113 samples over 8.96 seconds, with timestamps rebased to zero. We removed the VIN, original filename, calendar timestamps, programming metadata and other identifying fields. The retained source identifies BMW B58 and ECU Connect. Fuel, modifications, calibration and test conditions have not been independently verified.

The charts below connect recorded samples without smoothing. Use the table to check exact points rather than estimating values from a small phone display.

At 3.20 seconds, actual lambda is 0.091 below target. At 4.80 seconds it is 0.042 above target. A positive difference means leaner than the command at that sampled instant, not necessarily leaner than stoichiometric and not a diagnosis. The change in sign is a reason to inspect neighboring samples and operating conditions, not choose whichever single row supports a preferred conclusion.

At 4.80 seconds, the logged throttle is 75.3% while pedal is 100%. The excerpt does not include the intervention channels needed to assign a cause. It also does not include rail pressure, intake temperature or cylinder corrections, so it cannot answer whether a fuel-pressure event or timing correction accompanied the lambda difference.

The analyzer currently selects one interval from 2.00 to 6.96 seconds (63 samples, 2458–6155 RPM at the endpoints). That is a heuristic selection, not a guarantee of a clean single-gear pull. Notice that the 8.00-second row is outside it. The whole-file table and the selected-pull report answer different questions.

Load this example in the analyzer

4. Keep pressure units and targets honest

The retained manifold measurement is absolute pressure. At 3.20 seconds it is 18.33 psia, while the retained manifold target is 31.69 psi. Subtracting the same atmospheric reference from both preserves their 13.36 psi difference. It does not prove those channels are a valid control pair or explain the difference.

This excerpt lacks the pre-throttle idle samples used by the adapter to estimate atmosphere. Its gauge display therefore uses 14.696 psi, a stated assumption rather than a measurement of the test location. Do not compare that display with another run using a different pressure reference without reconciling the units.

TunerBench currently rejects this excerpt's pressure target for automatic comparison because it does not meet the adapter's tracking heuristic. That rejection is a tool limitation, not proof of a healthy or faulty boost system. A target can be valid even when actual pressure fails to reach it. Keep the raw columns and ask the calibrator to confirm the target definition before drawing a conclusion.

5. Carry the workflow to S58, not the numbers

For an S58 log, repeat the same evidence checks using that vehicle's logger definitions and calibration targets. Do not use this B58 excerpt as a reference boost curve, lambda prescription, pump limit or timing-correction allowance. Even two B58 logs can differ in hardware, fuel and software. An engine selector in TunerBench supplies report context; it does not certify the calibration or change the screening thresholds.

There is no S58 recording in this worked example. If an S58 file omits a relevant bank, cylinder or target, report that gap. Do not substitute an available channel from another bank or assume a missing cylinder recorded zero corrections.

6. Compare runs without declaring a winner

Use the analyzer's second-log comparison only after checking that the selected intervals cover comparable RPM ranges and conditions. An increase in timing, pressure or richness is not automatically an improvement. More timing can accompany different load; a different lambda target changes the meaning of the actual value. Review actual-versus-target behavior and supporting channels before attributing a change to a part or tune revision.

What to send your tuner

• The original unedited log through a private channel, with the engine, hardware, fuel and calibration revision.
• The exact time and RPM of the event, plus the actual and target channels you compared.
• Any symptoms, fault codes, gear changes and known traction or torque interventions.
• The missing channels and any pressure-reference or timestamp assumptions.
• A specific question, such as “Is this the correct manifold target for this calibration?”