So this started innocently enough. I've been running MHD on my E89 for a while now and started pulling datalogs more regularly. Figured I'd run them through Claude to get some interpretation since staring at CSVs isn't exactly my strong suit.
The AI was extremely thorough. Maybe too thorough.
From MHD logs:
The reassuring stuff:
BimmerLink also showed:
For context, the car drives absolutely perfectly. No limp mode, no CEL, no rough idle, no power loss, pulls hard, boosts clean. I've been doing canyon runs with no issues whatsoever. Before I started logging, I had zero concerns about this car.
Should I trust any of these numbers or is this a case of having data without the context to interpret it properly? Has anyone else run their N54 logs through AI and ended up spiraling like this? Are these BimmerLink values (knock voltage, uneven running ranges, misfire counters) actually saying something meaningful or is this just normal N54 behavior that looks scary on paper?
Curious what you guys think.
The AI was extremely thorough. Maybe too thorough.
From MHD logs:
- LTFT running at 14–18% on both banks
- Cyl timing correction showing -3° to -8° across all 6 cylinders
- Knock events at light load/cruise constantly
- Ignition timing Cyl 1 median: 1.5° BTDC (was 20-21° previously)
- Knock voltage Cyl 3 and 6 max: 4.39V (baseline was ~0.55V at idle)
- Uneven running Cyl 3: -4698 to +3357 µs
- Uneven running Cyl 6: -1260 to +4707 µs
- Lifetime misfire counters: Cyl 3 at 403, Cyl 4 at 308, Cyl 2 at 249
The reassuring stuff:
BimmerLink also showed:
- LTFT actually 3% (the MHD 14% was a misread — apparently the INA0S tune custom tables confuse MHD)
- Lambda 1.01 — perfect across multiple drives and conditions
- Leakage adaptation -1.75 kg/h — no vacuum leak
- At a actual WOT boost moment: timing correction was 0°
For context, the car drives absolutely perfectly. No limp mode, no CEL, no rough idle, no power loss, pulls hard, boosts clean. I've been doing canyon runs with no issues whatsoever. Before I started logging, I had zero concerns about this car.
Should I trust any of these numbers or is this a case of having data without the context to interpret it properly? Has anyone else run their N54 logs through AI and ended up spiraling like this? Are these BimmerLink values (knock voltage, uneven running ranges, misfire counters) actually saying something meaningful or is this just normal N54 behavior that looks scary on paper?
Curious what you guys think.
Scottish Borders
San Francisco, CA