Flight 006 — Post-Rebuild Shakedown

Setup

First real flight after a rebuild during which the board may have been held at soldering temperature longer than ideal. The question was narrow: did any joint or component suffer, and is there any electrical damage that would show up under load?

30.6 seconds armed, throttle peaking at 1482 of 2000 (roughly 48% stick). Deliberately gentle — the goal was a clean, comparable baseline, not performance.

Findings — airframe passed

Motor balance

Mean eRPM divided by mean motor command, over samples above idle:

Motor	Mean command	Mean eRPM	Ratio
0	674	1185	1.757
1	636	1077	1.694
2	651	1114	1.710
3	638	1135	1.779

Spread is under 5%. This is the measurement that answers the rebuild question: a cold joint, a damaged phase or a partially desyncing ESC shows up as one motor needing materially more drive for the same RPM — an outlier, not a uniform shift. Four consistent ratios means the power path to every motor is intact.

Noise floor

FFT of each filtered gyro axis, in dB relative to peak, above 50 Hz:

Axis	Median noise floor	Discrete peaks
Roll	−41.9 dB	none above −15 dB
Pitch	−38.2 dB	none above −15 dB
Yaw	−59.5 dB	none above −15 dB

No discrete peaks and no raised broadband floor. Broadband elevation is the signature of electrical interference or a damaged gyro, which was the specific worry after the rework — it is absent. This log is in fact cleaner than the pre-rebuild flights.

Saturation

No motor reached its ceiling at any point, and gyro standard deviation stayed modest (13.9 / 18.7 / 26.2 °/s across roll, pitch, yaw). Nothing suggests a tuning problem.

Open issues

Neither of these relates to the rebuild.

Battery discharged to 2.48 V. On a 1S pack that is deep over-discharge — normal cutoff is around 3.0–3.2 V under load. Repeated cycles this low will puff or kill the cell. Earlier flights showed 2.85–2.92 V, so this is getting worse, not better. Needs a voltage alarm and shorter flights.

Current sensor reported 1320 A on an earlier flight. Physically impossible on a 1S whoop. The signature is diagnostic: an instantaneous step from 1.0 A to 1320.2 A, held for exactly 41 samples (≈0.43 s), then gone. Real current draw ramps as motors spool; an instant discontinuity to an implausible constant is an ADC or scaling artifact. Cause is almost certainly miscalibrated ibata_scale / ibata_offset — not a short or any real electrical event. Flight 006 itself showed zero such samples.

Method

Decoded from btfl_006.bbl with blackbox_decode.

blackbox_sample_rate was 1/4, so the decoder reports ~75% of iterations missing — that is expected decimation, not corruption. Effective log rate was 1003 Hz, which puts a ~500 Hz Nyquist ceiling on the frequency analysis. That is why the noise-floor claim above is scoped to “above 50 Hz” and stops well short of the motor-noise band a full-rate log would reach.

Raw log kept locally, not committed — see the log index for the rationale.