Crafty — Blackbox Review, Rougher After a Crash

The pilot reported Crafty “didn’t sound very well” and some props looking worn. The flash held 39 logs and was completely full, so flights after btfl_039 were not recorded.

Parameter	Value
Logging	~1 kHz, bidirectional DShot
Tune	`btfl_001`–`002` before, `btfl_003`–`039` after the snappier Horizon change
Hover	~40 % stick, ~21 000 RPM, 7–8 A

Vibration jumps after the `btfl_029` crash

btfl_029 ends with the craft on its side (acc Z 0.47 g). The next two logs are failed takeoffs where the mixer drops motor 2 to idle and drives motor 3 to full, the signature of a craft tipping or snagged. From btfl_032 the airframe is noisier. PIDs, rates and filters are identical, so the change is physical.

Unfiltered gyro, roll and pitch averaged, throttle above 15 %:

Log	20–100 Hz	100–250 Hz	250–500 Hz
`btfl_020`	0.64	0.44	2.48
`btfl_028`	0.55	0.28	2.26
`btfl_029`	0.65	0.44	2.61
`btfl_032`	2.27	1.25	3.00
`btfl_034`	2.14	1.25	3.11
`btfl_036`	2.32	1.20	2.64
`btfl_039`	2.38	1.33	3.15

Filtered noise above 100 Hz rose from ~0.10 to ~0.33–0.37. The filters still work; there is more to filter. Low-throttle and bench logs are excluded because idle RPM moves the bands.

Motors 2 and 3 run rougher, but are not weaker

RPM jitter is the standard deviation of sample-to-sample eRPM change as a percentage of mean RPM, in steady hover:

Log	M1	M2	M3	M4
`btfl_019`	0.92	1.07	1.36	1.11
`btfl_028`	0.90	0.97	1.17	1.02
`btfl_029`	1.12	1.07	1.26	1.36
`btfl_032`	1.09	1.72	1.91	1.46
`btfl_036`	1.13	1.65	1.63	1.22
`btfl_039`	1.03	1.80	1.71	1.18

In Betaflight’s default Quad-X numbering motor 2 is front right and motor 3 rear left. RPM per unit of command, normalised to the mean, did not change across the crash (M3 ~0.94–0.97, M1 ~1.05–1.09), so nothing is dragging or weak. Rough running with normal output points at the rotating parts.

The later hover review revised this: motors 2 and 3 settled, and the lasting vibration is on motor 4.

Not a finding: motor 2 “at 8 400 RPM” in `btfl_035`

The set script reports an 82 s flight with a −39 % front/rear split. In fact the craft sat armed on the ground at zero throttle for ~72 s, airmode spinning the props and I-term winding up (pitch 129, roll 91), then took a throttle punch with that windup loaded. The averages describe ground idle, not a motor fault.

Avoid long armed idles on the ground.

Other observations

Battery sag: resting packs read 3.8–4.2 V, but several flights dip to 3.10–3.17 V under load.
Current sensor: the 1320 A full-scale glitch recurs in btfl_019 and btfl_035.
Centre of gravity: the front motors were 12.8 % faster on 12 Sep; clean hovers now sit between −2.7 % and +7.5 %.
Flash full: the flight that sounded worst may not be in the set.

Actions

Replace the props on motors 2 and 3, and any visibly worn ones.
With props off, spin motors 2 and 3 by hand: a wobbling bell means a bent shaft, a gritty feel a worn bearing.
Check the frame near those motors for cracks, and the FC mount and canopy for looseness.
Fly one pack in steady hover; 0.3–0.5 in the 100–250 Hz band means fixed.