Skip to content

Crafty — Hover Pack, Crash Pack and the Odd Sound

Two packs flown after the morning review. Pack 1 was a deliberate steady hover; pack 2 was flown actively, crashed twice and needed a turtle flip. The pilot also recorded flight audio because of an odd sound. The pilot calls this airframe the AIR75; it reports craft_name = Crafty and Crafty’s DeviceUID, so it is the same drone.

The flash was full again: btfl_005 stops at 52 s with no clean end.

LogPackArmedWhat happened
btfl_0011203 sSteady hover, one pack
btfl_00225 sShort hop, bump at 4.5 s
btfl_003217 sTurning flight, ends in a crash
21 s gap2Same power cycle, nothing recorded; most likely the turtle flip
btfl_0042156 sActive flight, ends in a crash (7.5 g)
btfl_005252 s+Rearmed 2.4 s later; logging stopped when the flash filled

All five logs share one tune, flown in Horizon. Pack 1 bottomed at 3.52 V, pack 2 at 3.20 V.

The sound is a once-per-revolution vibration, loudest on motor 4

The recording lines up with the logs to within ~0.1 s, so each motor’s measured RPM can be matched to the audio. Crafty flies 3-blade props, so a healthy whoop sounds mostly like blade pass (~1 050–1 100 Hz) plus electrical whine at 6× rotation (~2 100 Hz). This recording is dominated by 1× rotation, ~335–370 Hz, the tone of an out-of-balance spinning part.

Audio tone prominence, dB above the surrounding noise, at each motor’s own harmonics:

HarmonicM1M2M3M4
1× (imbalance), btfl_00414.412.812.817.3
1× (imbalance), btfl_00518.814.513.619.5
3× (blade pass), btfl_0044.16.13.64.5

The gyro agrees, and has history. Unfiltered roll and pitch gyro demodulated at each motor’s exact rotation phase in steady flight (background ~1–4):

LogM1M2M3M4
btfl_009 (12 Sep)8.64.18.112.6
btfl_028 (before the crash)6.93.98.016.5
btfl_036 (after the crash)6.44.95.527.3
btfl_0396.83.19.941.7
btfl_001 (hover)3.56.43.834.2
btfl_0047.82.87.027.7
btfl_00514.73.94.823.3

In the hover log the strongest gyro peak between 250 and 450 Hz sits at 369–373 Hz, exactly motor 4’s rotation speed. Motor 4 is front left. Its imbalance roughly doubled at the morning crash and has not recovered, while its RPM per command is normal (1.005), so it is not dragging. Motors 2 and 3 have settled.

The rattle

The pilot describes a rattle, not a hum. In the 2.5–9 kHz band there is no steady clicking (about 30 separate clicks in 214 s, mostly handling), and click energy at each motor’s rotation rate sits only ~2 dB above background for all four. The roughness follows loudness (r = +0.60) and only faintly follows motor 4’s vibration (r = +0.14 with loudness removed). The recording is HE-AAC at 48 kbit/s, which smears clicks, so audio alone cannot locate it. The best fit is a loose part shaken by motor 4’s imbalance, such as a prop worn loose on its shaft.

The front-left motor is visibly tilted

On inspection motor 4 leans forward a little, which fits its imbalance doubling at the btfl_029 crash. A few degrees of lean tips only about 2 % of the craft’s weight sideways, and steady-hover yaw I-term and diagonal mix wander between −1.7 % and +1.6 % with no step at the crash. The tilt therefore matters through vibration and a possibly loose mount, not through a steady push.

Actions. With the battery out, push the front-left motor side to side: movement or a click means a cracked mount or loose screws. Replace the frame if the plastic is bent, since it seldom stays straight. Fit a new prop, spin the bell with the prop off to check the shaft, and wiggle every prop for play. Then fly one hover: a fix shows motor 4 below ~15 in the table above.

In the hover, forward stick was held all the time

The FC’s own roll and pitch angle varied only ~1° in every 40 s block, yet holding that took continuous stick:

TimePitch stickRoll stickFC pitch angleFC roll angle
4–12 s+35 (7 %)+32+1.6°+1.5°
52–60 s+87 (17 %)+36+4.6°+1.7°
140–148 s+104 (21 %)+66+5.7°+3.2°
196–204 s+118 (24 %)+48+6.9°+2.3°

On the ground the FC read level (pitch +0.1°), but by the end of the pack it reported +5 to +7° nose-down while a quarter of forward stick was held. Most likely the FC’s idea of level drifts nose-up in flight and was being corrected without noticing.

This is the third attitude-reference issue in two days, after the bad Crafty calibration and the Air65’s 6.5° offset. It differs by starting at zero and growing in the air. Vibration can bias an accelerometer and motor 4’s imbalance is the obvious candidate, so fix that first. Then hover a pack with light hands and read the attitude in the Configurator right after landing.

What the active pack says about the flying

  • Very gentle attitude: under 11° of tilt 90 % of the time, under 15° 99 %. Throttle stayed between 1367 and 1465.
  • Flat turns: yaw stick was in use 60–85 % of the time, but real bank in turns was 3.4° against 3.1° flying straight. The nose rotates while the craft keeps sliding.
  • Both crashes were turns into something, not loss of control. In btfl_004 the craft was flying forward while yawing at 60–100 °/s, only 6° nose-down 50 ms before a 7.5 g impact.
  • Throttle does not follow tilt (correlation about zero in pack 2, +0.5 in the hover pack).

To practise: bank into turns with roll rather than steering on yaw alone, and leave more room before a turn.

Other observations

  • Centre of gravity: front motors 7 % faster than rear, pitch I-term −17. Slightly nose-heavy.
  • Current sensor: the 1320 A glitch recurs in btfl_001, 002 and 004.
  • Flash: 16 MB holds about seven minutes at this logging rate. Erase after every two packs.