What you are looking at
Both flights of the nosecone telemetry unit on 15 August 2026, rebuilt in 3D from the raw 100 Hz IMU and 50 Hz barometer logs. Same rocket, same motor, launched about an hour apart. Flight 1's ejection charge fired 0.1 s after apogee but the chute jammed in the tube — ballistic return at ~17 m/s. Flight 2, after a repack, deployed cleanly and came down at 5.1 m/s.
Attitude comes from integrating the gyro (bias re-estimated from 60 s of still pad data), position from double-integrating body acceleration rotated into the world frame, with altitude anchored to the barometer. The nose axis is taken from the burn-averaged thrust direction, which absorbs board-mount skew.
Why the trajectories are believable
The IMU-only altitude — integrated blind, never shown the barometer — is the validation channel:
| Flight 1 | Flight 2 | |
|---|---|---|
| IMU-only altitude at apogee | 85.9 m | 80.1 m |
| Barometer apogee | 84.6 m | 76.9 m |
| Worst ascent disagreement | 3.3 m | 3.9 m |
| Max speed, world-frame integration | 37.8 m/s | 37.4 m/s |
| Max speed, independent 1-axis method | 37.9 m/s | 38.5 m/s |
What the reconstruction settled
Flight 2 reached 8 m less apogee on an identical boost (6.1 vs 6.0 g). The orientation data answers why: its nose was 27.6° off vertical at burnout versus flight 1's 4.6° — it weathercocked into wind and spent the same energy arcing ~30 m downrange, crossing apogee nose-nearly-horizontal. Trajectory, not performance: drag-free energy heights agree to 0.4 m.
Validity limits
- Flight 1's solution is fully validated launch → ejection. The tumble that follows corrupts the attitude, so on the ballistic descent altitude is barometer-true and the horizontal path is qualitative.
- Flight 2's solution is fully validated launch → deploy. The 15.5 g opening jerk sat essentially at the accelerometer's ±16 g clip; under canopy only the barometer is trusted, so the descent is drawn as a vertical drop — real horizontal drift was not observable.
- Both logs go blind for ~3 s starting at exactly t+6.00 (the firmware's ring-buffer flush blocks sampling), and the barometer reads the ejection-charge bay transient, not altitude, for ~0.5 s after the charge fires. The descents are interpolated straight across both windows.
- No magnetometer or GPS: ground tracks are shape-true but the compass heading of each flight is arbitrary, including between the two flights.