docs: no real-flight dataset — all telemetry is SITL #11

Observation

Every test and demo in the repository runs against simulated sensor data:

• flight-software/run_cansat.py — SITL flight simulator (IMU + baro synthesis).
• flight-software/tests/test_mission_e2e.py — state-machine-only E2E.
• scripts/demo.py — TCP loopback AX.25 demo.

There is no real-RF captured telemetry, no real-GNSS track, no real-barometer ascent/descent curve anywhere in the repo. The AX.25 decoder has been fuzzed (hypothesis, 500 cases) but never exercised against real radio noise.

Why it matters

• Reviewers and competition judges ask for flight data.
• The HMAC replay-window + FDIR grayscale behaviour are only proven against clean simulated inputs. Real RF introduces bit errors that may reveal edge cases the simulator misses.
• A single real flight CSV would upgrade docs/testing/hil_test_plan.md from "plan" to "plan + heritage".

Definition of done

Any one of the following closes this:

• A CanSat flight CSV (≥ 1 minute continuous) stored under docs/characterization/flight/cansat_<date>_<site>.csv plus a companion *.md describing the flight (altitude, weather, launch vehicle, ground-station setup).
• A HAB flight PCAP + GNSS track under the same path.
• A sounding-rocket IMU / baro log.

Each should include:

• raw sensor output (not processed)
• ground station's decoded AX.25 frames (PCAP or JSON dump)
• flight-controller log (stdout)
• photos of the vehicle and the ground setup

Stretch

• A Jupyter notebook under notebooks/ that replays the dataset through the current ground-station/ decoder and produces the same plots as the live UI would — proving the repo can post-process real data.

Depends on a team actually flying a vehicle. Coordinates with #10.