Drone Horizon includes live flight telemetry and a buffered CSV blackbox recorder. Their purpose is to make the relationship between input and aircraft response inspectable. They are most useful when you begin with one question, such as why a turn overshoots or why a climb weakens.
The blackbox is flight data, not a video replay editor. The presence of a record action should not be interpreted as automatic cinematic video export.
Read the input before the response
The telemetry view separates raw and processed flight channels. Raw values help identify device behavior; processed values show what remains after mapping and response handling. If a centered physical stick produces a nonzero processed command, investigate calibration or the active profile before analyzing motor dynamics.
Next compare requested and actual body rates. A rapidly alternating requested rate can indicate alternating pilot input. A calm request with a different actual response asks a control-system question instead. These are different diagnoses even if both look like wobble in the camera.
Know the units
| Signal group | What it helps explain |
|---|---|
| Raw/processed axes | Mapping, neutral drift and shaping |
| Requested/actual rates | Command versus rotation response |
| Motor request, output, RPM and thrust | Available actuation and spool behavior |
| Velocity and acceleration | The path beyond the camera heading |
| Voltage, current and charge | Battery state during the maneuver |
| Mixer saturation/scales | Limits while combining collective and torque |
The live overlay labels its displayed units. CSV headers also encode units for many fields; do not assume every stored angular value is in degrees simply because the HUD displays degrees per second. Some recorded rate and attitude fields use radians.
Investigate one maneuver
Save a known profile and use a short, repeatable route. Start recording, perform the maneuver and stop the recording after the useful segment. Compare input, attitude, velocity and motor output around the event instead of scanning the entire file for the largest number.
For an overshoot, ask when the correction began, whether the requested rate changed as intended and when the trajectory actually turned. For a difficult dive recovery, compare when the aircraft became upright with when downward velocity decreased.
What the implementation records
The source includes flight-state samples, rate-controller terms, torque, aircraft motion, battery data, wind/wake values and per-motor signals. It writes CSV rows through a buffer and stores recording metadata. The settings expose a blackbox sample-rate range separate from the physics precision setting.
A lower recording rate is not necessarily a lower physics rate. It changes how frequently data is sampled for the record. Likewise, a high render frame rate is not the same quantity as either of those update frequencies.
Interpret assisted flight separately
In the DJI-style stabilized family, telemetry can also show target velocity, hold state, braking and gimbal information. Centered translation sticks can trigger a hold/braking task that does not exist in manual Acro. Identify the active family before comparing two recordings.
The simulated radio-strength HUD is a gameplay estimate, not a measurement of a physical Bluetooth or RF link. Do not use it to certify controller reception.
Keep conclusions proportional to the evidence
A log can show that a particular input produced a particular simulated response. It does not prove the model matches a real aircraft, and simulator PID units should not be copied blindly into flight-controller firmware.
Use physics fundamentals, rate settings and overshooting diagnosis to formulate a narrow question. A small, understandable comparison is more valuable than a long recording with no identified event.