A racing drone is a real-time physics experiment running at 8 kilohertz. Every millisecond, the flight controller reads angular velocity from a MEMS gyroscope, computes a correction, and updates all four motor speeds. Understanding what is actually happening in those 125 microseconds clarifies several important physical concepts.
PID Control as a Discrete Feedback System
The PID controller operates on the error between the commanded angular rate (from the pilot's stick input) and the measured rate (from the gyroscope). The three terms act as follows:
Proportional (P): Output is directly proportional to the current error. A higher P gain produces faster response but risks oscillation — the motor overcorrects, creating a diverging oscillation at the airframe's resonant frequency.
Integral (I): Output is proportional to the accumulated error over time. The I term corrects for steady-state offset — a drone with a slightly off-center battery will drift without I. Too high and the controller winds up, producing a slow, sustained oscillation.
Derivative (D): Output is proportional to the rate of change of the error. D damps the response — it predicts where the error is going and applies a countermeasure before the oscillation builds. In practice, D is limited by gyroscope noise, which is why filtering (particularly RPM-based notch filters) is critical to enabling high-D tuning.
Sensor Fusion: Gyroscope and Accelerometer
The gyroscope measures angular velocity and integrates to estimate attitude. The problem: gyroscopes drift. A small bias in the raw output, integrated over seconds, produces substantial heading error. The accelerometer measures the direction of the gravity vector, which gives absolute attitude reference — but accelerometers are noisy and respond to all linear accelerations, not just gravity.
A complementary filter merges both: high-frequency attitude changes come from the gyroscope (which is accurate over short time scales), while low-frequency drift correction comes from the accelerometer. The Kalman filter is a more optimal version of this, weighting each sensor by its noise covariance. In aggressive FPV flight, the accelerometer signal is corrupted by propeller vibration, which is why most Betaflight configurations run gyro-only mode with RPM filtering to suppress the vibration artifacts.
5.8GHz RF Propagation
The 5.8GHz band was chosen for analog FPV primarily because it was unoccupied spectrum available to low-power unlicensed transmitters. The physics imposes tradeoffs: higher frequency means shorter wavelength (about 52mm at 5.8GHz), which means smaller antennas, but also higher free-space path loss (inverse-square law with a frequency-squared penalty) and worse penetration through obstacles.
At 5.8GHz, concrete attenuates approximately 10–20 dB per meter. This is why FPV pilots experience video breakup in urban environments or behind tree lines. The 25mW–600mW power range represents the regulatory and practical limits — more power does not help proportionally once multipath and diffraction dominate.
Doppler shift on a fast-moving aircraft is measurable: at 200 km/h and 5.8GHz, the maximum Doppler shift (head-on approach) is approximately 1 kHz — negligible for analog video but relevant for digital FPV links, which must handle it in their physical layer design.
See the FPV & Drone Building collection for the full component lineup.