Flight Controller PCB: Sensors, Vibration and Power Filtering

A flight controller PCB is the board at the centre of a drone or an unmanned aircraft, and it has an unusual set of requirements. It has to read a set of inertial sensors accurately enough to estimate the vehicle attitude, run a control loop fast enough to stabilise it, and do both while the board is being shaken by the motors it is controlling. Layout errors that would be invisible on a static product are immediately visible here as drift or oscillation.

This article covers what goes on a flight controller board, and the layout decisions that determine whether the attitude estimate is usable.

What Is on a Flight Controller Board

The core is a microcontroller with enough processing power to run the attitude estimation and the control loops at a high rate. Around it sit the inertial measurement unit, a barometer, sometimes a magnetometer, the radio receiver interface, and the outputs for the electronic speed controllers and the servos.

The board also carries the power architecture: a regulator from the battery to the logic rails, a current and voltage sense circuit for the battery, and sometimes a power distribution function. That last part matters because the same board may be switching tens of amperes while trying to measure microvolt-level sensor signals.

The Inertial Sensors

The inertial measurement unit may be a single package containing accelerometers and gyroscopes, or several separate devices mounted at an angle to each other for redundancy. Whichever arrangement is used, the sensors measure the board rather than the vehicle, so the mounting and the layout are part of the measurement chain.

The sensor has to be placed so that its axes align with the vehicle axes, and the package has to be soldered without introducing stress. A sensor that is mechanically loaded by a warped board or by a poor solder joint will show an offset that changes with temperature, which the calibration cannot remove. The supply and the reference for the sensor are equally critical, and the mixed-signal layout rules describe the separation of the analogue supply and the return path that keeps the measurement clean.

Vibration and Mechanical Isolation

Vibration is the dominant error source on a multirotor, because the motors and the propellers generate a periodic disturbance at a frequency that the accelerometers cannot distinguish from acceleration. The mechanical mounting of the board and the damping of the sensor are therefore part of the design rather than an accessory.

The board layout can help by keeping the sensors away from the mounting points where the vibration amplitude is greatest, and by avoiding a geometry that has a mechanical resonance in the range the motors produce. Where the sensors are mounted on a soft mount, the layout has to leave room for the damping material and for the flex that carries the signals. Good vibration isolation depends as much on the mounting geometry as on the material, because a mount that is compressed unevenly tilts the sensor and introduces an offset that varies with the load.

The board itself is part of the mechanical system. A long, thin board fixed at two corners has a bending mode that the motor noise can excite, and the resulting deflection appears in the accelerometer output as a low-frequency signal that the controller cannot separate from real motion. Shorter unsupported spans, mounting points placed near the sensors and a stiffener where the board is large all raise the resonant frequency above the range the motors produce. The general quality characteristics of a well-designed board apply here, with the mechanical behaviour treated as a first-class requirement rather than a consequence of the outline.

Magnetic Sensing and Interference

A magnetometer measures the earth’s field, which is weaker than the field produced by the current flowing through the board. Any trace carrying a significant current near the sensor will corrupt the reading, and the error is proportional to the current, which means it changes with throttle.

The remedy is distance and geometry. High-current paths should be routed away from the magnetometer, and where that is impossible they should be arranged as a twisted pair or as a path whose fields cancel at the sensor location. The trace width and current calculation gives the geometry for the power paths, and their placement relative to the sensors is a layout constraint that has to be planned rather than adjusted later.

Power Filtering and Noise

The sensors and the microcontroller share a board with switching regulators and with the digital noise of the control loop itself. A clean supply is therefore a design goal rather than a given, and it is achieved with a combination of filtering, separation and a well-planned ground.

Each sensor should have its own local decoupling, placed as close to the pin as the layout allows, and the analogue supply should be filtered from the digital rail rather than taken from it directly. The ground should be a single continuous plane under the sensors, with the noisy return currents routed around it rather than through it.

Thermal Drift

Gyroscope and accelerometer offsets move with temperature, and the rate of change is large enough to matter during a flight. The board thermal design therefore affects flight performance: a board where the sensor sits in a thermal gradient will see the gradient move as the electronics warm up, and the calibration performed on the bench will no longer be valid.

Keeping the inertial sensor away from the regulators and from the power switching devices reduces the gradient, and giving the sensor area a uniform copper environment reduces the local variation. Where the design has room, a small copper island connected to the sensor’s own ground, isolated from the heat-producing regions of the board, is the usual approach.

Interfaces and Connectors

A flight controller connects to a large number of external devices: the receiver, the speed controllers, the GPS, the telemetry link and the payload. Each connection is a potential source of noise, and each connector is a mechanical feature that has to withstand vibration and handling.

Connectors should be placed so that cables can be routed without crossing the sensor area, and the signals that carry high current or fast edges should be grouped together and kept away from the analogue section. Where a connector carries power and signal in the same housing, the pin assignment should separate them, and the return for the sensor signals should not share a pin with the power return.

FAQ

Why does my quadcopter oscillate at high throttle? The most common cause is vibration reaching the accelerometers, which the controller interprets as motion. Damping the board, balancing the propellers and keeping the sensors away from the mounting points all reduce the effect.

Should the sensors be on a separate board? Sometimes. A small sensor board on a soft mount isolates the sensors mechanically and thermally from the power stages, at the cost of a connector and a flex in the signal path. Where the vibration environment is severe, that trade is usually worth taking.

Does the board layout affect the compass calibration? It does, because the calibration captures the magnetic field of the board as an offset. A layout change that moves a current-carrying trace closer to the magnetometer invalidates the previous calibration, and the vehicle has to be recalibrated.