EMC noise sources in a drone
Summary¶
Every wire carrying a changing current radiates an electromagnetic field. A drone concentrates several powerful noise sources — ESC switching, motor winding transients, battery lead current spikes, and switching-regulator noise — in a small volume alongside sensitive sensors that must be protected from that noise. Understanding the source, frequency, and coupling mechanism of each noise type is the prerequisite to choosing the right mitigation. There is no single fix: each source requires a targeted approach, and the approaches must be layered.
Concept¶
Maxwell's equations as the root cause¶
A changing current creates a changing magnetic field (Ampère's law). A changing magnetic field induces a voltage in any nearby conductor (Faraday's law). In a drone, this means every time motor current changes — on every throttle input, on every ESC MOSFET switching event, on every RPM change — a changing field propagates outward from the power wires and couples onto nearby signal wires, sensor supply traces, and PCB ground planes.
This is not a peripheral concern. The gyroscope is measuring changes in the order of 0.01°/s. The GPS is receiving signals at −130 dBm. The I2C bus operates at logic levels that a few millivolts of noise can corrupt. The noise sources in a drone operate at voltages and currents that dwarf these signal levels by many orders of magnitude.
ESC switching noise (48 kHz and harmonics)¶
The ESC switches its six MOSFETs at 48,000 times per second. Each switching event: - Creates a fast current edge in the motor phase wires (~50 A/µs) - Creates a voltage edge on the power bus (~50 V/µs with inadequate decoupling) - Radiates electromagnetic energy at 48 kHz and its harmonics (96, 144, 192 kHz...) extending into the MHz range
The 48 kHz fundamental is well above the gyro's useful signal band (0–100 Hz) but couples directly into the gyro supply voltage, appearing as shifted zero-rate output and elevated noise floor.
Motor winding transients¶
Motor windings are inductors. When the ESC switches the active phase, the collapsing current in the previous phase induces a back-EMF spike — the inductor resists the sudden current change. These spikes can reach several hundred volts per microsecond in the absence of clamping. The 1000 µF capacitor and TVS diode on the ESC power pads clamp these spikes on the power bus; what escapes as radiated field is attenuated by routing and separation.
Battery lead current spikes¶
Every throttle change requires a rapid change in motor current, which requires a rapid change in battery lead current. A 40 A step in 1 µs through a 20 cm battery lead radiates a significant electromagnetic field. The battery lead is an effective antenna at the frequencies corresponding to the current rise time. Twisting the battery leads together cancels most of the radiated field — two equal and opposite currents flowing in opposite directions cancel each other's fields at any external point.
Switching regulators (VTX 9 V BEC, 5 V BEC)¶
Both regulated rails on the Cinematic build come from switching regulators on the Matek H7A3-SLIM: the 5 V BEC and the switchable 9 V BEC feeding the VTX. Matek does not publish their switching frequencies. Their output noise has two parts: the switching fundamental and low harmonics (typically hundreds of kHz to a few MHz), handled by capacitance at the load and short wiring; and fast-edge harmonics in the tens-to-hundreds of MHz, which can radiate from the VTX power lead toward the GPS antenna and the receiver. Clip-on ferrites act only on that second part. → See ferrite-beads for the VTX-on/VTX-off test that decides whether they are needed.
USB/UART switching (minor)¶
Serial data transitions on UART lines are minor noise sources compared to the above, but at high baud rates (CRSF at 420,000 baud) the transition rate is sufficient to produce low-level radiated noise. Twisted-pair routing of differential signals and adequate separation from sensitive sensors mitigates this.
Reference¶
Noise source summary¶
| Source | Frequency | Amplitude | Primary coupling mechanism |
|---|---|---|---|
| ESC MOSFET switching | 48 kHz and harmonics | High (A-class) | Magnetic near-field, conducted via power bus |
| Motor winding transients | Broadband (DC–MHz) | Very high, brief | Magnetic near-field |
| Battery lead current steps | Broadband | High | Magnetic near-field |
| FC switching BECs (5 V, 9 V VTX) | Switching frequency (unpublished) and harmonics into the MHz range | Medium | Conducted along the VTX power lead, radiated from it |
| UART transitions (CRSF etc.) | Up to few MHz | Low | Electric near-field |
Sensitive victims and their susceptibility¶
| Victim | Susceptibility | Key frequency range |
|---|---|---|
| Gyroscope (ICM-42688-P) | Very high — supply noise shifts zero-rate output | 1 kHz–1 MHz |
| Magnetometer (QMC5883) | Very high — DC fields cause heading error | DC–1 kHz |
| GPS antenna (1.575 GHz) | High — wideband noise raises noise floor | 100 MHz–2 GHz |
| ELRS receiver (2.4 GHz) | Medium — spread spectrum provides processing gain | 2.4 GHz |
| I2C bus (400 kHz) | Medium — noise on SDA/SCL causes data errors | 400 kHz–1 MHz |
Procedure¶
Rationale¶
Why this article exists before the mitigation articles¶
A builder who only knows the mitigations — "twist the motor wires," "add ferrite beads" — follows instructions without understanding why. When something unexpected happens (elevated noise floor, compass drift, GPS multipath), they have no framework to diagnose it. Understanding the source, mechanism, and frequency of each noise type enables targeted diagnosis: "the compass is drifting intermittently — is it a DC field from a battery lead, or a high-frequency field from the ESC? Check the routing." This article provides that framework before the mitigation articles that follow.
Connections¶
requires: [] related: - vibration-isolation-theory - rpm-filter leads_to: - twisted-pairs - star-grounding - capacitor-placement-emc - [[power-signal-separation]] - ferrite-beads