Last time, we explained the physics of return current and the minimization of loop area. In this article, which is FC26, we will cover EMI sources. We explained the identification procedure for sources in the EMI article from the series before last (the early part of the overall FC design series), but this article will explain the circuit-level mechanisms of each source in greater depth within the context of the Hardware Design series.
Two Types of EMI—Revisiting Radiated and Conducted
EMI is classified into Radiated EMI and Conducted EMI based on the noise propagation path (we explained this in the series before last, but we will reorganize it here from the perspective of circuit design).
Radiated EMI is noise that propagates through space as electromagnetic waves. The area where loop current fluctuates acts as an antenna, becoming a problem in the frequency band above 30 MHz. As explained last time, the larger the loop area of the return current, the greater this radiated EMI becomes. The power of the radiated electromagnetic waves is proportional to the square of the loop area and also proportional to the square of the frequency.
Conducted EMI is noise that propagates through power lines or signal lines. The 150 kHz to 30 MHz band is the typical problem range, which is controlled by power supply impedance and filtering. The decoupling capacitors from the previous article and the filter design from the series before last are the primary countermeasures.
The two are not independent problems but are interrelated. The pattern where conducted noise is radiated via antennas (wiring or cables) is the most common in actual FC design.
Source 1—Switching Regulator (DC-DC)
The DC-DC converter explained in FC22 is the largest source of EMI on an FC board.
The switching node is a node that transitions at high speed between 0V and Vin due to the ON/OFF state of the MOSFET. This dV/dt (voltage change rate) is the direct source of radiated EMI.
V = L_parasitic × dI/dt
The steeper the voltage transition at the switching node (the larger the dV/dt), the more strongly the wiring connected to the node acts as an antenna.
Reconfirming the physics of the switching loop explained last time, the area of the DC-DC switching loop (input capacitor → high-side MOSFET → inductor → output capacitor → GND → input capacitor) determines the strength of the radiated EMI. Minimizing this loop area is the most important design point for DC-DC peripheral layout.
Conducted EMI spreads to other circuits through power lines. The input-side current of the DC-DC changes in a rectangular wave shape, and this high-frequency current component propagates through the VBAT line to other parts of the FC and the entire airframe. An input filter (π-filter, see the series before last) suppresses this.
DC-DC Switching Noise Spectrum
By understanding the spectral characteristics of the noise generated by the DC-DC, you can clarify the goals for filter design.
There is a strong spectral component at the fundamental switching frequency f_sw (several hundred kHz to several MHz), and its harmonics (2f_sw, 3f_sw...) continue up to several tens of MHz.
The spectral envelope is determined by the steepness of the switching edge. At frequencies higher than the reciprocal of the edge transition time 1/t_rise, the spectrum attenuates rapidly (20 dB/decade). On the other hand, in the band from f_sw to 1/t_rise, the spectrum remains relatively flat.
Increasing the gate resistance to slow down the switching speed (lengthening t_rise) reduces high-frequency components and improves EMI, but there is a trade-off where switching loss increases (the same mechanism as the switching loss in ESC12).
In FC power DC-DCs, a switching frequency of 500 kHz to 2 MHz is a common choice due to the balance between emissions (EMI radiation) and efficiency. To avoid frequency interference with GNSS (L1 band 1.575 GHz) or Wi-Fi (2.4 GHz/5.8 GHz), it is also necessary to consider selecting a switching frequency so that the fundamental frequency of the DC-DC and its harmonics do not fall into these GNSS/wireless communication bands.
Source 2—MCU Clock and High-Speed Digital Signals
High-performance MCUs like the STM32H7 have operating clocks that are themselves sources of EMI.
The 480MHz system clock and its harmonics can reach the GHz band. The steeper the rise and fall edges of the clock signal (hundreds of ps to a few ns), the stronger the higher-order harmonics become.
Noise radiated from the MCU package's power pins is mitigated by the decoupling capacitor design discussed previously, but if clock signals are routed, the traces themselves act as antennas.
Spread Spectrum Clocking (SSC) is an EMI reduction technique used in USB and HDMI standards. It intentionally modulates the clock frequency by about ±0.5 to 2.5% to spread the frequency, preventing power concentration at specific frequencies. While the STM32H7's RCC register does not have settings equivalent to SSC, designing with an external crystal oscillator that supports SSC is an option for industrial applications.
High-speed peripheral signals (SPI, SDMMC, USB HS, FDCAN) have steeper edges as the signals become faster, generating higher-order harmonics just like clock signals. If the return current paths for these signals are not secured (a violation of the principles in FC25), the problem is compounded.
Source 3—Shared Impedance of Power Lines
This section explains the mechanism of conducted EMI that occurs when multiple ICs share the same power line.
When an IC performs switching operations, it momentarily draws a large current from the power line. This current change causes voltage fluctuations through the impedance of the power line (wiring inductance + output impedance of the power regulator).
These voltage fluctuations also appear on the power pins of other ICs sharing the same power line. The phenomenon where noise generated by one IC propagates to the power terminals of other ICs through the same power line is called Common Impedance Coupling.
A typical example in FC design is the phenomenon where current fluctuations during high-speed processing by the MCU cause the power supply voltage of an IMU using the same 3.3V power line to fluctuate. This is the basis for the design principle explained in FC22: 'Separate the LDO for the MCU and the LDO for the sensors.' By using independent LDOs, the common impedance path through which current fluctuations on the MCU side propagate to the sensor side's power voltage is broken.
Decoupling capacitors mitigate common impedance coupling through local energy storage, but they may not be a fundamental solution. A combination with power system separation (independent LDOs) is necessary.
Source 4—Cables and External Wiring
Cables connected to the FC (telemetry, GPS, ESC signal lines) are long conductors and act as effective antennas.
The efficiency of a monopole antenna increases as the cable length approaches 1/4 of the electromagnetic wave's wavelength. As a reference for the relationship between frequency and wavelength in the EMI series, the 1/4 wavelength at 100MHz is about 75cm, and at 10MHz it is 750cm. Drone GPS cables (around several tens of cm) can function as effective antennas in the range of several tens to several hundreds of MHz.
The 'conduction to radiation' path, where noise from the FC leaks to the outside through cables, is one of the major causes of EMI problems in the entire FC system. Common-mode chokes (see the previous series) are the standard countermeasure to suppress this cable radiation.
Conversely, the path through which external noise (motor drive noise from ESCs, radiation from radio transmitters) enters the FC's power and signal systems via cables is also important. The EMI countermeasures from the previous series (TVS diodes near connectors, common-mode chokes) are measures against this entry path.
Source 5—Non-linearity of Contact Resistance and Solder Joints
A frequently overlooked source is the non-linearity of contact resistance in connectors.
The non-linearity of contact resistance causes Intermodulation Distortion (IMD) of noise signals, and there are cases where intermodulation products falling near the GNSS L1 band (1575MHz) degrade reception sensitivity.
In degraded connectors or oxidized solder joints, contact resistance increases and non-linearity becomes more pronounced. In particular, poor contact in connectors carrying large currents (battery connectors, motor current connectors) becomes a strong source of noise. Regular contact checks and the use of high-quality connectors are the countermeasures for this problem.
Solder quality also affects EMI. Cold joints (incomplete soldering) have high impedance and exhibit large fluctuations at high frequencies. In particular, the solder quality of antenna connections, which directly affects GNSS signals, must be strictly managed.
Generation sources specific to FC design—Intrusion from ESCs and motors
In drones where the FC and ESC are mounted on the same airframe, powerful noise from the ESC side affects the FC. While this was explained in the previous series, we will add details at the circuit level in the context of the Hardware Design series.
The ESC's three-phase inverter (detailed in ESC9 and later) switches MOSFETs at tens to hundreds of kHz, injecting strong high-frequency current into the battery line. In configurations where the FC and ESC draw power from the same battery, this noise propagates to the FC through the battery cables.
The FC input-side pi-filter, explained in the EMI countermeasures articles at the beginning of the FC series, is a countermeasure against this intrusion path, but it is necessary to accurately design the relationship between the filter's cutoff frequency and the motor drive frequency.
Common battery impedance becomes a noise propagation path. Because battery internal resistance and wiring inductance are finite, large-current switching on the ESC side causes fluctuations in the battery terminal voltage, which in turn causes the input voltage of the FC connected to the same battery to fluctuate. These fluctuations must be absorbed by decoupling and filtering at the FC input.
Differential serial communication like DroneCAN (see previous series) is adopted as a countermeasure against this noise environment, and when using single-ended DShot-type signals for signal communication between the FC and ESC, resistance to conducted noise on the signal lines becomes even more important.
Relative comparison of source strength
We will qualitatively organize the magnitude of the impact of each EMI source in FC design.
The largest sources are the DC-DC converter switching nodes (voltage transitions of tens of V/ns, large-area switching loops) and the injection of ESC motor drive noise into the battery line (large-current switching).
Medium-level sources include MCU clocks and high-speed peripheral signals (mainly radiated EMI), and conducted-to-radiated noise propagation via cables.
Relatively small but non-negligible sources include connector contact resistance nonlinearity, degradation of solder quality, and common impedance coupling.
Understanding this relative importance serves as a decision-making criterion for concentrating design resources (space, cost, design time) on the most effective EMI countermeasures.
Summary
EMI sources in FC design are organized into five categories: DC-DC switching nodes (the main source of radiated and conducted EMI), MCU clocks and high-speed digital signals (radiated EMI), common impedance coupling (conducted EMI), cable conduction-to-radiation conversion, and intrusion of ESC motor drive noise. By understanding the circuit-level mechanism of each source, you can prioritize effective countermeasures. In the next installment, FC27, we will systematically explain EMI countermeasures—countermeasure methods by source and implementation on PCB layouts.