Every Noisy Component You Need to Know in a Switched-Mode Power Supply

In Part 1 of this series of articles¹, we provide an overview of EMI filter design for switch-mode power supplies (SMPS). In this part, we will examine specific aspects of switched power converters. The goal is to help readers understand:

1. Emission spectrum of an SMPS
2. Noise sources in a typical SMPS
3. Coupling mechanisms of noise in an SMPS
4. Grounding considerations in switched-mode power supplies and
5. Input and output capacitors

Some readers may notice that this discussion does not directly address filters. However, years of experience in the field have shown that a deep understanding of these topics is essential to truly solving EMI issues. While filters are required for most switched-mode converters, failing to address points 1–5 makes designing an effective filter strategy inefficient, if not impossible. In this article, we will focus on the converter itself, with special emphasis on grounding, input, and output capacitors, as these often determine whether a filter will be effective.

Emission Spectrum of a SMPS

Figure 1 shows a typical emission spectrum for two fixed-switching frequency converters. The primary source of SMPS noise emissions is the switching frequency and its harmonics. Due to the asymmetry of the switching waveform—determined by the PWM duty cycle—both odd and even harmonics are typically present.

If the fundamental switching frequency is stable and well-defined, the resulting emissions form a narrowband spectrum that can extend well beyond 30 MHz, particularly when transition times are fast. Measurement bandwidth settings also influence how these harmonics appear in EMC testing (e.g., 9 kHz RBW from 150 kHz to 30 MHz, and 120 kHz RBW from 30 MHz to 108 MHz). This means that at lower frequencies, individual harmonics are clearly distinguishable.

In contrast, in the tens of megahertz range, they become harder to resolve, especially when the fundamental switching frequency is below 100 kHz. Designs in which the frequency is not stable will normally show modulation due to input or output ripple which has the effect of broadening individual harmonic lines so that an emission “envelope” is measured. Peaks in the emission profile are typical and can be caused either by resonances in the coupling path or by ringing on the switching waveform.

For low-power converters operating at 200 kHz switching frequency or higher (such as the 2.2 MHz converter in this case), emissions in the higher frequency range still appear in the narrowband. For example, in Figure 1, the blue trace demonstrates narrowband noise extending into the FM range.

Noise Sources in a Typical SMPS

In our previous article, we discussed hot loop areas and the switched node, which are typically associated with near-field magnetic field loops and electric field antenna-like structures. In this article, we will examine all key noise components in a typical SMPS.

Types of Power Electronics Switches

Silicon-based MOSFETs are the most common choice for SMPS, used in voltage ranges from 3.3V to 800V. Before the introduction of SiC MOSFETs, silicon-based MOSFETs dominated power applications.

One of the most critical factors affecting EMI performance is rise and fall time. The shorter the rise and fall time, the worse the EMI performance, as shown in Figure 2. These transition times can be controlled via gate drive resistors connected to the MOSFET gate. By using a diode in series with the fall time control resistor (R_{G_off}), designers can individually control rise and fall times. Increasing the resistor value slows down switching, reducing high-frequency EMI (particularly common-mode noise) but at the cost of increased switching losses.

P-N diodes, including the MOSFET body diodes, are another significant noise source due to reverse recovery charge (often referred to as Qrr in the component’s datasheet). In the time domain, this manifests as an overshoot during switching, while in the frequency domain, it contributes to high-frequency noise. Figure 3 illustrates this in a motor drive circuit.

To mitigate reverse recovery issues, Schottky diodes or fast-recovery diodes are often placed in parallel with MOSFETs. Since Schottky diodes do not have a PN junction, they theoretically eliminate reverse recovery charge (switching is essentially “instantaneous” with only a slight capacitive loading, which is much less of a concern). However, as we will discuss later, Schottky diodes are not always a perfect solution.

IGBTs, on the other hand, switch more slowly than MOSFETs and can withstand higher power, making them more suitable for high-voltage, high-power applications.

GaN and SiC transistors are becoming increasingly popular but introduce significant EMC challenges (again, due to the fact that they can be switched on and off much faster). We previously covered these topics in another of our articles²), and readers are encouraged to revisit that discussion.

Parasitic Effects in Switching Devices

Parasitic Capacitance and Resonance with Nearby Magnetic Component

Parasitic capacitance in switching devices is often overlooked, even though it is clearly specified in device datasheets. Engineers typically focus on R_DS(on), rise/fall times, and voltage/current ratings, but ignoring parasitic capacitance can lead to unexpected EMI issues.

The primary concern with parasitic capacitance is that it resonates with inductance in the system. A switching device can self-resonate, but, more commonly, it resonates with inductance from PCB traces and tracks. A proper PCB layout can minimize inductance caused by traces, but another significant resonance source is the isolation transformer, used in nearly all isolated power supply designs. The leakage inductance of the isolation transformer can strongly resonate with the parasitic capacitance of the switching device due to its close physical proximity. This often results in resonance peaks in the emission spectrum. 

Figure 4 illustrates how placing a flux band around the transformer reduces leakage inductance, improving EMI performance.

If power switches are already chosen, design engineers have no control over the parasitic capacitance of the device itself. Instead, they must optimize the layout and minimize the inductance of PCB traces and magnetics to reduce noise.

Another issue with parasitic capacitance is its impact on EMI when mounting switching devices on heatsinks. The larger the parasitic capacitance, the greater the common-mode current coupled into the heatsink. Additionally, the larger the heatsink, the greater the common-mode noise. In most cases, the dominant factor may be the heatsink size rather than the switching device itself. This highlights the importance of proper grounding and EMI mitigation techniques when integrating heatsinks.

Parasitic Inductance and Self-Resonance of Devices

Now that we understand parasitic capacitance, we turn our attention to parasitic inductance in switching devices. Just like in all EMC-related topics, geometry plays a crucial role in inductance.

One common package option of a power electronics switch is the through-hole device (such as TO-247). The long leads of these devices introduce significant inductance, which can negatively impact EMI performance. Engineers often prefer these packages because they allow for mounting on the PCB edge with heatsinks attached. However, if the heatsink is not properly grounded, it can worsen EMI issues. Additionally, mounting these devices at the PCB edge may exacerbate EMI problems due to unexpected return current paths. We have covered a detailed case study on this in another article³.

Another example of parasitic inductance affecting EMI is with Schottky diodes. In one case, an external Schottky diode introduced radiated emissions due to a high-frequency resonance between the diode inductance and the combined capacitances of both the Schottky and MOSFET. Since both MOSFET and diode capacitances vary with voltage, analyzing and mitigating these interactions can be challenging. 

Figure 5 illustrates this scenario, showing the frequency-domain measurements of the resonance phenomenon. A near-field loop probe placed close to the Schottky diode revealed the resonance issue described in time domain (Figure 6).

Coupling Mechanisms in a SMPS

Using an isolated SMPS as an example, noise can couple into a system through multiple paths:

1. Conducted coupling: Noise propagates via conductive paths through power and signal lines. While input and output filters provide some suppression, they often cannot block noise completely, allowing noise to directly couple into other connected systems. In EMC testing, a LISN is used to measure conducted emissions.
2. Near-field coupling: This occurs through magnetic and electric field coupling. One common issue is noise coupling onto input and output cable leads, effectively bypassing the filters, and reducing their effectiveness. This is a frequent reason why filters fail to work efficiently.
3. Radiated coupling: The power stage of the SMPS can radiate noise directly, which can be detected by far-field antennas in EMC testing.
4. Common impedance coupling: This occurs when an SMPS shares the same ground connection with another circuit. A star grounding scheme can sometimes introduce common impedance coupling.

For the reasons previously mentioned, it is advantageous to place the high-frequency components of an EMI filter near the power connector rather than on a PCB. In military applications, such filters were traditionally implemented in an “EMI doghouse,”⁴ but modern commercial/industrial designs increasingly incorporate shielded enclosures to enhance performance. We will explore this topic in greater detail in future articles.

As briefly mentioned earlier, mounting power electronics devices on a floating heatsink can introduce EMI challenges. Due to its size, the heatsink can act as an unintended antenna, coupling common-mode noise and either conducting or radiating EMI out of the system.

To mitigate this issue, it is highly recommended to ground the heatsink at multiple points to provide a low-impedance return path and minimize unwanted noise coupling. (More detailed recommendations on heatsink grounding strategies can be found in the sources referenced in endnotes 5, 6, and 7). A heatsink can unexpectedly couple noise into the common mode path. We will dedicate an article to this topic in a future publication to address this issue in greater detail.

Ground Potential Difference in Isolated SMPS

For non-isolated SMPS, the ground design strategy is relatively simple, as there is only one common ground. However, isolated SMPS inevitably create multiple ground potentials. If these different grounds are not properly connected in RF terms (e.g., using appropriately placed capacitors), a high common-mode voltage can develop between them. This voltage can drive common-mode currents across the isolation barrier, typically through the primary-to-secondary capacitive coupling of the isolation transformer.

This is a common cause of unexpected EMI emissions in isolated designs. In our Part 1 article, we demonstrated the impact of capacitors linking these isolated grounds.

It should be noted that filter performance often depends on both the grounding of the filter circuits and the grounding of the switched-mode power supply itself. In the following example, a design contains four switching converters, with the manufacturer incorporating three PCB mounting holes to electrically connect the board to the chassis. Ignoring whether three mounting holes are sufficient (the answer is no, they are not), the common-mode current behavior changes significantly depending on which combination of mounting holes is used for grounding. Despite using the exact same filter, the EMC performance varies greatly.

This highlights the importance of proper grounding in SMPS design. When a seemingly well-designed filter fails to perform as expected, one of the first areas to check is the system grounding.

Input and Output Capacitors in SMPS

One of the most frequent mistakes in SMPS EMI design is confusing EMI filter capacitors with input/output capacitors.

For example, in Figure 10, we illustrate a buck converter where C1–C4 are best understood as input capacitors rather than components of an EMI filter. While these capacitors do interact with the inductor (L1) part of a multi-stage filter, their primary role is to provide a stable voltage source for the converter. A common issue arises when engineers insert an inductor between the input capacitors and the switching devices, mistakenly believing an L-C filter is necessary for switching noise, or place the input capacitors too far from the converter, following an application note that suggests locating the filter away from the main stage to minimize coupling. These mistakes stem from a fundamental misunderstanding of the distinction between input capacitors and the filter stage.

Summary

This concludes Part 2 of our discussion on SMPS filters. As we explained in this article, understanding converter noise and coupling mechanisms is crucial before designing and laying out a filter. A solid grasp of these fundamentals also enhances simulation accuracy for those using a simulation-based approach to filter design.

Additionally, as we emphasized, filters are an essential part of SMPS design, but they cannot solve all noise issues. A key takeaway from this article is to first focus on good EMC design at the power stage before addressing the filter design.

In our next article, we will provide a step-by-step guide on designing effective filters.

References

M. Zhang, “Filter Designs for Switched Power Converters − Part 1: Overview,” In Compliance Magazine, September 2024.

M. Zhang, “GaN/SiC Transistors for Your Next Design: Fight or Flight?” In Compliance Magazine, October 2023.

M. Zhang, “EMC Design Techniques for Electric Vehicle DC-DC Converters,” In Compliance Magazine, December 2021.

K. Javor, online resource, “EMI Radiation Coupling in SMPSs.” 

B. Archambeault, PCB Design for Real-World EMI Control, Springer US, 2002.

M. Nave, Power Line Filter Design for Switched-Mode Power Supplies, Van Nostrand Reinhold, 2010.

K. Armstrong, “EMC Techniques for Heatsinks.”