EMC Considerations in Power Supplies and DC/DC Converters

Introduction

During the shutdowns caused by the COVID-19 pandemic, my car stood idle for several weeks and, being packed with electronics in standby mode, its battery eventually drained, making the car refuse to start. A trip to the accessory store led me to purchase a new “smart” and surprisingly cheap charger, which was hooked up and left to do its job. The job it did, but it also wiped out the house Wi-Fi. Despite the CE-marking and a range of certification stamps found on the device, it evidently had a massive level of radiofrequency (RF) emissions and was a prime example of electromagnetic incompatibility.

Whether the problem was due to radiated or conducted emissions, the charger should have met the published and mandatory standards for electromagnetic compatibility (EMC). These standards also include limits to mains harmonic emissions and “flicker,” along with immunity to the prescribed levels of magnetic, electric, and electromagnetic fields; line surges and transients; and static discharge. The globally used standards are in the IEC 61000 series, with parts 1–7 covering all aspects of requirements, test methods, and limits. Further, references are made to other documents for specific product categories along with their requirements, wherever available.

The filtering you should expect to see

What could the designer of that charger have done better? Looking at the conducted emissions first, the product, as a switched-mode power supply, can produce line-to-line differential-mode (DM) and line-to-ground common-mode (CM) noise (Figure 1). The DM input noise is attenuated by line-to-line “X” capacitors and series inductors, and, therefore, it can be readily reduced to low levels with high enough component values within size and cost constraints. The designers often try to keep the capacitor value below 100nF; however, as stated above, the component must be discharged to a safe voltage within a prescribed time, thereby forcing the addition of parallel resistance.

Moreover, if left permanently in circuit, the resistor’s constant leakage current can make compliance with the standby and no-load-loss standards problematic. Although the inductor(s) can be of high value, they pass the full AC running current; therefore, they must at times be physically large in order to avoid saturation. In this regard, the iron powder or gapped ferrite types are typically examples.

Although there is no direct statutory limit to DM noise, there are limits for CM noise, and the typical test method for CM uses a line impedance stabilization network (LISN) as called for in standards such as CISPR 32 for multimedia equipment. However, the LISN also registers half of the DM noise that is present, so there is a good reason to attenuate it. The CM noise from the line and the neutral-to-ground connection tends to take the form of a current source into the low 50Ω impedance of the LISN, and the “Y” capacitors from the line or the neutral-to-ground connection provide a local return path so that the noise does not circulate externally, thus registering in the LISN.

Then, a CM choke with coupled windings in each power line acts as a barrier between the converter and the supply. It can use high-permeability ungapped ferrite as the windings are phased so that the running current cancels magnetically, leaving a high impedance for the CM noise element. The CM chokes can be wound with controlled leakage inductance between windings, which then yields a combination of DM and CM attenuation.

The level of transient filtering depends on the installation overvoltage category

Along with attenuating emissions, an AC/DC input filter provides immunity to input overvoltages, which can either be high-voltage, low-energy transients and bursts or lower-voltage surges. The levels observed depend on the installation overvoltage category (OVC) at levels I through IV (with increasing severity) (Table 1).


Our charger should meet the OVC II at a minimum, which typically requires the addition of an input transient suppressor component such as a voltage-dependent resistor (VDR). Conversely, if it were OVC IV, you would expect to see high-energy-rated VDRs, along with, possibly, multiple gas discharge tubes.

Furthermore, if the charger were evaluated for compliance with the EU EMC directive (as indicated by its CE marking), it would also have to be immune to specified levels of applied electric, magnetic, and RF fields, along with ESD. Input filtering is not the solution here, but a good internal layout and design practice would also normally help in meeting the emissions limits.

Designs start with “lumped” components, but that is not the full description

It is practical to start any switched-mode converter design with the lumped components in the chosen topology and calculate performance to a first order. This can inform you if the approach is a sensible one; however, if EMC considerations are included, then “real” rather than “ideal” components must be used (Figure 2). The higher order or “parasitic” characteristics of components will often be those that cause EMC problems.

For instance, these could be stray capacitance to ground causing CM noise currents or a series inductance of connections causing radiation. Even the real components depicted in Figure 2 are simplistic. Often, the parasitic values are non-linear, such as the capacitor ESR, which varies strongly with frequency. Further, some parasitics have discontinuities in their characteristics. For example, the MOSFET total input capacitance alternates between effective values, depending on the switch state.

On top of their DC resistance, which merely varies with temperature, even wire and track connections have AC resistance that varies with frequency and material. This is due to the inherent inductance and the “skin effect” caused by eddy currents canceling in the center of the conductor. A rule of thumb is to say that the current at frequency f travels in a copper conductor to a depth of δ = 66/√f (Figure 3).

For example, a wire diameter of 0.4mm at 100kHz should show no skin effect. This is a good enough approximation in most cases, but δ is actually the depth at which the current drops to 1/e or 37% of the total (not zero) and strictly applies to sine waves (not the complex AC waveshapes often observed in converter designs).

Local coupling effects

The two main unwanted effects causing EMC issues are the inductive and capacitive coupling of signals, which result in conducted and/or ultimately radiated emissions. The inductively induced voltage from the current steps is quantified as E = -L.di/dt. Modern converter designs can produce current edge rates of 1000A/µs; therefore, only 10nH can produce a 10V spike. This inductance is only a few millimeters of tracking or wiring.

Similarly, the current is induced through stray capacitance I = C.dV/dt, and the voltage edge rates can be 50kV/µs, resulting in 500mA of displacement current through just 10pF, which is a typical value for the isolation capacitance of a transformer.

These refer to the current and voltage impulses. The steady RMS values from the fundamental frequency and the low harmonics of the waveforms are much smaller and are the values registered in EMC emission evaluations in the spectrum analysis. The RMS values could be obtained from Fourier analysis of the switching waveforms and, afterward, the currents and voltages at those frequencies obtained from simple impedance calculations (e.g., E = 2πfL.i or V = i/2πfC). The resonant converters make the calculations even simpler.

Near- and far-field effects

It is very difficult to quantify the effects of the fields at short distances from the source. As we have already seen, varying the electric or “E” fields induces displacement currents in the conductors through stray capacitances, whereas varying the magnetic or “H” fields induces voltages in the conductors. This is in the “near field” where effects at distance r from the source reduce proportionally to 1/r2 or 1/r3. Further away, in the “far field.” the effects turn into combined electromagnetic (EM) radiation, falling way as 1/r. This conclusion is drawn by assuming the radiation to be omnidirectional. The boundary between the near-field and the far-field depends on the physical dimensions of the source D and wavelength λ, although it can be approximated:

For source dimensions < λ, r = λ/2π
For source dimensions > λ, r = 2D2/λ

As regards the typical power converter fundamental switching frequencies, the source dimensions are certainly lesser than the wavelength, and r is in the range of many tens of meters; therefore, all the local effects are near-field. At the high harmonic levels, say in the order of GHz, the boundary is in the millimeter range for the sources of the order of millimeter size. The standards for EM emissions reflect this, with the prescribed limits typically being up to 1GHz, measured at relatively short, fixed distances away.

The role played by galvanic coupling

Unwanted coupling can be simply galvanic, where the current from a source flows in a connection and either drops excessive voltage or mixes with other current paths to produce “cross-talk.” The PCB tracks are often a culprit and can generate significant DC resistance: 35µm (1oz) of thickness copper 10mm in length and 1mm in width is nearly 5mΩ at 25°C, rising to 6mΩ at 85°C. The voltage drop across this resistance from the current flowing adds to any other power or signal current flowing through the same connection, potentially causing interference. The track impedance to AC is more complex and depends on the proximity to the adjacent tracks, ground planes, and other components.

For example, as Figure 4 shows, a track of width W and thickness t, over a ground plane or simple microstrip with separation H of a material with relative permittivity ε_r, has the following characteristic impedance Z₀:

Z0 = (87/√[ε_r + 1.41]). ln(5.98H/[0.8W + t])ohms

For a typical PCB, εr = 4, H = 0.76mm, and T = 35μm; therefore, a 1mm-wide copper track will have a characteristic impedance, Z₀, of around 65ohm. This value is significant, as any mismatch between this value and the source and the sink impedances for high-frequency current in the track will cause ringing on the switching edges.

Vias are not perfect either

The vias between the layers can also be characterized by their parasitic effects. As Figure 5 shows, if the outside diameter is D and the inside diameter is d, unfilled, and with length T, then the inductance is as follows:

L = 2T(ln(4T/d) + 1)nH

Meanwhile, the capacitance is as follows:

C = 0.55 ε_rTD(D - d)pF


For the typical unfilled vias, these values are 1.2nH and 0.33pF. Moreover, the DC resistance will be approximately 0.5mΩ, whereas the thermal resistance will be approximately 100°C/watt.

Sometimes, it is not possible to ideally separate the currents in the power paths of the converters. As Figure 6 shows, the classic buck topology is such an example, where a “star” connection for the common ground point is the best, but its position cannot be optimum due to the multiple current loops in operation with the energy storage and release phases of the circuit. Additionally, the best common ground point for a feedback signal is not necessarily the same as for the power path.

Conclusion

This article has touched on some of the design considerations necessary to achieve low interactions between the components and the connections in a power converter, which could help in achieving low conducted and radiated emissions as well as standard compliance. Moreover, certain real-life parasitic values have been provided to give an idea of the scale of the effects. The EMC Book of Knowledge [1], which was recently published by the power products manufacturer RECOM, has been the primary source for this article.

As for the battery charger, a teardown revealed no safety or EMI ground connection despite the metallic case, no VDR, a space where an “X” capacitor was not fitted, and a choke position that was strapped out. Perhaps the designer had the right ideas about getting the product through qualification, but cost-cutting overrode the good intentions.

References

[1] EMC ‘Book of Knowledge’ RECOM https://recom-power.com/rec-n-new--emc-book-of-knowledge-266.html

Additional information

EMC-RELATED SERVICES