Electromagnetic Interference (EMI) in Switching-Mode Power Supplies (SMPS)

Home | Glossary | Books | Links/Resources
EMC Testing | Environmental Testing | Vibration Testing


Electromagnetic interference (EMI), otherwise referred to as radio-frequency interference (RFI), the unintentional generation of conducted or radiated energy, is indefatigable in all switching-mode power supplies. The fast rectangular switching action required for good efficiency also produces a wide interference spectrum which can be a major problem.

Further, for proper operation of any electronic system, it's important that all the elements of the system be electromagnetically compatible. Also, the total system must be compatible with other adjacent systems.

As the SMPS can be such a rich source of interference, it's vital that this aspect of the design be carefully considered. Normal good design practice requires that the RF interference allowed to be conducted into the supply or output lines, or permitted to be radiated away from any power equipment, be minimized to prevent RF pollution.

Further, national, federal, and international regulations limit by law the permitted interference levels.

These regulations vary according to country of origin, regulatory authority, and intended application. The power supply designer will need to study the code relevant to the proposed marketing areas. In Common Market countries, IEC BS 800, or CISPR recommendations apply. The Federal Republic of Germany requires VDE 0871 or VDE 0875, depending on the operating frequency. In the United States the Federal Communication Commission's (FCC) rules apply, and similar limits are recommended in Canada under CSA, C108.8-M198

In general, the range of frequencies covered by the regulations spans from 10 kHz to 30 MHz. Domestic locations have more rigorous regulations than office or industrial locations.

IMG. 1 shows the FCC and VDE limits for conducted RFI emissions in force at the time of publication.

IMG. 1 Conducted-mode RFI limits as specified by FCC Part 15 (Subpart j) VDE 0871 and 0875.


There are two forms of propagation of interest to the power supply designer: electromagnetic radiated E and H waves and conducted interference on supply lines and interconnecting wires.

Radiated interference is normally minimized as a natural result of the layout and wiring practices required to reduce leakage inductance and improve performance.

Typically the high-frequency current loops will be short, and twisted pairs will be used where possible. Transformers and chokes with air gaps will be screened to reduce radiated magnetic fields (see Part 1, Sect. 4), screened boxes or equipment enclosures will often be used.

The techniques applied to minimize conducted interference will also reduce radiated noise. The following sections concentrate on the conducted aspect of power supply interference, as once the conducted limits have been met, the radiated limits will normally be satisfied as well.


Two major aspects of conducted interference will be considered: differential-mode conducted noise and common-mode conducted noise.

These will be considered separately.

Differential-Mode Interference Differential- or series-mode interference is that component of RF noise which exists between any two supply or output lines. In the case of off-line SMPS, this would normally be live and neutral ac supply lines or positive and negative output lines.

The interference voltage acts in series with the supply or output voltage.

Common-Mode Interference Common-mode interference is that component of RF noise which exists on any or all supply or output lines with respect to the common ground plane (chassis, box, or ground return wire).


It may seem out of place to be considering safety requirements at this stage; however, this is necessary because the safety agencies specify the maximum ground return currents, so as to minimize shock hazard in the event of ground circuit faults. This requirement not only requires good attention to insulation, but also puts a severe limitation on the value of capacitors which may be fitted between the supply lines and ground. This capacitor size limitation has a profound impact on the design of the line input filters.

The permitted limits for ground return currents vary among the regulatory agencies and also depend on the intended equipment applications. For example, medical equipment, as one might expect, has a very stringent, so-called "earth leakage current" limit.

The ground return current limits, as set by some of the major regulatory agencies, that are in force at the time of printing are shown in Table 1.

Tbl 1 Maximum Ground Leakage Currents Permitted by the Safety Regulations, and the Recommended Maximum Values for Y Filter Capacitors

IMG. 2 Ground leakage current test circuit, as specified by CSA 22.2, part 1. (Note: CT and RT values depend on the equipment and agency requirements.)

Tbl 1 gives the maximum value of decoupling capacitance that may be fitted in positions C1 and C2 in Img. 2, for each specification. These values assume zero contribution from insulation leakage and stray capacitance. To minimize inductor and filter size, the largest decoupling capacitor permitted by the regulations should be used. Since one side of the input is always assumed to be neutral (connected to the ground at the service entrance), only one capacitor will be conducting at any time. However, the total leakage current should be checked to establish the total contribution from all the capacitive and insulation leakage paths.

IMG. 2 shows the method of measurement for ground return currents. It is assumed that only one side of the supply could be "hot," and hence only one capacitor will be conducting to ground return at any time.


To meet the conducted-mode noise specifications, relatively powerful line filters will normally be required. However, as previously demonstrated, safety regulations severely limit the size of the capacitors fitted between the supply lines and ground plane.

Because of the limited size of the decoupling capacitors, the filter can't easily cure the severe common-mode interference problems which can occur as a result of poor wiring, bad layout, poor screening, or bad location of the power switching elements. Hence, good EM1 performance demands care and attention to all these aspects at every stage of the design and development process. There is no substitute for effective suppression of EM1 at the source.


IMG. 3 shows several of the more common causes of EM1 problems. Failure to screen the switching devices and failure to provide RF screens in the transformer are principal causes of conducted common-mode interference. This component of interference is also the most difficult to eliminate in the filter, because of the limited decoupling capacitor size.

The differential- or series-mode noise is more easily bypassed by the electrolytic storage capacitors and the relatively large decoupling capacitors C3 and C4 which are permitted across the supply lines.

Common-mode RF interference currents are introduced into the local ground plane (normally the chassis or box of the power supply) by insulation leakage and parasitic electrostatic and /or electromagnetic coupling, shown as Cp1 through Cp5 on Img. The return loop for these parasitic currents will be closed back to the input supply lines through the decoupling capacitors C1 and C2.

The prime mover for this loop current tends to a constant-current source, as the source voltage and source impedance are very high. Hence the voltage across the decoupling capacitors C1 and C2 tends to a voltage source proportional to the current magnitude and capacitor impedance at the interference harmonic frequency:

V_hi = I_i x X_c

where V_hi=harmonic interference voltage Ii=interference current at the harmonic frequency

X_c = reactance of C1 or C2 at the harmonic frequency

(It is assumed that the insulation leakage current is negligible.) This voltage source V_hi will now drive current into the series inductors L1, L2, and L3 and into the output lines to return via the ground line. It is this external component of RF current that will cause external interference, and hence it's this that's covered by the regulations and must be minimized.

IMG. 3 Example of parasitic RFI current paths in a typical off-line switching-mode power supply.

IMG. 4 TO3 mounting bracket and heat sink, with bracket configured to double as an RFI Faraday screen.


Consider the parasitic current loop A, B, C, D, and back to A, shown in the IMG.. Point A is the high-voltage switching transistor package.

For a flyback application, the voltage on this transistor may be of the order of 600 V and the switching frequency typically 30 kHz. Because of the fast switching edges, harmonics will extend up to several megahertz. Parasitic capacitive coupling (shown as Cp1 in the diagram) will exist between the transistor case A and the ground plane B.

The tenth harmonic of the switching frequency will be 300 kHz, well inside the RF band laid down in the regulations. If square-wave operation is assumed, the amplitude of this harmonic will be approximately 20 dB down on 600V, or 60V.

Assuming the leakage capacitance to be 30 pF, a current of 4 mA will flow into the ground plane at 300 kHz.

The current loop is closed back to the transistor by the filter capacitors C1 and C2.

To meet the most stringent safety regulations, the maximum capacitance allowed for C1 and C2 would be, say, 0.01 µF.

If the majority of the ground plane current returns via one of these capacitors, then the voltage V_hi across its terminals, nodes C to D, will be 180 mV. The inductors L1 and L2 now form a voltage divider network between point D and the simulated 50-_ supply line resistance RT. If the voltage across RT is to be less than 250 µV (48 dB up on 1 µV, the regulation limit), then L1 and L2 must introduce an attenuation of more than 50 dB at this harmonic frequency, an almost impossible task for inductors which must also carry the supply input current.

By fitting an electrostatic screen between the transistor and the ground plane, connecting it such that the RF currents are returned to the input source, the ac voltage across the parasitic capacitance Cp1 will be eliminated and the effective RF current from point A to ground will be considerably reduced (see Figs. 4 and 5). The demands now placed on the input filter are not so stringent.

Reducing the RF currents in the ground plane at the source is by far the best approach to EMI elimination. Once these interference currents have been introduced into the ground plane, it's very difficult to predict what path they will take. Clearly all the high-voltage ac components should be isolated from the ground plane, or if contact cooling is required, they should be screened (see Img. 4).

Transformers should have Faraday screens, which should be returned to the input DC lines, to return capacitively coupled currents to the supply lines (see Img. 5).

These RFI screens are in addition to the normal safety screens, which must be returned to the ground plane for safety reasons.

Capacitor C4 (Img. 3) reduces the differential- or series-mode noise applied to the terminals of L1. The major generator of noise in this part of the circuit is the input rectifier bridge (as a result of the rectifier reverse recovery current spikes). The series-mode noise generated by the power switching elements is best decoupled by a capacitor C5 close to the point where the noise is generated. In any event, the large electrolytic storage capacitors will usually effectively shunt away the majority of any series noise that appears between the high-voltage DC lines. In some cases, additional filter components L4, L5, and C6 (Img. 5) are provided to improve the series-mode filtering.

IMG. 5 Preferred positions for primary-to-ground-plane RFI screens.


IMG. 6 shows the standard LISN, used for the measurement of line-conducted interference, as specified by CSA C108.8-M1983 Amendment 5, 198 (Similar networks are specified by the FCC and VDE.) In principle the wideband line chokes L1 and L2 divert any interference noise currents from the supply into the 50-_ test receiver via the 0.1-µF capacitors C3 or C The line not under test is terminated in 0.1 µF and 50 _. It is normal to test both supply lines independently for common mode noise, as the user can connect the input in reverse or may have isolated supplies.

IMG. 6 Standard line impedance stabilization network (LISN) as specified for FCC, CSA, and VDE conducted-mode line interference testing.


The design approach used in Secs. 4 through 8 was to consider the line filter as an attenuating voltage divider network for common-mode RF noise. This approach is used in preference to normal filter design techniques, as the source and load impedances are not definable in the power line environment.

The interference noise generator, in switching-mode supplies, is very often a high voltage source in series with a high impedance; this tends to a constant-current source. To give good attenuation, one of the prime requirements is to convert the constant-current noise source into a voltage source. This is achieved by providing a low-impedance shunt path at the power supply end of the filter, Hence, power line filters will not be symmetrical or matched networks.

"Network analysis" shows that the greater the mismatch of the filter impedance to the source or terminating impedance, the more effective the filter is in attenuating the RF interference.

Referring to Img. 3, and assuming a constant current into nodes C and D, the attenuation into the external 50-_ test receiver would be 12 dB/octave provided that inductors L1 and L2 and capacitors C1 and C2 have good wideband impedance characteristics. Although capacitors meeting this criterion can be easily selected, wideband inductors are not so easily found and are difficult to design, as they must also carry the supply line currents without significant power loss.

Finally, as shown in Sec. 4, the safety requirements set a limit on the maximum size of the decoupling capacitors C1 and C2, so that any further increase in the attenuation factor of the filter is critically dependent on the value and performance of the series inductors L1 and L2. Some design criteria for the filter inductors will now be considered.


Inductor L1 in Img. 3 should be considered a special case. For the best common mode attenuation it must have a high common-mode inductance and also carry the 60-Hz supply current.

To provide the maximum inductance on the smallest core, a high-permeability core material will be used. It is normal practice to wind L1 with two windings.

These windings carry large currents at twice the line frequency, as the rectifier diodes only conduct at the peak of the input voltage waveforms.

In more conventional choke designs, this operating condition would require a low-permeability material or air gap in the magnetic path to prevent saturation of the core. However, in this application, the two windings on L1 are phased such that they provide maximum inductance for common-mode currents but cancel for series mode currents.

This phasing prevents the core from saturating for the normal 60-Hz differential line currents, as these flow in opposite phase in each winding, eliminating the 60-Hz induction. However, this phasing also results in negligible inductance for series mode noise currents, and additional non-coupled inductors L2 and L3 will sometimes be required to reduce series-mode noise currents.

This is one situation in which a large leakage inductance between the two windings on L1 can be an advantage. For this reason, and to meet safety requirements, the windings will normally be physically separated and a bobbin with two isolated sections will be used. As the low-frequency induction is small, a high-permeability ferrite or iron core material may be used, without the need for an air gap.

Where this type of common-mode inductor is used for the output filter in DC applications, the series-mode DC components will also cancel, and the same conditions prevail.

The performance of L1 for common-mode noise is quite different. Common-mode noise appears on both supply lines at the same time, with respect to the ground plane.

The large shunt capacitor C2 helps to ensure that the noise amplitude will be the same on both lines where they connect to the inductor. The two windings will now be in phase for this condition, and both windings behave as one, providing a large common-mode inductance.

To maintain good high-frequency rejection, the self-resonant frequency of the filter inductors should be as high as possible. To meet this need, the interwinding capacitance and capacitance to core must be as low as possible. For this reason single-layer spaced windings on insulated high-permeability ferrite toroids are often used. The effective inductance of the common-mode inductors can be quite large, typically several milli-henrys (mH).

When extra series-mode inductors are used (L2 and L3 in Img. 3), the common-mode inductor L1 can be designed to reject the low-frequency components only, and so the interwinding capacitance is not so important. For this application ferrite E cores can be used; these have two section bobbins, giving good line-to-line insulation. Inductors L2 and L3 must provide good high-frequency attenuation and would normally be low-permeability iron powder or MPP Permalloy toroids. Single layer wound chokes on these low-permeability cores will not saturate at the line frequency currents.

The inductance and size of the main common-mode choke L1 depends on the current in the supply lines and the attenuation required. This is best established by measuring the conducted noise with capacitors C1 and C2 in place but without inductors. The voltage and frequency of the largest harmonic are noted, and the inductance required to bring this within the limit can be calculated. It then remains to select a suitable core, wire size, and turns for the required inductance, current rating, and temperature rise.

It should be noted that the losses in L1 are nearly all resistive copper losses (I_2 R_Cu), as the core induction and skin effects are negligible. The design of L1 is an iterative process which is probably best started by selecting a core size for the current rating and required inductance using the "area-product" approach (see Part 3, Sect. 1).


Assume it has been established by calculation or measurement (Sec. 10) that a 100-W power supply operating from a 110-V ac supply requires a common-mode inductance of 5 mH to meet the EM1 limits. Further assume the power loss in the inductor is not to exceed 1 percent (1 W) and the temperature rise is not to exceed 30 K (all typical values).

For a temperature rise of 30 K at 1 W, the thermal resistance of the finished inductor (to free air) R0 is 30 K/W. From Table 2.19.1, at R0=30 K/W, a core size of E25/25/7 is indicated.

For a 100-W unit with an efficiency of 70% and power factor of 0.63 (typical values for a flyback SMPS capacitor input filter), the input current will be 2A rms at 110 V.

If the total loss (both windings) is to be 1 W, then I2 R=1 and the resistance of the total windings RCu must not exceed 0.25 _.

From the manufacturer's data, the copper resistance factor Ar for the E25 bobbin is 32 µ_. The turns to fill the bobbin and give a resistance of 0.25 _ can now be calculated:

Allowing 10% loss for the split bobbin, there will be 40 turns for each side.

The AL factor (inductance factor) for the E25 core in the highest permeability material N30 is 3100 nH. The inductance may now be calculated:

The largest wire gauge that will just fill the bobbin for this number of turns (from the manufacturer's data) is AWG 20. Since the inductance is marginal, the process can be repeated with the next larger core.


The design of the series-mode iron dust or MPP cored inductors is covered in Sects. 1, 2, and 3 of Part


1. Explain and give examples of some of the typical causes of conducted and radiated RFI interference in switching-mode power supplies.

2. What forms of electrical noise propagation are of most interest to the power supply designer?

Describe the difference between differential-mode interference and common mode interference.

Why is it important to reduce interference noise to the minimum?

5. At what position in the power supply is RFI interference best eliminated?

6. Why are line filters of limited value in eliminating common-mode line-borne interference?

Next: Faraday Screens

Prev: Questions/Answers about Switching Power-Supply Topology

top of page   Home

Home | Glossary | Books | Links/Resources
EMC Testing | Environmental Testing | Vibration Testing

Updated: Tuesday, 2012-01-31 3:39 PST