Improve EMC Testing Results with Realistic EMI Filter Modeling

A computer model that includes parasitic elements is essential when synthesizing a custom filter design or predicting how an off-the-shelf filter will perform in your circuit.

EMI filters are a rudimentary tool for controlling electromagnetic interference. EMC testing success often hinges on EMI filtering.

Selecting or designing a suitable filter seems like a straight-forward task. However, because EMC testing covers a broad range of frequencies, coming up with a good filter is not always easy. Filter components have parasitic elements that fundamentally affect performance, differently at different frequencies. The effect of parasitics can be dramatic.

Ideal versus Real

The two graphs below show predicted performance for the same EMI filter. The one on the left models the filter using ideal components, the one on the right uses realistic components.

At very low frequencies, the ideal model is adequate. However, over most of the frequency range using ideal components to model the filter produces performance predictions that are very different than the real filter.

A good computer model that includes parasitic elements is essential when synthesizing a custom filter design or predicting how an off the shelf filter will perform in your circuit.

Ideal Components

In order to understand the importance of parasitics, let’s first look at ideal components. Ideal components are simplified models of real components that include only the core element, but ignore the parasitics.

Most EMI filters are constructed using passive components: capacitors, inductors, and resistors; all are linear components. Their impedance changes linearly as frequency changes. When graphed on a log-log scale, which is convenient when evaluating frequencies that span several decades, their impedance is a straight line.

Ideal Capacitor

The impedance of an ideal capacitor decreases in proportion to increasing frequency.

Ideal Inductor

The impedance of an ideal inductor increases in proportion to increasing frequency.

Ideal Resistor

The impedance of an ideal resistor is constant with increasing frequency.

Real Components

Models of real components are constructed using ideal component elements in various combinations. Parasitic elements may be in series with or parallel to the core element. The addition of the parasitic elements allows the component to be more accurately modeled over range of frequencies to better simulate real component behavior.

Real Capacitors

Real capacitors have inductance and resistance in series with the capacitance, and leakage resistance in parallel. Values for the equivalent series inductance (ESL), equivalent series resistance (ESR), and leakage resistance vary by capacitor type and value.

When capacitive reactance and inductive reactance are equal, the capacitor self-resonates. At frequencies below its self-resonant frequency, capacitance dominates and the impedance of a real capacitor decreases with increasing frequency. Above the self-resonant frequency, the ESL dominates and the impedance increases. At the self-resonant frequency, the impedance equals the ESR.

Real Inductors

Real inductors have resistance in series with the inductance, and stray capacitance in parallel.  Values for these parasitics vary with wire gauge and inductor construction.

At the frequency where the inductive reactance and capacitive reactance are equal, the inductor self-resonates. Below the self-resonant frequency, inductance dominates and impedance increases with increasing frequency. Above the self-resonant frequency, capacitance dominates and impedance decreases.

Real Resistors

Real resistors have inductance in series with the resistance, and stray capacitance in parallel.  Values for these parasitics vary with resistor type and size.

Except for very small value resistors, resistor ESL is usually negligible. At low frequencies, resistance dominates, but at high frequencies the impedance of the resistor rolls off due to the stray capacitance. The frequency at which the capacitive reactance equals the resistance is the corner frequency. At the corner frequency the resistor impedance is one-half its resistance value.

Filter Insertion Loss

Insertion loss is a measure of how well an EMI filter attenuates a signal as it passes through the filter. Normally expressed in decibels, filter insertion loss is the ratio of the input signal to the output signal.

EMI filters are measured by connecting a signal source across the filter input terminals and then measuring the signal amplitude across the output terminals. Normally source and load impedance is 50 ohm. Differential mode insertion loss and common mode insertion loss are measured separately. The figure below shows the measurement setup for differential mode insertion loss.

Filter With Ideal Components

Insertion loss for a low-pass filter modeled with ideal components steadily increases as frequency increases.

To illustrate, the simple L-C filter below has a 500 kHz corner frequency and two-pole roll-off of 40 dB per decade.

Using ideal components yields an unrealistic prediction of filter performance. At high frequencies, predicted attenuation may be orders of magnitude greater than for the actual filter.

Filter With Real Components

Using component models that include parasitic elements results in a more useful prediction of filter performance. Consider the following adjustments to make the components more realistic.
The wire used to construct the 100 microH inductor has 25 m ohm resistance, and when wound on the inductor core there is 20 pF stray capacitance from one end of the winding to the other.

The 1 nF capacitor has 10 nH ESL, which includes internal, lead, and trace inductance.  Its ESR is 5 m ohm and its leakage resistance is 10 M ohm.

Insertion loss predicted using realistic models for the components provide a more realistic prediction of how the filter will behave over the frequency range we need it to operate.

The first dip is due to the self-resonance of the inductor. The second dip is due to capacitor self-resonance. At high frequencies, the filter becomes increasingly less effective, providing little attenuation above 100 MHz.

Get Real

It takes a little longer to dig up the information needed to model a filter using realistic components, but the difference is worth the effort. Compare the realistic filter prediction to the ideal filter prediction. At 100 MHz the real filter model predicts the filter has just 20 dB attenuation, whereas the ideal filter model predicts 90 dB. That is a 70 dB error if ideal components are used.

One of the easiest ways to improve EMC testing results is to select or design appropriate filtering for end circuits and power lines. Use real component models when building filter simulation models and your designs will more closely match your actual circuits.

For more information, check out these other helpful links:

EMI Filter Insertion Loss: How Circuit Impedance Affects EMI Filter Performance -

EMC Analysis: How to Calculate Filter Insertion Loss -