Bode Plots for Electromagnetic Interference Analysis

EMI Filter Performance – Actual versus Published

Bode plots are great. When it comes to EMI filters, Bode plots provide an intuitive way to compare the performance of one filter to another. However, when comparing filters based on the manufacturer’s published Bode plots, you have to be careful.

The performance of the EMI filter you select may be dramatically different than the data sheet when you install it on your system. The data sheet Bode plot may be correct, but only for specific controlled conditions. The difference between published and actual filter performance lies in the way EMI filters are measured.

Bode Plot Basics

In the most general terms, a Bode plot is a graph of system frequency response. For electromagnetic interference purposes, Bode plots are used to graph EMI filter attenuation. They are a convenient way to display filter performance versus frequency, offering a high-level summary at a glance.

Bode plots can also be used to graph filter phase shift. However, for EMI control purposes, filter phase shift is not usually important. Most often when selecting a filter, we are concerned with how well a noisy signal is attenuated as it passes through the filter and whether that attenuation is sufficient to control EMI.

Filter insertion loss

Insertion loss is the figure of merit that defines EMI filter performance. Insertion loss is the amount of loss a signal experiences when the filter is inserted in the signal path. The signal in this instance is unwanted noise voltage or noise current superimposed on power, control, status, or data lines.

Insertion loss is the ratio of the input signal to the output signal. More specifically, it is the radio frequency (RF) voltage, current, or power amplitude at the filter input terminals divided by RF voltage, current, or power amplitude at the filter output terminals, respectively.

Most often, filter insertion loss is specified in decibels as,

An EMI filter Bode plot graphs insertion loss versus frequency, with insertion loss on the y-axis and frequency on the x-axis. The chart below shows a typical Bode plot. Since the frequency range of interest for most EMI filters covers several decades, the x-axis is usually a logarithmic scale. When insertion loss units are decibels, the y-axis is a linear scale; when expressed as a ratio, it is logarithmic.

Typical Bode Plot

In the Bode plot above, the ratio of filter input signal to filter output signal is on the left-hand y-axis, and the equivalent value in decibels is on the right-hand y-axis. Below 10 kHz the filter provides no attenuation. Above 30 kHz the attenuation increases steadily in proportion to increasing frequency. The corner frequency of the filter is the frequency at which the filter reduces the input signal to one-half its value, i.e. the filter insertion loss is 6 dB.

Measured Filter Insertion Loss

When selecting circuitry, measured data is often more desirable than calculated data. However, when it comes to EMI filters, measurement data is only helpful as a starting point because the way a filter is measured and the way it is used are often very different.

Filter insertion loss measurements are made using a 50 ohm measurement system. Why? RF test equipment, such as signal generators, spectrum analyzers, and network analyzers have standardized 50 ohm source and load impedances. When filter insertion loss is measured, the signal source connected to the filter input, and the measurement instrument connected to the filter output, are both 50 ohm devices.

Insertion loss in a 50 ohm system

Most EMI filters are measuring using methods defined in MIL-STD-220. With the EMI filter removed from the test fixture, a baseline reading is made. Then a second reading is made with the filter installed in the test fixture. The ratio of the two measurements, usually converted to decibels, is the filter insertion loss.

Measuring the filter in a 50-ohm system tells you is how well the EMI filter works when connected to circuits that have 50 ohm impedance at all frequencies, but unfortunately it does not tell you how well the filter works when connected to circuits that are not 50 ohm. In practice, circuits rarely have 50 ohm impedance.

Calculated Filter Insertion Loss

To understand how the filter performs in your circuit, you either have to build prototype hardware and make measurements, or you have to simulate the filter and circuit in software. The software approach is usually faster and more cost-effective and accurate enough for making design decisions.

EMI Analyst software makes quick work of filter simulations. It provides a convenient means for calculating filter insertion loss, not only for 50 ohm systems but circuitry having any impedance.

To illustrate the importance of simulating filter performance with the correct source and load impedances, consider the following three three source and load impedance scenarios:

1. Standard 50-ohm source and 50-ohm load

2. High-Z source and a 50-ohm load

3. High-Z source with 50-ohm LISNs and a 2-meter cable

Standard impedance insertion loss

Let’s take a look at a hypothetical T-filter consisting of two 1 uH inductors and one 1 uF capacitor, as shown in the figure below. This filter configuration is a common low-pass topology, commonly used in situations where the source and load impedances are relatively small compared to the impedance of the filter inductors.

The signal source an ideal voltage source in series with a 50 ohm resistor. The load, a measurement instrument, is simulated as a 50 ohm resistor. The EMI filter is a ladder network containing the three filter elements. For simplicity, the inductors and capacitor are modeled as ideal elements. In a more rigorous analysis, parasitic elements would be included to simulate filter behavior more accurately over the frequency range.

The calculated Bode plot for the T-Filter in the figure above is shown in the graph below. This is the Bode plot for the T-Filter in a 50 ohm measurement system.

Filter Insertion Loss in a 50-ohm System (50-ohm Source / 50-ohm Load)

Non-standard impedance insertion loss

In real applications, the source and load impedance are almost never 50 ohm. What happens if we need to filter a switching power supply, which is more like a current source than a voltage source? And has source impedance much greater than 50 ohms?

Keeping all other settings from the previous example the same, but changing the source to be a current source with 1 M-ohm impedance, the insertion loss for the same filter changes significantly, as shown in the Bode plot below. At low frequencies, the filter provides slightly more insertion loss than before, but at high frequencies, it provides substantially less.

Filter Insertion Loss, 1 M-ohm Current Source / 50-ohm Load

In this configuration, the filter output is 0.30 (-10.3 dB) of the input at 10 kHz and 2.5 x 10^-6 (-112 dB) of the input at 100 MHz. Contrast that to 0.54 (-5.3 dB) and 4.0 x 10^-7 (-128 dB), respectively in the 50-ohm system. At 10 kHz, the filter is 5 dB better, but at 100 MHz is it 16 dB worse. The filter attenuates differently when source or load impedance changes.

Installed insertion loss

Now let’s look at a circuit that is closer to something you might encounter in practice. Use the same source and filter as in the previous example, but replace the 50 ohm load with 2-meters of wiring and a power source that consists of two Line Impedance Stabilization Networks (LISNs) and a dc power supply, a standard configuration for EMI testing of input power circuits.

The insertion loss of the filter for this real-world circuit configuration is dramatically different than its insertion loss in the 50-ohm system. As shown in the Bode plot below, the filter provides much more high-frequency insertion loss. At 10 MHz this configuration attenuates to 1.5 x 10^-6 (-116 dB) compared with 2.4 x 10^-4 (-72 dB) for the 50-ohm system. However, the Bode plot also shows that the filter amplifies noise between about 4 kHz and 20 kHz. Greatest amplification occurs at 15 kHz. Instead of being attenuated, RF noise emitted from the source at 15 kHz is 10 times bigger after passing through the filter.

Filter Insertion Loss, 1 M-ohm Current Source / Wire & LISN Load

The insertion loss curve from the 50-ohm system shows no such amplification. If the decision to use this filter were made solely on the 50 ohm datasheet curves, you might be unpleasantly surprised when the circuit has EMI problems near 15 kHz in the final installation.

Final Thoughts

Bode plots are a convenient way to display filter performance, but because insertion loss measurements are made using a standardized 50 ohm test setup, the filter attenuation may be very different in your circuit than it is on the data sheet.

A quick way to see how the filter is going to perform in your circuit is to model the filter using EMI Analyst. Use the same topology and component values, and then compare insertion loss for the standard 50-ohm setup with insertion loss when used in your circuit.

With EMI Analyst, you are not limited to modeling the just EMI filters. You can easily and efficiently model complete circuits to get a clear picture of their electromagnetic interference performance.

For more information about EMI Analyst, visit