Board Level EMI Filtering Guide

Board level emi filtering is the practice of stopping unwanted energy at the point where it would otherwise leave or enter the board. Filters work by providing a low impedance path for the unwanted frequency while presenting a high impedance to the signal, and their effectiveness depends far more on where they are placed and how they are grounded than on the component values chosen. A filter in the wrong location, or one with a poor return path, can be worse than no filter at all.

What Filtering Has to Achieve

A filter has to reduce the amplitude of an unwanted signal, whether it is emissions leaving the product or interference entering it, to a level that meets the requirement. The requirement is usually a limit at a specified frequency range, and the filter has to achieve it over that whole range rather than only at the frequency being tested.

The frequency range is what determines the approach. A filter that attenuates well at one megahertz may do very little at a hundred, because the parasitic elements of the components and the layout dominate at high frequency. Understanding the range and the mechanism that dominates within it is the first step in choosing a filter.

The filter also has to pass the wanted signal without distortion. A common mode choke passes differential signals and attenuates common mode currents, while a feedthrough capacitor shunts high frequency energy to ground. Choosing the wrong topology either fails to attenuate or damages the signal.

Placement and the Interface

The most important rule is that filtering belongs at the interface, at the point where the cable meets the board. A filter placed further inside the board allows the interference to couple into other parts of the circuit on its way, and it allows noise generated inside to travel out along the trace before being attenuated.

The connector area should be treated as a boundary with a clean reference plane beneath it. The filter components should connect to that plane with the shortest possible path, and the plane should be continuous. A filter whose return path is a thin trace rather than a plane has an inductance that cancels much of its benefit.

Where the board has several interfaces, each should be filtered at its own connector rather than sharing a filter further down. Sharing introduces coupling between the interfaces, which is precisely what the filters were intended to prevent.

Components and Topologies

A common mode choke presents a high impedance to currents flowing in the same direction on both conductors and a low impedance to the differential signal. It is the standard component for a differential pair or a supply entering a board, and its effectiveness depends on the symmetry of the winding and on the balance of the circuit it is protecting.

A feedthrough capacitor is used where a single conductor passes through a boundary, and its advantage is that the signal passes through the body of the capacitor so that the series inductance is very low. A conventional capacitor connected to a ground trace has a considerable inductance in the connection, which limits its performance at high frequency.

Series ferrite beads are useful for attenuating a narrow band and for damping resonances, but they have to be chosen carefully. A bead with high impedance at the frequency of interest may also have a high impedance at the signal frequency, and a bead that saturates with the DC current through it loses most of its effect.

Grounding and Return Paths

Grounding is where most filters fail. The attenuation of a capacitive filter depends on the impedance of the return path as well as on the impedance of the capacitor, and a long return with significant inductance limits the achievable attenuation. This is why the filter capacitor must sit directly on a plane, with the shortest possible connection.

A filter placed on a board with a split ground plane may connect two reference areas through itself, which creates a path for interference to flow from one to the other. Where two grounds must be separated, the filter is often the component that bridges them, and the arrangement should be deliberate rather than accidental.

Cable shields should be bonded to the board ground at the connector, ideally around the full perimeter. A shield terminated by a single wire has an inductance that destroys its effectiveness at the frequencies of interest, which is why connector shells and metal brackets are used rather than wire links.

Shielding and Its Limits

Shielding blocks radiated coupling, and it is used where filtering alone cannot achieve the requirement. A shield has to enclose the source or the victim, it needs an adequate number of openings rather than a large one, and it has to be bonded to the reference on all sides. A shield with a long narrow slot behaves as a slot antenna at the frequency where the slot is a half wavelength.

Shielding and filtering work together rather than as alternatives. A shielded enclosure with unfiltered cables will radiate through the cables, and a well filtered board with no shield may still couple radiated energy between its own sections. The two measures address different mechanisms and both are usually needed at the higher limits.

Shields add cost, weight and assembly steps, and they should be evaluated against a layout change or a filter improvement that might achieve the same result. Where the emission is coming from a single source, improving the layout of that source is often cheaper than enclosing the whole board.

Design Practice and Verification

The practical measures are consistent: locate filter components at the connector, keep the return path short and on a plane, use the topology that matches the type of current to be attenuated, and keep the filtered and unfiltered areas separated. None of these is expensive, and all of them require deliberate layout rather than an added component list.

Verification is by measurement rather than by calculation. Conducted emissions at the interface and radiated emissions in an anechoic environment are the standard tests, and they should be performed on a representative assembly rather than on a partly populated board. A measurement on a board that is missing its final enclosure and cable arrangement may not represent the product.

The measurements should be kept with the design, because a later change to a component or to the layout can invalidate them. Where a substitution is made, the effect on the filter should be assessed, particularly if the substitute has different parasitics or a different DC current rating.

Practical Rules

Filter at the interface, keep the return path on a plane and as short as the layout allows, and choose the topology for the type of current rather than for the component list. Bond shields around their full perimeter.

Verify by measurement on a representative assembly and keep the results with the design records. Filtering, grounding and shielding are one subject in practice, and the suppression design of the board as a whole decides how much work the filter has to do.

Related reading: our fabrication notes, board quality and design release notes cover the same ground.

FAQ

Where should a filter be placed? At the interface, at the connector. Anywhere further inside allows noise to couple into the circuit before it is attenuated.

Why does a feedthrough capacitor work better at high frequency? Because the signal passes through its body, so the series inductance of the connection is very small compared with a conventional capacitor on a trace.

Can a shield replace filtering? No. A shielded enclosure with unfiltered cables still radiates through the cables. The two measures address different coupling mechanisms.