Shielding Methods for High-Speed PCB Design

As data rates rise, the signals that carry the data become more sensitive and their energy becomes smaller at the same time. A fast edge spends a very short time at each level, and the receiver has to decide what it saw from a small amount of energy, which makes the link vulnerable to disturbance that a slower interface would have ignored. Shielding is the family of techniques that keeps that disturbance away from the signal, and on a board it is mostly a matter of geometry rather than of adding parts.

Where the Disturbance Comes From

Two kinds of field cause the trouble, and they behave differently.

Electromagnetic interference is the low-frequency case. Motors, fluorescent lighting and power wiring are the usual sources, and their fields are strong but change slowly. A trace that is exposed to such a field picks up a voltage along its length, which is why a long loop with a poor return path is the worst geometry for this exposure.

Radio frequency interference is the high-frequency case. Radio and television transmitters, radar and wireless equipment produce fields at frequencies that are comparable to the ones a fast digital edge already occupies. This is the more dangerous category for a high-speed link, because the interference falls inside the bandwidth of the signal and cannot be filtered out without filtering the signal as well.

Public networks add a third consideration that has nothing to do with noise. Data travelling in an unshielded pair radiates, and the radiated energy can be intercepted. A shielded link radiates less, so less energy is available to be picked up anywhere but at the intended receiver, and the time required to capture a meaningful amount of information becomes much longer.

<img src="https://www.gopcba.com/wp-content/uploads/2020/12/service_05.jpg" alt="via fence and ground stitching around a sensitive trace” />

The Principle Behind a Shield

A shield is a conductive barrier that a field has to cross before it reaches the conductor it is protecting. The interfering energy induces a current in the shield’s surface, and that current produces an opposing field, so the signal conductor inside stays quiet. The conductor itself must be inside the barrier and must not be exposed to the outside field through a gap.

The consequence is that a shield is only as good as its connections. A conductive surface that is floating, or connected at a single point, does not provide a low-impedance path for the induced current, and the current has to find a way around the discontinuity, which leaves a portion of the field to pass through. For high-frequency fields the return path has to be short compared with the wavelength of the disturbance, which is why the practical rule is to connect a shield at multiple points around its perimeter rather than at one convenient corner.

Shielding on the Board Itself

The most effective shield on a printed circuit board is the ground plane, and it is already there. Any trace routed over a continuous ground plane has a return path directly underneath it, and the area of the loop that can pick up an external field is reduced to almost nothing.

Where a plane cannot be used, the board offers three other structures.

A via fence is a row of vias that connects two ground areas through the board and forms a wall of vias along a route. A fence placed on both sides of a sensitive trace confines lateral coupling: the trace behaves as if it were in a channel, and crosstalk from neighbours is reduced. The fence has to be dense relative to the highest frequency of interest, which in practice means vias at intervals of a few millimetres or closer, and each via must connect to a ground plane rather than to an isolated region.

A guard trace is a conductive track that runs alongside a sensitive signal and is connected to ground at intervals. Used correctly it reduces coupling between adjacent nets, but it only works if the connections are frequent enough to keep the guard at ground potential along its whole length. A guard that is grounded at its two ends and nowhere else is a resonant structure that can make the coupling worse.

A shield can is a metal cover soldered over a section of the board, usually over an RF front end, a crystal or a switching regulator. It is the most complete form of shielding available to a board designer, and it also the most expensive, both in parts and in assembly. Its effectiveness depends on the ground connection around its perimeter, so the board needs a ground ring with vias under the can’s edge rather than isolated pads.

<img src="https://www.gopcba.com/wp-content/uploads/2024/09/d72fc51735905611378fa8e31dee527-1536×1152-2.webp" alt="shield can soldered over an RF section of a board” />

Designing a Shielded Section

Once a section is to be covered, the layout has to be arranged for the cover rather than around it.

Define the perimeter first. A ground ring along the whole outline of the shield, with vias connecting it to the ground plane at intervals, gives the can a proper termination. A perimeter that runs over a gap, or crosses a signal, leaves an aperture that leaks at high frequency.

Keep the signals that enter and leave the section away from the wall. Every trace that crosses the shield perimeter carries its own return current across the boundary, and the return has to pass through the ground ring rather than around it. Where a trace must cross, it should cross directly above the ground ring with a via to ground on each side, and the filter for that line should sit inside the shield.

Consider the height and the mechanical arrangement. A can that is too tall wastes volume, and a can that sits too close to tall components cannot be fitted. The frame and the lid have to be picked and placed with the rest of the surface mount parts, which means the pad geometry and the paste deposit for the frame are part of the assembly data. Our SMT assembly process handles shield frames as standard parts, but the footprint and the stencil have to match the part the design specifies.

Cables and Connectors

Shielding a board and leaving the cable unshielded wastes most of the effort, because the cable is the longest antenna in the system.

A braided shield is a woven mesh of fine conductors, and its effectiveness against low-frequency fields comes from the low resistance of the mesh. A foil shield, usually an aluminium layer with a drain wire, is a continuous surface and performs better at high frequency, because it has no apertures to leak through. The weakness of a braid is the openings between the woven strands; the weakness of a foil is that it is fragile and that the drain wire provides a long, inductive return path if it is terminated poorly.

For a field that contains both low and high frequency content, the combined construction is a foil layer over a braid, which covers both cases at the cost of stiffness and weight. Coverage percentage is the parameter that matters for braid, and a higher coverage means a smaller aperture and better performance.

The termination is at least as important as the shield. A shield connected to the chassis through a short pigtail behaves like an antenna at high frequency, because the pigtail is a long inductive path. The correct arrangement is a 360 degree connection to the connector shell, so the shield current can spread around the full circumference. Where such a termination is not possible, a shielded connector with a metal shell bonded to the chassis is the next best option.

What to Check Before Release

Shielding decisions are cheap while the layout can still be changed and expensive afterwards, because they affect the layer stack, the ground ring, the component placement and the assembly data at the same time.

The review worth running is a list rather than a measurement. Is every fast net referenced to a continuous plane? Is every shield perimeter ringed with ground and stitched with vias? Does any signal cross a boundary without a return path beside it? Is the cable shield terminated around its full circumference at the connector? A negative answer to any of those is a change that costs less before the data is released.

The layer arrangement that makes plane-based shielding possible is set with the fabricator, and our PCB capabilities page lists the constructions we build. Where the design includes a covered section, the assembly drawing should show the shield as a separate item with its own part number, and our quality management flow keeps the shield in the same traceability record as the board so that a missing can is caught at inspection rather than by the customer.

FAQ

Is a ground plane enough shielding on its own? For a trace routed over it, often yes. It does not shield a component that radiates, and it does not stop an external field from reaching a component mounted on the surface.

How often should a guard trace be grounded? Frequently enough that the guard stays at ground potential along its length. A guard grounded only at its ends can resonate and increase coupling.

Does a shield can need a ground ring? Yes. The can works by diverting current around the protected area, and that requires a continuous low-impedance connection around its perimeter.