EMI Shield Cans: Design, Grounding and Assembly

A shield can is the most direct way to stop a circuit from radiating or from being disturbed by a nearby one, but it only works if the board underneath is designed for it. A can placed on a footprint with no ground stitching, or with a paste pattern that never reflows, performs worse than no shield at all because it adds cost and gives false confidence. This article covers what a board level shield does, how the footprint is laid out, and where the limits are.

What a Shield Can Does

A board level shield is a metal cover that sits over a group of components and is soldered or clipped to a grounded perimeter on the board. Its purpose is to contain the electromagnetic field of the circuits inside it, and to keep external fields out. The mechanism is straightforward: a conductive enclosure connected to ground reflects and absorbs the field, and currents induced on its inner surface are returned to ground rather than escaping.

Two different problems look the same from the outside and are often confused. Radiated emissions come from the circuit and leave the product, which is what a compliance test measures. Susceptibility is the reverse: an external field couples into the circuit and causes a functional failure. A shield addresses both, but the layout work that makes it effective is different, so it is worth deciding which one is being solved before the shield is added.

Shields are also used to keep two radios on the same board from interfering with each other. A Wi-Fi module next to a cellular module will desensitise it without any external field involved, and a small can between them is often cheaper than redesigning the layout.

Materials and Construction

Most board level shields are made from tin plated steel, which is inexpensive, solderable and reasonably stiff. Nickel silver and phosphor bronze appear where better conductivity or spring properties are needed, and copper alloy or metallised plastic covers show up in low cost consumer products where the shield is clipped rather than soldered.

The common construction is a two piece design: a frame that is soldered to the board and a cover that clips onto the frame. The advantage is access. Components inside can be probed, reworked or tuned by removing the cover, which matters for anything that needs alignment or that is likely to need repair. A one piece can is cheaper and takes less board area because there is no separate frame land, but once it is soldered the parts inside are effectively sealed.

Single piece cans with a removable lid, and cans designed to be picked and placed like any other component, are also common. Check that the top surface suits a nozzle, because a domed or perforated shield may need a manual operation.

Height is the constraint that decides most of these choices. A can must clear the tallest component inside it with enough margin for the pick and place placement tolerance, the solder paste collapse during reflow, and any thermal expansion of the assembly. Two piece designs usually need more height for the same internal clearance because the frame and the cover overlap.

Footprint and Grounding Rules

The footprint is where a shield succeeds or fails. The metal is only half the enclosure; the other half is the ground plane under it, and the connection between them has to be continuous rather than a few pads in the corners.

The first rule is a continuous ground land under the entire perimeter of the can, not a set of discrete pads. A gap in the land is a slot in the enclosure wall, and a slot radiates. Where the can has fingers or a castellated rim, the board land should still be continuous so that every finger lands on copper.

The second rule is stitching vias. Place a row of ground vias through the land, spaced closely enough that the gap between adjacent vias is a small fraction of the shortest wavelength of interest. The usual working figure is that the via spacing should be no more than about one twentieth of the wavelength at the highest frequency that matters, which for a 2.4 GHz radio means roughly 6 mm, and in practice designers use 2 mm to 3 mm on the perimeter to keep margin. The vias connect the perimeter land to the internal ground plane, which turns the plane into the floor of the enclosure.

The third rule is component placement. Keep the noisy or sensitive parts well inside the can and away from the perimeter, so that the field has to travel before it reaches a seam. Keep traces that leave the shielded area short and, where possible, route them over ground, and filter the ones that must cross the boundary. A shield with twenty unfiltered signal traces passing through the perimeter is a shield with twenty slots in it.

Apertures, Vents and Openings

Every enclosure needs openings for at least the signals leaving it, and sometimes for ventilation or for a camera or a sensor. Each opening interacts with the field, and the rule that matters is the relationship between the size of the opening and the wavelength.

A long narrow opening behaves like a slot antenna and radiates efficiently when its length approaches half a wavelength. A round hole radiates much less effectively for the same area, and a group of small holes radiates less than one large hole of the same total area. That is the basis of the honeycomb pattern used on vented shields: many small holes, each well below the wavelength of interest, give airflow with far less leakage than a single large opening.

The direction of the slot also matters. A slot couples strongly to a field whose electric field is aligned across the narrow dimension, so a slot that runs parallel to the current flow on the enclosure surface is less of a problem than one that cuts across it. When a design has to have an opening, thinking about which way the current wants to flow on the can surface usually suggests the less harmful orientation.

For most boards operating below a few gigahertz, an aperture smaller than about 5 mm in its longest dimension is acceptable, and much smaller than that if the frequency is higher. Where a connector has to pass through the perimeter, the shield should be designed around the connector body so the opening is as small as the part allows, and the connector shell itself should be grounded to the board land.

Assembly and Reflow

A shield can is a large metal part on a small board, and that combination produces its own assembly problems.

The first is thermal mass. The can draws heat away from the perimeter during reflow, so the joints under it may not reach the liquidus temperature if the profile is set for the surrounding components. The usual fix is a longer soak and a slightly higher peak, with a thermocouple attached to the can rather than to a coupon. A joint that looks formed but never fully reflowed will crack later.

The second is paste volume. The perimeter land is long and narrow, and there is a tendency to under print the paste because the opening is thin. Most suppliers recommend a stencil opening that matches the land, and in some cases a small step or a thicker stencil in that region, because a joint with insufficient solder has no fillet and no fatigue life.

The third is placement accuracy. A can placed a fraction of a millimetre off centre can short to an adjacent pad, or can sit on top of a component inside it. That is why the components inside should be kept clear of the perimeter by a margin that covers the placement tolerance, and why the can should ideally have a locating feature or a chamfer that helps it self centre.

Inspection and rework also need planning. A soldered can hides everything inside it, so an X-ray or an impedance test that relies on probing is not available afterwards. If a design uses a two piece shield, the frame can be soldered and the board tested before the cover is fitted, which is a good reason to pay the small extra cost of the two piece design on anything complex enough to need debug.

When a Shield Is Not the Right Answer

A can is a fix at the boundary, and boundary fixes are expensive. If a clock trace is radiating because it is long, unfiltered and routed over a split plane, shielding the whole area hides the symptom while the underlying emission still couples into adjacent traces. The cheaper sequence is to shorten the return path, slow the edge rate where the timing allows, add a series termination, and keep the plane solid, then add the shield only if the measurement still fails.

Grounding quality is the usual limiting factor. A shield connected to a ground plane that is itself noisy, or connected through a long thin trace, will re-radiate the noise it was meant to contain. The perimeter land has to be a low impedance connection all the way round, not a single point, which is why the stitching vias matter as much as the can.

There is also a mechanical cost. A soldered can makes rework harder and adds height, weight and a part to the bill of materials, and it can interfere with a heatsink or a case. Where the radio already sits inside a metal enclosure, an internal can may add nothing at all.

Cost and Sourcing

A standard shield can is a low cost part: a stamped frame and cover in tin plated steel is among the cheapest components on most boards, and tooling is amortised quickly because the shapes are standardised. Custom shapes and custom heights cost more, and a can that has to be drawn rather than stamped costs more again.

The real cost driver is the board area around it. The perimeter land, the stitching vias, the component keepout and the height clearance all consume space that would otherwise carry routing, and on a dense board that is what makes engineers try to avoid the shield. For that reason it is worth deciding early, while the layout is still fluid, rather than trying to fit a can into a finished design.

FAQ

• Does a shield have to be soldered? No. Clipped or socketed shields exist and are easier to service, but the electrical connection is less reliable and the design has to control the contact force.
• How many stitching vias are enough? Enough that the gap between adjacent vias is a small fraction of the wavelength at the highest frequency of concern. On a radio board, 2 mm to 3 mm spacing is a common default.
• Can a shield be placed after reflow? Yes, with a socket or a clip, or as a manual assembly step. Machine placement during reflow is normally cheaper in volume.
• Do I need a shield on both sides of the board? Only if the emission or the sensitivity is on both sides. Shielding a single side is much more common and much cheaper.

Summary

A shield can works when the board underneath is designed to be half of the enclosure. That means a continuous grounded land around the perimeter, stitching vias tying that land into a solid plane, components kept clear of the seam, filtered traces crossing the boundary, and openings kept small relative to the wavelength. Assembly then has to solve the thermals of a large metal part on a small board, with enough paste on the perimeter and a profile that actually reflows the joints.

Get those elements right and a shield is a cheap, predictable fix that also meets the thermal and mechanical requirements of the assembly process. Get them wrong and it is an expensive part that hides a problem. The work belongs at the start, during layout and the capabilities review with the fabricator, together with the plan for how the boards will be tested once the cans are fitted and the circuitry is no longer accessible.