Board Level Shielding: Design Rules and Process Limits

Board level shielding puts a metal box around the part of a circuit that radiates or that is sensitive to radiation. It is a mechanical solution to an electrical problem, and it works only when the shield is bonded to a quiet reference with a low impedance at the frequency of interest. A can that is soldered to a noisy ground is decoration.

What a Shield Can Does

A board level shielding can is a conductive enclosure that surrounds a circuit and reflects and absorbs the fields it produces. The same enclosure also protects the circuit from fields that arrive from outside, so a shield serves both an emission and an immunity purpose.

The enclosure works by forcing the fields to induce a current in the metal, and that current returns through the ground connections. The shield is therefore part of the circuit, not an accessory, and its performance depends on how it is grounded rather than on the thickness of the metal.

The classic failure is a can that is grounded at one point, or through a long thin trace. The connection looks solid on a schematic and behaves as an inductor at the frequency of the radiation, so the shield becomes a resonator instead of a barrier.

Choosing the Shield Type

The one piece can is stamped from a sheet and soldered to pads on the board. It is cheap and it gives the best seam performance, but it cannot be removed without a rework station, so it is used where the circuit does not need to be adjusted.

The frame and lid type has a frame that is soldered to the board and a lid that snaps on. The lid can be removed for probing, which is a large advantage during development, and the seam between the frame and the lid is the price that is paid for it.

A printed shield is a copper area on an inner layer or a conductive coating on the housing, and it costs almost nothing. Its performance is lower because the copper is thin and the connections are far apart, but it is often enough for a low frequency problem.

The choice should follow the frequency and the amount of attenuation that is needed, not the cost of the part. A can that is oversized for the problem adds assembly steps, and it also traps heat around the components it covers.

Grounding and the Return Path

The shield is only as good as its ground. The current induced in the metal has to return to the source through a path with a low impedance, and at high frequency that means a wide, short connection rather than a single via.

A row of ground vias along the perimeter of the can creates a wall of low impedance connections, and the spacing between them sets the highest frequency at which the wall behaves as a solid barrier. The usual starting point is a spacing of a tenth of a wavelength at the highest frequency of interest.

The ground that the shield connects to must itself be quiet. If the can is bonded to a plane that carries the return current of a switching supply, the noise on the plane is coupled onto the shield and radiated from its surface, which makes the problem worse than it was without the can.

The principles are the same as those in the EMI suppression guide, and the layout work is the same work: keep the loop small and the reference solid.

Apertures and Seams

Every opening in a shield is an aperture, and every aperture leaks. A slot is far worse than a round hole of the same area, because the slot can act as a slot antenna when its length approaches a half wavelength.

Vents, cable entries and the gap between a frame and a lid are all apertures. The design rule is to make the longest dimension of any opening as short as possible, and to break a long opening into a row of small ones.

The seam between a frame and a lid is a special case, because it runs the whole perimeter. The lid has to make contact at many points along that seam, and a finger stock or a dimpled edge is used to guarantee the contact after the plating has aged.

An aperture that is unavoidable should be placed on the side of the can that faces away from the sensitive circuit, so that the leakage travels through the board before it escapes. The mixed signal immunity work should be done before the shield is designed, because the shield cannot fix a split ground.

Thermal and Mechanical Effects

A shield can traps heat. The components under it run hotter than the same components on an open board, and the temperature rise can be tens of degrees in still air. The effect has to be estimated before the can is specified, not discovered in a thermal chamber.

Metal cans conduct heat as well as electricity, so the shield can also be used as a heat spreader if it is bonded to a thermal pad on the board and to the housing above. That is a deliberate design decision and it requires a controlled gap filler rather than a snap fit.

Mechanically, the can adds stiffness and mass, and it changes the way the board behaves in a drop or a vibration test. A heavy can on a thin board can pull the board into a resonance that did not exist before, so the mechanical engineer should be part of the review.

Design and Verification

The layout has to reserve the space for the can and for its pads before the components are placed. A can that is added after the layout is finished will disturb the placement of every part near it, and the redesign costs more than the shield.

The pads should be on the ground plane with a row of stitching vias, and the pads should be wide enough to take the solder fillet. A narrow pad gives a joint that cracks in a thermal cycle and a shield that becomes intermittent.

Verification is done with a near field probe and a spectrum analyser before and after the can is fitted. The measurement shows whether the shield is doing the work or whether the problem is on a cable that leaves the enclosure, which is where a large fraction of real emissions come from.

A shield should never be the first fix for an emission problem. Filtering at the source, a smaller loop area and a better return path are cheaper, and the ferrite bead advice in the pool applies long before a metal box is justified.

FAQ

Does a shield can always reduce emissions? No. It reduces the fields produced inside it, but it can also couple noise onto a cable or radiate from its own surface if it is grounded to a noisy plane.

How many ground connections does a shield need? As many as the perimeter will take, spaced well under a tenth of a wavelength at the highest frequency of concern. A single connection is almost never enough.

Can a shield can be removed and refitted? Only the frame and lid type is designed for that. A one piece can has to be desoldered, and the pads and the board survive only a limited number of cycles.