Last month a client's wireless communication module failed EMC testing in the lab. Radiated emissions exceeded limits at several points around 900 MHz, and the first line of the corrective-action sheet read: "enclosure shielding is discontinuous."
This class of problem is unusually easy to localise structurally. When shielding effectiveness drops, eight or nine times out of ten it drops at a seam or a mating surface. The material itself is rarely the culprit. What follows is the fixed routine we use on these jobs, in three parts.
Part 1: Work Out the Wavelength First, Then Set the Seam and Mating Width
The first step in shield-cavity design is to establish which frequency band you are defending against. The rule of thumb in engineering practice is that leakage becomes noticeable once the long edge of a seam reaches 1/20 of the wavelength at the frequency in question. This is the common approach derived from EMC textbooks and the test bands covered by standards such as GJB 151B and CISPR 32. At 1 GHz the wavelength is 300 mm, so 1/20 is 15 mm. If you see an enclosure seam longer than 10 mm that also runs right alongside a high-speed trace or a crystal oscillator on the PCB, that is almost certainly your leak.
In structural terms, we control this with three moves.
First, make the cavity's mating surface a continuous full-perimeter contact, with a width of no less than 1 mm. That width is the lower bound for giving conductive foam or a metal finger gasket room to seat plus tolerance margin. In many projects the failure is that structural parts contact each other only at a few snap points with a gap in between, and that gap is the leak during testing. Wherever there is an opening, such as a button, a connector, or a vent, it must be handled with metal mesh, a perforated shield can, or a waveguide-below-cutoff hole array. The cutoff design rules for a perforated shield are well established: hole diameter and hole pitch must be far smaller than the working wavelength, and while the array can contain many holes, no individual hole may be large. For vents we commonly design the openings as narrow slots whose long axis runs perpendicular to the dominant radiation direction, and edge them with a conductive gasket or metal spring.
Second, the inner surface of the cavity must be conductive. For plastic enclosures used as shielding, the usual approach is metallic coating (EMI conductive paint) or vacuum metallisation, plus a metal shield can. For aluminium cavities, conductive anodising, electroless nickel plating, or tin plating are common. There is a trap here that we have walked into: a client switched the cavity interior's surface treatment from conductive anodising to ordinary anodising during trial production, purely for appearance consistency. Shielding effectiveness collapsed, because ordinary anodising is insulating, so the mating surface effectively has a resistive film across it. We now write this into the surface-treatment technical requirements on the drawing, annotated "insulating anodising or painting is not permitted on shielding mating surfaces."
Third, roughness and compression allowance have to be considered together. Machined aluminium faces at Ra 1.6 and painted faces at Ra 3.2 are the reference values we commonly use on production projects; only after compression do you get enough contact area between the conductive foam and the metal part. A shielding effectiveness of 60 dB corresponds to field-strength attenuation of a factor of 1000, which sounds like a lot, but a single resistive film or gap on the mating surface immediately knocks tens of dB off that figure. The cavity's own material thickness and rigidity therefore have to be sufficient, and it must not visibly warp under compression.
Part 2: Conductive Foam - Compression Is the Core, and Cavity Depth Is Back-Calculated From the Tolerance Chain
Conductive foam is the most common shielding gasket because it absorbs assembly tolerances and is inexpensive. Its selection and structural parameters come down to three things: compression, contact resistance, and environmental durability.
On compression, mainstream manufacturers' datasheets recommend a range of 25% to 50%. We typically set cavity depth at around 30% and then add assembly tolerance and material creep allowance on top. Compress it too lightly and the foam makes only point contact with the metal part: shielding effectiveness suffers and contact resistance runs high. Compress it too heavily and the foam's cell structure is crushed, so rebound force decays over the years until it is effectively gone. On one part we designed foam compression at around 12%. It passed testing. Half a year later, re-measurement showed shielding effectiveness at those points had dropped markedly, and the foam would no longer spring back. Since then, for any long-life product, we bias compression toward the middle-to-upper part of the range and require the foam supplier to provide compression set data. At 30%-50% compression, a permanent set within 15% is considered reasonable.
The cavity depth calculation is straightforward: foam free height multiplied by (1 minus the design compression ratio), then fold in structural tolerance, foam thickness tolerance, and surface roughness valleys to arrive at a total allowance. Our habit is to make the cavity 0.3-0.5 mm deeper than the theoretical value, then trial-fit foams of different heights to find the actual compression ratio. Foam positioning also matters: it must not be scraped during assembly, and adhesive dots must not land on the conductive face, because ordinary tape is insulating.
On contact resistance, conductive foam gaskets typically achieve a mating resistance on the order of 0.1 ohm. For critical shielding points we use a four-wire measurement to eliminate lead resistance. The foam's outer material must match the mating metal: if the mating surface is aluminium, the foam jacket is usually nickel-plated copper or aluminium-plated conductive cloth; if the mating surface is tin-plated, beware of galvanic corrosion between dissimilar metals, especially where the product must pass salt-spray testing. On an outdoor device we once selected a copper-plated foam jacket against a galvanised steel mating surface. After salt spray, contact resistance rose by an order of magnitude. Switching to a nickel-plated material was what finally passed.
Part 3: Grounding Springs - Look at Compression Travel, Plating, and the Tolerance Chain Together
The spring's job is to establish a reliable low-resistance bond between the PCB, the metal enclosure, and the shield can while absorbing tolerances. It is harder than foam, offers lower contact resistance, but is also more prone to failure when tolerances are miscalculated.
Spring selection is mainly about material and plating. Beryllium copper or phosphor bronze gives good springback and long fatigue life; stainless steel saves cost but trails copper alloys in springback force and contact resistance. Common platings are gold-over-nickel, tin, and silver. Gold-over-nickel resists oxidation and salt spray well at higher cost, so we reserve it for critical shielding bond points. Tin plating needs care: it oxidises readily after insertion and withdrawal wear, and loosens under long-term vibration. For working travel (the spring's compression) we usually take 20%-30% of the spring's free height. That keeps the spring in its effective working range without fatiguing it, though the manufacturer's force-deflection curve should always be checked. Keep the spring's load direction as close to perpendicular to the mating surface as possible; skewed compression makes contact resistance wander badly. Allowing roughly 0.2 mm of float in the spring's seating slot is common practice.
One more point that is easy to overlook: the foam, the spring, and the metal cavity must form a single continuous conductive path, with no insulator trapped anywhere along it. We revised the battery compartment of a blood-pressure monitor three times. The root cause was that in the first version a plastic battery retainer sat exactly between the shield cavity and the spring, which severed the grounding path. This kind of problem is invisible in CAD. It can only be found by checking the bond path segment by segment: from PCB ground to the spring, from the spring to the cavity, from the cavity to the enclosure. Every segment must conduct, and the contact resistance at each point must be measurable.
A final note on the test side: finishing the structural change does not mean the problem is solved. Re-test with the same setup and the same turntable, because antenna position and cable routing strongly affect the result. Many "the change made no difference" conclusions are actually cases where the test setup changed. Our internal process is to photograph the original setup and cable routing before the fix and reproduce them exactly on re-test, so that the effect of the structural change can be attributed cleanly.
FAQ: Shield Cavities and Conductive Gaskets
Q: How long can a shield-cavity seam safely be?
A: The common engineering rule is that the seam's long edge should not exceed 1/20 of the wavelength at the highest test frequency; at 1 GHz that is 15 mm. In product design we usually add margin and target between 1/20 and 1/50. The calculation is 300 mm divided by frequency in GHz, then divided by 20 or 50. This rule derives from EMC textbooks and the test bands covered by standards such as GJB 151B and CISPR 32. In practice, also factor in where the exceedance is: seams near the offending frequency get priority.
Q: What compression should conductive foam use?
A: Mainstream datasheets recommend 25%-50%. Structural design typically sets cavity depth at around 30% compression and leaves 0.3-0.5 mm for assembly tolerance and long-term creep. Too little compression raises contact resistance and compromises shielding; too much destroys the cell structure and decays rebound force over time. For long-service products, also require compression set data from the supplier.
Q: What contact resistance is acceptable for a grounding spring?
A: Most product specifications put shielding bond points on the order of 0.1 ohm; the exact figure follows the project's EMC specification. Critical points should be measured with the four-wire method. Elevated bonding resistance usually comes from insufficient compression, skewed load direction, oxidised plating, or an insulator trapped in the path.
Q: Can an aluminium cavity's interior use ordinary anodising?
A: No. Shield cavity interiors and mating surfaces must not use ordinary anodising or paint, because those layers are insulating and will substantially degrade shielding effectiveness. Use conductive anodising, electroless nickel plating, or tin plating instead. We have seen a production project switch surface treatment during trial production for appearance consistency, and every bonding point failed as a result. Drawings should call out the shielding mating surface's treatment separately and state that insulating treatment is prohibited.
Q: How do you choose between conductive foam and a metal spring?
A: Foam suits long seams and irregular mating faces and absorbs assembly tolerances. Metal springs suit flat mating and points with high-frequency shielding requirements; they offer lower contact resistance but are more tolerance-sensitive. Selection must also consider the environment: outdoor products or those with salt-spray requirements need matched plating to avoid galvanic corrosion between dissimilar metals, and products subject to long-term vibration need materials with slow rebound decay plus a check of the spring's fatigue curve.
Top comments (0)