Why board-level shields fail silently

A board-level EMI shield looks simple: a stamped-metal enclosure soldered over a noisy circuit. The reality is a precision electromagnetic structure where four independent failure modes can degrade shielding effectiveness (SE) from a designed-for 40 dB to a measured 12 dB — with no visible defect. This issue is the field guide: what the failure modes are, how to specify against them, and how to read a shield design before it reaches the chamber.

Form factor: one-piece vs. two-piece

One-piece shields (formed cans) are stamped in a single operation, soldered directly to a ground pour, and offer the best SE-per-dollar for circuits that do not require field access. The solder joint is continuous — or can be made continuous — which is the highest-performing seam geometry. They are optimal for production EMI containment and are the low-cost default.

Two-piece shields (frame + removable lid) sacrifice a small amount of SE at the lid-to-frame joint in exchange for access. The lid snaps or clips to the frame; the contact interface is a controlled-pressure mechanical joint, not solder. Properly specified — with a tin-plated lid surface and a compliant contact geometry that maintains > 100 g/cm of contact force across the lid perimeter — two-piece shields deliver 30–38 dB from 100 MHz to 6 GHz. They are the right choice for any module that will be tuned, reprogrammed, or accessed in the field.

The decision rule is direct: if the circuit will be accessed post-assembly, specify two-piece. Otherwise, specify one-piece. Do not specify two-piece shields to gain SE over one-piece — the one-piece will win. Specify them for access, then design the contact geometry to minimize SE penalty.

Material science

The material governs three things independently: skin depth (and therefore SE at a given frequency — though wall thickness renders skin depth irrelevant for practical gauges, see Design Brief #011), solderability (which governs fence-to-ground joint quality, the actual SE limiter), and spring force (which governs lid contact pressure in two-piece designs). Specify all three; do not let a single-material compromise all three simultaneously.

In the current metals environment — copper at $13,335/t, tin at $51,613/t, nickel at $18,985/t — tin plating thickness is a real cost lever. A 2 µm tin finish provides adequate solderability and corrosion protection; 10 µm is waste at today's tin price. Specify 2–5 µm.

The fence-pitch rule and why it governs SE

The SE of a board-level shield above 500 MHz is governed almost entirely by the fence-to-ground solder joint. The fence is the perimeter frame that solders to the PCB ground pour; if the solder joint is interrupted, the gap radiates. The physics:

SE_seam = 20·log₁₀(λ/2L)

where L is the longest unsupported gap in the fence. At 1 GHz, λ = 30 cm; for 30 dB SE, L must be ≤ 9.5 mm. The standard specification is fence-to-ground contact at ≤ λ/20 — approximately 15 mm at 1 GHz. For designs requiring 6 GHz performance, this tightens to ≤ 2.5 mm.

The practical implication: the PCB land pattern under the fence must be uninterrupted copper — no signal traces crossing the fence line, no vias breaking the pour within 1 mm of the fence. The most common SE failure in production is a trace routed under the fence by a layout engineer who did not understand the ground-continuity requirement.

🚫Trace routing under the fence is an EMC defect

Every signal trace that crosses the shield fence line creates an aperture equal to the trace clearance from the ground pour. A 0.2 mm clearance at 5 GHz has SE limited to 20·log(30 mm / 0.4 mm) = 37.5 dB — and that is the best case, when the trace carries no CM current. In practice, any trace exiting a shielded enclosure is a potential coupling path and requires an EMI filter or CM choke at the exit point.

The four silent failure modes

Solder bridging and void formation. A fence with incomplete solder fill leaves a gap; paste volume, reflow profile, and stencil design all matter. Specify reflow inspection (AOI or X-ray) on the fence solder joint.
Via absence under the fence. A continuous ground pour that has no vias to the inner ground planes is a surface-only shield — effective for surface-wave coupling but transparent to volume-mode radiation from inner layers. Stitch vias at ≤ λ/10 along the fence line.
Lid contact pressure loss. In two-piece designs, the lid contact force degrades with thermal cycling (spring relaxation) and mechanical shock. Specify a contact material with low creep (BeCu or phosphor bronze), and verify contact force at the temperature extremes of the operating range.
Surface-finish incompatibility. A shield stamped from nickel-silver and reflowed over an ENIG (gold) finish forms a tin-gold intermetallic at the joint — stronger than tin-tin but potentially brittle at the board-can interface. Verify the finish stack is specified for the solder paste and profile; mismatch is common and produces intermittent SE loss that is invisible without cross-section analysis.

POCONS engineering and production

POCONS USA (San Diego) custom board-level shields are manufactured in Korea under IATF 16949 — stamped, plated, and 100% dimensionally inspected before shipment. Korean manufacturing means the fence pitch, aperture geometry, material gauge, and plating specification are chosen for your specific frequency range and galvanic environment, not selected from a catalog.

One thing

SE is set by the fence-to-ground joint, not the metal. Specify the fence pitch (≤ λ/20), the ground pour integrity (no trace crossings, via stitching at ≤ λ/10), and the plating stack (2–5 µm tin, galvanically matched). Everything else is secondary.