When a shielded RF box fails an emissions or immunity test, the instinct is to ask for thicker metal. The walls are almost never the problem. Leakage happens through apertures: seams, vents, connector cut-outs, displays, and above all through cables that cross the shield. Some numbers to make that concrete.

1. Solid metal is an excellent shield
Absorption loss through a wall of thickness t is about A = 8.686 * t / delta dB, where delta is the skin depth.
- Copper at 1 MHz: delta about 66 um.
- Aluminium at 100 MHz: delta about 8.5 um, so a 1 mm wall gives over 1000 dB of absorption loss on paper.
Add reflection loss and a sealed metal box would exceed any test requirement many times over. Real enclosures reach 40 to 80 dB because of openings, not because of the material.

2. Slots: the longest dimension rules
A slot radiates like a slot antenna when its longest dimension approaches a half wavelength. A common first-order estimate for a single aperture of maximum length l (with l below lambda/2):
SE = 20 log10(lambda / (2 l)) dB

Examples:
- A 10 cm seam gap at 100 MHz: about 23.5 dB.
- The same 10 cm gap at 1 GHz: about 3.5 dB, effectively no shielding.
- A 5 cm gap at 1 GHz: about 9.5 dB.
Width matters far less than length. A long, hairline-thin gap between a lid and its base leaks almost as much as an open slot of the same length.

Multiple apertures add up. For n similar apertures close together, a common rule is that SE drops by roughly 10 log10(n) dB, so many small holes are better than one big one, but not free.

3. Seams and fastener spacing
On a bolted lid the effective slot length is roughly the screw spacing, because the surfaces only make reliable contact near each screw. To hold 20 dB at a frequency you need l below about lambda/20: at 1 GHz that means about 1.5 cm between contact points, and at 3 GHz about 5 mm. That is why RF lids use dense screws, spring fingers, conductive elastomer or knitted mesh gaskets.

Watch the surface finish. Standard anodizing is an insulator, so an anodized aluminium lid makes contact only where screws cut through it. Use a conductive conversion coating or mask the mating surfaces, and keep dissimilar metals in mind for corrosion.

4. Vents: waveguide below cutoff
Ventilation holes can be designed as short waveguides that cannot propagate below their cutoff frequency.
- Circular hole of diameter d: cutoff fc = 1.841 * c / (pi * d), about 17.6 GHz / d(cm). A 5 mm hole cuts off around 35 GHz.
- Well below cutoff, attenuation is about 32 dB per diameter of length for a circular tube, and about 27 dB per width for a rectangular one.
So a honeycomb vent whose cells are as deep as they are wide gives roughly 30 dB per cell at frequencies well below the cell's cutoff (attenuation falls off as you approach fc), and a vent made of thin perforated sheet (depth much smaller than hole size) gets only the much weaker aperture-array benefit.

5. Cables: the biggest aperture of all
An unfiltered conductor passing through a shield wall carries interference straight through, and on the outside it becomes an antenna. A shielded enclosure with an unfiltered DC cable can easily perform worse than an unshielded one.
- Filter every non-RF line at the point of entry: feedthrough capacitors or filtered connectors mounted in the wall, not on a PCB a few centimetres inside.
- Bond cable shields 360 degrees to the enclosure through the connector backshell. A pigtail of a few centimetres adds inductance that ruins shield performance in the hundreds of MHz.
- Keep RF connectors' outer conductors bonded directly to the wall, with no paint or anodize in the contact path.

6. Cavity resonances inside the box
A closed metal box is also a resonant cavity. For a rectangular box of dimensions a x b x d:
f_mnp = (c/2) * sqrt((m/a)^2 + (n/b)^2 + (p/d)^2)
A 30 x 20 x 5 cm box has its first resonance (TE110) near 900 MHz, with the next modes around 1.25 and 1.58 GHz. If your circuit has gain or spurs near these frequencies, coupling between sections can increase sharply. Fixes: internal walls that make smaller cavities, absorber sheet on the lid, or moving the resonance with geometry.

7. Practical checklist
- List every opening and its longest dimension, and compare it to lambda/2 at the highest frequency of concern (including harmonics of clocks and the LO).
- Treat seam spacing as an aperture length.
- Design vents as cutoff waveguides when the frequency range is high.
- Filter or shield every cable at the wall.
- Do a near-field probe sweep around seams and connectors on the first prototype; leaks show up quickly.

Interested to hear what gasket types people have found reliable over many lid openings.