RF Shielding: Why a Microwave Door Stops Waves but Not Light

Last updated: 12 August 2026

RF shielding is not decided by how thick the metal is. It is decided by the size of the holes in it. A microwave oven door blocks 12 cm microwaves using a mesh of 2 mm holes, while letting visible light straight through the same holes. A wave will not pass an aperture much smaller than itself. That single rule is most of what RF shielding is, and it explains the oven door as well as why a sensor sealed inside a steel cabinet often cannot reach its gateway.

Why can you see into a microwave but the microwaves cannot get out?

Because the two waves are wildly different sizes, and the door is built to exploit that. A domestic oven cooks at 2.45 GHz, which is a wave about 12.2 centimetres long. Visible light is also an electromagnetic wave, but its wavelength is around 0.0005 millimetres. The mesh you can see in the door glass has holes roughly 1 to 2 millimetres across. Those holes are tiny compared with a microwave and enormous compared with light.

So the microwave arrives at a 2 mm hole, finds nothing it can couple into, and is reflected back into the cavity. Light arrives at the same hole, finds an opening thousands of times wider than itself, and passes through untouched. One sheet of perforated metal, two completely opposite answers, and the only variable that changed is the size of the wave doing the asking.

That perforated sheet is a Faraday cage. The mathematics of how well a mesh screens a field is well studied, and the headline is consistent: RF shielding is set by the largest opening, not by the material budget.

How big does a hole have to be before it leaks RF?

The working rule engineers use is that an aperture starts to leak meaningfully once its longest dimension approaches about one twentieth of the wavelength. Below that the panel behaves like a wall and RF shielding holds. Above it, the opening stops being a hole and starts behaving like an antenna, which is the same physics that makes a slot antenna work in the first place.

Run the numbers for the bands industrial equipment actually uses:

• 868 MHz (LoRaWAN in the UK and Europe): wavelength 345 mm, so leakage becomes significant around 17 mm.
• 2.4 GHz (Wi-Fi, Bluetooth, and microwave ovens): wavelength 122 mm, leakage around 6 mm.
• 5 GHz (Wi-Fi): wavelength 60 mm, leakage around 3 mm.

A door seam, an oversized cable gland, a ventilation louvre or a keyhole is comfortably past every one of those thresholds. This is why RF shielding starts at the drawing of the enclosure rather than at the specification of the steel. Every other decision is detail by comparison.

One detail catches people out when they estimate RF shielding by eye. It is the longest dimension that governs leakage, not the area. A long thin slot leaks far more than a round hole of the same open area, which is why a 200 mm door seam is a much worse offender than a scattering of small vents.

Does thicker metal improve RF shielding?

Almost never, in this frequency range. The depth to which a radio field penetrates a conductor is called the skin depth, and at 868 MHz in steel it is a small fraction of a millimetre. A 1.5 mm panel is already hundreds of skin depths thick. Doubling it to 3 mm changes nothing you could measure with any instrument you would take to site.

This is worth knowing before anyone specifies a heavier gauge to fix a connectivity problem, because heavier gauge buys no RF shielding at all. The wall was never leaking. Effective RF shielding comes from continuity: conductive gaskets on doors, proper bonding between panels, waveguide-below-cutoff vents instead of plain louvres, and glands that maintain the shield around the cable rather than simply making a hole for it.

Why does a sealed IoT enclosure still lose signal?

Because it is almost never sealed at radio frequencies, and the RF shielding it does have was not designed by anyone. A fabricated steel panel has a door gap, a gland plate, a few unused knockouts and whatever tolerance the sheet metal shop worked to. The device inside is communicating through that unplanned collection of apertures.

Two consequences follow, and they pull in opposite directions. The first is that the link budget of a device in a cabinet is a property of the cabinet’s RF shielding, not of the device. Move the same sensor into a different panel on the same site and the result changes. The second is that the same apertures let interference in, so a partially shielded box is not reliably quieter than an open one.

Neither is under your control once the door is shut, which is the honest argument for putting the antenna outside the enclosure. An external antenna on a short, well-terminated feed removes the enclosure from the RF path altogether, and it does so for a fraction of the cost of chasing the problem with a more powerful radio. We cover the wider question of where signal actually goes in our guide to buying LoRaWAN sensors in the UK.

What actually fixes an enclosure that blocks the signal?

In the order worth trying:

1. Move the antenna outside. This takes the enclosure’s RF shielding out of the path rather than compensating for it, and it is usually the cheapest option on the list.
2. Move the device, not the antenna. If the sensor genuinely has to live inside, mounting it near a non-metallic window, a plastic gland plate or an open face can be worth 10 to 20 dB.
3. Fix the geometry outside the box. Raising a gateway two metres often beats every change you can make at the radio, because clearance is worth more than power.
4. Only then look at the radio. Transmit power is capped by regulation anyway. In the UK the 868 MHz band is governed by Ofcom’s short-range device conditions, so there is far less headroom available than people expect.

How do you tell whether an enclosure is the problem?

Test it rather than reason about it. Put the device in its intended position, note the received signal strength reported by the network server over a few uplinks, then repeat with the door open and again with the device sitting on top of the cabinet. Those three numbers bracket the answer in about ten minutes.

A door that costs you nothing means the RF shielding was never continuous and the enclosure is not your problem. A difference of 15 dB or more between open and closed means it is. The step from inside to on top tells you what an external antenna would recover, which turns the fix into a number rather than an argument.

Do it before the install rather than after, on the worst cabinet on site rather than the nearest one. RF shielding varies enormously between panels that look identical, because the variable is fabrication tolerance rather than design.

Why the oven door is the right mental model

The microwave door is worth keeping in mind because it is the one piece of deliberate RF shielding almost everyone owns, and because it makes the counterintuitive part obvious. Nobody looks at that door and concludes the glass is thick. The engineering is entirely in the mesh, and the mesh works because somebody matched the aperture to the wavelength on purpose.

Industrial enclosures are the same problem solved by accident. The metal is fine. The gaps decide the outcome, and nobody drew them with a wavelength in mind. Once you know the threshold for your band, walking up to a cabinet and looking at its seams tells you more about whether the link will hold than any figure on the device datasheet.

Frequently asked questions

Does a metal cabinet completely block LoRaWAN?

Rarely completely, but it can easily cost enough signal to break a marginal link. A fully continuous, gasketed enclosure will block it. A typical fabricated panel leaks through door seams and glands, so the device usually still works, unpredictably, and the result differs from cabinet to cabinet.

Is aluminium or steel better for RF shielding?

At 868 MHz and above, both are far beyond the skin depth at any practical gauge, so material choice makes no meaningful difference to RF shielding. Choose on cost, corrosion and mechanical grounds, then spend the attention on joints, gaskets and apertures.

Will a bigger antenna fix a sensor inside a cabinet?

Not usually. Antenna gain is directional, not a general amplifier, and it cannot recover a signal that has been reflected by the enclosure wall. Moving a modest antenna outside the box beats a high-gain antenna inside it almost every time.

How do ventilation openings avoid leaking RF?

Enclosures built for real RF shielding use honeycomb vents, which are effectively a bundle of short tubes acting as waveguides below their cutoff frequency. Air passes, radio does not. A plain rectangular louvre offers no such protection.

What frequency does a microwave oven use, and why that one?

2.45 GHz, one of the internationally designated industrial, scientific and medical bands. It is a licence-exempt allocation rather than a magic resonance of water, which is a common misconception. Wi-Fi and Bluetooth share the same band, which is why an oven can briefly disturb a Wi-Fi link.

Where this leaves you

RF shielding is a geometry problem wearing a materials problem’s clothes, and it is settled long before anyone powers the device on. Before anyone orders a thicker panel or a stronger radio, measure the longest gap in the enclosure and compare it with a twentieth of the wavelength. That one comparison predicts the outcome better than the rest of the specification put together.

If you are choosing hardware for a site where enclosures, distance or interference are in play, our level sensor comparison and IoT controller comparison both cover the mounting and antenna decisions alongside the specifications. Or talk to us about a site that is not behaving.