Why a radio still works inside a metal enclosure

Short answer: the steel is not leaking. At 2.4 GHz a continuous stainless wall blocks more than 100 dB, so no Bluetooth signal passes through the sheet metal itself. Your link survives because RF escapes wherever the box is not continuous: a seam or cutout wider than about 6 mm, or a cable that crosses the wall and re-radiates the signal outside. A working panel with an HMI window, a handful of connectors and a door seam was never a closed shield. Find the path that carries your link, then close it or design for it.

How much signal does a solid steel wall actually stop?

More than 100 dB: at 2.4 GHz the skin depth in 304 stainless is about 9 μm, so a 1.5 mm (16 gauge) wall is roughly 170 skin depths deep, and reflection plus absorption leave nothing a Bluetooth receiver can hear.

Skin depth is how far into a conductor the field travels before it has decayed to 1/e, about 8.7 dB. Copper manages about 1.3 μm in this band; non-magnetic stainless about 9 μm. Multiply roughly 170 skin depths by 8.7 dB and the absorption alone reaches four figures on paper. Surface reflections reduce the measured number, but any continuous sheet still ends up past 100 dB.

The practical conclusion saves you wasted effort: the wall never leaks. If the link gets out, the signal found a way around the metal, through an opening or along a conductor, and that is where the diagnosis belongs.

Why does the link still close at 50 ft?

Because the leak costs tens of dB, not 100: a 25 dB leak against a 33 dB link-budget margin at 50 ft still leaves 8 dB to spare, which is enough for the scanner to hold the connection.

Run the budget for a typical low-energy radio and the numbers explain the whole puzzle. Take a common BLE dongle and scanner pair: 0 dBm (1 mW) output and a -97 dBm receiver sensitivity, both ordinary datasheet values.

• Link budget: 0 dBm - (-97 dBm) = 97 dB.
• Free-space loss at 50 ft [15 m], 2400 MHz: 32.44 + 20 log10(0.015 km) + 20 log10(2400 MHz) = 63.6 dB, call it 64 dB.
• No enclosure: received level -64 dBm, margin 33 dB.
• One leak path costing 25 dB (a 6 mm seam at 20 dB plus 5 dB of cable loss): -64 - 25 = -89 dBm. Margin 8 dB, link holds.
• Walk to 100 ft [30 m]: path loss rises 6 dB, received -95 dBm, margin 2 dB. The link now drops whenever someone crosses the aisle.
• Through the wall alone, past 100 dB of shielding: -164 dBm. That is 67 dB below sensitivity; the scanner hears nothing.

For scale, Bluetooth Classic is a 10 m radio in most Class 2 devices, and BLE designs quote 30 m indoors and up to 100 m line-of-sight (Bluetooth SIG range figures). So the honest summary is not that the box halves the signal. Half the field is 6 dB, and halving the range in free space also costs 6 dB. Neither describes what a real enclosure path does: leaks cost tens of dB, and the link lives or dies on the margin that survives.

Which openings leak, and how wide is too wide?

Any slot with its long side over about 6 mm starts to leak at this frequency: the wavelength is 125 mm, the classic EMC rule keeps apertures under one-twentieth of that, and a slot near half a wavelength (62 mm) leaks almost everything.

The standard slot equation, SE = 20 log10(λ / 2L), puts numbers on it (Henry Ott's EMC engineering reference). Slot shielding drops to zero as the length approaches half the wavelength, which is why the table reads the way it does.

Three habits follow from the table. Holes spaced more than 62 mm apart leak independently and their powers add, so five 13 mm indicator holes lose another 7 dB (10 log10 5). A long slot leaks by its length, not its area, so a hairline seam 150 mm long behaves like a wide hole of the same length. And plastic is invisible to RF, so a viewing window is an open aperture the size of the display; an HMI in the door is a hole by design.

An elastomer door gasket seals water and does nothing at this frequency — only metal-to-metal contact closes a seam. Mesh screens obey the same 6 mm limit, and coarser mesh leaks at every opening. The longer version of this table, with the seam geometry, sits in how much RF leaks through enclosure holes and seams, and the cutouts, glands and HMI windows guide covers specifying them.

How does wiring carry RF out of a sealed box?

The dongle's field couples onto any conductor that leaves the box, and the outside run re-radiates it like a passive repeater: a stub as short as a quarter wavelength (31 mm on this band) is already an efficient antenna.

Inside the cabinet, the radio's near field wraps every cable. The part that crosses the wall as common-mode current keeps flowing on the outside run, and that run broadcasts. The 0.5 m USB tail on a typical dongle is far past the 31 mm where this starts. The leak is bidirectional too, so the same cable drags outside noise back in.

A metal-bodied gland closes its own hole, which is why the gland is rarely the leak; the cable through it is. This mechanism explains the strangest field reports: the link appears the day someone adds an external USB lead, and the signal moves when a bundle is re-routed. Bonding and isolation choices around that wiring are covered in isolate a metal enclosure from the signal path.

What does each symptom tell you to look at?

Five symptoms cover nearly every case, and each points at a different path: range swings with people and stock mean the margin is spent, while a link that appears only when a cable leaves the box means the wiring is doing the broadcasting.

• Range collapses from datasheet distances to 5-10 m, but never dies. Many small leaks together; the margin is spent but not exhausted.
• RSSI jumps when the door is touched or a bundle is moved. A seam or the wiring; the wall is never this responsive.
• Signal strongest at the hinge side, the door seam, or in front of the HMI window. The long slot or the window is acting as the antenna.
• Link worked until someone ran a USB or serial lead out of the box. Cable re-radiation, the passive-repeater path above.
• Interference to a second radio grew as connectors were added. Aperture count is up; each doubling of similar holes costs about 3 dB (10 log10 2).

How do you find the leak path?

Start with the cleanest test there is: power the dongle from an isolated supply inside the closed box with no conductor crossing the wall, and compare that range with the installed one — the difference is your leak.

1. Kill the cable path first. If the range collapses to nothing without the leads, the wiring was the path. If it survives, move to the apertures.
2. Test the window. Hold a bare stainless lid flat against the inside of the HMI window. A large drop marks the display cutout as the aperture.
3. Test the seam. Watch RSSI while you press the door, then while a helper clamps a braided strap across the seam. A 10 dB improvement marks the slot.
4. Count the remaining holes. Anything over 6 mm contributes; combine them with 10 log10 N for holes spaced over 62 mm apart.
5. Measure, then change one thing at a time. Log RSSI at fixed points — 5 m and 15 m on the same line — before and after each change, so you see the decibels rather than pass or fail.

Lock out the panel and verify absence of voltage before opening it to swap feeds or fit ferrites.

What actually fixes it?

Move the radio out of the box, or give it one designed feedthrough: an antenna on a bulkhead connector keeps the signal inside the coax while the enclosure stays a closed metal box, buying back every dB the seams and cables were leaking.

• Radio outside. A serial-to-Bluetooth converter or a panel PC with its own radio, mounted outside the enclosure, costs no dB and no new penetrations beyond its power feed. The placement options are walked through in radio signal through a metal enclosure: design options.
• One bulkhead antenna feedthrough. You drill one hole; the coax carries the signal inside its shield and the antenna radiates in the open. Running that lead alongside power is covered in coax and power in one weatherproof junction box, and bulkhead cable gland vs one gland for removable outdoor cables compares the entry hardware.
• Radio must stay inside? Clip a ferrite on every cable at the wall crossing; two or three turns through a clip-on core add tens to hundreds of ohms of common-mode impedance in this band. Keep outside stubs under 31 mm, and bond the door with continuous metal-to-metal contact — even a 31 mm fastener pitch is worth only 6 dB.
• Skip the dead ends. More wall thickness adds dB you already have, and an earth strap changes static, not shielding.

When should the enclosure be re-specified instead of patched?

Re-spec when the leak list is the layout itself: an HMI window in the door, more than about ten penetrations, or a required range past 30 m with the radio trapped inside — then one bulkhead feedthrough and a smaller window belong on the drawing, not ferrites.

Patch when one or two paths dominate and the fix is hardware: a ferrite, a strap, a re-routed lead. Re-spec when the fix would be geometry.

Planning a radio board's size, cutouts and entries before ordering is exactly what enclosure for a router board walks through. One line to carry out of this article: a metal enclosure is a fine cabinet and a poor secret-keeper, so plan the radio's path the same day you plan the cutouts.

Catalog dimensions to compare

Compare these catalog dimensions with your component layout. Open each enclosure for its rating, material and usable space, then check those specifications against your application.

Questions