How to test whether an enclosure actually blocks signal
Putting a radio inside a metal box and calling it shielded is the easy part. The useful question is how much signal still gets through, at which frequencies, and who measured it. The method follows: instruments, setup, the mistakes that produce fake numbers, and how to read a test sheet. You can run all of it yourself, with nothing from anyone else.
What attenuation actually is
Attenuation is a comparison, not a property you can see. Measure how much signal arrives at a receive antenna with the source sitting in open air. Then put the source inside the enclosure, seal it, and measure again with everything else in the same place. The difference between the two readings, in decibels, is the attenuation of that enclosure at that frequency.
A decibel is a ratio, not a unit of power. Every ten decibels is a factor of ten; three decibels is roughly a halving. A reduction of thirty decibels means a thousandth of the power gets through. That is a large drop, and it is not zero. Nothing you or anyone else can build is a force field. A shielded enclosure reduces coupling: it makes the path harder, and it does so unevenly across the spectrum.
That unevenness is the whole point. The honest description of any enclosure is a curve, not one figure. Signal leaks worst at particular frequencies, and the weak spots are almost always the same — long seams, corners, feedthroughs, and any cable that crosses the wall.
If you are still working out what actually needs protecting, the preparedness audit is a better starting point than any product page.
Why a number with no method behind it is worthless
A bare figure leaves out the four things that make it mean anything: the frequency range it was measured over, the instrument used and its own noise floor, the arrangement of source and receiver, and whether a baseline run was recorded at all. Two enclosures measured different ways cannot be compared, even when both sellers are honest.
Ask three questions before you trust any attenuation claim:
- Over what frequency range, and at which frequency was the worst case?
- Measured with what instrument, in what setup?
- Was the baseline run recorded, and can I see it?
Sheets answer those questions. Memories don’t.
The two instruments that can do this
You do not need a lab bench. Two handheld instruments cover most of what a workshop or a kitchen table can use, and both are ordinary retail items.
| Instrument | Typical retail price | Good for | Watch for |
|---|---|---|---|
| TinySA Ultra | ~$299.99 | A spectrum analyser with a signal generator built in. It sweeps a band and draws the level it hears across the whole range, so you see the leaky frequencies instead of guessing at them. Its range reaches into the low microwave bands. | It hears everything nearby — FM broadcast, cell towers, Wi-Fi, a neighbour running a welder. Keep the baseline run and the enclosed run close together in time so the ambient noise is the same for both. |
| NanoVNA-H4 | ~$129.99 | A vector network analyser. It measures through-path loss directly and hands you the loss curve from one sweep, with no separate receiver to hold still. Covers the HF and VHF bands comfortably, and up to roughly 1.5 GHz. | Wants a clean two-port setup and a calibration at the ends of your own cables. Fussier wiring than the analyser route, steadier readings once it is set up. |
Either way, buy a fixed attenuator — about $40 — and leave it inline ahead of the instrument input. One that drops the signal by roughly a hundredfold is the usual choice. A receiver front end that meets a transmitter directly is a dead instrument, and that pad is the cheapest insurance in the exercise.
Running the test, step by step
- Put the source inside the enclosure, at the centre if the leads reach, with its antenna in its normal orientation.
- Set the receive antenna outside at a fixed distance — a metre or two is plenty. Mark the spot on the bench or the ground. It must not move between runs.
- Sweep the band with the source in open air and no enclosure. That trace is your baseline. Record it before you do anything else.
- Sweep again with the lid on and sealed exactly as it would arrive. Same instrument settings, same cables, same antenna positions.
- Subtract the baseline trace from the enclosed trace at every frequency. The difference is your attenuation curve.
- Repeat with the source rotated two or three ways and keep the worst result. Orientation changes the answer more than most people expect.
- Write down the date, the instrument, the range, the cable layout, and the frequencies where the curve was worst.
Budget half an hour for one enclosure. The slow part is discipline, not the instrument.
The ways people fool themselves
- A cable acting as an antenna. A coax shield that is not bonded where it crosses the wall becomes a leak path. The numbers can look good because the signal is leaving along the cable, not through the box.
- A gap in the lid seam. RF does not care that a seam looks tight. Long thin slots pass signal at wavelengths tied to their length, so a seam that reads as closed can still be an opening.
- A feedthrough with poor bonding. A plain hole with a wire through it is a small slot antenna. This is where the worst reading usually comes from.
- Measuring in the near field. An antenna held centimetres from the surface reads the field hugging the metal, not a signal that has travelled anywhere. Back off and run it again.
- A lid that only touches on two sides. Two sides in contact means the other two are open. Clamp or screw down every edge you can reach before you believe anything.
- Skipping the baseline. Without it you have a loudness reading, not attenuation. It is the most common mistake, and it makes every later number meaningless.
- Changing one thing between runs. Moved the antenna? Re-run the baseline. Swapped a cable? Re-run the baseline.
What a credible test sheet contains
A sheet worth trusting is short and specific. It carries the date; the instrument make and model and any attenuator used; the frequency range swept; the worst-case reading across that range; the frequencies where that worst case fell; the source and antenna positions; and the name of whoever ran the test.
Ask for the sheet before you buy anything shielded, from anyone — including us. A seller who cannot produce one has told you something useful.
Each COLDSTART Shielded Case ships with its own per-unit attenuation test sheet, so the paperwork belongs to the enclosure in front of you. We state the bar plainly, as a promise and not a measurement: if it measures below 30 dB anywhere from 1 to 500 MHz, the order is refunded in full. The terms sit on the legal page.
If you would rather not buy an analyser, calibrate it, and spend a Saturday on baseline runs, that is a fair call and it is yours to make. The method above is yours to use either way, and it works on any enclosure — a salvaged electrical cabinet, a welded steel box, or ours.
COLDSTART ANCHOR is our answer to the same problem: a sealed aluminium case built to order in Cambridge, Ontario. Batch 01 is 100 units. The Shielded Case is $1,819 and the Core Kit is $519, and the price shown is the total payable. Inside is a plain-text archive of about 29 GB — full-text search, offline maps, field manuals and medical reference texts — plus a small local assistant that answers from the books and cites the book and page. About 22 hours of reading. No internet connection, no account, no expiry.
If you want the case and the kit rather than an instrument and a weekend of sweeps, reserve a unit from Batch 01. What sits in the archive is covered in the offline library guide, and COLDSTART covers both pieces.