Why Filtered I/O? What One Unfiltered Cable Does to a Shielded Enclosure

Of all the questions we field at JRE Test, this one might be the most common: “Why do I need those filtered connectors? Can’t I just drill a small hole and run my cable through?”

It’s an honest question, and the honest answer surprises most people: the hole isn’t the problem — the wire is. A small, empty hole in the wall of a shielded enclosure leaks remarkably little RF. Physics is on your side there, and we’ll show exactly why. But pass a single unfiltered conductor through that same hole and the game changes completely: the cable becomes an antenna with one end inside your chamber and the other end out in the RF-soaked world, and it carries signals right past every dB of shielding the aluminum provides. An enclosure that measured −100 dB sealed will typically be left with only 40 to 60 dB of isolation — the cable simply bypasses the rest.

This paper walks through that phenomenon step by step: why empty holes are nearly harmless, why conductors are not, why even a shielded cable doesn’t save you, and what a filtered I/O connector actually does at the enclosure wall. As usual we’ll keep the math light in the body and put the details in an appendix. For which filter suits which signal type, see our companion paper on understanding data signals through filtered I/O connectors — this paper is about why that whole product category needs to exist.

Start with the case everyone expects to be bad: a plain round hole in the enclosure wall, nothing passing through it.

An electromagnetic wave trying to squeeze through a hole much smaller than its own wavelength has a very hard time. Two effects gang up on it. First, a small aperture is simply a poor radiator — the wave can’t “grip” an opening much smaller than a wavelength, and the energy that gets through falls off extremely fast as the hole shrinks (with the fourth power of the hole size, per small-aperture theory — see Appendix A.1). Second, the wall has thickness, and the hole through it behaves as a tiny waveguide operated far below its cutoff frequency; the wave doesn’t propagate through so much as evanescently tunnel, losing tens of dB in even a sixteenth of an inch of aluminum.

Figure 2 puts numbers on it. A 1/4″ hole in a 1/16″ aluminum wall still provides roughly 50 dB of shielding at 2.4 GHz all by itself. A 1/8″ hole gives about 70 dB. This is not a loophole in shielding practice — it is shielding practice: it’s precisely why shielded enclosures can have ventilation holes and honeycomb air panels without ruining their isolation, and why the seams that worry us are the long thin ones (a long gap behaves like a slot antenna, which is a far better radiator than a round hole — that’s what door gaskets are for).

So if all you needed were an air path, a small hole would cost you a few dB and life would be simple. The trouble starts when something conductive goes through it.

A conductor has no cutoff frequency. That single sentence is the whole paper, so let’s unpack it.

The reason the empty hole protected you is that a wave in free space can’t fit through an opening much smaller than its wavelength. But RF traveling along a conductor isn’t a free-space wave — it’s a guided current, and a guided current is perfectly happy on a wire of any diameter at any frequency. DC, 60 Hz, 2.4 GHz, 28 GHz: the wire carries them all with no objection. The moment a conductor crosses the shield boundary, you have handed every signal in the environment a private tunnel through the wall.

Here’s the sequence, and it runs in both directions. Outside the enclosure, your cable is bathed in ambient RF — cellular, Wi-Fi from every access point in the building, broadcast, the works. Those fields induce currents on the cable, exactly as they do on any antenna. The induced currents travel along the conductor, through the hole — no cutoff, no attenuation, the shield never gets a vote — and the section of cable inside the enclosure re-radiates them like a little transmitting antenna sitting right next to your device under test. Meanwhile the same physics runs outward: whatever your DUT transmits couples onto the interior cable section, rides out, and radiates into the room from the exterior section.

Figure 1 shows this happening in a full-wave simulation at 2.45 GHz. Three identical enclosures sit near a transmitter. The sealed one: silent inside. The one with a small open hole: a faint whisper gets in, more than 25 dB down even in this deliberately unflattering 2-D model. The one with a wire through the same hole: the interior lights up, nearly 20 dB hotter than the empty hole — and all that changed is the presence of one conductor.

See it run

The same simulation, live in your browser. Watch the third panel: nothing about the box changed except that one conductor now crosses its wall.

On the bench, the result is the number we quoted at the top: an enclosure that delivers −100 dB sealed will typically measure only −40 to −60 dB with one unfiltered cable through the wall. And the leakage is fickle. Coupling depends on cable length and routing, and it peaks wherever a cable section happens to sit near a resonant length (a half-wavelength is just 2.4″ at 2.4 GHz — every cable has resonances somewhere). Move the cable, change channels, and the leakage shifts. If you’ve ever seen a chamber that “worked fine yesterday” fail isolation today after someone re-dressed the cables, this is why.

A common follow-up: surely a shielded cable fixes this — the signal is inside a grounded braid, after all.

Unfortunately, from the enclosure’s point of view, a cable shield is just one more conductor passing through the wall. The braid’s outer surface carries induced common-mode currents exactly as a bare wire does, and those currents ride through the hole regardless of what’s happening on the inner conductors. Worse, if the shield is “grounded” by a pigtail — a couple inches of wire from the braid to a chassis screw — that pigtail is an inductor at RF, and the shield currents largely ignore it.

The only way a shielded cable crosses an RF boundary cleanly is if its shield is bonded to the enclosure wall in a full 360° ring right at the penetration — which is exactly what a bulkhead coax connector does, and why coax feedthroughs (SMA, N-type) on a JRE I/O plate don’t wreck the isolation. The general principle, and it’s worth stating as a rule:

Every conductor that crosses the shield boundary must be treated at the boundary — bonded, filtered, or eliminated. No exceptions, including grounds, shields, and power.

A filtered I/O connector is that rule made into hardware. Figure 3 shows the idea.

The filter mounts in the enclosure wall, metal body bonded 360° to the shield, so it becomes part of the boundary rather than a hole in it. Inside, each signal line passes through a network that does two jobs at once: it presents an easy, low-loss path to the frequencies your data actually uses, and it presents a brick wall to RF — shunting those currents into the grounded enclosure wall the instant they arrive. The RF energy riding the outside of your cable never gets a conductor to carry it past the shield; it’s dumped into the metal at the doorstep. Your data, which lives at much lower frequencies (or, for the high-speed differential interfaces our patented filters handle, in a different electrical mode entirely), sails through untouched.

That last distinction — frequency-based low-pass filtering for signals that stay below the interference (USB 2.0, and Ethernet up to 1 Gbps), mode-based filtering for the fast differential ones that don’t (USB 3, USB-C, 10-gigabit Ethernet, HDMI), and fiber optics when you’d rather have no conductor at all — is the subject of the data signals companion paper, which covers the filter families and how to choose among them. The point here is simpler: whatever the flavor, the filter’s job is to make sure nothing conductive crosses your shield untreated.

And “nothing conductive” means nothing: power lines get filtered power entry modules, antenna lines get bulkhead coax, data gets filtered data interfaces, and anything that can move to glass — fiber optic feedthroughs are the gold standard, since light carries no RF current at all.

Figure 4 pulls the story together as a budget. Start at −100 dB with a properly sealed enclosure. An empty vent hole barely dents it. One unfiltered cable knocks you down to the −40 to −60 dB neighborhood — you’ve spent half your shielding on a single mistake, and it’s the loudest path in the whole system, which means nothing else you improve will matter until it’s fixed. Route that same cable through the proper filtered interface instead and you’re back in the −80 to −95 dB range, limited now by the filter’s own specification rather than by an accident of cable routing.

Two footnotes to the budget, both covered in detail elsewhere: filters in parallel add up (each one is an independent, well-controlled leak, and ten identical filters cost you about 10 dB over one — see the effect of adding multiple I/O filters), and the only numbers that count are measured ones (see measuring and verifying shielding isolation for the transmitter-inside method we use).

A few field-tested habits that keep the boundary honest:

• Never leave a “temporary” cable through an open port. The most common isolation complaint we diagnose traces to exactly this — a USB or power lead run through a connector opening or a vent during setup and forgotten. One conductor is all it takes.
• Cap what you’re not using. Unused filtered connectors are fine — the filter is still in the wall. Unused openings (removed connectors, spare punch-outs) need conductive covers, properly gasketed.
• Keep exterior leads short and dressed. The outside section of every cable is the pickup antenna. Shorter leads, routed away from transmitters and against grounded metal, pick up less. If isolation readings dance when you move a cable, that cable is telling you something.
• Use the I/O plate as designed. JRE enclosures concentrate every penetration on a machined, gasketed I/O plate so each crossing is bonded or filtered in one controlled place — and so upgrades (adding a USB filter, swapping to fiber) don’t require surgery on the shell. See the options in the interface connector guide.
• Verify after changes. Any time the I/O configuration changes, re-check isolation with a known source; the measuring and verifying paper shows the quick way.

A shielded enclosure is an RF boundary, and boundaries are only as good as their crossings. Empty openings are cheap to cross safely — small holes below cutoff leak almost nothing, which is why vents and honeycomb panels exist. Conductors are the opposite: a wire has no cutoff frequency, so a single unfiltered cable through the wall acts as an antenna on both sides and typically drags a −100 dB enclosure down to −40 or −60 dB, with the exact figure at the mercy of cable routing. Shielded cables don’t escape the rule; their braids are conductors too, and only a 360° bond at the wall tames them. Filtered I/O connectors exist to enforce the rule in hardware: every signal crosses the boundary through a device that passes the data and shunts the RF into the shield right at the wall.

The metal does the shielding. The foam does the damping. And the filtered I/O does the gatekeeping — deciding, conductor by conductor, what gets to cross the wall.

For filter selection by signal type, see understanding data signals through filtered I/O connectors; for the full enclosure line and I/O options, visit the JRE Test catalog, and find more application notes on our resources page.

A.1 Leakage through an empty round hole

Two mechanisms set the shielding effectiveness (SE) of a small empty hole of diameter d in a wall of thickness t.

Small-aperture (Bethe) coupling. For a hole much smaller than the wavelength λ, the fraction of incident power that couples through scales as (d/λ)⁴. Expressed as shielding effectiveness relative to the open aperture:

SEₐ ≈ 10·log₁₀[(27π³/64)/(ka)⁴], where k = 2π/λ and a = d/2

Below-cutoff attenuation through the wall thickness. The hole is a circular waveguide whose lowest mode (TE₁₁) cuts off at fᶜ = 1.841·c/(πd) — about 27.7 GHz for a 1/4″ hole. Below cutoff, fields decay evanescently at approximately

SE_depth ≈ 32·(t/d) dB (for f ≪ fᶜ)

— the well-known “32 dB per diameter of depth” rule. The two terms add. Selected values for a 1/16″ (1.6 mm) aluminum wall:

These are per-hole values for normal incidence; arrays of holes (vent panels) and off-normal angles modify the totals, but the scaling is the story: small empty holes are excellent shields, and they get better fast as they shrink.

Why seams are different. A long thin gap of length L behaves as a slot antenna and leaks according to its length, not its width — approaching zero dB of protection as L nears λ/2 (just 2.4″ at 2.4 GHz). This is why a hair-thin gap along a door can out-leak a drilled hole by orders of magnitude, and why continuous gasketing gets the engineering attention it does.

A.2 Why a conductor has no cutoff

The empty hole protects you because free-space propagation through a small aperture is forbidden below cutoff. But a conductor through the hole supports a guided mode — a transverse electromagnetic (TEM) wave between the wire and the surrounding wall metal, exactly like the center conductor and shield of a coaxial line. TEM modes have no low-frequency cutoff: the wire-plus-hole is effectively a working coax feedthrough for every frequency at once. The shield’s aperture physics is simply bypassed.

The coupling strength on each side is antenna physics. A straight conductor of length ℓ exposed to a field at its resonant frequencies picks up power on the order of what a dipole of that length would; resonances occur near ℓ ≈ nλ/2. At 2.4 GHz, λ/2 = 61 mm — shorter than nearly any real cable section — so a practical cable is always near some resonance in the bands that matter. Between the two antenna couplings (outside pickup, inside re-radiation) and the lossless ride through the wall, the end-to-end path commonly leaves only 40–60 dB of isolation, dominated by geometry rather than by anything the enclosure can control.

The same analysis applies to a cable shield’s outer surface: skin effect isolates the braid’s outside from its inside, so the outer surface is an independent conductor — one more wire through the wall unless it’s terminated into the enclosure metal in a full 360° bond at the penetration.

A.3 About the simulation in Figure 1

Figure 1 is a 2-D finite-difference time-domain (FDTD) solution at 2.45 GHz: a 300 × 200 mm enclosure with 4 mm walls in a free-space domain with an absorbing border, illuminated by a nearby point source, with field magnitude time-averaged after steady state. The “wire” is a thin perfect conductor entering through a 2 mm opening, with a short exterior pickup section parallel to the wall and an interior section ending in a bend — a fair cartoon of a real cable dressed along surfaces.

One honest caveat: in a 2-D model, a “hole” in a wall is geometrically a slot of infinite extent in the third dimension, which leaks far more than a real 3-D round hole — that’s why the middle panel shows −26 dB where Appendix A.1 predicts ~50 dB for a true 1/4″ hole. The comparison to take from Figure 1 is the relative one, which the 2-D model captures faithfully: adding a single conductor through an opening raised the interior field energy by roughly 20 dB, from a whisper to a roar, with no other change to the enclosure. In three dimensions the contrast is even starker, because the empty hole starts from a much quieter baseline.

References

• H. A. Bethe, “Theory of Diffraction by Small Holes,” Physical Review, vol. 66, 1944 — the foundation of small-aperture coupling theory.
• Standard EMC practice for waveguide-below-cutoff attenuation and slot leakage (see e.g. H. W. Ott, Electromagnetic Compatibility Engineering).
• JRE Test white papers: Understanding data signals through filtered I/O connectors; The effect of adding multiple I/O filters; Measuring and verifying shielding isolation — all in the resources library.