Digital signal cables through a shielded room wall

Short answer: bond each cable shield 360 degrees to a stainless gland plate bolted through bare wall metal, and continue every pair in at least 12 in [300 mm] of steel conduit bonded at both ends. Sleeve every spare hole while you are at it. That combination typically holds 40-60 dB to 1 GHz on a penetration like this. A certified 60 dB to 10 GHz survey needs the copper gone: fiber through a waveguide-below-cutoff sleeve. Steel wool packed around the jacket adds nothing measurable and starts corrosion. The worked case below is an RS-485 thermostat pair through a 3/8 in [9.5 mm] hole in 16 gauge steel, in a room that has to measure that target once the repair is done.

What the bare hole is worth

A round hole in sheet steel behaves as a circular waveguide running below cutoff. Below cutoff it attenuates a field by 32 × t/d dB, with t the wall thickness and d the hole diameter. The cutoff frequency works out to 1.8412c/(2πa) for hole radius a. A 3/8 in [9.5 mm] hole therefore cuts off at 18.5 GHz, and every survey frequency sits below it. Thickness is the problem: 16 gauge sheet runs 0.060 in [1.5 mm], and 32 × 1.5/9.5 is 5 dB. The bare hole buys about 5 dB against the target.

The cable is worse than the hole. A pair whose braid is cut at the wall is two antennas joined by a gap, and RF current rides the jacket straight across the barrier. What an enclosure hole leaks at each frequency puts numbers on that leak. Steel or copper wool adds no continuous conductor between braid and wall, so it moves the reading by nothing you can measure, and it holds moisture against the joint. Crevice corrosion follows within a season in damp air.

What you need

Lock out the HVAC control circuit and prove the pair dead with a meter before you open the wall.

• A stainless wall-mount enclosure to bridge the hole, sized for one gland per pair plus spares; the catalog line runs from 6 x 4 x 4 in to 60 x 36 x 16 in outside (the stainless size finder)
• Clamp-type cable glands sized to the jacket; an M20 clamp gland typically takes 6-12 mm, and the maker's seal range states it exactly
• Steel conduit with bonding bushings, 1/2 in trade size or whatever fits the jacket
• A stainless bonding strap and stainless hardware; plain steel against stainless corrodes at the joint
• Snap-on ferrites, one per cable
• A bonded metal box or copper-mesh cage for the thermostat inside the room, with a removable lid for setpoint work
• Tube stock for the spare-hole sleeves: 1 in [25 mm] per 3/8 in hole, per the worked example below

Making the penetration, step by step

1. Move the penetration onto a bonded box. Mark the gland positions and a bonding stud, and order the holes machined rather than drilling stainless on the wall. Custom cutouts come off a drawing and are delivered in 10 business days from your approval of that drawing; the boxes are rated NEMA 4X / IP66 by the manufacturer; not UL listed. The custom stainless cutout page lists what the drawing needs, and when a stainless box wants a gland plate covers plate versus drilled holes. If a hole must grow before the box arrives, cutting stainless in the field keeps the cut and the seal intact.

2. Bond the box to bare metal. Scrape paint and oxide from under every fastener and along the full strap run. NEC 250.12 requires paint, lacquer and enamel to be removed from threads and contact surfaces at grounding and bonding connections, and a strap sitting on paint is exactly the joint it targets. The grounding and bonding guide covers the rest of the bond.

3. Clamp each shield 360 degrees at its gland. The clamp cone must grip the braid around the whole circumference onto the plate; shield termination at an enclosure entry shows the hardware. A pigtail to a ground screw leaves the braid floating at RF. One inch [25 mm] of pigtail wire is about 25 nH, which is roughly 160 ohms at 1 GHz, so the braid sits 160 ohms off the wall where it matters. One pair per gland keeps the clamp sealing; when two cables can share one gland draws that line. For the bond at the devices themselves, RS-485 shield bond page walks through direct versus RC ties.

4. Continue each pair in steel conduit. Run at least 12 in [300 mm] of bonded conduit on each side of the wall, and carry it all the way to the device where the geometry allows. Bond both ends. That keeps the braid inside steel from wall to device, which is where most of the 40-60 dB at the low end of the band comes from.

5. Add a snap-on ferrite per cable near the gland. A ferrite is worth 50-300 ohms of common-mode impedance between 30 and 300 MHz, per its datasheet impedance curve, and the differential pair passes it untouched: a bus at 38.4 kBaud has a 19.2 kHz fundamental, so ten harmonics deep you are still under 200 kHz, two decades below where the ferrite bites.

6. Cage the thermostat. A device with a microcontroller and a display radiates and receives on its own. Put it in a bonded metal box or copper-mesh cage with a removable lid, and bond the cage to the wall. The feed-through then serves a shielded box instead of a bare antenna.

7. Sleeve every spare hole. Fit a 1 in [25 mm] tube on each 3/8 in opening, per the arithmetic below. If you convert the pair to fiber, the glass goes through the same sleeve with nothing metal in it; protecting fiber runs in an enclosure covers the rest of that route.

Sizing a sleeve for a 3/8 in hole to meet the target

Inputs and where each comes from:

• d = 3/8 in [9.5 mm] hole diameter (the existing penetration)
• t = 1.5 mm [0.060 in] wall thickness (16 gauge steel)
• 60 dB at 10 GHz (the shielding target of the room)
• Rule: below cutoff, a circular tube attenuates A = 32 × (t/d) × √(1 − (f/fc)²) dB, with cutoff fc = 1.8412c/(2πa)

Cutoff first, with a = 4.76 mm and c = 3 × 10⁸ m/s:

fc = 1.8412 × 3 × 10⁸ / (2π × 0.00476) ≈ 1.85 × 10¹⁰ Hz ≈ 18.5 GHz

At the band top the correction factor is √(1 − (10/18.5)²) = 0.84. Solve the rule for tube length:

t = 60 × 9.5 / (32 × 0.84) ≈ 21 mm ≈ 0.83 in

Any sleeve 0.85 in long or more therefore meets the target at the top of the band. Fit a 1 in [25 mm] tube and it delivers 32 × (25.4/9.5) × 0.84 ≈ 72 dB there, and about 85 dB at the low end. For scale, the same arithmetic on the bare 1.5 mm wall gives 32 × 1.5/9.5 ≈ 5 dB, which is what the hole-and-wool arrangement was worth.

The decision this leads to: a sleeve only works while it is empty. Copper in the tube carries the energy around the metal and the 72 dB evaporates, so sleeves go on spare holes and on fiber. EMP facility practice sizes its waveguide penetrations by the same arithmetic, for a cutoff of at least 1.5 GHz and at least 80 dB at the low end; only the targets differ.

Where each option lands

The decision splits on the target. If the room must pass a measured survey across the whole band, the copper pair loses at this wall however well you bond it, because braid leakage climbs steeply above the low frequencies. Convert to fiber and run the glass through the sleeve. If the goal is suppression and the lines stay copper, the bonded gland plate with 12 in of conduit at both ends is the repair. IEEE 299 is the standard that defines how enclosure shielding effectiveness gets measured, so write the post-repair survey against it.

Checks when the repair is done

• Bond quality: measure plate to wall with a 4-wire micro-ohmmeter; a sound RF bond reads single-digit milliohms. A DMM only proves continuity.
• Path check: with the bus idle and the equipment isolated, short the pair at the far end during the survey and watch the reading. A clear drop means the cable was the ingress path and the new bond is holding.
• Full sweep: run the survey per IEEE 299 across the whole specified band, not only the old problem frequency.
• Follow-up: re-measure the milliohms after a few thermal cycles and re-torque anything that has crept. Bonds corrode month by month, which is how a neglected room gets here.

Mistakes that cost the rating

• Wool around the jacket: no continuous conductor, so nothing gained, plus a moisture trap.
• A pigtail instead of a 360-degree clamp: 25 mm of lead puts the braid 160 ohms off the wall where it matters.
• Conduit bonded at one end only: a 12 in [305 mm] stub is a quarter-wave resonator near 250 MHz, and it re-radiates what you were blocking.
• Paint or gasket under a foot or a strap: the bond reads fine in October and corrodes by spring.
• An oversize or burred hole under a gland: the seal cannot bridge the gap, and the wall leaks at the seal instead.

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