Isolate a metal enclosure from the signal path

Short answer: isolate every penetration that carries signal — plastic-bushing jacks, an insulated-shaft pot on nylon shoulder washers — and bond the case to the low-level side at exactly one point. Your hand brings about 100 pF to the panel, the human-body model figure, and at speaker level even 1 pF between case and circuit tries to push half a volt into a guitar amp's input. That is the howl: the case rides the output signal on its jack sleeves, your palm couples it to the pickups, and the amplifier's gain closes the loop. Keep the metal for strength, heat and shielding; keep the circuit off it.

Why does the box howl when you touch the case?

Your hand closes a capacitive feedback loop: the case rides the speaker signal through its metal jack sleeves, your body couples about 100 pF of that to the guitar pickups, and the amplifier's gain turns the coupling into a running howl.

A passive attenuator has no earth connection, so nothing holds the case at a quiet potential. Wire the output jack through a metal bushing and the whole panel swings with the speaker signal. A guitar pickup is a high-impedance source, and a typical tube-amp input sits its grid on a 1 MΩ resistor, so a whisper of coupled current is already a signal.

Worked example, assumptions marked. A 100 W amplifier into 8 Ω delivers V = √(P × R) = √(100 × 8) = 28.3 V RMS at the speaker, and the case swings through that potential with the output sleeves. Take your flat palm on the lid as 100 pF of coupling to the pickups — the human-body model value, 100 pF behind 1.5 kΩ; substitute your own geometry and amplifier. At a howl pitch of 3 kHz, the coupling reactance is 1/(2πfC) = 1/(2π × 3000 Hz × 100 pF) = 531 kΩ.

The coupled current is 28.3 V ÷ 531 kΩ = 53 µA, and across the 1 MΩ input impedance the divider tries to develop 53 V. A guitar front end reaches full output somewhere between tens and hundreds of millivolts, so the loop closes with a hundred to a thousand times the drive it needs. Cut the coupling to 1 pF and the same arithmetic gives 28.3 V × (1 MΩ ÷ 53 MΩ) ≈ 0.53 V, still past full drive.

That arithmetic explains commercial metal boxes: their output sleeves never touch the case. A case with no forward signal on it gives your hand nothing to couple, and every step below reaches that state.

What do you need to build a quiet metal box?

Four parts do the whole job: isolated-bushing jacks, a potentiometer with an insulated shaft, a nylon shoulder-washer set for anything metal that still passes through the wall, and a meter that reads resistance to 20 MΩ.

• Two isolated 1/4 in jacks, input and output, the plastic-bushing kind.
• One pot with a plastic shaft; if the pot you want has a metal bushing, add a nylon shoulder washer and a flat nylon washer for under the nut.
• One bonding jumper: 18 AWG wire, two ring terminals, a star washer. The jumper carries signal-level current only, so heavier wire buys nothing.
• A handheld meter reading resistance to at least 20 MΩ; common meters reach 20-50 MΩ.
• The connector datasheets, for hole sizes and the isolation claims.

Buy the jacks as 'isolated' and read the datasheet rather than the photo. Isolation lives in the plastic bushing between the sleeve-and-barrel block and the mounting thread. A metal-bushing jack lands its sleeve on the panel the moment you tighten its nut, and no wiring inside the box changes that.

Which part of the circuit should the case carry?

Bond the case to the quiet side and to nothing else: the input sleeve on a passive box, the equipment grounding conductor on powered gear. A case held at the reference potential turns stray coupling into current that returns harmlessly to ground instead of landing on the input.

On the passive attenuator the input sleeve is the bond point. With the case tied there, the panel becomes the shield for the low-level side, the way a metal wall blocks interference once it is held to one reference — the radio signal design options article works through that shielding behaviour. Your palm then touches a node with no signal on it, and the pF-scale capacitance from the output sleeves dumps into the reference, where it makes nothing audible.

Powered devices bond to their own reference. Mains-connected gear bonds the case to the equipment grounding conductor: NEC 250.96(A) asks metal parts serving as equipment grounding conductors — raceways, cables, equipment, enclosures — to be bonded together, and the bond lets them safely conduct any fault current likely to be imposed on them. The grounding and bonding guide and grounding an enclosure mounting plate cover the bond, the plate and the terminals.

This is also why a battery-powered pedal full of metal jacks stays quiet. Its case sits at circuit ground through the jack bushings, your hand touches a reference, and no speaker-level signal rides the panel.

How do you isolate each penetration?

Test every part before it goes in, mount it on its insulators, then retest the assembly: each isolated jack and the pot must read open beyond your meter's 20 MΩ top range to the panel, and the single bond must read like your leads alone.

1. Bench-test each jack. Measure from sleeve to mounting thread: a metal-bushing jack reads near 0 Ω, an isolated jack reads beyond the meter's 20-50 MΩ top range, and anything in between goes back in the drawer.
2. Cut the holes to the datasheet size. A common 1/4 in jack takes about a 3/8 in [9.5 mm] hole; the exact value comes from the connector's datasheet, and the right size lets the plastic bushing clamp without metal-to-panel contact.
3. Mount the pot on its insulators. A plastic shaft settles the shaft coupling on its own; a metal-bushing pot gets a nylon shoulder washer in the hole and a flat nylon washer under the nut, then repeats the open-circuit test between track and panel. Some pots tie a track to the case internally, and this test catches those too. The insulated jack mounting article shows the washer stack in detail.
4. Give the big entries the same treatment. Where a metal conduit or a bonded fitting meets the wall of a larger enclosure, an insulating bushing with a thermoplastic liner keeps the fitting off the panel. That liner class will not chip, crack, swell or shrink, and it shrugs off chemicals; in an insulated grounding bushing, the liner is rated flame retardant to 105 °C. The same bushing family also protects conductor insulation from abrasion under vibration. The cutouts and glands guide covers the layout side.
5. Bond the case once, at the point the previous section picked: ring terminal, star washer on bare metal, one fastener. On a coated box, scrape the finish back to bare metal under the washer first; NEC 250.12 requires paint, lacquer and enamel to be removed from threads and contact surfaces at grounding and bonding connections.
6. Retest the assembled box. Every isolated penetration still reads open to the panel, and the bond reads at your leads' own resistance: touch the meter tips together, note 0.1-0.5 Ω, and confirm the bond adds nothing measurable.

How do you prove the box stays quiet when it is done?

Three checks settle it: open-circuit readings beyond 20 MΩ at every isolated penetration, a bond that reads like the leads alone, and ten seconds with your palm flat on the lid at playing volume while the guitar idles.

• Sleeve to panel on every jack, and pot track to panel: open beyond 20 MΩ, with the box buttoned up.
• Input sleeve to panel: your leads' resistance only, about 0.1-0.5 Ω.
• Listening test: guitar plugged in, amp at playing gain, palm flat on the lid for 10 s. Silence, or a small hum shift, passes.
• Amplifier-side check: if the howl survives with the guitar lead out of the box, the loop lives in the output stage. Its output stability network, a resistor and capacitor in series across the output, corrects oscillation there — not hand-coupled feedback.

Measure inside a tube amplifier only with the mains lead unplugged and the filter capacitors discharged.

Which mistakes put the case back into the circuit?

Hardware, not wiring, undoes most builds: a metal anti-turn washer under the pot nut, a shoulder washer pushed out of an oversize hole, or a second fastener that bolts a grounded bracket to an isolated part.

• Swap metal anti-turn washers for fibre or nylon; the metal kind shorts the pot case to the panel right next to the washer stack.
• Hold the datasheet hole size. An oversize hole lets the jack shoulder or the washer flange bridge to the panel edge, and a hole that already came out wrong gets plugged and redrilled — the wrongly drilled hole fix shows the sealed way back.
• Leave 3 mm [1/8 in] of air between an exposed metal pot shaft and any bonded bracket or the panel skirt.
• Trace every fastener through the wall. One ground lug bolted through an isolated zone re-bonds the case, and the open-circuit retest catches it.
• Bare the metal at the bond point on painted or powder-coated boxes; the finish is an insulator exactly where the bond needs metal.

What do you order when heat or strength rules out plastic?

Order the metal box with every cutout placed for isolated hardware: a 304 or 316L stainless enclosure built to your width, height and depth, cutouts laid out in the cutout designer, delivered in 10 business days from your approval of the drawing.

Heat is the usual reason plastic leaves the list. A −10 dB L-pad passes 10^(−10/10) = 0.1 of the input to the speaker — 10 W of a 100 W amp — so the other 90 W burns in the pad's resistors, inside the box. Metal walls take that; the resistor wattage rating matters as much as the isolation parts.

Metal boxes for this build: 304 stainless with a #4 brushed finish for most indoor and outdoor sites, 316L for salt air, brine and chemical areas, built from one piece to the size you set. Our stainless wall-mount and HMI boxes are rated NEMA 4X / IP66 by the manufacturer; not UL listed — and any connector you fit sets the finished assembly's rating at that part's own seal. Place the jack, pot and gland cutouts in the cutout designer, check the spacing, and send the layout with the quote; custom stainless enclosures take material, dimensions and cutout sheet from there.

Pick the row that matches the build and order around it. The case ends up a shield instead of an antenna, and your hand on the lid is just a hand on the lid.

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