Short answer: yes. Give the metal a defined tie to your PCB ground, and let the supply's isolation class pick the form. On an earthed box, bond chassis to ground at one mounting boss with a 0 Ω link. On a floating Class II product fed by an isolated 20 V adapter, tie them through a 1 MΩ resistor in parallel with a 10 nF capacitor and a TVS diode: static bleeds off at 8 mA at 8 kV, 100 MHz noise sees 0.16 Ω, and no DC path exists. Land the USB-C shell on the ground plane in every build. Fit both footprints, the 0 Ω link and the three-part network, so you can change the decision after EMC testing with a soldering iron instead of a new board.
What the tie actually does
Left alone, the chassis is a piece of metal coupled to the world through only about 50 pF, sitting at whatever voltage it collects. A person crossing carpet in dry air carries 10 kV or more and hands it to the box on first touch. The only equalization path is then a spark from a screw head to your nearest trace.
The tie buys you three things. It drains collected charge through a part you chose rather than an unplanned arc across a standoff. It references the box to your plane at RF, so the metal shields your emissions instead of re-radiating them. It also pins the chassis potential, so a nearby drive or radio induces current on the box rather than across your signals. The analog-noise side of grounded versus floating enclosures is a separate question; this decision runs on ESD and EMC.
Three ways to make the tie
Match the row to your supply and your safety class; the safety section below works out which class you are in.
| Build | Tie between PCB ground and chassis | ESD | EMC |
|---|---|---|---|
| Earthed box, Class I | 0 Ω at one boss; chassis on protective earth | Discharge dumps straight to earth | Best shielding; seams carry the current |
| Floating box, Class II | 1 MΩ + 10 nF + TVS in parallel | TVS takes the first peak; resistor bleeds | 0.16 Ω at 100 MHz; no DC path |
| Metal behind a plastic shell | No tie; board floats inside | Plastic takes the hit; keep 4 mm to the metal | Shields only when bonded to the plane |
The hybrid network, with values
Put three parts in parallel between your plane and the chassis, within 10 mm [0.4 in] of the boss that touches the metal:
- 1 MΩ, high-voltage rated — the bleed. 8 kV across 1 MΩ is 8 mA, gentle enough for any circuit and quick enough to empty a 50 pF chassis in 50 µs (τ = 1 MΩ × 50 pF).
- 10 nF, 2 kV rated — the RF short. It presents 0.16 Ω at 100 MHz and 0.016 Ω at 1 GHz, so the whole band sees a bolted joint.
- Bidirectional TVS, ESD class — the clamp. ESD-rated parts are built for 8 to 30 kV discharges lasting microseconds or less.
A Microchip application note on ESD (AN-139) starts this network at 1 nF rated 2 kV from chassis to supply ground, so the discharge returns to the supply first. Anything from 1 nF to 10 nF works, and the larger capacitance buys you the lower RF impedance.
Lay an unpopulated 0 Ω footprint across the resistor position as well. If the EMC lab wants a hard ground, you fit the jumper; if a safety submission wants the metal clean, you pull the network. Which boss to use matters as much as the parts — choosing the chassis ground point walks that choice.
What the ESD test throws at the box
IEC 61000-4-2 is the immunity test your product will meet, and it is not gentle. At its top severity it fires 8 kV contact and 15 kV air discharges from a generator equivalent to 150 pF discharging through 330 Ω. The first peak at 8 kV contact reaches about 30 A in 0.8 ns. Stored energy is small — ½ × 150 pF × (8 kV)² = 4.8 mJ at contact, 17 mJ at 15 kV air — but nanoseconds of current through the wrong millimetre of trace is what kills ports.
An enclosure rating — NEMA 4X / IP66 included — covers hose-directed water and corrosion, and it says nothing about ESD paths or seams at radio frequency. A watertight gasket is an insulating seam until you give the current a metal-to-metal way across it.
Worked example: an 80 W board in a 10 x 8 x 4 in box
Take a 20 V, 4 A USB-C input (20 × 4 = 80 W), a Wi-Fi module behind a plastic window, and LED wiring leaving the board, all mounted in a 10 x 8 x 4 in 304 stainless wall-mount enclosure, rated NEMA 4X / IP66 by the manufacturer and not UL listed — or its 316L twin of the same size where salt air or sanitizer reaches it.
- Untied, air discharge. A 15 kV air discharge lands beside a screw head 4 mm from a PCB pad. The stress is 15 kV ÷ 4 mm = 3.75 kV/mm, past the roughly 3 kV/mm that dry air holds, so it arcs and pours up to 17 mJ into a trace you never chose. Nylon standoffs invite exactly this.
- Tied, air discharge. The TVS clamps plane-to-chassis in nanoseconds and the 10 nF carries the fast edge. The 1 MΩ then drains the residue at 8 mA, emptying the chassis in about 50 µs. Your plane rides tens of volts for under a microsecond, which filtered I/O shrugs off.
- Tied, contact discharge on the connector. The 8 kV contact lands on the USB-C shell, flows shell → plane → TVS and capacitor → chassis, and returns through the test ground. The 30 A peak splits between the two parts, and the arc never wanders.
- LED wiring. One metre of wire is a quarter-wave antenna near 75 MHz, right where switching harmonics live. Filter those wires at the connector, on the plane, before you blame the box.
Safety: the class picks the DC path
IEC 62368-1, the product-safety standard for equipment like this, gives you two routes. Class I bonds accessible metal to protective earth; your 0 Ω tie plus an earthed supply fits it, and the adapter ahead of you then needs only basic isolation. Class II carries no protective earth anywhere, and everything touchable stays separated from the mains by double or reinforced insulation. A floating product therefore needs an adapter whose mains-to-20 V barrier is rated for reinforced isolation.
Within those rules the hybrid tie is functional, not protective. Your 20 V DC island sits well inside the touch-safe band (ES1, up to 60 V DC), so chassis metal sharing its reference stays safe to touch. If your submission still wants the circuit and the metal fully separate, put a plastic shell between them and bond the metal to nothing.
Protective earth and ESD ground get confused often enough that the difference has its own walkthrough in ESD grounding versus protective earth. And if a box you build ever carries mains itself, bonding becomes a fault-current question — the enclosure grounding and bonding guide covers that case. Your only mains here lives inside the adapter, so opening the box is low-voltage work; unplug it anyway before you pull the board.
The Wi-Fi window
Metal detunes a nearby antenna whether it is grounded or not, so the tie barely moves your link budget. What you control is distance. At 2.4 GHz a quarter wavelength is 31 mm [1.2 in]; hold the module antenna at least that far from any metal edge, and cut the plastic window with about 25 mm [1 in] clear around the antenna on all sides. Grounding the box does buy one thing: a repeatable reference, which shows up as unit-to-unit consistency in the antenna match. Where the antenna is a PCB trace, the metal under it acts as a partial ground plane — antenna ground planes on a metal enclosure works through that trade.
The USB-C shell
Bond the shell to the ground plane in every build, earthed or floating, at 0 Ω, right at the connector. The shell is where a contact discharge lands and where the cable's common-mode current wants to return; giving it the plane keeps both off your signals. Land the shell tabs through vias straight into the plane rather than along a trace. USB and Ethernet shield grounding in an enclosure covers the shell-versus-chassis question in depth.
What to put on the drawing and the order
| Item | Specify |
|---|---|
| Enclosure | 304 stainless indoors; 316L for salt air or sanitizer; 10 x 8 x 4 in for the example build |
| Door window | Plastic insert, ≥ 25 mm [1 in] clear around the antenna |
| Bonding boss | One boss within 10 mm [0.4 in] of the connector cutout, bare metal |
| Board tie | 1 MΩ at 2 kV, 10 nF at 2 kV, bidirectional TVS in the 8 kV contact class, plus an unpopulated 0 Ω footprint |
| Connector | USB-C shell tabs to the plane at 0 Ω, stitched with vias |
Order the box as a custom enclosure and you set the window, the cutouts and the boss positions in one drawing; it arrives as one piece, delivered in 10 business days anywhere in the world, counted from your approval of the drawing. If the size is still open, the enclosure selector narrows it by inside dimensions, and stainless wall-mount enclosures lists the standard sizes.
The decision in one breath: earthed box, 0 Ω at one boss. Floating box on a reinforced-isolation adapter, 1 MΩ plus 10 nF plus a TVS with the 0 Ω footprint waiting. Metal behind plastic, no tie and 4 mm of clearance. Shell to plane, always.
Enclosures in this article
| Enclosure | Material | Outside H x W x D / display | Rating |
|---|---|---|---|
| 10 × 8 × 4 in | 304 stainless | 10 x 8 x 4 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
| 10 × 8 × 4 in | 316L stainless | 10 x 8 x 4 in | Rated NEMA 4X / IP66 by the manufacturer; not UL listed |
Questions
What power and voltage rating does the 1 MΩ bleed resistor need?
Peak power looks brutal — 8 kV across 1 MΩ is 64 W — but the whole contact discharge carries only about 4.8 mJ, so a standard 0.25 W part absorbs the energy fine. Voltage rating is what bites: a 1206 is typically working-voltage rated near 200 V, so fit a high-voltage thick-film part rated around 2 kV, or two standard resistors in series to share the stress.
Why must the tie components sit within 10 mm of the chassis contact?
Trace inductance runs about 1 nH per millimetre, so 20 mm of trace adds roughly 20 nH. At 100 MHz that is 12.6 Ω in series with a capacitor whose whole job is 0.16 Ω; the trace wins and the RF short is gone. Keep the parts at the boss, pour plane under them, and keep the loop area small.
How do I handle this in a plastic enclosure with no chassis?
In a plastic box there is no chassis to tie, so the connector shell becomes your reference: bond it to the plane and give the plane a solid pour. The roughly 50 pF plate that was your chassis disappears, so collected charge leaves mainly through the cable — a 1 MΩ bleed from plane to shell at the connector gives it the same slow exit.
What should I look for when choosing the TVS diode?
Pick a bidirectional part, because the chassis can swing either polarity during a strike. Set the working voltage above the largest steady ground-to-chassis offset you expect — a few volts in a signal box — and set the ESD rating at or above the 8 kV contact and 15 kV air levels of IEC 61000-4-2.
