A sheet metal enclosure is not an EMI shield. It is an EMI shield only if its seams are bonded. A SanJun customer shipped a batch of industrial control enclosures to a European integrator last year; the first three passed FCC Part 15 Class A in the integrator's chamber, and the fourth failed radiated emissions at 180 MHz by 12 dB. The hardware was identical. The fourth enclosure had been powder-coated across the seam area, including the panel-to-enclosure bond zone. This guide covers why shielding is a seam problem rather than a material problem, the three bond strategies and where each belongs, how gasket material and seam geometry interact, what finishing does to a bond, how apertures and vents leak, and what to check before an enclosure goes to a test chamber.
Key takeaways
- The panel is never the problem; the seam is. Cold-rolled steel gives roughly 60 dB at 1 MHz and 30 dB at 1 GHz. A 10 mm gap in the seam can drop attenuation by 40 dB in the gap's resonance band.
- Paint, anodizing and chromate are non-conductive. If any of them sits in the bond zone, the seam is open electrically — which is the single most common cause of a failed emissions scan.
- Seam geometry beats gasket selection. Keep the assembled gap at 1 mm or less and give the gasket at least 5 mm of conductive overlap on both sides. A flat seam matters more than a thicker gasket.
- Apertures leak by their longest dimension, not their area. Keep the longest dimension of any opening below roughly λ/20 at the highest frequency of interest.
- SanJun supplies the metal, the machining and the supporting evidence. We are a sheet metal fabricator: we do not issue listings or certificates of compliance, and we do not run UL 9540A.
Why is EMI shielding a sheet-metal seam problem?
Because the metal works and the joint does not. A continuous sheet of cold-rolled steel attenuates electromagnetic radiation well — on the order of 60 dB at 1 MHz, falling to about 30 dB at 1 GHz. Those numbers are not the difficulty. The difficulty is that an enclosure is not one continuous sheet: it is panels, a door, a cover, gland plates, I/O cutouts and a display window, joined by seams.
Every seam is a slot. A 10 mm gap radiates like an antenna, and in the gap's resonance band the attenuation can fall by 40 dB — which is why a design that looks sound in CAD fails in the chamber. Three variables decide whether a seam behaves:
- Conductive continuity across the joint. The panel has to be electrically bonded to the enclosure body at every seam, with surface resistance below about 5 mΩ across the joint.
- What is sitting in the bond zone. Powder coat, anodizing and chromate conversion are all non-conductive. Any of them between the two halves of a seam is an open circuit.
- Gasket compression where IP and EMI have to coexist. An EMI gasket does both jobs — sealing and bonding — but only if the geometry lets it compress correctly.
The practical consequence is that shielding is a process-control discipline, not a materials-specification exercise. The physics is settled; the failure mode is always the same three things: the bond got painted over, the seam warped, or the gasket cold-flowed.
What does the metal itself actually attenuate?
Enough that you never need to worry about it, which is why spending money on exotic shielding materials is usually the wrong first move. Solid sheet attenuates by reflection and absorption, and both depend on frequency, thickness and the material's conductivity and permeability.
| Material | Approx. attenuation at 1 MHz | Approx. attenuation at 1 GHz | Where it is the right choice |
|---|---|---|---|
| Cold-rolled steel | ≈ 60 dB | ≈ 30 dB | Default for industrial enclosures; magnetic absorption helps at low frequency |
| Galvanized steel | ≈ 60 dB | ≈ 30 dB | Same shielding as CRS; the zinc layer is conductive and bondable |
| Aluminium | ≈ 55 dB | ≈ 28 dB | Weight-sensitive enclosures; anodizing breaks the bond if unmasked |
| Stainless steel | ≈ 55 dB | ≈ 30 dB | Washdown and coastal; lower conductivity than copper but comparable in practice |
Two caveats keep this table from being misused. The figures describe a solid, continuous sheet with no apertures — they say nothing about the assembled enclosure. And they are attenuation of the material, not compliance with a limit: the emissions limit your product must meet comes from the EMC standard applicable to your market and product class. On OEM enclosures the one we see named most often on drawings is FCC Part 15 Class A or Class B, and which class applies is decided by where the product is sold and installed, not by the enclosure.
What are the three bond strategies?
Three strategies cover almost every OEM enclosure, and the choice is driven by how much shielding you need and whether the same seam also has to seal.
| Strategy | How it works | Typical shielding | Best for | Cost |
|---|---|---|---|---|
| Bare-metal masked seam | Mask the seam before coating; bolt the panel down with self-clinching fasteners onto bare metal | 40–70 dB, falls with frequency | Indoor enclosures, low-cost programmes, cosmetic finish not critical at the seam | $ |
| Conductive gasket in the seam | Continuous conductive gasket provides both IP sealing and EMI bonding | 60–100 dB depending on material | Outdoor enclosures, marine, medical — anywhere IP54 or better coexists with an emissions limit | $$–$$$ |
| Bonded cover with finger stock | Beryllium-copper finger stock on a milled flat seam; lowest inductance | 100 dB and up | RF test equipment, military, satellite ground gear | $$$ |
The first strategy is the one people skip because it looks crude, and it is the right answer more often than anyone expects. A masked bare-metal seam with self-clinching fasteners — see our fastener selection guide — will pass Class A on a well-designed indoor product at no gasket cost at all. The second is the workhorse for anything that also has to keep water out. The third is specialist, and on commercial OEM work it is almost always the wrong place to spend money.
Which gasket material suits which job?
Four gasket families are common on sheet metal enclosures, and the selection is a three-way trade between shielding, sealing and cost. Note that the highest-shielding option in the table does not seal at all.
| Gasket material | Shielding effectiveness | IP rating | Cost | Best for |
|---|---|---|---|---|
| Knitted wire mesh (Al / Sn / Cu) | 80–100 dB | IP54 | $ | Indoor cabinets, telecom racks |
| Conductive elastomer (silver-filled silicone) | 60–90 dB | IP65 and up | $$$ | Outdoor, medical, marine |
| Finger stock (BeCu) | 100 dB and up | IP20 — no sealing | $$ | High-frequency, shielded rooms |
| Conductive foam (Ni/Cu over PU) | 50–70 dB | IP54 | $ | I/O panels, low-cost programmes |
The most common pick for an OEM enclosure that needs both jobs is knitted wire mesh over a foam core: one part, roughly 80 dB, and IP54 in a single gasket. Where the enclosure is outdoor and has to hold IP65 or better, a silver-filled silicone gasket is the right answer and worth the cost. Compression matters as much as material — 20–30% compression for knitted mesh and 15–25% for conductive elastomer — and the gasket has to be specified with the compression ratio in mind, not just the part number. Sealing geometry is covered in our gasket sealing to IP65 guide.
What seam geometry actually works?
Two geometry rules decide whether the bond survives production, and both are cheap on the drawing and expensive after welding:
- Maximum seam gap ≤ 1 mm at assembly. Above that, a gasket cannot span the gap without compression-set failure inside a year. Tighten the tolerance on panel flatness rather than on the gasket.
- Minimum contact overlap ≥ 5 mm. The gasket has to see at least 5 mm of conductive surface on both sides of the seam. Less than that and it slips out under vibration.
The single most common failure we see is a seam that looks perfect in the model and has 2–3 mm of warpage after welding. The gasket spans the warpage at install; six months later it has cold-flowed and lost half its compression, and the enclosure that passed pre-compliance fails its next scan. The fix is not a thicker gasket — it is a flatter seam. Weld sequencing, fixturing and post-weld stress relief are the levers, and they are covered in our welding guide and weld distortion control guide.
A medical device customer had exactly this: a 1.6 mm stainless door that passed emissions at first article and failed fourteen months later at 240 MHz. The gasket was intact. The door had 2.4 mm of bow across a 600 mm seam, measured after painting. Adding a stiffening return flange — not changing the gasket — brought it back inside the limit. That is a stiffener and rib design problem wearing an EMC costume.
Does welded or modular construction shield better?
It depends on what you are trading, and the answer is different for the frame and for the removable panels. A welded seam is a continuous conductive path with no gasket to age, no fasteners to loosen and no compression to lose — it is the best shield you can build and the worst enclosure to service. A bolted seam gives you access and costs you a gasket, a flatness tolerance and a maintenance item.
Three patterns cover most OEM work:
- Fully welded — best shielding, no gasket cost, no service access. Used where the electronics are not field-serviceable and the enclosure is replaced as a unit.
- Fully modular, bolted with gaskets — full access, gasket at every seam, and the shielding depends entirely on seam flatness and compression. The one that fails in the field when flatness was not toleranced.
- Welded frame with gasketed removable panels — the usual compromise. The frame is welded and continuous; only the door and the service panels are gasketed, so the number of seams that can degrade is small.
The distortion risk sits with the welded option: a long weld on thin sheet pulls the seam out of flat, and the flatness you needed for the gasket is gone before the part is painted. That is a weld distortion problem, and it is why welded frames are sequenced and fixtured rather than simply welded up. The full trade-off, including cost and service access, is in our welded versus modular construction guide.
What does finishing do to the bond zone?
This is where most OEM enclosures fail EMC testing, and it has nothing to do with the gasket. Three finishes sit between the two halves of the seam and open it electrically:
- Powder coat. Fully non-conductive. Specify a masked zone at least 5 mm wide around the seam, and leave it as bare zinc-plated or cold-rolled steel.
- Anodizing on aluminium. Type II and Type III are both insulators. Mask before anodizing, or machine the bond face afterwards to expose bare metal.
- Conversion coatings. Alodine and chromate films are thin and slightly conductive, and are usually acceptable as a bond surface — but check the resistance is below about 10 mΩ across the joint rather than assuming it.
- Zinc plating. Conductive, but it corrodes. For long-term EMI stability, specify zinc plus clear chromate, or a tinned finish.
Masking is free before coating and expensive afterwards. It belongs on the drawing as a dimensioned zone, not in a note that says "keep bond area clean", because the applicator will read that as "mask as little as possible". Our surface finishing guide covers which finishes are conductive and which are not.
How do apertures, vents and windows leak?
An aperture does not leak in proportion to its area. It leaks according to its longest dimension, because a slot behaves as an antenna at the frequency where that dimension approaches a significant fraction of a wavelength. The working rule: keep the longest dimension of any opening below roughly λ/20 at the highest frequency of interest. At 1 GHz that is about 15 mm; at 10 GHz it is about 1.5 mm.
| Opening | Why it leaks | What to do |
|---|---|---|
| Cooling vents | A row of slots is a long aperture; the overall length, not each slot, sets the resonance | Use a honeycomb waveguide vent or a perforated pattern with a small pitch — see the perforated vent guide |
| Display windows | A large single aperture straight into the electronics bay | ITO-coated or fine-mesh transparent shield, bonded around the full perimeter |
| I/O cutouts | Unshielded connector openings and an unbonded panel | Conductive gasket around the bezel; bond the I/O panel to the enclosure body |
| Cable entries | The cable shield is terminated nowhere, so the shield stops at the gland | 360° shield termination at the gland, not a pigtail — see the cable gland guide |
| Door gaps and louvres | Long, narrow, and directly in the seam | Gasket the door; use a shielded louver or a labyrinth with a conductive path |
Honeycomb vents deserve a special mention because they solve the problem that usually forces a compromise: they can reach 70–80% open area — enough for real airflow — while each cell acts as a waveguide below cutoff, so the panel shields and cools at the same time. That is why they appear on dense server racks and outdoor inverter cabinets, and why they cost more than a perforated sheet.
How do you bond and ground the enclosure?
Shielding and grounding are different jobs that share hardware, and confusing them is how an enclosure ends up bonded for EMC and unsafe for fault current. Three functions get named separately on a drawing:
- Protective earth (PE) — carries fault current. It needs a dedicated stud and a conductor sized for the fault, not a seam that happens to be conductive.
- Functional ground (FG) — the signal reference, often kept separate from PE to avoid ground loops.
- Equipotential bonding (EB) — bonds every exposed metal part that could otherwise sit at a different potential during a fault.
The enclosure-side hardware matters: a riveted ground lug, a weld stud or a PEM clinch nut, each with a masked contact face. Relying on painted contact between bolted panels is the classic mistake — a paint film under an IP-rated gasket measured 4–6 MΩ across a seam on a project we reviewed, which is a fine seal and no bond at all. Product safety standards such as IEC 62368-1 and component standards including UL 467 and IEEE 837 are what these terminations get specified against; our grounding and bonding guide covers all three classes and the termination methods in detail.
What should be checked before an enclosure goes to a chamber?
Six checks, and the last one is the one that pays for itself:
- List every seam and opening. Each seam, each I/O cutout and each display window is a potential leak, and the list is the specification.
- Specify gasket part number and compression ratio — 20–30% for knitted mesh, 15–25% for conductive elastomer.
- Validate surface resistance at the bond, targeting below 5 mΩ across each seam joint.
- Verify seam flatness after welding and finishing — gap ≤ 1 mm.
- Plan I/O panel and window shielding, including 360° cable-shield termination at every gland.
- Run a pre-compliance scan. A pre-compliance chamber catches most issues for a small fraction of the cost of a formal certification failure — including the cost of a redesign after tooling.
We build to the agreed procedure and supply the supporting evidence: mask-zone callouts, gasket part numbers, bond-resistance measurements and first-article records. What we do not do is certify. Where a project needs a witnessed third-party test or a certificate naming a specific standard, that is agreed at RFQ and run by the named laboratory — the boundary is set out in our enclosure certification guide.
Frequently Asked Questions
How much EMI shielding does a sheet metal enclosure give?
The material gives roughly 60 dB at 1 MHz and 30 dB at 1 GHz for cold-rolled steel, and comparable figures for aluminium and stainless. Those numbers apply to solid continuous sheet. The assembled enclosure performs to whatever its seams and apertures allow, which is why shielding is designed at the joint rather than in the material.
Why does a powder-coated enclosure fail emissions testing?
Powder coat is non-conductive, so any coating in the bond zone opens the seam electrically. The panel is then only coupled to the enclosure body through fasteners and stray capacitance, and the seam radiates. The fix is a masked zone at least 5 mm wide around every seam, left as bare metal.
What surface resistance counts as a good bond?
Below about 5 mΩ across each seam joint is the target we design to. Thin conversion coatings such as Alodine or chromate are usually acceptable as a bond surface, but check the resistance rather than assuming it — the acceptable upper bound in practice is around 10 mΩ.
Can one gasket provide both IP sealing and EMI shielding?
Yes. Knitted wire mesh over a foam core gives roughly 80 dB and IP54 in one part; silver-filled silicone elastomer gives 60–90 dB with IP65 and above. Finger stock shields best of all but provides no environmental sealing, so it is used where shielding alone is required.
How large can a vent be before it leaks?
Governed by the longest dimension rather than the open area. Keep the longest dimension below roughly λ/20 at the highest frequency of interest — about 15 mm at 1 GHz and 1.5 mm at 10 GHz. A honeycomb vent reaches 70–80% open area while each cell acts as a waveguide below cutoff, so it cools and shields at once.
Does anodizing break an EMI bond on aluminium?
Yes. Type II and Type III anodizing are both electrical insulators. Either mask the bond face before anodizing or machine it afterwards to expose bare metal. Chromate conversion coating is thin and slightly conductive, and is usually acceptable.
Is a bonded enclosure the same as a grounded enclosure?
No. Bonding for EMC gives a low-impedance path across a seam; protective earthing carries fault current and needs a dedicated stud and a correctly sized conductor. A seam with a paint film under the gasket may seal to IP65 and still measure megohms, which is no ground at all.
Does SanJun run or certify EMC testing?
No. We build the enclosure, do the machining and supply supporting evidence, and we build to the agreed procedure. We do not issue listings or certificates of compliance, and we do not run UL 9540A. A witnessed third-party test or a certificate naming a specific standard is agreed at RFQ and run by the named laboratory.
Shielding is process control
The physics of shielding has been settled for decades. What fails enclosures is process: a bond zone that got painted, a seam that warped by two millimetres, a gasket that cold-flowed because the flatness tolerance was on the wrong feature. Every one of those is decided on the drawing and verified at first article, which is why the checklist matters more than the gasket catalogue.
Send us the drawing, the target emissions class and the IP requirement. We will return a seam specification — bond strategy, mask zones, gasket part numbers with compression ratios, and a pre-compliance test plan — with a quotation. For the airflow side of the same problem, see our server chassis airflow, EMI and strength guide.
Need an enclosure that passes its emissions scan first time? Send your drawings and target class for a seam specification and quotation within 24 hours.