Shielding of Boxes and Enclosures (Part 2)

Our former . EMC articles reviewed the principal conduction and radiation coupling mechanisms, as they affect equipment/system susceptibility, and the last one (EMC Article #5, June Issue) was addressing Shielded Cables. The present article is focusing on the shielding of equipment boxes, from the smaller hand-held devices up to large cabinets or even entire rooms.

Someone may wonder why treating separately cables shielding and box shielding? Against an EM field a shield is a shield, no matter if it is a tube or a cube … In fact, there is a significant difference: in a shielded cable, the wires are closely coupled with their tubular envelope, such as it is the mutual inductance that does the cancelling effect. In a shielded box there is no such close coupling: it is the portion of the field that goes through the barrier that gives a measure of the shield effectiveness. This long ”Box Shielding” article has been split in two parts.

 

Part 1 (Sept. issue of EE) addressed the basic approach for successful shielding: defining the objectives and selecting the proper material and thickness. This Part 2 describes the methods/hardware for obtaining the desired Shielding Effectiveness by controlling the various leakages.

4. Shield degradation caused by apertures

All the SE figures given above assume a plain, homogeneous barrier. Real life housings are never made like continuous metal cubicles: they have slots, seams and other apertures that inevitably leak. As for a chain, a shield is only as good as its weakest link; therefore it is important to know the shield’s weak points in order to match realistic objectives.

  • At low frequencies, what counts is the nature of the metal (conductivity, permeability)and its thickness.
  • At high frequencies, where any metal would provide SE of hundreds of decibels, such figures are never seen because seams and discontinuities completely spoil the metal barrier (Fig.5).
Figure 5. Field attenuation by a plain metal barrier (increases with F) compared with the attenuation through an aperture (decreasing with F).

4.1 Attenuation of one single aperture

A slot in a shield can be compared to a slot antenna that, except for a 90° rotation, behaves like a dipole. When slot length reaches λ/2, no matter how small the height (h), this non-intentional antenna behaves as a perfectly tuned dipole, i.e., it reradiates on the “exit” side all the energy that excites the slot on the incoming side. It may even exhibit a slight gain of about 3 dB. Below this resonance, the slot leaks less and less as frequency decreases. The equivalent circuit for a slot is an inductance (Fig. 6), until it resonates with the edge-to-edge capacitance, when = λ/2.

Figure 6. Effect of a discontinuity in a shield. A slot with length ℓ behaves as a dipole of same length. At low frequency, slot impedance ZAB is shorted by the low metal impedance: reflection is significant.

A simplified expression gives aperture attenuation below λ/2 resonance (from Ref. 9). It is the worst-case far-field attenuation, for the worst possible polarization (in general, actual attenuation will be better).

where,

= length (largest dimension) of aperture in mm

h, d = height and depth of the aperture (no unit given since it is the ratio /h or d/ that counts)

 

The first three terms in Eq. (6) represent the reflection loss of a square aperture, due to the mismatch of the incident wave impedance (377Ω for far-field conditions) with the slot impedance. The 100dB constant represents the 0.5λ/ℓ attenuation ( 20log 1 50.103/) of a square aperture, with an added √2 margin for diagonal polarization.

Figure 7. Attenuation (far field) for a few square apertures. If the aperture is rectangular, use the correction (fatness factor) on the right hand graph.

The fourth term is the “fatness factor” of the slot, accounting for the effect of h. Notice that h plays only a secondary role by the logarithm of /h. A slot 100 times thinner will not radiate 100 times less than the equivalent square aperture, but only 5 times less. Typical values of this factor are:

 

0 dB for = h (square aperture)

10 dB h/ = 0.1

15 dB h/ = 0.01

18 dB h/ = 0.001

 

The last term, 30 d/, directly given in decibels, is the guided wave attenuation, as it would happen in a real waveguide below its operating frequency. It has only some effect if depth “d” is a significant fraction of ℓ. For ordinary sheet-metal enclosures where d = metal thickness, this term is negligible. For small holes, or artificially lengthened holes (Fig.8) the added attenuation is significant.

Figure 8. Additional attenuation offered by lengthened holes (Waveguide below resonance).

4.2 Effect of multiple apertures leakages

A question often arises: how can we estimate the combined effect of several apertures, wether they are similar or not? Following are some guidelines for the most frequent cases.

 

a) Several apertures, scattered and not identical:

Compute A(dB) for each aperture. The global SE will be dominated by the poorest attenuation( greatest leakage). A more accurate prediction can be made by adding the leakages. Example: assume that calculations using Equ.6 has given:

 

Aperture #1: 26dB (a relative leakage of 0.05, meaning 5% of field gets through),

Aperture #2 : 14dB (a relative leakage of 0.2, meaning 10% of field gets through),

 

The total leak is: 0.2 + 0.05 = 0.25, that is a total attenuation = – (20 log 0.25) = 12dB. As predictable, the total attenuation is slightly less than A#2, the worst one.

 

b) N apertures, identical but scattered (not adjacent):

Compute A(dB) for one hole, and subtract 20 Log N. This worstcase assumes that all openings are re-radiating in phase, which is not entirely true. If there are many apertures, such as the result is approaching 0dB, total SE must be clamped to 0 dB: slots cannot cause negative loss and amplify the field!

 

c) N apertures, identical, not scattered and adjacent, with thickness of ribs t < or h,

Compute A(dB) for only one aperture. DO NOT substract 20 or 10 LogN . This is because when identical holes are separated by thin ribs, mutual cancellation occurs by the edge currents (see Fig. 9, bottom).

Figure 9. Some box typical leakages

4.3 Alterations of the ideal ”hole-in-the-wall” model

The above calculations for metal and aperture SE are assuming a rather academic situation where:

  • the metal wall has quasi infinite dimensions, or at least very large versus the source-to-shield distance such as the current density in the plain shield (without any leakage) would be uniform,
  • the reflected wave does not encounter any opposite wall, causing multiple reflections.

Reality is different:

a) Electronic cabinets have finite dimensions, causing current concentrations at the edges.

 

b) For radiated emissions, sources can be rather close from the box walls openings, such as a good part of the concerned frequency range is in a near field condition.

 

c) the box may behave as a cavity excited by internal sources, if one or several of the circuit frequencies meet the natural box resonances.

4.3.1 Effect of source proximity on aperture leakage

As mentioned, aperture SE in the near field departs significantly from its far-field value of Eq. (6). The wave impedance will differ from 377Ω, affecting the reflection term, hence a greater SE for a predominantly electric field, but lower SE for a predominantly magnetic field.

 

Discarding the case of pure E field, that is academic for virtually all radiated EMI problems, when the radiating source within a box is in nearfield conditions (i.e. distance D(m) < 48/F(MHz)), the most severe condition would be that of a near-field magnetic source. At worst, the attenuation of a slot against an ideal H field loop is given by:

Notice that this last expression is independent of frequency, as long as the near-field criterion and quasi perfect H field exist. It can be regarded as the worst low boundary of aperture SE against H-field sources.

4.3.2 Effect of box natural resonances

For a rectangular metal box with dimensions , w, h, the natural resonance frequencies of the waveguide occur every time the path length becomes equal to an odd multiple of λ/2.

 

At these specific frequencies, an empty metal box could exhibit resonances with a Q factor as large as 10 (20dB), resulting in a negative SE (an apparent “gain”). Hopefully, equipment boxes are never empty but filled with PCBs, components and cables behaving as scattered lossy elements, such as the measured Q stays within 0 – 10dB, with typical values of 6dB.

 

This doubling of the inside field results in an apparent 6dB drop of the expected SE at every self-resonance frequency. For box self-resonant frequencies that are less than twice below the cut-off frequency for the largest opening, the box SE will drop to 0dB faster than expected. Some shielding products manufacturers are offering lossy ferrite composites that can be used like an anechoic coating inside the enclosure.

5. Reducing apertures leakage to reach a shielding objective

A conductive housing is naturally an efficient barrier. All the talent of the designer should be aimed at not spoiling this barrier with excessive leakages. Leakages (i.e., poor SE) are caused by:

  1. Assembly seams at mating panels, covers, etc… (a frequent cause of SE spoiling)
  2. Cooling apertures
  3. Viewing apertures
  4. Component holes: fuses, switches, keyboards, shafts
  5. Cable or miscellaneous conduit penetrations

5.1 Mating panels and cover seams

The general, simple rule is: all metal parts should be bonded together: a floated item is a candidate for re-radiation. For cover seams, slots, and so forth, how frequently they should be bonded depends on the design SE objective. Figure 7 shows that a 10 cm leakage is worth about 20 dB of shielding in the neighborhood of 100 MHz. If the goal is closer to 30 or 40 dB, seams or slots should be broken down to 3 cm or 1 cm. For permanent or semi-permanent closures, this means many screws or welding points or an EMC conductive gasket.

 

For covers, hatches and such, this means flexible contacts or gaskets. In any case, it is always a safe practice to designing fold-over shapes to the cover edges. With a sufficent overlapping, a waveguide “labyrinth” is formed that is adding some penetration loss. By doing this, one could complement the use of gasket, or even avoid them.

Figure 10. Leakage reduction by frequent seam bonding (for moderate shielding needs).

 

The following is a sequential list of these solutions. As efficiency increases, cost increases as well.

 

• If only minimal SE is needed, in the 0 to 20 dB range, the simplest technique is to have frequent bonding points and, for covers, short straps made of flat braid or copper foil as in Fig.10. This solution bonds only on the hinge side, but provided that no sensitive or noisy cables and devices are located near the opposite side of the hinge, this can be sufficient. For the unbonded opposite side, a wise precaution is to use several grounded fasteners. The λ/20 criterion of Fig. 10 means that, for a maximum EMI frequency of 100 MHz, the distance between ground straps should stay within 15 cm for a 20 dB shielding objective, and up to 45 cm if a 10 dB reduction is sufficient. However, for emission shielding, this criteria would imply that the emission source inside is at a distance greater than 45cm from the leaky seam, which may not be the case (see previous “effects of source proximity”).

 

• If bonding only the hinged side leaves an excessive length of leaky seam, more bonding points are necessary. In this case, Fig.11 shows an example of a soft springs that can be scattered along the cover edges. For durable performance, the spring contact riveting must be corrosion free. Several types of such fingerstocks are available, such as lowpressure, knife edge and medium-pressure styles. Provided an adequate control of pressure through tight manufacturing tolerances, they are extremely dependable. A third technique, shown in Fig.12 is an interesting alternative taking minimal surface preparation. The grounding buttons, mounted by press fit or a threaded stud are compliant to gap variations.

Figure 11. Maintaining shield integrity by evenly spaced, flexible bonding points ( from Ref 11)
Figure 12. Left: Press-Fit grounding buttons (LAIRD co.) Right: soft grounding pads (Chomerics/Parker div.)

 

If a higher grade of shielding is required (20 to 60 dB), a continuous conductive bonding of seams is necessary, since an SE of 40 dB at 300 MHz (λ/2 = 50 cm) would require screws or rivets spacing of less than 1cm! Such continuous conductive joints are available in several forms and stiffnesses (Fig.13). Metal braid or mesh-type gaskets provide higher shielding, close to or beyond the upper side of the required SE range.

Figure 13. Compressible RF gaskets and mounting styles.

Hollow elastomer gaskets are less expensive to use because their elasticity compensates for large joint unevenness and warpage. The counterpart for this is a lesser contact pressure, hence higher resistivity; it is a solution for the lower side of the SE range. A good quality mating surface can be made by applying conductive tape over the metal surface before painting, protected by masking tape during the painting process, after which the masking tape is peeled off.

 

The contact resistance of such conductive tapes after hard compression must not exceed few mΩ per square. For applications with long term exposure to harsh environment, beware that the conductive adhesive backing of these tapes does not behave well with aging, with a tendency to polymerization of the glue after several years.

Figure 14. Fingerstocks with 100% perimeter coverage (Courtesy of LAIRD Co.)

Finally, if an even higher hardening is needed, the ultimate solution of Fig.14 is the most efficient because 100% of seam becomes a good conductive joint; it is the one favored for shielded rooms. Besides its direct cost, it adds the need for a strong locking mechanism ensuring even pressure on all of the spring blades. This method is applicable to both rotating (hinged) or slide-mating surfaces. Whatever the choice, conductive elastomer core, mesh or spring fingers, all gaskets requires an adequate design of covers and box or frame edges to provide:

  • a smooth seating plane or groove, with well conductive surface finish, for the gasket
  • proper mechanical tolerances avoiding gasket overpressure at some places (lower tolerance gap) causing permanent gasket crushing, and underpressure at others (higher gap) resulting in insufficient contact.

Good surface conductivity is paramount to an effective bonding. Mating areas must be paint-free, bare metal is generally treated against corrosion. These treatments are not all good conductors:

  • Anodized aluminium is non-conductive
  • Bichromate olive-green, and most aluminium treatments make poor, unstable contacts
  • Alodyne provides a fair conductivity, but the process has been banned due to its toxicity. It has been replaced by neutral chromate treatments, like CHROMITAL/Surtec.
  • Zinc, nickel or cadmium plating provide a good conductivity

 

For metallized plastic housings, the seam treatment needs only to be proportionate to the box skin SE, which is generally more modest (typically less than 50 dB below 100 – 200 MHz). If the conductive coating is resistant to abrasion, mating edges can be designed to provide an electrical continuity, without the need for gasket. This is done by using tongue-and-groove or other molded profiles for assembly (Fig. 15). The flexibility of plastic provides the necessary contact pressure of the conductive surfaces.

Figure 15. Metallized plastic box design. Conductive coat should extend far enough into the tongue-and-groove shape to make a continuous contact, but not too far (avoid ESD problems). Bottom: avoid long, protruding screws, likely to be not, or poorly grounded. They can become re-radiating antenna for ESD or high RF.

5.2 Shielding for cooling apertures

Several techniques can restore shield integrity at convection or forcedair cooling vents (Fig.16):

Figure 16. Methods for shielding cooling apertures

a) Break large openings into several smaller ones. This has the advantage of virtually no cost if the holes are made during stamping or molding of the housing. It also can put the source at a relative greater distance, compared to the aperture size, eliminating some proximity effect. The SE improvement is simply, ΔdB = 20 log N, if N is the number of identical holes that are replacing one larger aperture. This is done by replacing long slots with smaller (preferably round) apertures. If some depth can be added to the barrier such that d/ℓ >1, the waveguide term becomes noticeable, improving SE.

 

b) Install a metal screen over the cooling hole. This screen has to be continuously welded or fitted with a conductive edge gasket having an intrinsic SE superior to the overall objective.

 

c) Install a honeycomb air vent if an SE greater than 60 dB is required above 500 MHz and up to several GHz, along with a low aerodynamic pressure drop.

5.3 Shielding for viewing apertures

LCDs, alphanumeric displays, tactile keyboards, meters and the like are often the largest openings in an equipment box, offering the poorest SE of the whole housing. On the other hand, high-frequency sources are seldom mounted right on or behind display panels. Compared to the typical RF “hot plate” of a filter mounting plate or I/O connectors area, experience shows that most equipments can tolerate rather large, unshielded apertures on their user’s display panel, while a 10 times smaller slot in the cable entry zone would leak significantly. In a sense, the intrinsic SE of any aperture being calculable, its radiation still depends on whether it is excited. Since one does not know in advance how RF currents will distribute on the box skin, it is safe keeping the assumption that viewing apertures are as prone to leak as any other one. The shielding solutions are:

 

a) Finely knitted or woven wire mesh, on top of, or sandwiched in the glass, plexiglass or other material. Densities of up to 12 wires/cm (knitted) or up to 100 wires/cm (woven) are obtainable. The performance can be derived from Fig.17. The denser mesh offers more SE, but this is at the expense of transparency. A modern alternative is the litho-photographic deposit of a thin copper mesh.

Figure 17. SE of wire mesh for cooling or viewing apertures, in far-field conditions (Distance in meters > 48 /FMHz)

b) Transparent conductive film, where a thin film of gold or indium tin oxide (ITO) is vacuum deposited on the transparent substrate. The film thickness has to be low (10-3 to 10-2 microns) to keep an 80 to 60% optical transparency, but thinner film means more the surface resistance. Typical transparent coatings have surface resistivities ranging from of 50 to 5 Ω/sq, translated in far-field SE of 10 to 30 dB. Near E-field SE would be better. One recent promising technique (Ref. 8) is the deposit of a silver nanoparticles emulsion, creating a random mesh pattern.

 

c) Shielding the display from the rear side: the display is shielded behind the box panel by a doghouse, which is equipped with feedthrough capacitors for connecting wires.

 

In all three of the solutions described above, an EMI gasket is needed at the shield-to-box joint. Often, one such fitting is already provideded by the shielded window vendor.

 

5.4 Shielding the component holes

Holes for potentiometers shafts, switches, lamps, fuseholders, etc. generally are small. But their mere presence in the middle of metal pieces that have picked up CM current from inside the box will enhance the radiation phenomenon. The shaft, lever or fuse will act as a monopole, exiting via a coaxial line, capable of transmitting radio signals. As far as FCC, CISPR and other civilian limits are concerned, component holes are seldom a problem because of the relatively small leakage. With MlL-Std-461 or TEMPEST emission limits, component holes can be significant contributors to EMI radiation. The solutions are:

  • Use nonconductive shafts or levers, and increase the hole depth with a piece of metal tube to create a waveguide attenuation.
  • Use grounding washers or circular springs to make electrical contact between the shaft and panel.
  • Use shielded versions of the components.

5.5 Shielding of cable penetrations, connectors and nonconductive feed-throughs

Last but not least, this breach in box skin integrity is a serious concern, since cables are the largest potential RF carriers in the entire system. Wether to shield or not the cable penetrations depends on the cable entry hole contribution to SE, as follows:

 

a) Calculation of box SE shows that the cable exit naked hole is tolerable. However, even if the box SE is correct, the cable can behave as a radiator. If the cable needs to be shielded for radiation and/or susceptibility, its shield must properly terminate at the barrier crossing via a 360° clamp, ultra-short strap or, best of all, a metallic connector shell. If the cable is not shielded but still is a threat, each one of its wires must have been filtered: there is no point in shielding the hole.

 

b) Calculation of box SE show that the cable naked hole is not tolerable. We must use shielded cable/shielded connectors, creating a sort of shielded enclosure for the whole interconnect cabling system. A trade-off would be to use unshielded cable and to block aperture leakage with a shielded and filtered connector receptacle. This would recreate a recessed shield barrier behind the cable hole. The interface of cable shields at box penetration is a topic indissociable from cable shielding.

 

Some other exit/entry ports exist for nonconductive lines such as pressure sensors, fluid lines, fiber optics, etc. If the tube is nonconductive and the SE of the naked hole is not sufficient, this type of leakage is easily reduced by using the waveguide effect. For fiber optics, transmitters, and receivers, metallic packages are available with appropriate tubular fittings.

5.6 Aggravated effect of box leakages near a cable penetration

When a cable exit is close to a slot leakage, the slot attenuation is locally less than its theoretical far-field value. In emission situations, the exciting source inside can couple to the first centimeters of the outer cable segment (Fig.18) by a mechanism that is closer to magnetic or capacitive crosstalk. Using a current probe, CM currents will be found on the cable, even after I/O filters or ferrites have been installed, turning the cable into a secondary antenna. Such leakages in a “hot plate” area must be controlled very carefully.

Figure 18. Excitation of I/O cables by a nearby slot ( from Ref. 11)

5.7 specially hardened equipments housings

Several vendors of ready-to-use racks and cabinets offer EMI-shielded versions. Even a standard steel or aluminium cabinet with some simple precautions (paint-free and zinc- or tin-plated contact areas, metal-mesh air filters) provides some degree of shielding.

 

Equipped with EMI gaskets, shielded air vents, 100% welded frame joints and piano-hinged doors for tight seam tolerances, these cabinets offer SE > 60dB up to 150MHz, 40dB at 500MHz, at a cost increase of $350-$500 (2010 prices) compared to standard version. A word of caution when dealing with emission problems: some SE data shown on shielded cabinet ads are measured per MIL-Std-285 method with a radiating source outside, 30cm from the doors. This may not replicate cases where the source is inside. Proximity effects can cause lower SE than expected, especially when inner devices and cables are near the cover seams.

Figure 19. Example of a commercially available shielded cabinet (Source: Equipto, Aurora IL, USA)

6. Shielding components for mass-production, consumer products

Since late 1990’s, the technical evolution has brought a huge number of miniature, popular devices using high speed digital circuits and wireless RF techniques, operating at >1 GHz. This has urged the development of new shielding hardware. These shielding items have to be economical, lend themselves to mass production techniques like a production of 10,000 or more devices/day, and provide performances which were barely attainable by the costly military electronics of the 1980’s. Such SE hardware includes:

  • Heat-formable, shrinkable films: They are generally polymer-fiber films coated with a metal mesh with low fusion point (3M #6100) or a conductive ink grid (G.E. Lexan). When heated at 150 -200°C after diecut, the film conforms itself to the 3-D shape of the plastic housing that needs to be shielded. With thickness of 0.2 to 0.8mm, foil resistance is in the range of 0.1-1Ω/sq., and the textured nature of the metal content prevents resonant cavity effects.
  • Thin, form-in-place gaskets: Conductive caulking can be applied in a regular cord-gasket with diameter as small as 0.3 -1mm, acting both as EMI and weather gasket. They can be deposited with an automatic dispenser, or printed in a single operation like an ink, conforming to very intricate shapes.
  • PCB component shields: Five-sided cans, stamped from tin-plated steel or brass, are available off-the-shelf in standard shapes, with heigths as low as 3mm. They can be wave-soldered to a printed ground belt around the specific component, or PCB zone that needs to be shielded. The PCB ground plane acts as the sixth side of the enclosure.
Figure 20. Partial shielding at the IC/module level (from LAIRD Co)

7. Large shielded rooms and shielded buildings

Certain applications like EMC testing, medical investigations (NMR) or protection against eavesdropping (Tempest) require rooms, or even entire buildings that are fully shielded with high requirements of 60 to 100dB. Such shielding has the same basic aspects as reviewed before, but with additional constraints: the shielded zone must keep all the access and conveniences of ordinary facilities: single or double doors, windows, air conditioning ducts, lighting etc… that require special hardware elements.

 

Shielded rooms up to 30m² of floor area are available as self-supporting cages, made of prefabricated galvanized steel wall panels, door frames etc… that are assembled on site. Larger rooms are shielded by installing copper foil layers on the walls and ceiling, and prefabricated or customized shielded doors and windows.

8. Summary of radiated EMI control via box shielding

 

1: When the best affordable measures have been taken at PCB and internal wiring level, the equipment housing is the ultimate barrier against radiated emissions/susceptibility.

 

2: Until the last hole or slot is checked, the best metal box could appear to be useless as a shield.

 

3: For metal housings:

  1. Bond all metal parts (a floated item is a candidate for re-radiation).
  2. Avoid long seams and slots: a 30 cm seam is a total leak at 300MHz and above.
  3. Use gaskets, or waveguide effect: design fold-over shapes for the cover edges

4: For plastic housings: use conductive coating ≤ 2 Ω/sq, then treat like a metal housing. Avoid long, protruding screws inside.

 

5: Maintain or restore shield integrity at: cooling holes, viewing apertures, component / cable penetrations

 

6: Beware of noisy circuits or cables close to seams and slots: they degrade an otherwise sufficient SE

 

 

Michel Mardiguian
EMC Consultant, France

 

Part one of this article was published in Electronic Environment, and at electronic.nu

References

  1. Casey, K.F, Shielding of wire-mesh screens, IEE/EMC Transactions, Aug 1980 Vol 30
  2. Leferink, F. Shielding Basics, IEE/EMC Symposium, 2010, Ft Lauderdale
  3. Mohr, R. “Schelkunoff approach to shielding”, IEEE/EMC Symposium, Hawaii 2007
  4. Ott, H., Electromagnetic Compatibility Engineering, Wiley, 2009 (replacing ”Noise Reduction Techniques”/Wiley, out-of-print)
  5. Schelkunoff, S. Electromagnetic waves, Van Nostrand, Princeton, 1943
  6. Schultz, R. Shielding theory and practice, IEE/EMC Transactions, Aug 1980 Vol 30
  7. J. Muccioli, in Radiation from Microprocessors, IEEE/EMC Sympos, 1990 and 1997
  8. White, D.R.J, Mardiguian, M. Electromagnetic shielding, ICT Inc. Gainesville, VA, 1988
  9. White, D.R.J, Electromagnetic shielding materials, Don White Consultants, Gainesville, VA, 1988
  10. Mardiguian,M. Controlling Radiated Emissions”, Springer NY, 2014