Conformity Magazine

See this article in our Digital Edition
Download a pdf of this article (right click to save the pdf)

It is not difficult to do a good job of shielding when you can use the ï¿½iceboxï¿½ approach (everything inside and zip it up tight). But it is a lot less obvious how to handle shielding when this is not feasible.

A typical case is with a complex system that cannot be fully enclosed, for one reason or another. You may have diagnostic equipment that is frequently handled by the operator, so parts of the equipment need to be open during operation. You may have an internal RF or optical source that must penetrate the enclosure. You may have cables connected to the patient. In such cases, you may be forced to use selective shielding techniques and arrange the internals so that you minimize the need for shielding effectiveness.

Letï¿½s take a look at the various aspects you may be facing. In this article, we wonï¿½t be concerned with shielding effectiveness of materials or appropriate gasket selection. These factors are important, but much more often we are concerned about the openings and penetrations to the shield.

The shielding approach needs to be planned from the beginning. All too often, the problem is such that major structural changes are required in order to resolve the problem, changes that could have been avoided with a little forethought.

The Basics of Shielding
We need to start with the basics of shielding, no matter how you expect to contain the interference. Shielding is needed to block RF fields, whether to keep internal emissions from escaping the enclosure, or to keep external RF sources from penetrating the enclosure.

Figure 1 shows a shielded enclosure with openings and penetrations. For effective shielding, you need to keep the openings small enough to prevent RF from passing through. Shielding effectiveness (SE) is given by:

SE = 20*log(Î»/2L)

Where Î» is wavelength and L is the longest dimension of the opening. (Wavelength, Î» = 300/f, where Î» is in meter and f is in MHz.)

As can be seen, SE is zero at L = ï¿½ wavelength or longer. We consider 1/20 wavelength to be a practical maximum opening, giving you a 20 dB shield.

Figure 1: Entire enclosure shielded and wire penetrations shielded or filtered

From Table 1, we see that at lower frequencies, the 1/20 wavelength criteria is easy to meet, but above 100 MHz, openings become increasingly critical. As a crude rule of thumb, we figure frequencies above ten times the clock frequency to be the approximate upper limit for the 1/20 wavelength criteria. Above that, the emissions and immunity needs are diminishing rapidly.

Frequency	Î»	Î»/20
1 MHz	300 m	15 m
3 MHz	100 m	5 m
10 MHz	30 m	1.5 m
30 MHz	10 m	50 cm
100 MHz	3 m	15 cm
300 MHz	1 m	5 cm
1000 MHz	30 cm	1.5 cm

Table 1

Note that the concern is the longest dimension of the opening; a round 10 cm hole will leak about the same as a 10 cm slot. If you canï¿½t get the slot length small enough, you need to consider using EMI gaskets to close the seams. If you have a ventilator opening, you may need to consider using a conductive screen or perforated metal to reduce the maximum dimension to an acceptable level. The number of holes is not a key parameter; itï¿½s the biggest hole that sets the pace.

The second aspect of shielding is the penetrations, namely the data and power lines that penetrate the enclosure. Any interference currents on the wire will freely pass through the enclosure boundary, from either direction. Your solution is to shield the cable or filter the wires in the bundle. Where isolation is required, as would be the case with a patient connected cable, there are severe limitations as to how the cable can be shielded or filtered.

Note that this concept is completely general. You are free to select the boundary of the shield, at the entire box, at the circuit board, or at the chip. If the icebox approach isnï¿½t feasible, you will be forced to go inside the box and selectively shield certain functional modules, say, the power supplies, the microprocessor modules, and the analog input circuits.

Selecting the Shielding Material
Now that we have a handle on what constitutes a shield, we need to look at the shielding materials. For most shielding needs, a thin conductive shield will be adequate (low frequency magnetic field is the only notable exception). Most metals provide ample shielding effectiveness.

But for various reasons, plastic is often the enclosure material of choice, and this means making it conductive. Conductive plastics (plastic loaded with metal filaments) have been available for many years, but have not achieved widespread use. The shielding effectiveness is usually adequate, but it is hard to get the necessary conductive contact at the boundaries. More commonly, a conductive coating is used, such as electroless plating or vacuum plating. This provides you with good shielding, but does require that the mating surfaces conductively close all around the perimeter.

In order to work, you need to design the mating surfaces so that they are rigid enough to make contact without excessive flexing, and you need to bring the conductive material right up to where the surfaces mate. A simple dovetail requires bracing to keep the mating surfaces straight; tongue and groove works better. Whichever method you use, continuous mating is key, and may require EMI gasketing to finish the job. This isnï¿½t as onerous as you might think with the gasket selection available today (all cell phones are gasketed!).

If you are using a conductively coated plastic for the outer skin, the coating provides numerous discharge points for ESD. If the shield is well done and complete, this is of little consequence, but if the shield terminations are haphazard, discharge currents can be troublesome. Generally, this is a problem only where the plastic is exposed to direct touch, and this is usually only with the outer shell.

Planning Your Shield
Now that we have the shielding basics, we need to decide where to put the shield. If you can shield the entire enclosure with an icebox, you can effectively eliminate emissions and RF susceptibility, leaving only internal self compatibility to worry about. In practice, this concern is a good start with designing your internal shield.

Lacking a good outer shield, you will need to implement one or more internal shields, along with shielded or filtered cables. You might choose to shield the power supply, the microprocessor, and the analog section, an approach shown in Figure 2. The question is what do we need to shield. Any electronics present a potential problem. You need to start by identifying the critical circuits, including internally noisy sources and circuits that are vulnerable to external interference.

Figure 2: Individual modules shielded and wire penetrations shielded or filtered

For emissions:

Clocks, data and address busses
Video data
Switching power devices of all types, including PWM drives
SCR and Triac power controls
Florescent lamps
RF heaters

For immunity:

Analog (Op-amps and voltage regulators) for RFI
Digital for ESD and power transients

Where is the shield needed?
Once you have identified those circuits that need attention, you need to select the shield boundaries. This is not a clear decision, but a little thought will narrow the choices. Each shield needs to accomplish the following:

Enclose the selected circuit up to the highest vulnerable frequency, whether emissions or RF immunity. Modules might be a power supply, a microprocessor board, or even a chip on the circuit board. Where noisy and vulnerable circuits exist in close proximity, they may need to be isolated from each other.
All lines entering the shield need to be either filtered or shielded. Power and audio frequency data lines can usually be filtered, but high speed data lines will probably need to be shielded.

Once you have decided which elements need to be shielded, your shield boundaries would consider which would involve the fewest cables and which are physically the easiest to implement.

Designing an Enclosure
If you follow the above rules for shielding and filtering, you will have little problem meeting EMC requirements. Realistically, you will probably take some short cuts (not that we advise such an approach!).

If we continue with the approach that we have identified and addressed the high risk modules and lines, what do we do with the low risk elements? We canï¿½t cover all possibilities, even if we wrote a book on it. But we can adopt an alternate approach that will minimize the demands placed on the shield

Ground
In principle, shields donï¿½t need to be grounded. In practice, when using multiple shields in an enclosure, things go much easier if you have a good ground system. This doesnï¿½t involve earth ground, but is basically a local reference we call ground, usually (but not necessarily) connected to the enclosure which, in turn, is often connected to earth ground. But a shielded enclosure doesnï¿½t need to be connected to earth ground to work. Suffice it to say that, for an enclosure, ground is a substantial amount of metal, preferably planar, that circuit modules can be bolted to.

The textbook case of a good high frequency ground is a ground plane, which provides a low impedance path between two points. It is hard to overemphasize the merit of a ground plane; at higher frequencies (say, 100 MHz), the impedance of even a short wire length (a couple of centimeters) is easily 1000 times as high as that of a plane.

But we donï¿½t have to have a flat plane to achieve a low impedance ground system. You can bend it like an L, splice it like a T, shape it to meet form factor requirements. You can cut holes to allow wire harnesses or mechanical members to pass. If you have to splice several metal members, make sure the mating surfaces are conductive, and use lots of fasteners - continuous welds are ideal, but admittedly not usually feasible. The key is to have a substantial wide surface.

You could use the metallic base as a start, with metal risers bolted down as needed. Mechanical engineers like to use various metal stock like Tinker Toys; unfortunately, they make poor ground paths. Better to have larger sheet metal stampings.

Movable metal members are unacceptable for any ground, even if wired down with a green wire. Screw threads, latches and bearings are unacceptable as ground. Anodized, paint or other nonconductive coatings are unacceptable for mating surfaces or ground contact, but permitted elsewhere.

The bottom line is, to get a good ground, use planar metal members and bolt them together at every opportunity. Do not use wires for RF grounds. If you need to use a flexible metal interconnect, use a flat metal strap, with length to width radio no greater than 5:1.

Shielding
Assuming you are going with some number of shielded modules, these modules should be bolted to the ground plane. Do not use green wire for this purpose. If the module cannot be mounted onto a ground plane, bring the ground plane up to the module. Safety grounds are always permitted, but they cannot be relied upon to provide high frequency grounding.

Wire Routing
We strongly recommend putting your wire harnesses and cables under control drawing. When we open an enclosure and see a ratï¿½s nest of wires, we immediately draw two conclusions: first, that we have significant parasitic antennas, suitable for both radiated emissions and susceptibility; and, second, that there will be poor repeatability EMI performance between any two similar boxes. You will get better and more repeatable performance by following these basic rules:

Route all wires, including signal and power cables, along the ground surface to minimize loop areas. Avoid hanging them in midair. If you are using part of the metallic outer skin as a ground surface, make sure it is very well grounded to the above mentioned ground structure.
Group cable typesï¿½Noisy power or data lines should be separated from vulnerable data lines. Avoid long parallel runs, or shield one or both cables.
Use twisted pair to route power lines, switches and indicators.
Avoid daisy chaining power to various modules. It is particularly important to separate noisy and sensitive loads.

Power and Data Entry
Mount connectors and power entry at the enclosure shield boundary (this would also normally be in intimate contact with the ground mentioned above). This includes any filters, transient protection and connector mounts. This could be to the ground plane, or to a plate directly connected to ground.

From a systems standpoint, all external data cables (those destined to another piece of equipment) and power cables should enter the enclosure in contiguous locations, to minimize facility ground currents passing through the enclosure. This would be to the ground plane or to the plate mentioned above.

Avoid having any enclosure seams in the immediate proximity of the connectors. Donï¿½t skimp on the shield here, of all places.

Shielded Cables
Shielded cables, whether internal or external to the enclosure, need to be designed appropriately, or they wonï¿½t work as intended.

The cable shield needs to be circumferential to the connector. Pigtail wire grounds and drain wire grounds are not acceptable. Foil shielded cables are difficult to terminate adequately, and we avoid using them.

Internal cable shields should be clamped directly to the ground plane as close to the cable end as possible. Do not use pigtails to ground the shield. If the cable connection is too far from the ground plane, peel the cable sheathing back so you can make a direct connection, and leave the cable shield extended to the end of the cable. Cable shields may be grounded up to 1/20 wavelength from the end.

Except for low frequency electrostatic shields, the cable shield needs to be grounded at both ends. Single point grounded shields are unacceptable for cable lengths longer than 1/20 wavelength.

External patient connected cables cannot be grounded at the patient end, and the shield termination at the equipment end will need to go to isolated ground. This places additional burden on signal filtering.

Summary
EMI shielding for medical electronic devices poses some unique problems. Patient isolation requirements make it difficult to effectively shield or filter signal lines. Complex equipment can be difficult to shield effectively.

A combination of selective shielding, module grounding, cable routing and shielding will help minimize the demands placed on the enclosure shield. n

William D. Kimmel, PE and Daryl D Gerke, PE, are principals of Kimmel Gerke Associates, and can be reached through their web site at www.emiguru.com.