An MEM 1996 Designer's Guide Feature
Some 12 common EMI-related mistakes account for most typical EMI problems in medical devices; with careful planning in the design stage, they can be easily prevented.
Although some EMI problems are quite subtle, most are readily understood and arise as a result of the device designer's not considering the basic principles of electromagnetic compatibility and planning for EMI control during product development. The following article describes 12 EMI-related mistakes that are commonly made by medical electronics designers. Perhaps 75% of the problems we have encountered in our consulting practice fit into one of the categories described. Accordingly, if these issues are adequately addressed during the design stage, manufacturers can reasonably expect to avoid most EMI problems.
These beryllium-copper gaskets are used for shielding sophisticated electronics. Photo courtesy of Tecknit, Cranford, NJ.
Until recently, the principal concern in EMI control was to reduce emissions from electronic equipment, specifically for the purpose of protecting licensed communications systems. In the United States, the Federal Communications Commission regulates emissions from equipment but has exempted medical electronic equipment from its test requirements. As knowledge of EMI's potential to damage sensitive circuits has increased, however, immunity from external sources of interference has received more attention, especially in the field of medical electronics, where FDA has taken a strong position.1, 2
In addition, devices exported to Europe need to meet the requirements for both emissions and immunity set forth in the IEC 1000-4 series of standards developed by the International Electrotechnical Commission.3-5 These standards will be superseded by the IEC 601 standards, which will become mandatory in 1998.6 For most medical electronics, the IEC immunity requirements, which cover radio-frequency interference (RFI), electrostatic discharge (ESD), and various power disturbances, will be the most difficult to meet. All of these regulatory requirements should be kept in mind when considering the following common EMI problems.
Mistake #1: Ignoring Capacitor and Inductor Resonance
When attempting to achieve EMI control, designers need to be alert for parasitic or second-order effects, which become especially important at higher frequencies. Parasitic elements in the components cause capacitors and inductors to resonate. As seen in Figure 1 below, lead inductance in series with capacitors results in series resonance. Interwinding capacitance in inductors results in a parallel resonance. Above these resonance frequencies, which are quite low, typically in the 1-50-MHz range, the components switch roles, with capacitors becoming inductors and inductors becoming capacitors. When these components are used to filter critical circuits, resonances sharply restrict their effective frequency range. In fact, most filters are completely nonfunctional at higher RFI test frequencies.
Figure 1. Typical resonances for capacitors with 1/2-in. leads shown as impedance versus frequency.
Solution. To avoid problems, designers should select components based on their functionality at the frequency range of interest and mount them carefully to maintain the integrity of the filter. Capacitors should have a low effective series resistance (ESR), which is a typical characteristic of ceramic versions. Surface-mount and feed-through components have better high-frequency performance than through-hole capacitors. Leads and printed circuit board (PCB) traces should be as short as possible, and lead inductance can be minimized by using fat traces, if necessary.
When inductors are needed, ferrites are the best choice for EMI control because they have high resonant frequencies, becoming very lossy at resonance and absorbing unwanted energy. In contrast, wound inductors, with either an air core or a permeable core, resonate at low frequencies. Furthermore, they are open-flux-path devices and need to be placed carefully in order to minimize coupling to adjacent components and PCB traces. If a wide frequency range is needed, small and large inductors, including ferrites, can be wired in series to minimize the impact of resonances.
Mistake #2: Using Double-Sided Circuit Boards
Ground impedances and loop areas become a significant concern at frequencies above 5 MHz, which includes both RFI and ESD frequencies. Regardless of the circuit board design, it is impossible to keep the ground impedances and loops sufficiently low to control ESD (300-MHz-equivalent frequency) on a two-sided board (see Figure 2 below). Once ESD is on the board, the situation is irreversible and it is practically impossible to maintain the circuit's functionality. A similar situation exists for RFI immunity on sensitive analog circuit boards, where open-loop areas create fine antennas.
Figure 2. The high impedances on a two-sided circuit board make it susceptible to damage from ESD and RFI.
Solution. Although interference currents can be intercepted outside board boundaries, it is much more effective to use multilayer circuit boards to drastically reduce ground impedances and loop areas. As a rule of thumb, multilayer boards are 10 times less sensitive to RFI than two-sided boards. With increasing attention being paid to EMI, and more sensitive high-performance applications being implemented, it has become very risky to even attempt to use double-sided circuit boards for medical electronics. In fact, if a double-sided board cannot be fully shielded, it is unlikely that satisfactory EMI performance can be achieved.
Mistake #3: Failing to Control Bandwidth on I/O Lines
Exposing input and output (I/O) lines to high levels of high-frequency energy generates excessive emissions and cross talk. A failure to control the bandwidth on input lines makes them susceptible to RFI. Low-level analog signal lines are especially vulnerable. If high-frequency energy is input to sensitive analog chips, it is demodulated and the erroneous signals are output to the next stage. The argument often made by designers that amplifier bandwidth is only 10 kHz is not valid. High-level RFI (>1 V) will easily overdrive amplifiers with millivolt or microvolt sensitivity, regardless of the bandwidth. Interference from a 150-MHz radio is capable of upsetting input lines, output lines, reference lines, and power/ground lines.
Solution. Designers can limit bandwidth to that needed to meet system requirements by using filtering (see Figure 3 below). Filter selection should be based on the products' effectiveness at the highest frequency of concern. Low-frequency analog circuits should have high-frequency filtering even if the perceived bandwidth of the amplifier is low. RFI on off-board signal lines, most notably from patient-connected sensors, must be prevented from reaching the chip. Therefore, high-frequency filtering should be used on all lines exposed to external interference, including low-level input lines, reference lines, and power/ground lines. Even high-level analog output lines are not immune, and should be filtered if connected to the outside world.
Figure 3. Controlling bandwidth minimizes emissions and enhances device immunity.
Mistake #4: Failing to Protect Reset Lines
Reset lines are notoriously subject to interference events such as ESD and electrical fast transients (EFTs), which may cause a loss of data or a loss of function at a critical time. Why are these lines so vulnerable? The principal reason is that they are ultimately connected to exposed locations.
At a minimum, every microprocessor requires a power-up reset. Thus, any power glitch that gets past the power supply is a potential culprit in an inadvertent reset. The additional reset lines that are available to the device operator are susceptible to ESD transients, which can also trigger a reset incident. The voltage supervisor, which is intended to provide a clean reset and to shut down when the power supply voltage dips, typically to 4.7 V for a 5-V supply, can also be the source of an EMI problem because it is now the most sensitive chip on a digital board.
Solution. Medical electronics designers must take extra pains to protect all reset lines from transients (see Figure 4 below). Filters should be used wherever a reset line enters the board (especially from switches). The voltage supervisor can be protected by decoupling it at Vcc and at the reference input and adding a ferrite in series with the power feed to the chip.
If a problem does occur, it may be traceable to the power supply--power transients can trick some power supply regulators, resulting in a slow voltage sag. Installing a filter after the power supply will not be effective in such cases; transients have to be intercepted before they reach the regulator.
Figure 4. Reset lines and the voltage supervisor need protection against power glitches.
Mistake #5: Failing to Consider Component Placement
The functions of circuit-board components extend beyond their boundaries, resulting in coupling to adjacent circuits and structural elements. For example, the magnetic field from an open-flux-path inductor extends well beyond the boundary of the inductor itself, coupling to adjacent inductors, wires, and circuit-board traces. Electric field coupling also occurs from the high voltage side of filter components and from fast switching devices to nearby elements. Common coupling paths are from an inductor or ferrite filter to ground, circuit-board traces, heat sinks, and connectors. Capacitive coupling between a heat sink and switcher in a power supply is a classic example, and involves not only switcher energy but also external interference transients. Nowhere is the problem of stray capacitive and inductive coupling paths more important than in medical electronics, where the common-mode noise that results from this coupling is extremely difficult to filter.
Solution. During product development, device designers should identify potential coupling paths and then take steps to minimize them through proper component placement. Spacing between switching elements and adjacent parts should be maximized, Faraday shields should be inserted at transformers and heat sinks, loop areas should be closed, and magnetic coupling paths should be oriented orthogonally.
Mistake #6: Letting an Autorouter Make All Routing Decisions
While it may be unfair to generalize about autorouters, most are not attuned to the need for careful trace routing on sensitive PCBs. In fact, autorouters have been known to make decisions that even a neophyte performing manual routing wouldn't consider. The problem lies with the routing algorithm, which blindly goes down the net list and routes accordingly. The last signals to be routed may follow some very circuitous paths. The result is often excessive circuit-board emissions caused by poor clock routing (see Figure 5 below), differential signal lines separated on the board, and I/O lines not routed directly.
Figure 5. Examples of poor trace routing created by an autorouter.
Solution. Designers can partition a board so as to facilitate good trace routing. The clock source should be close to receivers, the I/O lines should be close to connectors, and clock chips should be positioned away from the board and the I/O ports. Critical lines, which include clock lines, I/O lines, balanced lines, buses, edge-triggered strobes, and sensitive analog lines, should be routed manually. Then, after the autorouting is complete, a visual inspection should be performed to ensure that no poor decisions were made.
Mistake #7: Providing Inadequate Returns
When several signals share one return line, noise will accumulate, possibly causing signal failure. The high impedance of return lines means that the aggregate of the signal line currents will add up to significant voltages on the shared return lines. In addition, any interference currents will also contribute noise voltages to the lines. Finally, the large loop areas that often accompany a ribbon cable encourage excessive current pickup and emissions.
Solution. To prevent signal return problems, designers can use multiple return lines and distribute them throughout the cable (see Figure 6 below). A common rule of thumb is to use one return per signal line for high-speed (100-MHz) designs and one return for each five signal lines for medium-speed (10-MHz) designs. The effectiveness of the return can be maximized by placing critical lines adjacent to return lines. Dc voltage supply on a cable can be considered a satisfactory high- frequency return if the voltage is well decoupled.
Figure 6. Grouping signals on one return line results in high ground impedance and large loop areas. A better design is to distribute returns throughout the cable.
Mistake #8: Using a Single-Point Ground for a High-Frequency Application
Single-point grounds are highly regarded for eliminating ground loops. The implicit assumption behind the use of such grounds is that the velocity of light is infinite relative to the frequencies and dimensions of the equipment in question. This assumption works well for audio frequencies, but becomes inapplicable at higher frequencies where the speed of light plays a significant role. As frequencies increase, inductance in wires and alternate stray capacitive paths also become significant. At 150 MHz a 0.5-m cable is 1/4 wavelength long and acts as an antenna, providing maximum voltage at one end even when the other end is grounded (see Figure 7 below).
Figure 7. Single-point grounds are ineffective at high frequencies, as seen here for a shielded cable.
Solution. Generally, any cabling longer than 1/20 wavelength should be considered distributed and should be grounded at both ends to avoid standing waves. If low-frequency ground loops are a problem (rarely an issue within a single system), then designers can solid terminate the cable shield at one end and capacitively ground it at the other end with a 0.01-0.1-µF capacitor.
Mistake #9: Terminating Cable Shields Improperly
Although commonly used in audio-frequency applications, pigtail grounds can reduce the effectiveness of shield terminations. A 1-in. pigtail has an inductance of 20 nH, which produces an impedance of ~35 W at 300 MHz. This means that high-frequency energy, including RFI up to 1 GHz and ESD with frequency components ranging up to ~300 MHz, can easily penetrate the shield (see Figure 8). Similarly, high-frequency energy can emanate from the opening. A drain wire is, by any measure, the worst way to terminate a shield, because it has long leads and encourages coupling.
At high frequencies, termination connections become a critical issue. Even when backshell terminations are used, five mating connections must be made correctly or the shield will be ineffective. These mating surfaces are the joints between the cable and the metal backshell, the two backshell halves, the backshell and the connector, the connector and its mating connector, and the mating connector and the bulkhead. If any of these is not tight, the shield will leak, just like a garden hose with a leaky connection.
Solution. The only way to eliminate inductive coupling paths is to terminate all connectors circumferentially, such as with a metal backshell. Pigtail terminations should be avoided, or if this is impossible, the pigtail should be short and fat. A drain wire should never be attached to ground via a connector pin--it is an invitation to trouble.
This aluminum, honeycombed vent panel traps and grounds energy that escapes from an electronics enclosure while allowing adequate ventilation. Photo courtesy of Tecknit (Cranford, NJ).
Mistake #10: Failing to Close All Slots and Seams
Shielding is defeated by the presence of seams and other apertures, such as ventilation holes and access panels. Such openings act as slot antennas and are very efficient at coupling energy in and out when they are longer than a 1/20 wavelength of the highest threat frequency of interest. For the ESD frequency of 300 MHz, for example, this amounts to 5 cm. The longest dimension of the opening is the prime parameter, so a small slit can be as bad as a large circular hole. Using ventilation screens to close an opening is ineffective if they are fastened only at the corners. Seams that are not mated with gasketing must be assumed open between fasteners, even if both surfaces are conductive.
Solution. Device designers should attempt to keep openings to a minimum, using the 1/20 wavelength rule. When used, fasteners should be close together and provide conductive contact, but a better alternative is to close seams with conductive gasketing. Screw threads, hinges, latches, and bearings make poor contact and should not be relied upon for shielding purposes. Sometimes these requirements for openings can be relaxed, but that should never be permitted in the immediate area of very noisy or sensitive circuits or in the immediate proximity of wires and cables.
Mistake #11: Failing to Terminate Penetrations
Cables, cable shields, and power lines that penetrate the device enclosure can act as carriers of RF energy, capable of conducting that energy into or out of the enclosure regardless of the quality of the enclosure shield. Depending on the frequency of the interference, the energy may be conducted directly to a circuit or reradiate and couple to adjacent circuits.
Solution. All wires must be terminated at the enclosure shield boundary. Typically, the cable shield would be bonded directly to the conductive enclosure. For unshielded lines, direct termination is not possible, so a capacitive termination is recommended, typically between 0.001 and 0.1 µF. Where such termination is not possible, the alternative is to provide a very high impedance in series, either common mode or individually; however, this technique is of limited effectiveness.
Mistake #12: Failing to Protect the Keypad from ESD
If a membrane keypad is unshielded or poorly terminated, exposure to an ESD event can upset the device's internal electronics (see Figure 9). Unshielded keypads are subject to indirect discharges and to direct discharges that penetrate the dielectric, while shielded keypads are vulnerable to direct discharges penetrating the dielectric and to discharges at the side of the membrane.
Solution. A high-voltage dielectric can be used to prevent discharges from passing through the keypad, and metal parts can be recessed so discharges cannot reach them. The remaining threat will be indirect discharges, which are a much less critical concern than direct discharges.
Using a conductive shielding layer in the membrane is often the cause of, rather than the cure for, ESD problems. If the metallization layer runs to the edge of the membrane, it will be a discharge reception point. Therefore, the metallization must either be recessed so that direct discharges will not occur, or it must be conductively terminated to a conductive enclosure at the perimeter. Full continuous termination is preferred, although termination at four corners may be adequate. The single ground tab that is commonly used is almost always unacceptable.
Basic design issues that are not addressed early can turn into complicated system problems. EMI-related problems can be solved during the design phase for a few dollars in component costs, but may cost tens of thousands of dollars to solve when the project is nearing the production and test phase. If they are not uncovered until the device is in the field, the results may be disastrous.
Consequently, it is critical to plan for electromagnetic compatibility at the start of the project, identifying the risk areas and working around those conditions. Test data should be collected during the prototype phase to identify and solve problems before the device enters production. Designers can use the following list, which summarizes the issues discussed above, to ensure that their electronic products will be safe from the effects of EMI.
The most critical rule to remember, however, is also the simplest--plan for EMI!
1. Bassen HI, Witters DM, Ruggera PS, et al., "CDRH Laboratory Evaluation of Medical Devices for Susceptibility to Radio-Frequency Interference," in Designer's Handbook: Medical Electronics, 3rd ed, Santa Monica, CA, Canon Communications, pp 44-49, 1994.
2. Kahan JS, "Medical Device Regulatory Requirements for Electromagnetic Compatibility," Medical Device & Diagnostic Industry, 17(9):86-92, 1995.
3. Electrostatic Discharge Requirements, IEC 1000-4-2, Geneva, International Electrotechnical Commission (IEC), 1995.
4. Radiated Electromagnetic Field Requirements, IEC 1000-4-3, Geneva, IEC, 1995.
5. Electrical Fast Transient/Burst Requirements, IEC 1000-4-4, Geneva, IEC, 1995.
6. Medical Electrical Equipment, Part I: General Requirements for Safety, IEC 601-1-2, 1st ed., Geneva, IEC, 1993. n