Considering EMC techniques during the design phase will minimize EMC problems and avoid disasters uncovered at EMC test time. At this time, the remedial measures are usually very painful.
In the typical case, where the electronic equipment is contained within a single enclosure, enclosure shielding along with cable filtering or shielding are commonly employed to meet EMC requirements.
Complex systems, however, may need to be handled differently. There may be multiple elements within the system. The equipment may require human access during operation, or there may be patient connections requiring isolation measures. There may be internal self compatibility issues. In such cases, a systems EMC approach is appropriate, addressing internal modules on a case-by-case basis: There may be an assortment of digital electronics, analog electronics, power electronics, electric motors, actuators and relays, RF heaters, lasers, and high intensity lamps.
There will always be some digital electronics, with a microprocessor and supporting electronics. Digital electronics are sources of clock emissions and are vulnerable to transients generate internally and external to the system.
There may be analog sensors connected to the patient or to internal sensors. As would be expected, the input amplifier stage is most exposed to and most vulnerable to RF sources, whether internally or from outside sources.
Power electronics may be confined to the power supply, or there may be electric motors, actuators and relays, lasers and high power lamps. Power switching will usually cause conducted emissions, but may also interfere with internal electronics. Voltage regulators are vulnerable to RFI, whether internal or external.
Effective EMC design starts with assessing the external EMC environment, then assessing sensitive and interference sources within the system, then mapping out a design approach.
ASSESS THE ENVIRONMENT
The first step in EMC design is to assess the EMC environment, starting with regulatory requirements (IEC 60601), and other applicable requirements (such as vehicle based equipment). There may be nearby high energy threats that pose a problem greater than that covered by regulatory standards, including MRI, diathermy, high intensity lamps, power lasers, and others.
The equipment may have internal EMC issues as well, affecting self compatibility. Power circuits, RF sources and internally generated ESD need to be considered.
In any EMC design, circuit and circuit board design is important, but usually needs to be supplemented by shielding the enclosure and shielding or filtering cables as in Figure 1. The key issues are in keeping the openings small, so as to achieve the necessary shielding effectiveness, and to filter or shield the wire entry points so that adverse energy doesn’t enter or leave by wire.
Figure 1: Illustrating the essentials of EMC control – enclosure shielding, filtering or shielding cables.
Depending on the nature of the component, shielding may not be needed, but if it is needed, the openings need to be kept to less than 1/20 wavelength of the highest frequency threat. Filters are appropriate for power entry and audio frequency analog devices. Filtering for high speed digital may not be feasible, necessitating shielded cables.
These issues are important, and are well covered in detail in various publications. This article will discuss the overall systems aspects.
BREAK INTO PIECES
With complex systems, we might also shield the entire enclosure and filter or shield the cables. This is certainly worthy of consideration, but it does have some limitations. First, it assumes the entire enclosure can be wrapped up and enclosed during operation, so that the only issues are regulatory in nature. But what happens if the operator must have access to some of the internal components during operation, and that there are no internal compatibility issues? What happens if we have noisy internal components degrading performance of sensitive components in close proximity?
In such a case, it may be appropriate to break the system into multiple simple elements, and handle the system as a number of simple individual elements.
First to note, the designer is free to select the boundaries of EMC control, shielding and filtering each one individually. This can be done at various places, from the entire enclosure to the chip – the key is to place a boundary between the source and the recipient of the interference. Figure 2 illustrates the two approaches. In a), we have the traditional approach, where a number of modules are contained within the enclosure. In this case, the internal modules would not be shielded. In b) we have shielded the individual modules and filtered or shielded the interfaces, as appropriate. The overall enclosure would not necessarily be shielded.
Figure 2: Shielding all or part of the enclosure
ADDRESS EACH MODULE AS NEEDED
First order of business is to identify the threat for each module. For emissions, the usual case is periodic internal sources. Computer clocks, along with associated harmonic frequencies, constitute the primary regulatory radiated emission limits, but usually don’t pose a threat to nearby electronics. Lower frequency switching power sources may be an emission limit issue, but may also pose a threat to sensitive internal electronics. These will include SMPS power supplies, VFD and PWM motor drives, florescent and high intensity lamps, relays and motor starting/stopping transients, high frequency heaters and pulsed lasers.
For immunity, low level analog input devices, notably op amps and voltage regulators, are the most sensitive. Scientific sensing devices can be very sensitive to RF interference, and patient connected devices are especially problematic, as patient isolation concerns dominate. They may be affected by external RFI sources, as cited for regulatory purposes, or by internal interference sources, notably RF heating devices. At higher levels, digital electronics will also start to become affected.
Electromechanical devices are rarely affected by interference, either locally or external, but they may well be sources of interference. Shielding is usually not needed, as the interference tends to be low frequency conducted.
Electrostatic discharge can be a problem to any part that is touchable by human operator or patient. Internal ESD may be generated by belts and webs.
SELECTING THE FIXES
Once the problem areas have been identified, we need to prescribe remedial measures. Let’s go through an assortment of electrical and electronics devices typically encountered.
By and large, electromechanical devices are not vulnerable to, nor do they generate much, interference in the radiated frequencies. They are substantial sources of conducted interference at conducted frequencies, notably those electronic modules sharing the same power bus.
Different electric motors interfere in various ways. Three phase, split phase and synchronous motors are about as clean a load as you will find. Starting loads will result on a low frequency transients or voltage sags. Turnoff transients from inductive kick are high frequency transients, as characterized by the electrical fast transient (EFT). These will be primarily conductive in nature, fixable by power line filters, transient suppressers, and possibly power supply design.
Brush type DC motors, including universal motors, share the transients effects mentioned above, as is typical of heavy starting loads and inductive loads. In addition, they generate substantial brush noise – this is primarily a conducted emission issue, best handled by line-to-line capacitors across the brushes or as close to them as feasible.
Modern electronically driven motors all share a common problem – they have the starting and stopping transients as the above motors, but they also generate substantial high power switching interference – they all provide higher efficiency at the expense of high frequency emissions. The switching frequencies are generated electronically (Figure 3) and drive substantial currents down the lines to the motors.
Figure 3: Power switching circuits generate substantial output noise
These include VFD, PWM and brushless DC motors. Although the frequency may vary, and may not even be constant, they all generate lots of switching noise. Motor windings pose an inductive load, tending to even out the low frequency current pulses. The high frequency switching components don’t go through the windings, rather, they capacitively couple to the rotor or stator. From there, the current returns to the switching source by one or more paths, often back through the power source, creating an emission problem.
The preferred solution is to block the high frequency currents from leaving the drive electronics. Lacking that, provide a low impedance return path back to the switching source common.
Analog sensors are mostly low frequency/low level sources. As such, they are vulnerable to RF sources. Do not assume that the sensor will ignore the radio source – amplifiers looking for 10 microvolts of input signal (as for EKG and EEG) will go nonlinear with RFI which may produce volts of interference to the amplifier input, resulting in a demodulated signal.
The choices are to shield the signal path or filter the signal input to the op-amp. Cable shielding is not always achievable, but is most effective. Signal filtering is feasible for low frequencies, as there is usually ample spacing between signal bandwidth and RF sources.
Such an approach may be reasonable with diagnostic equipment, but is much more difficult to deal with patient connections. Isolation requirements restrict filtering and shielding to be terminated specifically to the isolated input area – the idea is to amplify the weak input signal before crossing to the un-isolated area of the electronics, and that takes considerable care.
Not all analog sensors are low frequency, a particular case in point is ultrasound, operating at perhaps one MHz, and having receiving signals well below one microvolt, and carrying multiple channels. High frequency filtering is still a viable approach, for frequencies above, say, 10 MHz, but frequencies in the signal band pass cause problems. And, considering the signal strength, any stray noise is a potential threat. Unfortunately, it is not possible to shield from the input amp to the patient leads, but you have to start there – run the shield out as far as possible to block the lower frequencies and filter the higher frequencies.
In any case, it is important to keep local noise from digital and power electronics from getting into the analog section – fortunately, patient isolation pretty much forces you to do so. If you have sensitive inputs, the solution may be EMC design at the circuit board level.
High Speed Digital
In most cases, high speed digital modules will require shielding of the card or card cage. High speed digital, notably computer clocks, will generate emission noise. These are best handled as close to the source as possible, typically with moderate filtering on the signal lines and good decoupling on the power rails. Filtering high speed signals is limited by signal bandpass – that which can’t be filtered will need to be handled by shielding the enclosure, and, if you need to drive high frequencies on the cables, you need to shield the cables, as well.
Where digital signals must leave the module and cannot be filtered, the remaining option is to shield the cables from module to module. Cable shields need to be properly terminated at both ends – pigtails and single point grounds are not allowed for high frequency threats.
Digital circuits are vulnerable to transients originating by power switching transients within or external to your system. These are mostly carried by conducted paths, either through the power feed, or by ground bounce. ESD, however, is generally propagated to the vulnerable circuit due to injection into the enclosure or to any internal members touchable by humans.
Digital circuits are not as vulnerable to RFI as analog circuits, but they can be affected if the RFI is sufficient. Modern electronics are being powered by ever lower supply voltages, and this is increasingly impacting RF immunity.
Power handling varies with the nature of the power source. Equipment installed in a fixed location will generally be entirely AC powered. Mobile equipment may be powered by internal batteries or vehicle batteries. Equipment using rechargeable batteries may or may not be operable while charging at a wall outlet. Chargers may be pulsed, necessitating testing of the charger, as well.
50/60 Hz line power presents the most common problems – both conducted emissions and immunity, but increasingly, higher frequency radiated emissions and immunity from switching power supplies is occurring.
Distribution of switched power, notably motor drives, needs to be treated, as well via the line power cord, and also to other internal circuits sharing the power bus.
In most cases, power EMC is handled with power line filtering, but where do you put the filter? For a simple system, there is essentially one module to power, with one power supply to filter. But for complex systems, there may be several internal power supplies, perhaps individually filtered, and some number of electromechanical devices that need some kind of protection. Once we have taken care of the internal compatibility problem, we may have installed some local filters and transient protection.
The first inclination is to tack on a big power line filter on the front end, and that may well do the trick. There is a possible drawback, however – stacking up power line filters may lead to unpredictable results. Even under the best circumstances, the behavior of the power line filter is somewhat unpredictable. Most off-the-shelf power filters will have shunt capacitance at both the input and output, and are tested at 50 ohms input and 50 ohms output. For regulatory testing, the input impedance is also 50 ohms, but is uncontrolled in actual installations, creating an indeterminacy. It’s worse at the load, which is also uncontrolled during test as well as in actual installation. For this reason, the manufacturer specified filter attenuation should be de-rated by at least 20 dB.
Adding some internal filtering produces the situation as shown in Figure 4. The capacitors stack up to alter the frequency response of the filter network, generally degrading the overall filter performance, and possibly starving the power or creating an unstable condition in a power supply. This is not really feasible to predict in advance. Most internal power problems will be show up during prototyping, but may not show up until EMC testing.
Figure 4: Schematic of cascaded EMC filters
Preferably, filter each module individually, then bus the power leads together with minimal filtering before exiting the system enclosure. The exit filter would preferably be a minimal high frequency filter, targeting frequencies above the effectiveness of the internal power line filters. This will minimize possible filter interaction.
Similar situations occur with battery powered systems with the possible exception of external conducted interference – if the equipment is never operated from line power, then conducted testing is not applicable. And, of course, if the equipment is powered by a vehicle battery, then automotive or avionics conditions will apply, generally with significantly poorer power quality, especially transients.
ENCLOSURE DESIGN CONSIDERATIONS
Once the individual modules and their interconnect have been addressed, it is necessary to turn to the enclosure and internal cabling. Much can be done to promote internal compatibility by careful design. The key issues in mechanical design is grounding, shielding and power distribution.
For this discussion, ground is defined as the housing for the entire enclosure. Connection to earth ground may or may not exist, but this is mostly irrelevant for EMC issues.
Ground impedance plays a key role in internal module interconnect. Modules are typically bolted down to a structural member. Often, internal EMC problems arise due to ground bounce – a shift in ground between two modules creates a signal error. Reducing ground impedance in the support members in the system can be instrumental in error reduction.
The most important aspect of enclosure ground is to conductively mate all structural members, module shields, etc., together at every possible interval. Use sheet metal planes wherever possible – wire paths are unacceptable as a high frequency ground. Mate planes together at every opportunity – holes and bends are okay (Figure 5). Mating surfaces must be conductive – hinges, latches, screw threads and bearings are unacceptable connections.
Figure 5: Low impedance grounds need not be planar
Bolt modules directly to ground. Do not attempt to implement single point ground except where you have a low level analog input. This is difficult to do in a retrofit situation, but readily achievable during initial design.
Power and Data Cable Routing
A significant amount of EMC problems involve crosstalk in internal power and data cabling. The worst case is tie wrapping sensitive cables with noisy cables.
Start with the basic assumption that similar signals and power are mutually compatible – the problem occurs with bundling noisy power lines with digital signal lines and either of those with sensitive analog lines. Ideally, these would be grouped into separate bundles, but this may be unrealistic for complex systems. If cables cannot be so isolated, then be alert for mixing sensitive and noisy lines – either separate the critical lines, shield selected lines in the bundle, or augment with better filtering.
Minimize daisy chaining of power lines, particularly where noisy sources share power with sensitive electronics. Avoid daisy chaining of signal lines entirely. Avoid tie wrapping power lines with signal lines.
Note that crosstalk is minimized by routing cables next to a continuous ground plane, if possible, especially the critical lines. Best to put cable routing under control drawing – EMC performance can vary widely with cable placement.
Since EMC behavior is sensitive to cable position, EMC performance with uncontrolled cable routing may vary considerably from unit to unit.
Use a systems approach to handle complex medical equipment. Protect each module, then bolt them together.
Start by identifying key emitters and receptors. Find appropriate fixes for each module, including module shielding, cable shielding and signal/power filtering, as appropriate for each.
Design the enclosure, starting with a well built ground – bolt it together, no wire grounds. Route the cables adjacent to ground.
|Daryl Gerke, PE and Bill Kimmel, PE
are partners in Kimmel Gerke Associates, Ltd., an engineering consulting and training firm that specialized in EMI/EMC design and troubleshooting issues. Both are degreed engineers (BSEE), iNARTE Certified EMC Engineers, and registered Professional Engineers (PE).
Daryl and Bill have prevented or solved hundreds of EMI problems in a wide range of industries – computers, military, medical, industrial controls, automotive, avionics, railroad, telecomm, facilities, and more. They have also trained over 10,000 engineers through their public and in-house training classes. They just celebrated 25 years in full-time practice as EMI/EMC consulting engineers. For more EMI information, visit their web site at www.emiguru.com.