An MD&DI February 1997 Feature Article
Electromagnetic interference can be minimized by using care when selecting and mounting supply components, and when designing filter elements.
When designing power supplies for patient-connected medical electronics, one of the biggest challenges is controlling electromagnetic interference (EMI) while maintaining the low leakage currents necessary for safe patient connection. In most electronic devices, EMI is easily controlled by integrating filters between the line power and the power supply. Y-capacitor-type filters, which connect both line and neutral to enclosure ground, are especially effective against common-mode interference.
However, where this option is restricted by leakage-current requirements, designers must rely on the only available option: installing the high-impedance series elements known as inductors, which are far less effective than capacitors at controlling EMI. The high impedances described for such filters in textbooks are not realized in actual practice, at least at higher frequencies--parasitic paths become efficient above 100 MHz, and carelessly designed power supplies and filters fail to perform to expectations. Accordingly, designers must pay attention to minimizing effects of parasitic elements.
After describing the threats to medical power supplies from EMI and the coupling paths involved, this article presents design recommendations that will help to minimize EMI without sacrificing safety.
EMISSIONS AND IMMUNITY
In the European Union (EU), both immunity to external interference and emissions from medical electronic equipment are regulated. EMI immunity requirements, which are set forth in the International Electrotechnical Commission (IEC) 601 series of standards, include tests for radio-frequency interference (RFI), electrostatic discharge (ESD), electrical fast transients (EFTs), and surges.1 Of these four tests, those for EFTs and surges are most important for designers of power supplies.
At present, there are no mandatory EMI immunity requirements in the United States, but FDA has been showing increasing interest in the issue and is working with the IEC to establish uniform requirements.2 The agency has not yet shown a similar interest in regulating emissions, and the Federal Communications Commission has exempted medical electronics (and several other equipment categories) from emissions testing. Nevertheless, for reasons of safety as well as market access, it makes sense for U.S.-based manufacturers to comply with the European requirements for both immunity and emissions, even for domestic products.
Figure 1 shows the sources of the threats to medical device power supplies from both external interference and internal emissions. The first threat involves emissions that originate in the power supply itself. All ac supplies generate rectifier noise, including rectifier switching and snap-off, and switching-mode power supplies add switching noise to the system. Such interference can either be propagated back to the power line or forward to the load. Interference that travels to the power line will show up in a conducted test and may also show up in a radiated test, especially with switchers running above 100 kHz.
The second threat is the interference generated in the device electronics. High-frequency clock noise is the most notable but noise is also generated by electric motors, especially brushless motors and those with variable-frequency drives, and relay or other power switchers. In this case, the power supply simply acts as a conduit. The power supply designer cannot control the interference source, but must attempt to intercept the noise that is conducted from the device to the supply.
The third threat is external interference that enters the power supply and either attacks the supply directly or passes through it to attack the device electronics. All types of EMI (power disturbances, RFI, and ESD) can interfere with power supplies, but power disturbances such as EFTs and voltage surges are directed primarily at the supply. A high-frequency, low-energy test is used to determine immunity to EFTs, which will easily pass through the power supply and attack the electronics within. The surge test is low frequency and high energy, simulating a surge caused by a lightning strike. While the required immunity levels are similar, the threats are significantly different.
RFI can also assault a power supply directly or can be intercepted by the power cord and conducted to the supply (generally, lower frequencies will be conducted and higher frequencies will be radiated). If the energy is conducted, it may attack the power supply itself or simply pass through to attack the device electronics. ESD does not usually affect power supplies, but when it does, its effects are much like those of EFTs. The static discharge attacks the supply's regulator, causing temporary voltage sags (or surges), or passes through the supply to get at the digital electronics.
EMI coupling paths, which connect the interference source with a receptor, may be conducted, radiated, or a combination of the two. Differential-mode conductive paths, with signal or supply current or interference current propagating down a wire and coming back via an adjacent return line, are the norm. Common-mode paths occur when current travels down both the signal or supply line and the return line in phase; the return is an unintended path, often to a chassis ground or earth ground.
Interference almost always originates as differential mode, but both electric and magnetic field coupling paths are predominantly common mode, which means that the farther EMI gets from its source, the higher the percentage of common-mode coupling paths. Low-frequency energy (
Because common-mode interference, which is very difficult to filter, relies primarily on parasitic coupling paths, it is important for designers to keep those paths to an absolute minimum. This involves controlling both the parasitics in the component itself and parasitics between the intended current path (traces and components) and other system members, which include other components and traces, structural members, and perhaps a heat sink. When addressing this challenge, the following generalizations should be kept in mind:
- Because of the length of their leads, capacitors have series inductance, which reduces their effectiveness at high frequencies. Capacitors will resonate at a surprisingly low frequency, becoming inductive above resonance.
- As a result of interwinding capacitance, inductors have shunt capacitance, which reduces their high-frequency effectiveness. In addition, open-flux-path inductors have a magnetic field extending well beyond the inductor envelope.
- Coupling paths from the intended current path to heat sinks and circuit board traces can be a significant concern.
(These issues have been addressed in detail in previous articles.3, 4)
COPING WITH INTERNAL INTERFERENCE
A clear-cut rule for designing for EMI is to control interference as close to its source as possible. The farther EMI is from the source, the harder it is to contain the energy. In the case of power supplies, internally generated high-frequency noise from the bridge and switcher is best controlled right at the source. Properly applied, filtering at these locations can be differential mode. For the switcher, the high-frequency currents can either be blocked immediately at the switch or returned immediately to their original path. For the bridge, it is not sufficient to include only a line-to-line filter at the input or output because imbalances in the bridge currents will leave unequal snap-off spikes, which will end up as common-mode voltages; filtering to a neutral common point will help to confine these spikes to the immediate area of the bridge.
Placing ferrites in-line may seem counterintuitive to some power supply designers, who believe switches should be slammed against the rails as fast as possible to keep losses down. But such rapid switching is the primary source of internally generated noise. The application's actual switching requirements should be considered carefully before rejecting use of in-line filters. Increasingly large ferrites can be tried until only a minimal change in rise time is perceivable on the switching waveform. In most supplies, there is really no need to pass 30-MHz interference from a 100-kHz switcher.
The best way to block high-frequency currents from passing from a medical device to the power supply is to insert high-frequency ferrites in series in all lines connected with the device (including power return). The low-frequency filter elements used in the power supply design will probably not suffice to control the higher noise frequencies.
Once everything possible has been done to control internal emissions at the source, the issue of coupling to adjacent circuit elements should be addressed. The magnetic field from an open-flux-path inductor extends far beyond the footprint of the inductor, so it is very difficult to provide adequate isolation. Using toroids or, even better, pot cores, in the power supply is recommended, but it is important to make sure they perform well at frequencies well above the supply's switching frequency. Inductors resonate at surprisingly low frequencies: Any inductor used in a 100-kHz supply will almost certainly resonate below 10 MHz, rendering it ineffective at controlling interference at 30 MHz or above. Inductor resonances should be checked on a network analyzer before installation. Similarly, it is important to ensure that differential-mode filter capacitors are effective at higher frequencies. Although the new tantalum units have been described as very good at high frequencies, it is probably better not to rely on them to filter current above 5 MHz.
It is also critical to mount filter components so as to minimize coupling to printed circuit board traces, adjacent circuit components, and mechanical elements. Capacitors should be mounted with short leads, and closely spaced inductors and transformers should be placed orthogonally, even if they are toroids. Designers need to consider which is the high-noise side of these inductors, as capacitive coupling to adjacent traces and heat sinks can be considerable.
COPING WITH EXTERNAL INTERFERENCE
Although the steps described above are directed specifically at controlling and preventing internally generated interference, they are also effective for external interference. However, they are not likely to provide sufficient protection, especially against conducted EMI.
External conducted interference can be controlled by filtering or transient suppression. Transient devices such as arc suppressors (crowbars) and clamps (metal-oxide varistors and zener diodes) are designed to handle high-energy transients, such as voltage surges, but their use in medical devices needs to be tempered by safety considerations. As with Y-capacitors, transient suppressors connected between line or neutral and ground present a possible leakage path. Therefore they should only be employed if their steady-state leakage current is low enough to meet the leakage-current limits for patient-connected devices and their breakdown is sufficiently high, typically at least 500 Vrms.
If at all possible, it is preferable to use filtering alone, without transient suppressors. Generally, this means that the power supply itself must be built to withstand as much interference as possible. The designer will need to select supply components capable of withstanding any high-voltage transients that can't be blocked. High-frequency filtering is effective against EFTs, but during surges, where the high voltage is of fairly low frequency, it is likely that excess voltages will get into the supply.
Among the supply components, the converters are most at risk from external interference. Any disturbance that reaches the feedback circuit in a regulator can cause out-of-tolerance supply voltages. RFI, being a continuous wave, will cause an output-voltage error unless filters keep it out of the feedback path. Generally, this will require placing a high-frequency capacitor at both the input to and output from the regulator (see Figure 2 below). As shown in the figure, capacitors must go from input and output to circuit common, not to chassis ground. Power line transients will also attack the feedback circuit. Even brief transients can cause a long-term sag or surge in the supply voltage. Designers need to keep such currents away from the regulator.
Figure 2. Isolation of the power supply regulator with capacitors at its input and output.
EMI problems in medical electronic equipment power supplies can be prevented if designers heed the following advice. Select supply components carefully to ensure their functionality at the frequency range of interest and design filter elements to intercept interference as close to its source as possible. Also, mount components carefully to minimize coupling paths to other components, especially those that connect to the outside world.
1. "Electromagnetic Compatibility--Requirements and Tests," IEC 601-1-2, Geneva, International Electrotechnical Commission (IEC), 1993.
2. Kahan JS, "Medical Device Regulatory Requirements for Electromagnetic Compatibility," Med Dev Diag Indust, 17(9):86-92, 1995.
3. Kimmel WD, and Gerke DD, "Selecting Components to Minimize EMI," Med Dev Diag Indust, 17(1):212-218, 1995.
4. Kimmel WD, and Gerke DD, "Filtering Analog Signals in Medical Devices," Med Dev Diag Indust, 17(2):88-92, 1995.