Originally published August, 1996
William D. Kimmel and Daryl D. Gerke
Electromagnetic compatibility is an important consideration in the design and operation of today's sophisticated medical electronic equipment, particularly as portable systems proliferate. Electronic devices can both emit and be damaged by electromagnetic interference (EMI) and must be protected from its harmful effects. Issues of patient and operator safety must also be addressed. Previous articles have covered such means of achieving EMI control as filtering, cable shields, and enclosure shielding (MD&DI, February, July, and November 1995, respectively). This article focuses on grounding.
Perhaps no topic in electronics is as misunderstood as grounding, which usually evokes an image of a long braid snaking off to a ground post set into a concrete floor. As the following discussion makes clear, an earth ground is not essential to EMI control and is almost never needed. In the overwhelming majority of medical electronic applications, good grounding involves achieving a sufficiently low-impedance return path for the highest interference frequency of interest. If it were possible to achieve zero impedance, all other grounding issues would become meaningless. Since it isn't, device designers need to seek ways of maximizing the effectiveness of the grounds that can be implemented.
WHAT IS A GROUND?
Succinctly put, a ground is a return path for current. Its purpose is to close the current loop, not to lead it into the earth. If an interference current is diverted successfully into earth ground, it will simply come out elsewhere in order to return to its source. The only time earth ground is necessary is for lightning.
Confusion arises because the term ground is used for a variety of applications and means different things to different people. Facility engineers, for example, look at a ground as a return for lightning strikes. In this application, the ground needs to be able to handle currents up to 100,000 A for a few milliseconds. Because the approximately 1-microsecond rise time produces significant Fourier frequency components up to about 300 kHz, inductance can become an important concern. In contrast, electricians look at a ground as being a return path for fault currents, which may involve tens or hundreds of amperes at 50 or 60 Hz. At this frequency level, inductance is not important, so a length of 4/0 wire connected to the nearest building steel works just fine--an earth ground may be present, but is not needed for electrical safety.
These two cases are the most commonly known uses of grounding, but the grounding requirements for EMI control in medical device applications are vastly different. EMI can cover a very wide range: currents from microamperes to amperes and frequencies from direct current to daylight. The duration of an event can range from nanoseconds, in the case of a transient, to years, in the case of a continuous wave. For the specific case of electrostatic discharge (ESD), transients are measured in nanoseconds (giving Fourier frequency components up to 300 MHz), and currents range to 10 A or even higher. The edge rates and current magnitudes are such that significant voltage bounce will occur across even the smallest length of wire or circuit-board trace. Whatever the condition, however, device designers must provide a way for the interfering current to return to its source, and that rarely involves earth ground.
GROUND LOOPS AND SINGLE-POINT GROUNDS
Whenever grounding is an issue, design engineers inevitably turn to ground loops and single-point grounds. What do these terms mean and when are the techniques appropriate?
A ground loop exists whenever there is more than one conductive path between two points. This condition allows interference currents to mix with signal currents, which may lead to ground interference. Figure 1(a) shows the effects of a ground loop when stray interference currents divide and flow through signal ground. This problem can be eliminated by having a zero-impedance ground. Lacking such a ground, separate ground paths can be provided. As shown in Figure 1(b), by breaking the ground loop, the device designer has created a single-point ground. The need for a single-point ground originated in telephony, where it was almost impossible to get impedances low enough to prevent power line frequencies from intruding as a hum, and the technique is still useful in a number of low-level, low-frequency analog applications.
However, a single-point ground is not suitable for handling the higher frequencies encountered in modern computing devices. Figure 2 shows the effect of a standing wave on a cable shield that has been grounded to its enclosure at a single point. If the shield were exposed to an incident interference of 150 MHz (a popular land mobile radio frequency) with a wavelength of 2 m, the cable, which is represented here as being a 1/4 wavelength of the interfering frequency, or 0.5 m, would act as an efficient antenna, with standing wave voltage on the shield as indicated in the figure. In the immediate proximity of the ground connection, the shield voltage is near zero, but at the unterminated end, the voltage is at a maximum, and with stray capacitance, there is ample coupling to the signal lines.
The fundamental assumption behind the principle of single-point grounding is that the velocity of light is infinite. Any time designers need to consider the velocity of light, notably at computer speeds, the single-point ground technique doesn't work. A useful rule of thumb is that a single-point ground is appropriate if the longest dimension of interest is less than a 1/20 wavelength of the highest-frequency threat. Thus, single-point grounds are appropriate for handling EMI with audio frequencies in most cases but inappropriate and unachievable for radio frequencies used in digital electronics.
Consider, for example, the case of a designer who wanted to use a single-point ground for two freestanding cabinets located about 10 ft apart. Based on the common assumption that the inductance of a wire is 20 nH/in., the minimum inductance for the single-point ground path would be about 2.5 µH. Using the formula for impedance
Z = 2¼fL
where f is frequency in megahertz, L is inductance in microhenries, and Z is in ohms, the impedance at 100 MHz would be 1600 , which is hardly a short circuit. Using the rule of thumb that capacitance between freestanding equipment and ground is ~100 pF and the formula
where C is capacitance in microfarads, the impedance with two 100-pF capacitors in series with a ground plane is 30 . This is not a short circuit either, but is certainly a lot lower than that of the intended single-point ground path.
ACHIEVING GOOD GROUNDS
Achieving a low-impedance ground for a medical electronic device is easy in concept-- use a ground plane. At 50/60 Hz, the impedance of a grounding wire will be primarily resistive, but above audio frequencies inductance begins to dominate and at radio frequencies the inductive impedance of even a short wire or circuit-board trace is enough to cause problems. To determine the requirements of a particular application, the designer needs to know what voltage the device can tolerate, the magnitude and frequency of the anticipated interference current, and the impedance of the path. Given these data, Ohm's law can be applied to find out when problems will occur.
A lightning strike, for example, might result in 10,000 A flowing in an I-beam with 10-V transients across even short lengths. Two interconnected devices grounded to that I-beam at different points may easily experience upset. Or suppose a 1-in. length of wire or circuit-board trace were subjected to a 10-A ESD event. Assuming an inductance of about 20 nH, the voltage drop across the wire or trace could be calculated using the equation
where E is voltage drop across wire, L is inductance in nanohenries, di is magnitude of current transient (assumed to be 10 A), and dt is rise time (assumed to be 1 nanosecond). For these typical conditions, E = 200 V. Thus, it can be seen that a length of wire as short as 1 in. makes a poor ground for ESD purposes.
Because ordinary wire is not a satisfactory ground in many circumstances, the common wisdom is to use a flat strap instead. This approach is indeed appropriate, but the rationale behind it is widely misunderstood. To achieve low inductance, the key factor is not the strap's flatness but its length-to-width ratio. To ensure that the inductance of a ground strap is sufficiently low, its width must be at least one-fifth or, better yet, one-third of its length. If a designer cannot achieve this ratio, there will not be a satisfactory high-frequency current return path.
Circuit-Board Grounds. It is almost impossible to get good low-impedance grounds on two-sided circuit boards, so it is critical to keep ESD currents and high-level radio-frequency interference off such boards. On the other hand, it is easy to achieve low impedances with the ground plane underneath the traces on multilayer boards. Circuits built immediately above the ground plane are well protected, regardless of the threat. Our observation is that EMI control is always problematic with double-sided boards, while electronic devices with multilayer boards are rarely harmed. If a manufacturer is adamant about using double-sided boards, the product development budget should include additional funds and three months should be added to the schedule for test and redesign. Even then, there will be a high probability that EMI control will not be achieved.
Probably nowhere in electronics do designers face such a difficult challenge as that posed by sensitive analog input circuits. The circuits can be fairly well protected on an isolated ground plane; the problem involves interconnections to an unisolated ground or to the wires and cables that connect the sensor to other equipment. For an isolated ground, it is important to minimize the amount of external EMI currents that reach the ground plane. Once the sensitive input signal has been captured and amplified, or perhaps digitized, crossing the boundary to unisolated circuits is the remaining design problem. Any interference currents that are diverted to the isolated ground become common-mode interference and must be handled by an isolator component, of whatever type. Although some fairly effective isolators are available, they have their limits, so it pays to minimize the common-mode currents in the first place.
Interconnect Grounding. Once the designer has coped with the circuit-board ground, the next consideration is the interconnects within the equipment, such as the connections between the mother and daughter boards and the ribbon cables between modules. EMI problems are frequently the result of high-impedance interconnects. Again, designers need to keep the ground impedance low, either by connecting the circuit boards or modules to a common ground plane or by providing a very-low-impedance ground interconnect via the cable, usually by allocating as many connector pins to grounds as possible. Although the connector space is an important concern, so is functionality. For high-speed (100-MHz) interconnects, there should be one ground line for each signal line. For lower speeds (~10 MHz), one ground line for each five signal lines may be sufficient. Anything less is inviting trouble.
External Grounding. Finally, designers need to consider the interconnections between various pieces of equipment. If a low- impedance ground plane can be implemented between enclosures, and multipoint grounds are used for cable shields, problems should be minimal. However, if cables run long distances or if sensitive low-frequency analog signals are being transmitted, audio-frequency interference may be a concern. In such cases a single-point ground may be needed as well as the multipoint ground required to control high-frequency interference. A hybrid ground with a capacitor termination at one end, typically 0.010.1 µF, and a hard termination at the other end can provide an open circuit at audio frequencies and a short circuit at radio frequencies, thus combining the best of both worlds.
Medical electronics designers can base their decisions on how to implement grounding for EMI control on three principles:
* An earth ground is not necessary for EMI control (although it may be needed for safety). What is needed is a low-impedance current return path, usually a conductive plane or a shield.
* Single-point grounds are usually appropriate only for handling audio-frequency interference and are not achievable at radio frequencies. The 1/20-wavelength criterion can be applied to determine if a single-point ground is acceptable.
* Ground impedance must be kept acceptably low at the current frequency of the anticipated interference event. At high frequencies, inductance gives rise to high impedances, so use of ground wires is generally not acceptable. A wide ground strap or plane can be used to reduce impedances.
William D. Kimmel and Daryl D. Gerke are principals in the EMI consulting firm Kimmel Gerke Associates, Ltd., based in St. Paul, MN.
Figure 1. Schematics showing ground loop currents: (a) unbroken and (b) broken (thereby providing a single-point ground).
Figure 2. Effects of a standing wave on a single-point-grounded cable shield.