Protection of Electronic Circuits from Overvoltages

MATHEMATICS

Part 1

Symptoms and Threats

1

Damage and Upset

A. NATURE OF ELECTRICAL OVERSTRESS PROBLEM

Electrical overstresses (e.g. from lightning, electromagnetic pulses from nuclear weapons, and switching of reactive loads) can cause failure, permanent degradation, or temporary malfunction of electronic devices and systems. The characterization of these overstresses and the design of effective protection from them is of great importance to manufacturers and users of industrial, military, and consumer electronic equipment.

Electrical overstresses have received increasing attention during the period between 1960 and the present (1988). This trend can be expected to continue. There are several reasons for this trend: (1) devices are becoming more vulnerable; (2) vulnerable systems are becoming more common; and (3) awareness of the existence of overstresses has increased.

Modern semiconductor integrated circuits are much more vulnerable to damage by overstresses than earlier electronic circuits, which used vacuum tubes and relays. Progress in developing faster and denser integrated circuits has been accompanied by a general increase in vulnerability. At the same time that electronic circuits were becoming more vulnerable, they were also becoming more widely used. (As an example, consider desktop computers and videotape recorders: these were nonexistent items in 1960 but are quite common now.) Therefore, there are now more systems to protect from overstresses. Finally, as awareness of overstresses increases, users of vulnerable systems request appropriate protective measures.

In general, techniques for protection against transient overvoltages can be divided into three classes:

shielding and grounding

application of filters

application of nonlinear devices

Shielding, while important, is not sufficient protection against electromagnetic fields from lightning or nuclear weapons, because compromises in the integrity of the shield must be made (e.g., windows in aircraft; long lines must enter the shielded volume to supply electric power and carry communication signals). Various shielding and grounding techniques are covered in detail in books by Ott (1976), Morrison (1977), Ricketts et al. (1976), and Lee (1986) and in government reports by Lasitter and Clark (1970) and Sandia Laboratories (1972). The design of filters is covered in many electrical engineering text and reference books. The emphasis in this book is on the third class of techniques, nonlinear transient protection devices, although some information on filters is included in Chapters 13 and 19.

B. ORGANIZATION OF THIS BOOK

This book is divided into four parts:

symptoms and threats

nonlinear protection components

applications of protection components

validating protective measures

1. Symptoms and Threats

Transient overvoltages in electronic circuits can arise from any of the following causes: lightning, electromagnetic pulse produced by nuclear weapons (NEMP), high-power microwave weapons (HPM), electrostatic discharge (ESD), and switching of reactive loads. These sources are described in Chapter 2. These transient overvoltages can be coupled to vulnerable circuits in several different ways:

direct injection of current–for example, a lightning strike to an overhead conductor

effects of rapidly changing magnetic fields–for example, induced voltage in a conducting loop from changing magnetic fields owing to nearby lightning or NEMP

effects of rapidly changing electric field–for example, charging by induction from ESD

changes in reference (ground) potential due to injection of large currents in a grounding conductor that has nonzero values of resistance and inductance

A discussion of surveys of transient overvoltages in specific environments is given in Chapter 3. The propagation of transient overvoltages from their source to the vulnerable equipment is discussed in Chapter 4. Chapter 5 discusses standard overstress test waveforms that are simplifications of the environment. Chapter 6 gives a brief sketch of protective circuits and devices, which is useful in preparing the reader for the following two parts of the book.

2. Nonlinear Protection Components

Chapters 7 to 14 contain a discussion of properties of various components that are useful to protect circuits and systems from overvoltages. Spark gaps, metal oxide varistors, and avalanche diodes are emphasized. However, other components, such as semiconductor rectifier diodes, thyristors, resistors, inductors, filters, and optoisolators are also discussed. Chapter 15 explains why minimization of parasitic inductance is critical in practical transient overvoltage clamping circuits.

3. Applications of Protective Devices

Chapters 16 to 20 are concerned with application techniques to protect circuits and systems from damage by transient overvoltages. Specific applications of the components in Part 3 are discussed in the context of signal lines, dc power supplies, and low-voltage mains. Protection of circuits and systems from upset is covered in Chapter 21.

4. Validating Protective Measures

Chapters 22 to 24 discuss how to validate protective measures against damage by overstresses. Preparation of a test plan, high-voltage laboratory procedures, and safety are discussed.

C. NOMENCLATURE

There is no general agreement on a name for electrical overstress. The Institute of Electrical and Electronics Engineers (IEEE) in the United States has adopted the word surge to denote an overstress condition that has a duration of less than a few milliseconds. American engineers also use the word surge to mean something quite different: an increase in the rms voltage for a few cycles, which is called a swell by Martzloff and Gruzs (1987). To avoid misunderstanding, the author favors the use of overvoltage, which is translated from the German Überspannung. To emphasize the brief nature of the event, one may say transient overvoltage. The term electrical overstress is more general, because it includes excessive current or energy as well as voltage. To be precise, it is necessary to say electrical overstress, because there are other kinds of adverse environments–for example, extreme temperature. Because this book is only concerned with electrical overstresses, the modifier electrical is omitted.

Overstresses can cause two different kinds of adverse outcomes in sensitive electronic circuits and systems: damage or upset. Damage is a permanent failure of hardware. A damaged system may fail completely or partially. The only way to recover from damage is to replace defective components. Upset is a temporary malfunction of a system. Recovery from upset does not require any repair or replacement of hardware. An example of damage is a charred printed circuit board after a lightning strike. An example of upset is loss of the contents of the volatile memory in a computer when there is a brief interruption of power. A system or circuit is said to be vulnerable to damage but susceptible to upset.

Components and circuits that protect vulnerable devices and systems from damage by electrical overstresses are members of a class of devices called terminal protection devices (TPDs) or surge protective devices (SPDs). The term TPD is used by the U.S. military; SPD is used by the engineers in commercial practice. A surge protective device that is intended for electrical power systems is called an arrester.

The word mains is used in this book to refer to low-voltage ac power distribution circuits inside of buildings. In this context, low-voltage means less than 1 kV rms, and ac means a sinusoidal waveform with a frequency between 50 and 400 Hz.

Definitions of these and other specialized terms are contained in the glossary in Appendix A.

D. DAMAGE AND UPSET THRESHOLDS

Because many modern semiconductor devices (small signal transistors, integrated circuits) can be damaged by potential differences that exceed 10 V, the survivability of modern electronics is limited when exposed to transient overvoltages. Modern electronic technology has tended to produce smaller and faster semiconductor devices, particularly high-speed digital logic, microprocessors, metal oxide semiconductor (MOS) memories for computers, and GaAs FETs for microwave use. This progress has led to an increased vulnerability of modern circuits to damage by transient overvoltages, owing to the inability of small components to conduct large currents and to breakdown at smaller voltages.

Smaller devices make a more economical use of area on silicon wafers and decrease components cost. These smaller devices also have less parasitic capacitance and are therefore faster. However, devices often fail when the current per unit area becomes too large. The magnitude of transient currents is determined principally by external circuit parameters (e.g., nature of the source, characteristic impedance of transmission lines, resistance and inductance between the source of the transient and the vulnerable circuit, etc.). Smaller devices obviously have less area and are thus more vulnerable to damage from a current of a given level. When breakdown is considered, smaller devices that have less spacing between conductors will break down at lower voltages.

A logical approach to transient protection would be (1) to determine the threshold at which damage occurred, (2) to determine the worst-case electrical overstress that could arrive at a particular device, and (3) to design and install a protective circuit that would limit the worst-case overstress to less than the damage threshold. This simple, scientific approach has become a practical nightmare. First, as is described below there are apparently no simple criteria for determining the maximum overstress that a part can withstand without being damaged. Second, as described throughout this book, a protective circuit that can survive a worst-case overstress is often extremely expensive, massive, and bulky.

1. Damage Threshold

Failure of transformers and motors is caused by breakdown of insulation. The most important parameter in insulation breakdown is the magnitude of the peak voltage, although the rise time may also be important. Once the insulation has broken down, some of the winding is shunted with a low-impedance arc. Transformers and motors can withstand voltages that are much greater than those that cause failure in semiconductor devices. Therefore, recent concerns about damage caused by overvoltages has focused on semiconductors and has tended to ignore damage to transformers and motors.

The value of voltage, current, or power that is necessary to cause permanent damage to semiconductor devices is known as the damage threshold or failure threshold. In general, the value of the threshold is a function of the duration of the overstress. Such information is essential during the design of protective circuits, since the protection must attenuate overstresses to below the damage threshold.

The damage threshold is defined as the minimum power transfer through a terminal such that the device’s characteristics are significantly and irreversibly altered. The damage threshold is a function of the waveshape and is particularly sensitive to the duration of the transient.

The most widely used model for damage threshold was presented by Wunsch and Bell (1968). They showed that the maximum power, P, that could be safely dissipated in a semiconductor junction was given by

P = kt-1/2

where t is the time duration of a pulse and k is the damage constant. Devices with a larger value of k are able to withstand larger transients. The inverse square-root dependence on the pulse duration was derived by Wunsch and Bell (1968) for adiabatic heating of the junction. This relation is approximately valid for pulse durations that satisfy

0.1 µs < t < 20 µs

This simple model is known as a thermal model, since the mechanism for damage is melting of the semiconductor by excessive energy deposited in the bulk semiconductor.

Ideally, the value of k would be a constant for a device with a particular model number. Much effort has been devoted to finding the proper value of the damage constant, k, for hundreds of different silicon diodes and transistors. Several conclusions are clear from this effort.

A relation of the form

P = At-B

fits the empirical failure data better than the form where B is one-half. Efforts to predict the values of k, A, or B from parameters on the specification sheet (e.g., thermal resistance) have not been particularly successful, so the value of k or the values of A and B must be determined by experiment.

Enlow (1981) described variations in the mean failure threshold for samples of 100 transistors from each of five manufacturers for four different 2N part numbers. Even for devices of the same model number and same manufacturer, the standard deviation for the failure threshold was often about 25% of the mean value. When failure thresholds for devices of the same model number but different ± σ). The mean for the IDI devices is 4.86 σ from the mean of the TI devices; the mean for the TI devices is 49.3 σ from the mean of the IDI devices:

97 = 452 — (4.86 × 73)

452 = 97 + (49.3 × 7.2)

These two distributions of failure thresholds are clearly distinct. Measurement of Texas Instruments 2N718 transistors tells nothing about IDI 2N718 transistors.

Kalab (1982) investigated failure thresholds for 2N1613 and 2N4237 transistors from 30 and 29 manufacturers, respectively. One transistor of each model from each manufacturer was opened and inspected with a microscope to determine the geometry. Sixteen different geometries were used for each model–evidently some manufacturers used the same pattern as other manufacturers. For the 2N4237 transistor, the ratio of smallest to largest chip area was a factor of 20. Clearly the model number does not specify how these transistors are fabricated. Twenty samples of each model from each manufacturer were pulsed with a rectangular waveform with a duration of 1 µs and a polarity that reverse-biased one junction. The ratio of minimum to maximum failure power for the collector-base junction for these 1800 transistors of each model was about 3 x 10⁴. Clearly the collector-base failure power cannot be specified by model number.

These damage thresholds are a statistical concept, not precise numbers that are applicable to a particular piece part. When damage thresholds are being determined, the device either fails or it does not fail. If it does fail, one has an upper bound for the damage threshold but no information about the effect of slightly smaller stresses. After the device fails, the experiment cannot be repeated for that particular piece part.

By testing a large number of devices, one can obtain a statistical distribution of damage thresholds and fit various models to these data. Such effort is expensive, and the results are not applicable to components of the same model number from a different manufacturer. Worse yet, there are apparently no discussions in the literature about whether such statistical distributions are applicable to different production lots of the same model number and same manufacturer.

One of the major reasons for this variation in damage thresholds is inherent in the device specifications. Specifications for electronic components give maximum or minimum values for various parameters that are important in the original application of the device, but they do not specify how the device is to be constructed. Therefore, parts from different manufacturers with the same model number will probably have different electrode geometries and different compositions. Also, manufacturers will revise their process for a particular model from time to time without changing the model number of the parts. Such unspecified features can be important in determining transient performance and damage thresholds.

The simple thermal model for damage in semiconductors ignores effects caused by different failure mechanisms such as second breakdown, metalization failure, and breakdown of gate oxide in MOSFETs.

Chowdhuri (1965) showed that diodes that were conducting when a transient voltage was applied had a breakdown voltage that was about half that for a diode that had no initial bias. This would be expected to affect the failure thresholds for these devices. However, most empirical studies of failure thresholds are made with an initially unbiased device. These thresholds are not necessarily valid for devices that are conducting when the transient occurs. This is an important point, since during normal operation of a system, many vulnerable devices may be conducting when the transient occurs and thus have different failure thresholds.

Moreover, most empirical studies of failure thresholds of transistors apply the transient pulse to the base-emitter junction (with a polarity that will reverse-bias this junction) while the collector terminal is an open circuit and there is no initial current in any of the transistor’s terminals. This is certainly not the usual way to operate a transistor, and one would expect that failure thresholds obtained in this way would not be representative of failure thresholds during normal operation.

Devices that are initially conducting may enter the second breakdown region of operation as a result of transient overstress. Once the transient pushes the device into second breakdown, the dc power supply in the system may kill the semiconductor. In this scenario, the transient does not necessarily need a large energy content for the device to fail, since the transient only initiates the failure process.

An insidious possibility is that a stressed part may suffer an adverse change of parameters (called degradation) but still be capable of performing its intended function. A degraded part may have values of its parameters that are either within or outside the minimum and maximum limits, although it may be of more concern when the values of the parameters are outside the limits on the specification sheet. A degraded component may fail later when subjected to some small stress that a virgin component would survive. This late failure is, at least in part, due to latent damage caused by an earlier overstress.

In 1984 a committee of the U.S. National Academy of Science examined the methodologies of calculating the hardness of electronic systems. The committee was skeptical of the assurance one can have in the hardness of systems protected by designs based on mathematical models of a system (Pierce, 1984, p. 21). The lack of reliable information about the damage threshold of individual transistors in an electronic system was part of the reason for this lack of confidence. The committee specifically noted that uncertainties in the damage thresholds of military specification components persist (Pierce, 1984, p. 45).

The author wonders if determinations of damage thresholds are useful for hardening. Certainly one can spend large sums of money testing components and calculating statistical parameters such as mean and standard deviation. Research cited above shows that there are large variations in damage thresholds for a specific model of transistor. Large, complicated systems have many component parts. If damage thresholds are to be used to design overvoltage protection circuits, one must concentrate on devices with extremely low damage thresholds, because these devices are the weak link in the chain that composes the system. It is difficult to characterize extreme values with statistical analysis, because these outlying data do not fit standard statistical distributions. Indeed, the outlying points may not be members of the same population as the data nearer the mean value.

One attack on the problem of determining damage thresholds involves more research in pure and applied solid-state physics regarding failure mechanisms in devices. With more knowledge one might be able (1) to design devices with greater damage thresholds or (2) to control production parameters to eliminate devices with low damage thresholds without testing every piece part. Although such research is certainly worthwhile, it is of no help to today’s circuit designer.

Another attack on the problem of determining damage thresholds is to use the absolute maximum ratings given in the specification sheet. These parameters are usually steady-state ratings. It is well known that components can tolerate, for a few microseconds, values of current and power that are factors of 10² (or more) greater than their maximum steady-state ratings (Wunsch and Bell, 1968; Alexander et al., 1975). Therefore this approach is very conservative. If protecting a circuit to the absolute maximum ratings imposes a considerable hardship on the designer, stresses of a factor of 2 above these steady-state maximum ratings are easily tolerable for a suitably brief time (e.g., a few microseconds). Such an approach also avoids the problem of specifying in advance the transient waveforms to which the component will be exposed, although it will be necessary to have an estimate of the worst-case peak current and a few other parameters.

Such an approach has been endorsed by U.S. Department of Defense Military Handbook 419 (p. 1-50, 1982). In the absence of data from manufacturers or laboratory testing, the following surge withstand levels were given as typical:

Integrated circuits: 1.5 times normal rated junction and supply voltage

Discrete transistors: 2 times normal rated junction voltage

Diodes: 1.5 times peak inverse voltage

Under this approach, components should be protected against the following conditions:

rise time on the order of few nanoseconds

continuation of stress at reduced level for 0.5 second

polarity reversals

estimate of worst-case peak current and total energy in transient

The brief rise time is typical of the threat from NEMP and ESD. The continuation of stress for 0.5 second is typical of continuing currents in cloud-to-ground lightning, as explained in Chapter 2.

Some will object that using the absolute maximum ratings as an estimate of the damage threshold is much too conservative, since components are known to withstand greater stresses for brief periods of time. However, it is often possible to design economical protective circuits that can protect components from stresses that are greater than the absolute maximum ratings. The use of these circuits can provide substantial assurance that equipment will survive exposure to adverse electrical environments.

2. Upset Threshold

Nearly all circuits that protect vulnerable devices from damage allow a small fraction of the incident transient, called the remnant, to propagate to the protected devices. In a properly designed protection circuit, the remnant will have insufficient energy, current, or voltage to damage protected devices. However, there is still concern that the remnant could be misinterpreted by the system as valid data. Such misinterpretation may cause upset. The threshold for upset is usually within the normal range of input voltages to the system. Therefore, one cannot discriminate against upset on the basis of voltage levels alone. Some ways of dealing with the problem of upset are discussed in Chapter 21.

Some engineers associate large magnitudes of voltage with damage and large magnitudes of dV/dt with upset. This simple association can be misleading. Overvoltages corrupt data, which can lead to upset. Large magnitudes of dV /dt can propagate through the parasitic capacitance between primary and secondary windings of a transformer and damage components in a power supply. One of the most common causes of upset is an outage on the mains, which has a zero value of dV /dt.

2

Threats

There are several different sources of overvoltages that are of interest to electronic circuit designers: lightning, electromagnetic effects of detonation of nuclear weapons, high-power microwave weapons, electrostatic discharge (static electricity), and switching reactive loads. Properties of these sources are described in this chapter. This discussion is deliberately concise: interested readers will find a more detailed account in the references cited.

A. LIGHTNING

The physics of the lightning discharge have been reivewed by Uman (1969, 1987), Golde et al. (1977), and Uman and Kreider (1982). There are two common forms of lightning: cloud-to-ground lightning and intracloud lightning.

Cloud-to-ground lightning begins when a highly ionized plasma, called the stepped leader, propagates from a thundercloud toward the ground. When the leader is within about 50 m of the ground, another electrical discharge, called a streamer, propagates upward from the ground and establishes a highly conducting path between the ground and the stepped leader. At this time an arc current, called the return stroke, flows from the ground up the ionized channel into the thundercloud. The return stroke produces the intense luminosity that is seen as lightning. At the end of a return stroke, there is often a continuing current on the order of 100 A in the channel. This continuing current can have durations between a few milliseconds and a half-second. After a few tens of milliseconds or more, another leader can travel down the same ionized channel toward the ground and produce a second return stroke. This process can repeat itself several times. The entire event is called a lightning flash. One flash typically contains between three and five leader-return stroke sequences. Most lightning research has been directed at understanding the return stroke process in a cloud-to-ground lightning flash. Some of the major parameters in cloud-to-ground lightning flashes are summarized in Table 2-1.

TABLE 2-1. Major Parameters in Cloud-to-Ground Lightning Flash

Use of the term worst case to describe natural phenomena is risky. Since only a trivial fraction of lightning strokes are measured by scientists, it is quite possible that rare events may have more severe properties. Events with large magnitudes of currents or charge transfer often exceed full-scale limits of electronic measuring and recording instruments. Some reports of large magnitudes in the literature were obtained by inference from indirect techniques.

Rise times of the order of 1 µs are commonly reported in the older literature. These values for the rise time are too large, owing to inadequate bandwidth of recording devices (e.g., tape recorders, oscilloscopes) and electronic signal processing circuits. Even when oscilloscopes with adequate bandwidth were used, the sweep rate was usually set to a relatively slow rate in order to capture most of the return stroke waveform. Therefore, data on submicrosecond rise times could not be obtained. Recent measurements with faster electronics, rapid analog-to-digital data conversion, and storage in semiconductor digital memories have revealed rise times on the order of 0.1 µs. These data may still suffer from limited bandwidth.

For practical reasons, direct measurements of lightning currents are usually done with instruments located on tall towers or tall buildings that are struck by lightning many times during each year. However, the presence of the tall object may alter the properties of the lightning. In particular, many of the lightning events are upward-propagating discharges that are not preceded by a stepped leader.

The combination of a 20 kA peak current and a 0.2 µs rise time.implies a value of dI/dt of 10¹¹ A s-1. This large value of dI/dt implies that transient protection circuits must use radio frequency design techniques, particularly considerations of parasitic inductance and capacitance of conductors, which are discussed in Chapter 15.

Although peak currents of the order of 10 kA in cloud-to-ground lightning are certainly impressive, one should recognize that the bulk of the charge transferred by lightning flashes occurs during the continuing currents, which are usually between 50 and 500 A, for a duration between about 0.04 and 0.5 seconds. The continuing current is responsible for much of the damage by direct strikes, including arc burns on conductors and forest fires (Fuquay et al., 1972; Brook et al., 1962; Williams and Brook, 1963).

Kreider et al. (1977) have measured the electric field from stepped leaders and report a mean rise time on the order of 0.3 µs and a full width at half maximum of about 0.5 µs. They estimate that the peak leader currents are typically between 2 and 8 kA when the leader is near the ground. This is of particular importance to aircraft and missiles in flight, since these vehicles may be near a leader.

In contrast to cloud-to-ground lightning, rather little is known about the properties of intracloud lightning. Measurements of lightning strikes to instrumented aircraft often show peak currents of a few kiloamperes (Thomas and Pitts, 1983). The presence of the aircraft probably triggers many lightning events that would otherwise not occur. Nevertheless, data obtained in this way are certainly relevant to assessments of lightning hazards to aircraft.

B. ELECTROMAGNETIC PULSE (EMP) FROM NUCLEAR WEAPONS

Detonation of nuclear weapons produces an intense electromagnetic pulse (EMP) which is a threat to many electronic systems, both military and civilian. There are three different types of EMP, which depend on the location of the weapon and observer:

high-altitude electromagnetic pulse (HEMP): the weapon is detonated above the atmosphere, and the observer is on the surface of the earth or in the lower part of the atmosphere within line of sight of the detonation

surface burst EMP: the weapon is detonated at ground level, and the observer is on the ground within a few kilometers of ground zero.

system-generated electromagnetic pulse (SGEMP): the weapon is detonated above the atmosphere, and the observer is also above the atmosphere and within line of sight of the detonation

When a nuclear weapon is detonated, a very large flux of photons (gamma rays) is produced. The interaction of these photons produces the various types of EMP.

1. High-Altitude Electromagnetic Pulse

Detonation of nuclear weapons above the atmosphere (altitude greater than 30 km, typically between 100 and 500 km) produces HEMP that illuminates all objects within line of sight of the weapon on the surface of the earth or in the lower atmosphere of the earth. A burst 300 to 500 km above Kansas would illuminate most of the continental United States.

Gamma rays from the weapon interact with air molecules through the Compton effect to produce pairs of electrons and positive ions. The electrons are ejected with large speeds and are turned by the earth’s magnetic field to produce HEMP. The volume where the charge is separated and the electromagnetic field is produced is known as the source region. For HEMP, the source region is in the upper atmosphere above the observer on the ground or in the lower atmosphere, as shown in Fig. 2-2. The physics of the generation of HEMP has been reviewed in a number of references (Sherman, 1975; Longmire, 1978, 1985a; Glasstone and Dolan, 1977, pp. 514—540; Lee, 1986; Longmire et al., 1987).

The analytical expression for the unclassified HEMP waveform is given by Eq. 1, where E is the electric field in units of volts per meter, t is time in seconds, and H is the magnetic field in amperes per meter (Sherman, 1975; Stansberry, 1977).

(1)

(2)

This HEMP waveform given in Eq. 1 is plotted on two different time scales in Fig. 2-1. As shown in Fig. 2-1, the HEMP waveform is a pulse with a rise time (10—90% of peak) of about 5 ns, a full width at half-maximum of about 0.2 µs, and a peak electric field of 50 kV/m.

It is important to recognize that Eq. 1 is a simple model that is representative of some worst-case features of HEMP: it has the minimum rise time and maximum duration that are likely to be observed. It is not an accurate description of the waveform at any particular location. For example, the 0.2 µs duration does not occur at the same location where the 5 ns rise time is observed (Sherman, 1975, pp. 13, 22; Longmire et al., 1987). The details of the HEMP waveform vary depending on type of weapon, explosive yield, altitude of burst, and location of the burst and observer. The use of Eq. 1 is justified as a threat waveform for use in designing protection, because the location of the detonation is unknown in advance.

The energy density, w, is given by Eq. 3:

(3)

where E and H are given by Eqs. 1 and 2. The value of w is 0.9 J/m². The physical interpretation of w is that a loop antenna with a cross-sectional area of 1 m² oriented perpendicular to the direction of propagation of the EMP wave would provide a pulse with an energy of 0.9 J. This is sufficient energy to destroy most small-signal high-frequency transistors or integrated circuits. This computation is not representative of the energy delivered by HEMP at any particular location, because a simple threat waveform was used for E and H rather than actual values of fields at a specific location. The value of 0.9 J/m² may be considered an approximate upper bound for actual values of w.

Fig. 2-1. Free space electric field of HEMP waveform as a function of time.

Equation 1 describes the electromagnetic field in free space. It does not describe the current or voltage waveform at the end of a long cable that is illuminated by HEMP. Klein et al. (1985) have calculated the current in an overhead high-voltage transmission line that would be caused by HEMP. They found that the initial peak current occurs at about 0.03 µs, which is much longer than the peak of the free-space HEMP waveform.

Fig. 2-2. HEMP production (drawing not to scale).

2. Effects of HEMP

Little unclassified information is available on the effects of HEMP on modern electronic systems. Weapons tests in the 1950s and 1960s were mostly concerned with damage from blast and thermal radiation from detonations in the lower atmosphere or at the earth’s surface. At long ranges, the magnitude of EMP from such tests is small.

Most of the unclassified data on HEMP comes from the STARFISH PRIME test, 8 July 1962 at about 23 h Hawaii local time. A 1.4 megaton device was detonated at an altitude of 400 km above Johnson Island (Glasstone and Dolan, 1977, pp. 45, 523). Some disruption occurred in civilian systems in Honolulu, Hawaii as a result of HEMP from this burst. In particular, hundreds of burglar alarms were activated by HEMP, and some fuses were opened in long series strings of street lamps. Longmire (1985b) has calculated the expected magnitude of HEMP at Honolulu. The peak electric field was calculated to be 5.6 kV/m, and the energy density about 0.01 J/m². The peak electric field was about 10% of the value given in Eq. 1. The implication is that STARFISH PRIME at Honolulu was far from a worst-case event.

Modern systems may be much more vulnerable to damage by HEMP than circuits in 1962, owing to the increasing use of integrated circuits. One anecdotal report² states that waves from an EMP simulator damaged electronic ignition systems in automobiles, so that the cars that were parked near the simulator would not start after the tests. Modern telephone systems use semiconductor switches instead of electromechanical relays; there is no doubt that the semiconductor switches are more vulnerable to damage.

HEMP may be a particular threat to systems that depend on electrical power or data that are carried on long lines. This includes nearly every civilian system and many military systems. The peak current in overhead lines may be on the order of 10 kA (Vance, 1975; Glasstone and Dolan, 1977, p. 530). Considerable shielding from HEMP can be obtained by burying the cable in the ground. However, depending on soil conductivity and depth of burial, peak currents are predicted to be between a few hundred amperes and a few kiloamperes. Such currents are not negligible.

There are two principal areas of concern about HEMP and the electric power transmission network. The abnormal voltages and currents caused by HEMP may trip circuit breakers on these lines, causing a momentary interruption of power. Some of the circuit breakers will probably automatically reclose, so that the power will be restored. Still this momentary outage could produce malfunction (upset) in computers and electronic systems. One must be concerned not only with abnormal currents and voltages on the transmission lines themselves, but also with the effects of HEMP on electronic control circuits for electric power transmission. It is possible that HEMP could destabilize the electric power grid and cause a nationwide outage. It is difficult to ascertain the effects of HEMP on large systems, such as the electric power grid. Large systems cannot be tested in a laboratory simulator; computer calculations that are unsupported by experimental confirmation are notoriously unreliable.

HEMP may also adversely affect aircraft and missiles in flight. The electromagnetic field changes produce skin currents that excite resonances in the aircraft. The resonance frequencies are generally between 1 and 20 MHz. The electromagnetic field from skin currents, in turn, induces currents in cables inside the wing and fuselage. In an independent process, transducers and antennae on the exterior of the aircraft are illuminated directly by the electromagnetic field from HEMP. Electronic circuits that are connected to these transducers or antennae may be exposed to relatively large transient currents.

Nuclear warheads in antiballistic missile (ABM) systems pose a special threat. Detonation of the ABM’s warhead above the atmosphere will produce HEMP that may adversely affect operation of electronic systems at the launch site on the ground.

3. Magnetohydrodynamic HEMP

Nearly all discussions of HEMP concentrate on overvoltages that have durations of the order of no more than a few microseconds. However, there is a late-time effect, known as magnetohydrodynamic (MHD) EMP, that persists for hundreds of seconds after the detonation. The MHD effect is generated by distortion of the earth’s geomagnetic field by slowly moving high-conductivity material in two places: (1) the source region in the upper atmosphere and (2) near the point where the weapon was detonated. The magnitude of MHD electric field is on the order of 10 V/km, which is much smaller than the submicrosecond HEMP described above. The component of the MHD electric field that is parallel to the earth’s surface causes abnormal currents in long conductors, such as the electric power transmission system (Legro et al., 1986). These currents provide a dc bias in transformers and produce saturation of the magnetic field inside the core of transformers.

Electric fields of similar magnitudes and frequencies are produced by severe solar magnetic activity at higher latitudes, which also produce aurora. There have been reports of upset of power transmission by tripping of circuit breakers and saturation of transformer cores during geomagnetic storms (Albertson et al., 1973; Albertson and Thorson, 1973; Kappenman et al., 1983). Saturation of the magnetic field inside the core of transformers can result in severe harmonic distortion of the power waveform and possible overheating of the transformer. Albertson et al. (1973) report reductions in rms voltage to as low as 45% of normal during geomagnetic storms.

A power system can be protected from some of the effects of geomagnetic currents, for example by connecting a capacitor in series with the neutral of a three-phase system to block dc current. This might be done for power systems at extreme latitudes, where severe geomagnetic storms are more common. It will not be economical to install this kind of protection on power systems at lower latitudes, so these systems will remain susceptible to upset by MHD EMP.

4. Surface Burst EMP

The EMP from surface bursts, in contrast to high-altitude bursts described above, is confined to a relatively small region, about 3 to 8 km in radius centered about ground zero (Glasstone and Dolan, 1977, pp. 517—518). This region will also be affected by blast and thermal radiation. However, surface burst EMP is still an important problem for hardened targets (e.g., underground missile silos and command bunkers). Combat troops who are far enough from ground zero to survive the blast may have their electronic equipment destroyed or degraded by EMP. Furthermore, EMP from surface bursts will produce large transient currents on both overhead and buried cables such as power lines and telephone lines. These transient currents could travel far from the region of the burst and cause damage to facilities that would not be affected by the blast or heat from the burst. There is no standard waveform for surface burst EMP, because this phenomenon is strongly dependent on the type of weapon, explosive yield, altitude of burst, and distance between the burst and observer. In general the electric field has a rise time of a few nanoseconds to a positive peak, followed by a negative peak after a few tens of microseconds. Surface burst EMP has relatively more energy at frequencies below 100 kHz than does HEMP.

The physics of the generation of surface burst EMP has been reviewed by Longmire (1978), Glasstone and Dolan (1977, pp. 517—518, 535—536), Gilbert and Longmire (1985), and Lee (1986, pp. 33—40). A schematic drawing of the source region is shown in Fig. 2-3. Because most of the large electromagnetic fields are confined to the source region, EMP from surface bursts is often called source region EMP (SREMP).

The electric field at a distance of about 1 km from ground zero of a 10 megaton detonation is sufficient to trigger lightning that travels from the ground upward (Uman et al., 1972). One lightning stroke had a continuing current that provided enough illumination to be recorded on photographic film for 75 ms.

French researchers measured the current in a field of wires placed on the sand beneath a surface burst in the Sahara Desert and connected to earth about 3 km from ground zero (Ferrieu and Rocard, 1961). The current had a peak value of 150 kA, which decayed to zero at 150 µs. The current had a second peak of 56 kA of opposite polarity to the first. No other results were given.

Little unclassified information is available on the effects of surface burst EMP on modern electronic systems. If any electronic systems were exposed to surface burst EMP before the test ban treaty of 1963, those systems contained vacuum tubes and not semiconductor integrated circuits. It is now known that integrated circuits are many orders of magnitude more vulnerable to damage by EMP than vacuum tubes.

Fig. 2-3. Surface-burst EMP production mechanism.

5. System-Generated Electromagnetic Pulse

When nuclear weapons are detonated in space, gamma rays and subatomic particles travel away from the explosion. When these particles impinge on the metal container of spacecraft, an intense electromagnetic pulse is produced inside the space vehicle. This pulse is known as system-generated electromagnetic pulse (SGEMP), because the pulse is produced by the interaction of the system and the incident particles.

The most important part of SGEMP is produced when the photons (X-rays and gamma rays) from a nuclear detonation impinge on materials in a spacecraft (Higgins et al., 1978; Lee, 1986, p. 42). The interaction between the photons and atoms in the spacecraft occurs through three different processes–the photoelectric effect, the Compton effect, and pair production, all of which produce electrons. There are two effects that occur as electrons are ejected from the conducting walls of the spacecraft: a dipole electric field is created by the ejected electrons and the positive ions that remain in the conducting material; a magnetic field is created by the current in the conducting material as an equilibrium charge distribution is sought. Unclassified estimates are that the magnitude of the electric field are between 100 kV/m and 1 MV/m (Glasstone and Dolan, 1977, p. 522).

The incident particles from the detonation and electrons produced by interactions can directly effect electronic circuits inside the spacecraft. These effects are often known as TREE, an acronym for transient radiation effects on electronics. In particular, each electron can ionize many atoms of material inside the system and produce either damage or upset of the electronic system. The flux of neutrons from a nuclear detonation can also damage microelectronic circuits. Because the current is injected throughout the system, there is no input port for this overstress. Therefore, conventional circuits for protection against overvoltages are of no use in TREE.

One cannot rely on a Faraday Cage to shield the interior of the spacecraft from SGEMP, since time will be required for an equilibrium charge distribution to occur in the conducting shield. Most of the time delay will be caused by the finite speed of light in the conducting material. Furthermore, a perfect Faraday Cage would make a space vehicle useless: one must have penetrations for antennae, optical sensors, and possibly panels of solar cells for electric power.

Mitigation of SGEMP could be done in several different ways. One way is to attenuate the incident photon flux with a