The energy of transients

To properly size transient protection, you need to predict the energy content of the expected transients and surges. This calls for a knowledge of the source impedance and likely amplitude of the transients. Some research has been done on mains transient amplitudes and in the telecom and automotive environments. For guidance on expected transients in mains-supplied environments, see “Guide on the Surge Environment in Low-Voltage AC Power Circuits”, IEEE C62.41.1–2002.

The actual energy delivered by a test generator (which is intended, to some degree, to mimic real-life transients) into a defined resistive load can be calculated. For the ESD and EFT waveforms, these can be the calibration loads of 2Ω and 50Ω respectively. For the surge and ring waves, a load which matches the output impedance can be chosen, and the voltage or current waveform is assumed to be maintained into this resistance with half the open circuit (or short circuit, for current) amplitude. The standard surge test for commercial equipment is specified in IEC (EN) 61000-4-5. This gives levels from 500V to 4kV peak voltage, with a voltage waveform of 1.2μs rise time and 50μs half-time, and a current waveform of 8μs rise time and 20μs half time. There are three levels of source impedance, depending on the mode of coupling: the most severe is 2Ω line-to-line on power supply inputs or directly to cable screens, which would give a worst-case peak current of 2000A at the highest surge level. The other impedances are 12Ω for line-to-protective earth, on power supplies, and 42Ω line-to-earth on signal lines.

In these cases the energy in Joules (watt seconds) is shown in the graph and is given by

where V(t) and I(t) are the open circuit voltage and short circuit current waveforms, respectively.

The graph is for comparative purposes only – the real energy delivered to a particular load in an EUT can only be calculated if the load impedance characteristics, and the actual waveshape applied to this load, are known accurately. A load such as a surge suppressor will be non-linear and will also have a time or frequency dependence.

Clamping

Incoming over-voltage transients are clamped at the circuit interface by using non-linear components: metal oxide varistors (MOVs), transient voltage suppressors (TVS - avalanche devices, i.e. zeners, including foldback types) and spark gaps (GDTs - gas discharge tubes). These are placed in parallel with the transient source and their effectiveness depends on the ratio of their dynamic slope impedance Z_S to the transient source impedance Z_T.

The choice of device is dictated by circuit operating parameters: leakage current, capacitance and threshold voltage are important with respect to normal circuit operation; clamp voltage, follow-on current, energy capability and response time are important when the device is faced with a transient. Response time is particularly critical for transients with fast rising edges, as a slow device will let through the leading edge which may well be the most damaging part of the transient. For these purposes device inductance becomes as important as slope resistance and short leads or chip packages are preferred.

The device must be sized so that it stays high-impedance up to the maximum continuous operating voltage of the circuit, with a safety margin for tolerances, and to be able to absorb the energy from any expected transient. For most smaller components the "knee" or threshold voltage is defined at a current of 1mA through the device (also referred to as the breakdown voltage or V_BR by manufacturers). The first requirement means that the maximum transient clamping voltage can be 2-3 times the continuous voltage, and circuits that are protected by the suppressor must be able to withstand this. The procedure for specifying a device looks something like this:

define the maximum normal operating voltage at which the device must take less than 1mA; for example, a 24V DC supply may have a +25% tolerance, so this comes up to 30V;

select the voltage rating for the TVS device such that it remains above this voltage even with the specified tolerance and temperature coefficient; with a 10% tolerance this would typically mean at least a 33V device;

from the expected peak surge voltage and source impedance, derive the peak current that the protection device can expect to see; e.g. for 1kV/42Ω, this is 24A, not taking into account any other impedances;

check the peak varistor voltage that can be expected at this current; for a 1206 MOV device, this might turn out to be 100-120V, or for a zener TVS it might be 50-60V;

confirm whether the downstream circuit can be expected to withstand such a voltage, whether the device's capacitance is acceptable to the circuit's normal operation, and whether the device's energy or peak current rating is adequate. Be prepared to iterate a few times, and with a particularly difficult interface, you may want to combine different devices.

Actual waveforms

The waveforms seen across actual devices when they are exposed to a surge are shown here. The applied stress is the IEC 61000-4-5 surge, from 300V to 1000V peak (each of the family of curves shows 300V - blue, 500V - purple, 750V - green and 1kV - red) and from various source impedances as defined in the standard.

MOV devices have a high slope resistance and a greater peak clamping voltage, but are very robust and can take high energy repeatedly. Even small parts (1206 and smaller) are able to withstand severe surges without destruction, even though they may experience up to three times their rated stand off voltage. For power supply protection, where likely source impedances are low but the supply components could be conservatively rated, these would be the appropriate component.

TVS (Zener) devices have a low slope resistance and a very flat clamping profile at low energies, but are more easily destroyed by higher energy surges. (Note the graph which demonstrates the destruction of the 33V TVS part when subjected to 1kV from a 12Ω impedance.) These would generally be the preferred part if the source impedance is high but the downstream circuit withstand voltage is not, as in signal circuits.

Ordinary zeners will work but aren't characterised for surges; again, for signal circuits with some series impedance, they may well be adequate.

When the expected transient source impedance is low (less than a few ohms), it is worthwhile raising the impedance of the input circuit with a series resistor, if possible, or an inductor if not. The expected surge amplitude needs to be carefully investigated with this approach, as the bulk of the surge voltage will be dropped across the series component, which must be sized to withstand it without damage. Also, non-lossy series inductors should be used with caution: a transient can cause an inductor/capacitor resonant combination to ring, which will actually increase the peak voltage after the filter.

Layout of suppressors

A high transient current flows through the device when it operates and this dictates careful layout. Short and direct connections to the suppressor (including the ground return path) are vital to avoid compromising the high-speed performance by undesired extra inductance. Transient edges can have very fast risetimes (a few nanoseconds for switching-induced interference down to sub-nanosecond for ESD) and any inductance in the clamping circuit will generate a high differential voltage during the transient plus ringing after it, which will defeat the purpose of the suppressor, since it appears in series with the suppressor's clamp voltage.

The component leads must be short (suppressors are available in SM chip form) and they must be connected locally to the circuit that is to be clamped. Any common impedance coupling, via ground or otherwise, must be avoided. As the purpose of transient suppressors is deliberately to conduct the transient current, inadvertent common impedances will degrade their effectiveness. Tracking should run from the interface terminals directly to the suppressor, and from there into the circuit to be protected. This creates a designed-in current path for the transient, through the suppressor, not through the circuit. Spurs into the suppressor should be avoided or configured for minimum inductance.

Clamping device capacitance

For fast transients such as an ESD event, the clamping device capacitance may affect the result as much as the clamping action itself. The device capacitance varies with voltage and vary from about 10s of pF at about 48V bias to about 100s of pF at lower voltages. A few nanohenries of inductance (often introduced by layout) and 100s of pF of capacitance may resonate at about 100s of MHz. This could cause some oscillatory response in some applications. 

Combining suppressors

Devices can be combined with each other and with series chokes or resistors to offer a total suppression performance that would be unobtainable from any single device. For example, telecoms applications require signal line protection not just from straightforward transients, but also from local lightning strikes and shorting between the signal line and AC power lines. Military applications may require protection against NEMP, whose rise time is much faster and duration longer than for lightning.

These can only be dealt with by providing coordinated protection; primary protection, offered by gas discharge tubes (GDTs) or carbon spark gaps, will remove the major part of the incoming energy but leaves an initial spike due to the device's slow response time. Secondary protection, provided by a semiconductor device and/or a MOV, is faster and needs only to deal with the residual energy. It is of course necessary to place inductive or resistive impedance between each device, and the analysis and optimisation of this type of very non-linear circuit is best achieved by a circuit simulation package.

Very often, you will combine suppressors with RF filtering. For this you can use a 3-terminal varistor/capacitor device - effectively a capacitor with a defined breakdown voltage and energy capability - which will provide both functions in one package.

Not clamping

It is entirely reasonable to design an interface without clamping-type surge suppressors altogether, if practical. Ways of doing this include:

if the surges are only expected in common mode (with respect to chassis earth) then ensure an adequate separation distance and electrical isolation between all circuit components and the chassis. Air gap breakdown occurs at about 3kV/mm, creepage across the surface of insulators significantly less - roughly half this.

if high-ish series impedances and low bandwidths are feasible, then a simple R-C time constant may be all that is needed; 10kΩ and 0.1μF make 1ms, which will attenuate a 1kV 1.2/50μs pulse to about 60V, although be aware that most of the 1kV will be dropped across the resistor, which probably rules out an 0402 part (!)

in a few applications, it may be feasible simply to use heavily over-rated components, such as 800V bridge rectifiers or MOSFETs where you only need 50V.

Click here for Questions