Why does a low clamping voltage not guarantee a low voltage at the device?

Because the protected part sees the clamp plus whatever the loop inductance develops, and during an electrostatic discharge that second term dominates. An 8 kV contact discharge reaches about 30 A in roughly 0.8 nanoseconds, which is 37.5 amps per nanosecond. Across 1 nH that is 37.5 V, and across 5 nH it is 187.5 V. A TVS specified to clamp at 9 V therefore delivers about 46 V to the device with a tight 1 nH loop and about 196 V with a sloppy 5 nH one. The datasheet number describes the component; the layout decides the result.

What is the difference between a TVS, a varistor and a gas discharge tube?

Speed, energy and leakage, roughly in opposition. A TVS diode clamps fast and tightly, has low capacitance in its signal-line variants, and handles the least energy. A metal oxide varistor absorbs far more energy, clamps less tightly, and degrades cumulatively with every event until it eventually fails, often short. A gas discharge tube handles the most energy of the three and has almost no capacitance, but takes microseconds to strike and then crowbars nearly to zero volts, which can hold a mains circuit in conduction after the transient has passed. Serious mains protection usually stages them: a gas tube or varistor for the energy, then a TVS for what gets past.

Where should the protection physically sit?

At the connector, before the transient reaches anything else, with the shortest possible loop from the protected line to the reference the surge is returning to. Protection placed after a run of trace lets that trace act as an antenna and injects the transient into everything sharing the reference along the way. The loop that matters includes the return path, so a TVS with a short signal pin and a long ground via is not a tight loop. This is the same geometry argument as the rest of EMC: the enclosed area decides the voltage.

Transient protection: TVS, MOV, GDT and layout

Author Hugues Orgitello Last updated 30 August 2026 ~4 min read

Guide, EMC fundamentals

A protection component is selected from a datasheet and defeated by a layout. The arithmetic below shows a TVS specified to clamp at 9 V delivering nearly 200 V to the part it was fitted to protect, entirely because of the loop it sits in. This page separates the two transient problems, which need opposite answers, and then explains why geometry decides both.

Two different problems

Electrostatic discharge is fast and nearly energy-free. IEC 61000-4-2 discharges a 150 pF capacitor through 330 ohms, giving a current that rises in under a nanosecond and decays in tens:

Test level	First peak	di/dt at 0.8 ns rise
2 kV	7.5 A	9.4 A/ns
4 kV	15 A	18.8 A/ns
6 kV	22.5 A	28.1 A/ns
8 kV	30 A	37.5 A/ns

Surge is the opposite. IEC 61000-4-5 uses an 8/20 microsecond current waveform that rises in microseconds and can reach hundreds or thousands of amps, carrying joules rather than microjoules.

The rates of change differ by roughly six hundred times. A 500 A surge rising in 8 microseconds is 0.0625 A/ns; an 8 kV ESD is 37.5 A/ns.

That single ratio explains the whole design split. ESD is an inductance problem. Surge is an energy problem. A component sized for one is usually wrong for the other.

Why the clamping voltage is not what the device sees

The protected part sees the clamp voltage plus the voltage the loop inductance develops, V = L di/dt.

At 8 kV ESD, with a TVS clamping at 9 V:

Loop inductance	L di/dt	Device actually sees
1 nH	37.5 V	46.5 V
3 nH	112.5 V	121.5 V
5 nH	187.5 V	196.5 V
10 nH	375 V	384 V

A 9 V clamp delivering 196 V. The datasheet is not wrong; it describes the component, and the component is a small part of what the transient passes through.

For surge the same 5 nH develops only 0.31 V at 500 A, because di/dt is six hundred times lower. The inductance term matters for ESD and is negligible for surge, which is why ESD protection is a layout exercise and surge protection is a component-rating exercise.

The three device families

	TVS diode	Varistor (MOV)	Gas discharge tube
Speed	Fastest	Fast	Slowest, microseconds to strike
Energy	Least	High	Highest
Clamping	Tightest	Loose, rises with current	Crowbars to near zero
Capacitance	Low in signal variants	High	Very low
Ageing	Negligible	Degrades with every event	Limited strike count

TVS diodes clamp fast and tightly. Signal-line variants have low enough capacitance not to disturb high-speed lines, which matters because a protection part on a USB or Ethernet pair must not spoil the signal it protects.

Varistors absorb far more energy but clamp loosely, and the clamping voltage rises with current. They also degrade cumulatively: each event slightly lowers the trigger voltage, so an MOV that has absorbed years of small surges eventually conducts at working voltage. Mains varistors are therefore fused or thermally protected, because the end state is often a short.

Gas discharge tubes handle the most energy and add almost no capacitance, but take microseconds to strike, so something else must survive the interval. Once struck they crowbar to a few volts, which on a mains circuit can sustain follow current after the transient is gone.

Staging, and the series element that makes it work

Serious protection is staged: a gas tube or varistor for the energy, then a TVS for the fast residue that gets past while the first stage is still striking.

This only works if there is impedance between the stages. Two clamps in parallel do not share: without a series element the fast, weak TVS sees the full transient first and fails before the slow, strong device has done anything. A small series resistance or inductance lets the upstream part take the energy and develops the difference the downstream stage needs to stay out of the way.

The series element is not a refinement. Without it, a two-stage protector is a one-stage protector with an expensive decoration.

Layout decides the result

Three rules, all consequences of the arithmetic above.

Protect at the connector. A transient allowed to travel along a trace before being clamped has already coupled into everything sharing that reference. Protection placed after a run of board is protecting the wrong side of the problem.

Minimise the loop, including the return. The inductance that matters is the whole path from the protected line, through the clamp, to the reference the surge is returning to. A TVS with a short signal pin and a long ground via is not a tight loop. This is the same argument as return current paths: enclosed area decides the voltage.

Do not route protected and unprotected sides together. A clean line running beside the dirty side of the same connector picks up capacitively what the clamp just removed.

Key takeaways

ESD and surge are opposite problems. Their di/dt differs by about six hundred times: ESD is inductance, surge is energy.
The clamp voltage is not what the device sees. At 8 kV, 5 nH of loop adds 187 V to a 9 V clamp.
TVS, MOV and GDT trade speed against energy against leakage, and varistors degrade with every event.
Staged protection needs a series element, or the fast device fails before the slow one strikes.
Protect at the connector, with the smallest loop including the return path.

Frequently asked questions

Why does a low clamping voltage not guarantee a low voltage at the device?: Because the protected part sees the clamp plus whatever the loop inductance develops, and during an electrostatic discharge that second term dominates. An 8 kV contact discharge reaches about 30 A in roughly 0.8 nanoseconds, which is 37.5 amps per nanosecond. Across 1 nH that is 37.5 V, and across 5 nH it is 187.5 V. A TVS specified to clamp at 9 V therefore delivers about 46 V to the device with a tight 1 nH loop and about 196 V with a sloppy 5 nH one. The datasheet number describes the component; the layout decides the result.
Are ESD and surge the same problem?: No, and treating them alike is why protection is often fitted in the wrong place. Electrostatic discharge is fast and carries little energy: nanosecond rise, tens of amps, microjoules. Surge is slow and carries a great deal: the 8/20 microsecond current waveform rises in microseconds and can reach hundreds or thousands of amps. Their rates of change differ by a factor of roughly six hundred, so ESD is an inductance problem solved by layout and a fast clamp, while surge is an energy problem solved by a component that can absorb joules. A part sized for one is usually wrong for the other.
What is the difference between a TVS, a varistor and a gas discharge tube?: Speed, energy and leakage, roughly in opposition. A TVS diode clamps fast and tightly, has low capacitance in its signal-line variants, and handles the least energy. A metal oxide varistor absorbs far more energy, clamps less tightly, and degrades cumulatively with every event until it eventually fails, often short. A gas discharge tube handles the most energy of the three and has almost no capacitance, but takes microseconds to strike and then crowbars nearly to zero volts, which can hold a mains circuit in conduction after the transient has passed. Serious mains protection usually stages them: a gas tube or varistor for the energy, then a TVS for what gets past.
Why is a series element needed between stages?: Because two clamps in parallel do not share. Without impedance between them the fast, weak device sees the transient first and at full current, and fails before the slow, strong device has struck. A small series resistance or inductance lets the upstream part take the energy and develops the voltage difference the downstream stage needs in order to stay out of the way. The series element is not a detail of the design; without it a staged protector is a single stage with a decoration.
Where should the protection physically sit?: At the connector, before the transient reaches anything else, with the shortest possible loop from the protected line to the reference the surge is returning to. Protection placed after a run of trace lets that trace act as an antenna and injects the transient into everything sharing the reference along the way. The loop that matters includes the return path, so a TVS with a short signal pin and a long ground via is not a tight loop. This is the same geometry argument as the rest of EMC: the enclosed area decides the voltage.