1. The current from the transient has to go somewhere. If your power supply already has a very low impedance, even at surge frequencies (ESD has a rise time in the nanoseconds) and for large currents (an 8kV ESD spike is some 16A peak), you don't need to worry.
There's also the matter of surges, for circuits that are connected to very long wires that may be subject to induced lightning transients, or full blown strikes (which includes telephone and mains lines, but Dave's recent home security module teardown illustrates another example). These surges are much longer duration and lower impedance, and so contain far more energy. Even a very low impedance power supply is unlikely to stomach this much energy (a small 8/20us pulse with a peak of 10A will charge a 100uF capacitor by maybe 2V), so it needs to be dissipated with a TVS instead.
2. There could be, but note that a spark gap won't fire unless the voltage across it gets high enough to trigger. If this never happens, the spark gap was wasted time, effort, space and (if using GDTs) money.
A very robust system might be constructed thusly:
- Pin to the outside world
- Spark gap
- Series impedance (air cored inductance, resistance: as much of each as is tolerable)
- MOV or TVS (the bigger the better, given limits of capacitance)
- More series impedance (including ferrite beads if needed for EMC)
- TVS or clamp diodes
- Filter cap(s) or band limiting parts (for bandwidth reduction (e.g., radio antenna input, scope front end, etc. etc.) or EMC)
- Series resistance
- device (transistors, ICs, etc.).
Very fast and large pulses cause enough voltage drop across all the series impedances to cause breakdown of the spark gap. This occurs for anything from ESD (fast, short) to sufficiently powerful surges (slow, but lots of peak current and voltage).
What's left after the spark gap ranges from short blips of manageable amplitude (the absorbed ESD spark might go from 8kV, 16A and 50ns down to 2kV, 16A and 5ns, followed by a "drool" of maybe 100V and 100ns) to nothing at all (a sufficiently small surge will never spark, and goes through the first series impedance unaffected).
The second stage is capable of absorbing lots of energy (MOVs are a cheap source of energy capacity, but high in capacitance; avalanche TVSs of comparable ratings are very expensive; thyristor type TVSs have good ratings but can latch on and blow fuses), but at a relatively high voltage drop (plan on a peak voltage up to three times the MOV rating).
The third stage reduces the excess voltage of the second stage, bringing it down to manageable levels for the device.
The final stage is always the device itself being protected. No practical surge protection system will fully control the transient, and some will always appear at the device. If the device can ride out some voltage, but must carry zero current (example: unprotected MOSFET gates), the transient protection must reduce the transient to no more than that voltage. If it can shunt some current, it also becomes part of the protection system (e.g., input clamp diodes, CMOS outputs). Some circuits don't mind much either way (a discrete audio amplifier built from BJTs and resistors will draw some current when driven beyond normal range, but rarely anything destructive by the time the protection system is doing its job; such circuits, historically speaking, have rarely if ever bothered with ESD protection at all, anyway).
Finally, the purpose of properly designed protection is to limit those transients to the device ratings, as sufficient for the equipment specifications (how much a beating it can take, and whether it must remain functional during or after such events).
- Consumer junk rarely uses TVSs at all: they can use chips with built-in ESD protection, and the attached power supply handles surges.
- Professional stuff usually has adequate protection (one MOV or TVS, some series impedance as applicable).
- Two stage protection is occasionally seen in special purposes. Example: CRT monitors, where internal breakdown in the CRT can deliver over 20kV at impedances much lower than an ESD test, and where the protected circuit (video amplifier) must have absolutely minimal loading capacitance.
- Three and four stage protection are the gold standard, but rarely if ever used simply because it's not necessary. Some examples might possibly include military gear that has to survive rough handling, machine model ESD, radar bombardment and jamming, and nuclear EMP, where operation must continue without any reaction or interruption during such events. But I haven't seen such equipment up close, so I don't know what they use.
Tim