ESD occurs when an imbalance in the amounts of positive and negative electrical charges on the surface of an object is released. The most dramatic example of ESD is lightning, which splits trees as easily as it lights up the sky. The amount of energy released when you touch a metal object can be quite large. The buildup of energy with nylon clothes can easily reach 21,000 volts. About 750 volts are required to produce a visible spark with ESD, while a mere 10 volts or so can ruin a computer chip.
The actual amount of energy in a given ESD event depends on the types of materials involved (wool fabrics generate less than nylon), the humidity (low humidity offers less resistant to the discharge), the amount of physical energy (friction) involved, and how quickly the energy is released.
ESD does not have to be seen (a spark) to do damage to electronic components. Voltages lower than 10 volts can damage some parts.
Over the years, engineers have produced smaller and smaller components, which operate at lower and lower voltages. Their goal is to reduce size, cost, and operational heat production. Those are worthy goals, but because of reduced component size, they present a smaller target with less resistance to power surges. This makes the parts more susceptible to damage from ESD.
The amount of damage and resulting problems caused by ESD can be divided into three categories:
This is sometimes referred to as "frying" or "smoking" a part because of the heat (and sometimes the noise and smoke) generated during the failure. Mishandling and misapplication of a power source, cable, or test instrument are the most likely causes. Care in opening, installing, cabling, and testing are the best ways to prevent this type of ESD damage.
An ESD can produce an erratic fault in a component. This kind of problem is very difficult to detect and repair, because the failure is intermittent. It is easy to blame the operating system or a program operation for the data loss or system crash. The best way to diagnose and correct this type of problem is to remove and replace suspected parts until the failure stops appearing.
This type of failure weakens the actual transistor. The part will seem normal in most operations and will frequently pass quality control and conformance tests. Like upset failures, these can be very difficult to isolate.