Information traveling across a computer network or another type of telecommunications network typically comes in packets. Packets are smaller, "bite-sized" pieces of a larger chunk of information. Although telecommunication technology is generally reliable and seamless from the end-user's point of view, that is only because the computer does the dirty work. It constantly sifts through the packets, looking for fraudulent and corrupt data, discarding it when found. A burst error is a string of corrupt data, measured as the length between — and including — the first and last error signals.
For example, imagine sending a packet containing all of the letters of the alphabet, A through Z. If the recipient's computer "opens" the packet and finds that the first letter in the sequence is "Q" and the last letter in the sequence is "R," that is a burst error. The "burst" of data in the packet is corrupt.
Although in the example the first and last letters are defined as corrupt, that does not mean that every letter within the packet is damaged. Imagine that every other letter is as it should be; only position one, "A," and position 26, "Z," have been damaged. The number of correct bits of information between the damaged ends is called the guard band. In this case, the guard band would be 24, because there are 24 correct letters separating the two damaged ones.
Measuring the length of a burst error is simple. It is defined as the number of individual bits separating the very first occurrence of the error from the last occurrence, including the initial and final incorrect bits. In the previous example, the length of the burst error would be 26.
The causes of a burst error can vary widely. It is not always possible to measure them accurately. Generally, this corruption can occur through any number of sources, including signal degradation, packet loss, other types of network failure, or sending failure on the part of the computer. In networking, as in the real world, sometimes things go wrong. Fortunately, most forms of networking provide built in error-checking mechanisms, allowing a receiving computer to compare the actual received data against an impression of the data that was sent, allowing it to recognize whether something has gone wrong along the way.