The electrical pulses which are sent through a network are normally sent at very specific intervals of time. However, every piece of equipment has a certain amount of tolerance within which it operates. If a signal passes through several pieces of equipment or cable, these tolerances add up and cause the resulting signal to shift slightly in phase compared to what was originally sent. This makes it difficult or impossible for the next device on the network to "lock on" to the signal and causes errors.
Figure One below shows a "normal" signal on a network. Note how the pulses are repeating at specific times. Let's assume that the figure shows 50 cycles of the same pattern. Since the signal is so tightly timed, overlaying them on one display still results in a clean signal.
Signal Without Jitter
Figure Two shows the same signal as Figure One, except with the introduction of a significant amount of jitter. Note how instead of arriving at the exact same interval, some of the pulses are arriving earlier than they should, and others are arriving later. This signal is difficult or impossible for the reciever to "lock on" to and decode properly. Some pulses which are binary 1's will be seen as 0's, and vice-versa.
Signal With Significant Jitter
If a system has enough jitter that it is no longer reliable, there are ways to correct the situation. The normal approach is to use a repeater, bridge, or switch with a buffer of some type to accept the signal from one side and send it back out at a tightly controlled speed. This results in a clean output signal which can reliably be recieved by the next piece of equipment on the network. The specific mechanism for accomplishing this will vary depending on the type of signal being "cleaned up."