Fiber-optic test is an important part of building high-speed digital-communications systems. The bottom line is this: Are the bits coming across correctly? When a bit is a one, does the equipment see it as a one? When a bit is a zero, does it see it as a zero and not make any mistakes?
The tests are based on fundamental parts of the eye diagram. Take the shape of the eye diagram. Ideally, the eye diagram should be as open as possible. A "one" should be as far away from the zero level as possible. At the other end, the receiver has to decide: Is the bit a one or is it a zero? If the eye is wide open, the receiver isn't going to call a zero a one, or a one a zero. So a key element used to characterize an eye diagram is its shape. This is done is by mask testing. An industry-defined polygon is proportionally sized to the power level of the signal and must fit within and around the eye diagram.
Subtle differences exist between eye diagrams for different standards. For example, at the 622-Mbit/s SONET rate, a six-sided polygon sits in the middle of the eye diagram. At 2.5 Gbits/s, it's shaped like a rectangle (see opening art). The concept is the same, but the shape is a little more forgiving as the user closes in on higher rates. Fibre Channel and Gigabit Ethernet have their own masks, which are six-sided. The eye diagram's width is proportional to the bit rate. Approaching higher rates, the width of the mask is adjusted to match the data rate.
A designer might expand the size of the mask until errors occur. That establishes compliancy by a certain percent. The figure below shows how a scope constructs an eye diagram. Greg LeCheminant, Agilent Technologies.