A.1 Eye-window measurement

The output interface optical eye-window (EW) measurement involves measuring the open eye-window on a bit-by-bit basis, using a BERT (bit error rate test) test set. The bit error rate (BER) is measured at various T sub _d’s (decision points) within the eye pattern to ensure conformance to the eye-window specification.

The eye-window is given by:

EW = | T_d (max) - T_o | + | T_o - T_d (min) |

For each position of T_d from T_d(min) to T_d(max), a BER measurement is taken, giving the probability of error at the T_d position. In effect, T_d is swept across the eye pattern, measuring the probability of error at each point in the eye. The range of T_d values that result in a BER ≤ 10 ^-12 establishes the eye-window, and the smallest range from T_o must be ≥ half the appropriate eye-window specification.

In practice, a BERT test set is used to generate and sweep the decision point (using the BERT clock in conjunction with a precise delay generator), to make the bit-by-bit error count and to calculate the measured BER. The center of the baud interval (T_o) pattern is the midpoint between positioning T_d to the left and right edges of the eye to achieve a BER > 10^-2 while transmitting a square wave pattern. Subsequent measurements are made while transmitting allowed 8/10 code patterns. The measured BER at T_o, T_d (max), T_d (min) must be ≤ 10^-12 and the values of both (T_d (max) - T_o) and (T_o - T_d (min)) must be greater than or equal to half the appropriate eye-window specification. All measurements are made with respect to a linear phase, low-pass filter with a 3 dB cutoff frequency of 800 MHz for single mode fiber optic channel links and 300 MHz for multimode fiber optic channel links. It is important that the BERT retiming data latch be significantly faster than the timing resolution of interest.

A common practice used to save time is to measure the eye-window at higher probabilities (for example, 10 ^-6) and then extrapolate to the eye-window at a 10 ^-12 probability.