Agilent Technologies J-BERT N4903, User Manual

Page: 283 / 434

Considering the BER bathtub curve, the sampling point is in the middle
of the descending line.
If random data is used, the average bit error rate at the left-hand side
of the jitter region is 0.5.
The bit error rate at the right-hand side of the jitter region is 0. We
therefore expect a bit error rate around 0.25.
N O T E
The Spectral Jitter measurement should only be used in conjunction with data that
has an equal distribution of ones and zeros over time. Otherwise, the results are
hard to predict and may be not reproducible.
Signal Processing
If the error signal is obtained as explained above, an analysis in the
frequency domain reveals the absense or presence of deterministic
jitter. Dominant frequency components become visible and their
contribution to the total jitter can be measured.
The modified error signal is subject to a fast Fourier transformation
(FFT).
FFT requires that the data record to be processed has a length that is a
power of two (2
n
, such as 2
17
, 2
18
, 2
19
, 2
20
, and so on).
From that data record, the FFT calculates pairs of frequency/power
values. The number of pairs is half the number of samples. If you have
chosen a record length of 2
17
(which means 131,072 bits or 128 Kbit),
the result contains 65,536 pairs. The maximum frequency is half the
data rate used for the test.
The results are displayed in the Spectral Jitter measurement graphical
and numerical result window.
Advanced Analysis
5
Agilent J-BERT N4903 High-Performance Serial BERT 283