Teledyne LeCroy WaveExpert 100H Operators Manual - Manual (Page 221)

Wave Expert
WE-OM-E Rev A
219
contain over 3x1014 cells if all of the samples were used to accumulate signal data. Clearly, this is
far too much storage to be practical, so the grid is actually a construct consisting of only the vertical
slices around the edges and the digitizer levels within the horizontal slice. The waveform samples
from multiple acquisitions are summed within the zones around the data transitions so that
statistics can be accumulated.
As shown in the figure, a waveform is acquired by the CIS timebase, and the data samples are
examined around each transition within the horizontal slice defined by the fit region. A best-fit line is
drawn through the samples on each edge and the intersection of this line with the center of the fit
region is used to measure the average displacement of each edge from its ideal location in time,
given by the selected number of CIS timebase samples in each unit interval. The average
displacement for each transition is plotted versus each bit, and the peak-to-peak value of this plot
gives the data dependent jitter. The distribution of samples around the center line for each edge is
caused by random and uncorrelated (to the data pattern) jitter. A separate Rj + BUj histogram is
measured for each edge, and all of these histograms are summed to measure the overall RJ + BUj
distribution. Note that these histograms are summed simply by adding the bins, since they are all
zero mean.
The Rj + BUj histogram is converted to a probability density function by normalization to an area of
1 and analysis on a normalized Q-scale. This analysis breaks the measured distribution into two
normal (Gaussian) distributions whose mean values are separated by the bounded, uncorrelated
jitter (BUj). The standard deviation of each Gaussian is set so that the tails fit the distribution on the
left and right sides. The average of the standard deviations, by convention, gives the random jitter
(Rj).
Total jitter is computed by convolving the DDJ, BUj, and Rj distributions and then determining the
cumulative distribution function (CDF) by integrating the convolved distributions. The CDF plots the
probability of an edge transition occurring at a displacement greater than a given value as that
value is varied from the nominal crossing time to some maximum. The plot of the CDF from 0 to 1
bit interval gives the bathtub curve and the opening of this curve at the selected bit error rate is the
complement of the total jitter for that BER.
Initial Setup for Jitter Measurement
Jitter analysis is accessed by selecting Analysis from the menu bar, then Serial Data from the
drop-down menu. Two related measurements are available from the Serial Data Analysis dialog:
Mask Test and Jitter. Mask testing allows you to measure the signal eye pattern against a
compliance mask from the relevant standard. Jitter testing measures the timing variations of the
signal transitions relative to a stable timing reference at the nominal bit rate. Jitter is expressed as
peak-to-peak total jitter at the selected bit error rate, and as random (Gaussian), deterministic, data
dependent, and uncorrelated components.