HP 64700 series, User Manual

Page: 298 / 312

1
Statistical Errors
Caused by
Sampling Process
The timing analyzer calculates statistics on the sampled data in its
memory, not on all of the data generated by the system under test.
The data in memory may misrepresent the actual data. Misleading
data can be captured when you trigger your trace on some
occurrence that causes the timing analyzer to capture samples at
misleading points in the data flow of a system under test.
Use of the "trigger on anything" specification may not overcome all
measurement bias problems. Consider the case where the timing
analyzer is measuring an interval of time between positive edges
occurring on a probe line. Suppose there are two intervals on that
line, and they are occurring alternately (one is 10 us long and the
other is 20 us long). Interval measurements are made by marking
"x" on the first positive edge of the selected label and marking "o"
on the next positive edge after "x". The random beginning of a new
trace will probably occur twice as often during the 20 us interval as
during th 10 us interval. Because of this, the timing analyzer will
appear to be finding twice as many 10 us intervals as 20 us
intervals, but in the system under test there are equal numbers of
10 us and 20 us intervals.
One possible approach to solving the problem of misleading data
in the above example is to find another line with a uniform
squarewave operating at twice the frequency of the combined
intervals. Such a squarewave will have as many positive edges
preceding 10 us intervals as 20 us intervals. By triggering the
interval measurements on positive edges in that squarewave, and
marking "x" and "o" on the first interval after each trigger, the
timing analyzer will measure as many 20 us intervals as 10 us
intervals.
D-6 Accurate Timing Measurements