Table of Contents
Advertisement
Note
I mproving the
1
Accuracy of Mean
Value
Measurements
D-4 Accurate Timing Measurements
The time interval being measured must not be synchronous to the
sampling clock of the timing analyzer. Typically, this is not a
problem in the timing analyzer unless the sample rate is extremely
low. The timing analyzer halts its interval sample clock between
each measurement; therefore, the probability is low that the time
interval being measured is synchronous with the timing interval
sample clock.
The accuracy of the displayed mean value of a single interval
depends on the number of executions in the series used to
determine the mean value.
Assume the timing analyzer is measuring a stable, repetitive time
interval approximately 100 us long. Using a 20 MHz sample rate
(50 ns sample period), you capture 51.2 us of timing data,
calculated as follows:
51.2 us = 1024 samples X 50 ns sample period.
A single measurement will have an accuracy of plus or minus the
sum of the sample period plus the skew specification, or:
+ /-(50 ns + 4 ns)
= + /- 54 ns
By making one hundred measurements in a repetitive series, the
accuracy of the mean value displayed will be improved by a factor
of the square root of the number of traces included in the series, as
shown in the following:
+ /-((50 ns/sqrt(100)) + 4 ns)
= + /- (5 ns + 4 ns)
= + /- 9 ns
Advertisement
Table of Contents
