Agilent Technologies HP 8719D Service Manual page 447

In any measurement, certain measurement errors associated with the system
add uncertainty to the measured results. This uncertainty defines how
accurately a device under test (DUT) can be measured.
Network analysis measurement errors can be separated into two types: raw and
residual. The raw error terms are the errors associated with the uncorrected
system that are called systematic (repeatable), random (non-repeatable),
and drift errors. The residual error terms are the errors that remain after a
measurement calibration.
The error correction procedure, also called measurement calibration, measures a
set of calibration devices with known characteristics It uses the measurement
results to effectively remove systematic errors, using the vector math
capabilities of the analyzer. The residual systematic errors remain after error
correction, primarily due to the limitations of how accurately the electrical
characteristics of the calibration devices can be dehned and determined. Also,
the random (non-repeatable) and drift errors, cannot be corrected because
they cannot be quantified and measured during the measurement calibration
and device measurement. However, the effects of random errors can be
reduced through averaging. Random errors, that occur during a measurement
calibration, are part of the error correction and become systematic errors when
the calibration is turned on. For this reason, it is best to use a large number
of averages during measurement calibration to reduce to the effect of the
random errors. The averaging may then be reduced for device measurement.
The residual systematic errors along with the random and drift errors continue
to affect measurements after erro.correction, adding an uncertainty to the
measurement results, Therefore, measurement uncertainty is defined as the
combination of the residual systematic (repeatable), random (non-repeatable),
and drift errors in the measurement system after error correction.
The following measurement uncertainty equations and system error models
(flowgraphs) show the relationship of the systematic, random, and drift errors.
These are useful for predicting overall measurement performance.
Determining System Measurement Uncertainties A-l