HP 8712ES User Manual page 251

Calibrating for Increased Measurement Accuracy
Measurement Calibration Overview
The analyzer has several methods of measuring and compensating for
these test system errors. Each method removes one or more of the
systematic errors using an equation called an error model. Measurement
of high quality standards (short, open, load, and through, for example)
allows the network analyzer to solve for the error terms in the error
model. The accuracy of the calibrated measurements is dependent on the
quality of the standards used for calibrating. Since Hewlett-Packard
calibration standards are very precise, great accuracy is achieved.
When you perform a calibration, the analyzer calculates the difference
between measurement data of known calibration standards and ideal
measurement data to create error correction arrays. The network
analyzer uses these error-correction arrays to correct subsequent
measurement data.
To see all of the calibration choices available for measurements, refer to
Table 6-1, "Calibration Choices," on page 10 and Figure 6-4 on page 6-11.
The full two-port error model includes all six of the systematic error
terms for the forward direction and the same six (with different data) in
the reverse direction, for a total of twelve error terms. This full two-port
vector-error correction accounts for all major sources of systematic error
and gives the most accurate measurements of two-port devices. By
default however, the analyzer's Isolation on OFF softkey is turned
off, resulting in a two-port error model that includes only ten error terms.
This is done because the isolation calibration, which creates the eleventh
and twelfth error terms, adds noise to the error model, and should only
be used when measuring high-isolation devices and high-dynamic range
devices. The selection of the Isolation on OFF softkey affects both the
default two-port and the user two-port calibrations. The data for the
isolation calibration is always collected (though not always used) in a
two-port calibration when the analyzer is measuring the response of the
load-standard with each port. It should be noted that because it corrects
for ten or twelve error terms, the two-port error correction is the only
calibration that can be used to measure all four S-parameters.
ES User's Guide 6-5