Electronics engineering student at TU Dresden.

Interested in electronics, ham radio, UNIX like OSs and software development (Rust, Julia).

Fulltime GNU+Linux user since 2012.

GitHub: knightshrub

Creative Commons License All text is licensed under CC-BY-SA 4.0 International License

GNU GPLv3 Code is licensed under GPLv3

[ 2018-12-18 ]

VNA OSL calibration

Categories: Electronics

A simple application of Mason's gain formula can be used to derive the VNA OSL calibration routine.

When making measurements with a VNA, it is important to calibrate the test equipment before the measurements in order to remove sources of error arising from non-ideal directional couplers, test cables and connectors. This is done using a so called cal-kit which contains devices whose S-parameters are well known (i.e. they have been accurately characterized for example by using a more accurate VNA than the one that is being calibrated).

In the case of a one-port measurement the calibration is often done using three calibration standards: open, short and load. This calibration routine is thus referred to as OSL calibration.

Error model

The error model for a one-port VNA can be represented by the following signal flow graph (SFG).

b0 ------> a1 ------> b2 ------> a3
            |          ^          |
        S11 |          | S22      | SDUT
            |          |          |
            v          |          v
a0 <------ b1 <------ a2 <------ b3

where S11, S21, S12 and S22 are the S-parameters of the error network that sits between the ideal VNA and the device under test (DUT) SDUT. The error network thus represents all sources of measurement error within the VNA itself and in the connection to the DUT.

In order to find the reflection coefficient measured by the VNA

from the SFG, Mason's gain formula (MGF) can be used.

Masons's gain formula

The following section has been directly copied from Wikipedia:


Steps to find MGF solution

  1. Make a list of all forward paths and label them Gk
  2. Make a list of all loops in the SFG and label these Li. Make a list of all pairwise non-touching loops and calculate their gains LiLj etc. Continue until there are no more loops left.
  3. Compute the determinant and the cofactors.
  4. Subsitute into the equation for G.

Solution using MGF

According to the steps outlined above and noting that the input is b0 and the output is a0, the value of SM can be found as follows:

With regards to the VNA error model, the terms are usually referred to as follows

This equation can be rearranged such that it is linear in three unknowns:

Using the three calibration standards a system of three linear equations can be formulated.


Once all three measurements have been made the system of equations can be solved for the parameters S11, S22 and S'.

S21S12 can be calculated as follows:

Now all unknown parameters of the error network have been determined.

Correcting DUT measurements

The reflection coefficient of an arbitrary DUT SDUT using the measurement SM is now given by