EC-22.2 Scattering Parameters, and the Network Analyser

The standard treatment of scattering parameters: why waves replace voltage and current, the two-port matrix, reference planes and the network analyser, September 2026

What this is and why it exists

Once a component is a noticeable fraction of a wavelength across, voltage and current stop being measurable quantities.

Measuring a voltage needs a well-defined pair of terminals and a probe that disturbs nothing. Neither survives at microwave frequencies. So high-frequency components are described by what they do to travelling waves instead.

That description is the scattering matrix, and it is the language of the whole field. Every data sheet, every simulator and every measurement speaks it. Reading one fluently is the most portable skill in this module.

The vocabulary

  • Port — a place where a wave can enter or leave a component.
  • Incident wave — the wave arriving at a port.
  • Scattering parameter — the ratio of a wave leaving one port to a wave entering another.
  • Insertion loss — the reduction in a signal passing through a component.
  • Isolation — how much a signal is attenuated in the unwanted direction.
  • Reference plane — the exact point along the connection at which the data applies.
  • Reciprocity — the property that transmission is the same in both directions.
  • Directional coupler — a component that separates forward and reverse waves.

The mental model

Think of a component as a black box with waves going in and out. For a two-port component there are four ratios worth knowing. Two of them describe what is reflected back at each port. Two describe what is transmitted from one port to the other.

Those four have everyday names. The reflections are what a matching discussion calls return loss. The forward transmission is what a catalogue calls gain or insertion loss. The reverse transmission is what an amplifier data sheet calls isolation. Attaching the everyday names to the matrix entries is what makes data sheets readable.

Structural properties show up as relationships between the entries. A passive component made of ordinary materials transmits equally in both directions, so two of the entries are equal. A lossless component cannot emit more than it received, which constrains the magnitudes. A glance at measured data can therefore say whether a device could be passive, which is a useful sanity check.

The reference plane is where beginners lose time. Measured data applies at a specific point along the connecting lines. Using it at any other point changes the phase, because the wave has travelled further. A simulation that ignores this produces a result that no measurement will reproduce.

A vector network analyser does one thing well. It generates a wave, separates the incident part from the reflected part using a directional coupler, and compares the two in both amplitude and phase. Everything the instrument reports comes from those comparisons, repeated at each frequency in a sweep.

Finally, components combine. Chaining two of them is done by converting each matrix into a form that multiplies, multiplying, and converting back. Simulators do exactly this. Doing it once by hand for two components makes clear why cascading works, and why the reference planes have to agree.

What you should now be able to explain or do

  • Say why voltage and current become unmeasurable at microwave frequencies.
  • Name each entry of a two-port matrix and give its everyday name.
  • Read reciprocity and losslessness off a set of measured parameters.
  • Explain what a reference plane is and what happens when it is ignored.
  • Describe what a network analyser measures and how it separates the waves.
  • Combine two components' matrices to predict the response of the pair.

Check yourself

Because a voltage measurement needs defined terminals and a non-disturbing probe. At microwave frequencies neither exists, while incident and reflected waves remain well defined.

Isolation. It says how much of a signal entering the output finds its way back to the input. That decides stability and how well stages are separated.

The phase is wrong by however far the wave travelled. A well-matched component can appear badly matched, and a simulation will not reproduce any measurement.

It separates the wave travelling towards the device from the wave coming back. Without that separation, the instrument could not distinguish incident from reflected.

Go deeper

Back to Scattering Parameters, and the Network Analyser: work through the checklist