s-params to capacitance

To derive the input capacitance (\(C_{in}\)) of a device or circuit from its scattering parameters (S-parameters), you must convert the S-parameters into Admittance parameters (Y-parameters) or Impedance parameters (Z-parameters). Because capacitance directly relates to imaginary admittance, the Y-parameter method is the most direct and universally applied approach.

YouTube tutorial math derivation:

Cin=IM(Y11)/2.pi.f

Active Devices (e.g., MOSFETs): If you are extracting the gate capacitance of a transistor, remember that S-parameters are small-signal parameters taken at a highly specific DC bias point. The resulting capacitance will shift dynamically depending on your applied gate-source (\(V_{GS}\)) and drain-source (\(V_{DS}\))

AI generated, and seems straightforward. I didn't watch the YT, but the formula AI gives in terms of S-params , which I can't display in ASCII, is correct and links to that video. It's 9-1/2 mins long, which can be an eternity when trying to decipher broken English.

Simon tells me that scikit-rf does that—he’s using a network analyzer to make Zobel compensation networks for our thermal Faraday shields.

It doesn’t have Rexx bindings, so I’m going to have to bite the bullet and learn Python.

Cheers

Phil Hobbs

Belay that. It produces big opaque behavioral things.

Still useful, but not at all intuitive.

Cheers

Phil Hobbs

It's trivial to get the node capacitances from S parameters, but being small signal measurements they are only valid at the operating point at which the parameters are measured. Trying to fit an assumed discrete model to a two port using some LMS type of algorithm and s params measured over a wide range of operating points only works reasonable well with some types of structures. The other problem is the quality of the original s params, there are a lot of metrics used to quantify that (see the touchstone analyser in Simbeor for an example of that), and in practice parameters given on data sheets are often unreliable.

Most tools will spit out a SPICE model from S params (from example measurements on a PCB trace) but they usually include W elements. If you're working on the cheap I think only ngspice supports them.

I mean the commercial tools that people actually use, ADS etc.

My mistake, ngspice has an RLGC model, but without frequency dependent loss, and has it's own lossy model.