Click here to go to our main page on S-parameters
Click here to go to our page on ABCD parameters
New for April 2023. Often in Microwave engineering, you obtain measured S-parameter data on a device that is embedded in a network that you cannot take apart, so the measurement reference planes are not right at the device you are interested in. The hardware that you don't want in the measurement could be RF coax cables, or an evaluation board, or other test fixture that the part is mounted to. The process of mathematically removing the network is called de-embedding. The stuff that you want to remove from the ports are sometimes referred to as "error boxes", and we will use that terminology here. S-parameters of the pieces of the fixture are measured or simulated somehow (which can be a complicated subject, for another day).
Before we dive into this, you can create custom calibration standards that will move the reference planes of the measurement to where you want them, and almost all vector network analyzers support this. One example is creating "TRL" standards, measuring a through, a reflect, and one or more lines (which are longer than the through). That is not what we are talking about here.
Computationally, removing the effects of the fixture is done using ABCD matrices, which are way easier to manipulate than S-parameters. But if you don't enjoy that math or don't have time for it, you can use Microwave Office or Keysight's Advanced Design System (ADS) to remove the error boxes using "negation". In ADS, the element that does this is called the "de-embed" element.
Below, Anurag Bhargava explains how ADS is used to remove the "error boxes". The test fixture's S-parameters are represented as a two-port network (the same file is used twice, once on the DUT input and once on the output, such symmetry is not often the case and should not be expected). The "de-embed" blocks reference test fixture S-parameters, and "negates" them to remove their effects. In Microwave Office, that circuit element is called a "negation element". Anurag sets up a six-port network, where the DUT's "raw" S-parameters (which are known in this example but would be unknown in the real world) are connected from port 1 to 2. Embedded data is connected between ports 3 and 4, and the de-embedding example is connected between ports 5 and 6. In the port 5-to-6 circuit, embedded data in the center, and fixture networks (the error boxes) are placed into ADS's de-embed elements surrounding it. The conclusion of the video is that the raw data (port 1 to 2) and de-embedded data (ports 5 and 6) lie on top of each other, which is what we like to see. This example proves out how negation works in an ideal situation. Whether you get accurate de-embedded data in real life depends on how accurately you know the S-parameters of the error boxes.
Learn ADS in Five Minutes, Tutorial 63, by Anurag Bhargava
Anurag's YouTube channel is called "learn ADS in five minutes", and contains scores of short videos on relevant topics. If you are just learning ADS, or want to expand your ADS skills, you should check them out!
This video does not explore how the error boxes were created; we'll help you with an example on that later. Also, it seems like there could be a possible de-embed error on port 6. If the same error box is used to de-embed both port 1 and port 2 of the embedded DUT, it should be flipped around on port 2 but it is not. Like in this really crummy sketch:
1 Error Box 2-------------- 1 Embedded DUT 2--------------2 Error Box 1
Scikit-RF
These comments came from "hash" on our message board. Thanks!
Concerning the April (2023) newsletter video deembedding, we have also documented this topic on the scikit-rf documentation (with the flip thing!):
https://scikit-rf.readthedocs.io/en/latest/examples/networktheory/Balanced%20Network%20De-embedding.html
as well as more advanced methods:
https://scikit-rf.readthedocs.io/en/latest/tutorials/Deembedding.html
Comments always welcome to improve the documentation :) !
By the way, new merchandises are available for scikit-rf supporters ;)
https://scikit-rf.org/merch.html
And yes, we bought one of those cool shirts. Blam!