De-embedding research in cold test of output window of Gyro-TWT (gyro traveling wave tube)

Abstract

In many S-parameter measurements, one would desire to make the measurement with some other setup than what one has. There may be a test fixture required between the normal coaxial calibration planes and the DUT (Device Under Test); it may be useful to see the DUT performance with a certain matching network in place, and may be desired to see what the subsystem performance would be when the given DUT is inserted, etc. One way of handling these chores within the instrument itself is through embedding and de-embedding. The de-embedding technique, it is the process of mathematically subtracting networks from the measured result. This technique has been around for many years and has been subject to a number of refinements to improve accuracy and applicability <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[1]</sup> . The purpose of this article is to discuss the implementation in the broad band TE <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">10</sub> -TE <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">01</sub> cruciform mode converter <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[2]</sup> , and to provide a number of examples and procedures to show how they can be used to get the required measurement results. This literature aims to provide two kinds of de-embedding techniques to conclude the best suitable method of the converter. Since the S-parameter data for the DUT is available from the measurement and the S-parameter data for the network to be de-embedded is available (either from an s2p file or from a circuit model), if multiplying one transfer matrix by another is equivalent to connecting in the new network, then multiplying by the inverse of that T matrix is equivalent to removing it <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[3]</sup> . The first method is according to the principle above to do the de-embedding work. Based on the principle, this passage has proved that whether the principle is suitable to the microwave measurements through two kinds of methods. It turns out to be that the methods mentioned above can effectively reduce error and is more accurate to the actual value of the DUT when the data which needs to be de-embedded is accurate, but when the data is resonant, a big error may occur. Except from the above two methods (from one principle), we can also get the actual value by using the other principle, that is to modify the error model parameters to get the actual value. Through repeated measurement, we finally get the required results to estimate the properties of the DUT. Also, through comparative analysis, we find the different results of the S parameter before and after de-embedding, the de-embedding results are obviously better than the ones before de-embedding. Therefore, the de-embedding technique can help us get much more accurate information on the DUT. All in all, the theory and the experiment both illustrate the feasibility and effectiveness of the de-embedding technique.