De-embedding procedure for multi-port networks using thru-reflect-line and thru-reflect-match calibration

The context of the present thesis is the measurement of the scattering parameters of a device-under-test (DUT) employing vector network analyzers (VNA). The de-embedding of the DUT S-parameters from the total measurements that include the test-fixture contribution is investigated in detail. The fixtures are required to physically connect the DUT to the instrument and make the data acquisition possible, but their presence is making the measured S-parameters look completely different from the actual DUT ones. The scope of the present work is to provide a strategy to de-embed the DUT from the total, or fixture-DUT-fixture, measurements. In literature, the techniques to pursue this goal are many and are considering most of the time the two-port case. In this thesis, a simple and effective de-embedding strategy for multi-port measurements is developed. The most general topic of VNA calibration in S-parameters measurements is investigated first from a mathematical perspective. Then, the most classical state-of-the-art de-embedding strategies are reviewed. These strategies are considering two-port networks, such as filters or amplifiers. An environment able to perform the de-embedding in this classical way is implemented with a MATLAB ® script, that takes in input the measurements on the calibration structures and provides the DUT S-parameters. These methods are validated in a SPICE environment and by measurements on real devices. The advantages, disadvantages as well as limitations concerning other de-embedding strategies are studied. A criterion for the definition of the calibration structures is developed, considering the working bandwidth and the available reference impedances. The core of this work is on the extension of the state-of-the-art de-embedding strategies for a more general case. The proposed solution can be applied to a multi-port DUT with an even or odd number of ports. The DUT can be embedded within test-fixtures for which no hypothesis is made on their symmetry or type. The overall strategy can be used in a bandwidth that extends from the ultra-low frequency (ULF) to the GHz range. After the validation of this de-embedding technique, another one that minimizes the number of built calibration structures and following associated measurements is proposed.