Limited-Information Signal De-Embedding for a Grain Bin Electromagnetic Imaging System

Abstract

Measurements taken with a Vector Network Analyzer (VNA) are often corrupted by the presence of 2-port networks such as cables or amplifiers between the VNA and device-under-test (DUT). These 2-port devices are an essential part of the signal path, but distort the desired measurements. These distortions are commonly removed by de-embedding the 2-port networks between the VNA and DUT. Herein we consider the case of a grain bin electromagnetic imaging system where the DUT (the grain bin) is connected to the VNA by a set of amplifiers, a switching matrix, and locally installed cables. Further, for cost savings, the VNA only measures the S21 and S11 scattering parameters. Applying traditional de-embedding techniques to such a system would require a full S-matrix measurement, as well as knowing full 2-port parameters of all system components between the VNA and DUT. As these grain imaging systems have cables cut to length on site and are located in remote locations, it is not possible to obtain full 2-port measurements of the networks between the VNA and the DUT. Herein, we present an algorithm for using de-embedding techniques on such a system that allows us to approximate the various network parameters needed, and show that these approximations allow us to de-embed the grain bin measurements. Experimental results from one bin show that de-embedding reduces the average S21 errors in magnitude to less than 0.5 dB, with average phase errors of less than 0.15 radians. The average magnitude errors are similar to the measurement error between two different VNAs. A second experiment on a grain bin where the authors did not have physical access, shows that de-embedding the signal improves the computational model of the signals (a key indicator for future imaging success).