A Methodology to Correct in-Fixture Measurement of Impedance by a Machine Learning Model

It is usual to measure the characteristic impedance of the transmission line in the factory by the time-domain reflectometer (TDR) with the test fixture, but the insertion loss of the test fixture would raise the measurement error, especially when the transmission line designed with unconventional impedance that is deviated from the system impedance of the measurement instrument. The traditional calibration that requires the standard kits is not generally implemented in production line due to the complicated operation and high cost. Therefore, a method is proposed to correct the measurement by a generalized transformation that is a numeric model trained by the machine learning from the measurements corresponding to the well-chosen features of the test fixture. Based on the precise circuit model of test fixture presented in this paper, a large amount of data used for training can be produced by circuit simulation instead of real measurements. An experimental instance is given to demonstrate that the measurement error could be suppressed from around 20% to below 3%.