Separation of Cyclostationary Signals and Interference in Transmission Lines of Printed Circuit Boards Based on Independent Component Analysis

Abstract

Introduction. When developing and testing high-speed communication links of digital electronic devices, the pulse shape of electrical signals and interference in transmission lines is commonly controlled using eye diagrams. Partial pulse responses are estimated by modern multi-channel testing and measuring equipment using specific testing procedures. Aim. To develop an approach to blind identification of a wireline communication signal measured by a digital oscilloscope in microstrip transmission lines.Materials and methods. The study was carried out using a communication signal model composed of a superposition of the information signal, intersymbol interference, and crosstalk interference from an adjacent transmission line, with the symbols of these signal sources being assumed statistically independent. The implemented blind identification method makes it possible to separate three components of the signal and evaluate their partial pulse responses. The proposed procedure for signal separation includes preliminary processing of the data measured by principal component analysis followed by an analysis of independent components based on fourth-order statistical characteristics.Results. The performance of the proposed signal separation method is demonstrated using the results of parallel and independent data processing measured in microstrip transmission lines in an experimental setup with two closely spaced printed circuit boards. A comparison of blind signal separation methods is carried out based on second-order statistics, fourth-order cumulants, and independent component analysis. The proposed method of independent component analysis demonstrated the highest efficiency in terms of minimum bit error probability for blind signal separation with arbitrary partial pulse responses, which was confirmed by simulation results.Conclusion. The developed approach to blind identification of wireline signal parameters measured in high-density integrated digital electronic devices extends the range of systems for circuit design and testing of multi-channel high-speed communication lines.

Keywords

• crosstalk
• blind identification method
• principal component analysis
• cyclostationary process
• fourth order cumulant