Linear predictive coding without Yule-Walker approximation for transient signal analysis: Application to switching noise

Abstract

We are investigating countermeasure techniques against electric noise on telecommunication and related equipment. There are many types of electric noise, and we focus on the characterization of transient noise such as the switching noise of switching power supplies. We are developing a method of transient signal analysis based on linear predictive coding (LPC) for the purpose. Unfortunately, as standard LPC contains Yule-Walker (YW) approximation, which replaces a given time series with a periodic time series for simplification, it is not suitable for transient signal analysis. Therefore, our method of transient signal analysis does not contain YW approximation. Instead, we use the local linearization technique, which was developed in the field of quantum mechanics. Our method makes it possible to obtain instantaneous frequencies and instantaneous decay rates simultaneously with high precision from a small number of samples. We discuss our LPC method avoid using YW approximation and using the local linearization technique and its application to repetitive ringing signals, which mimics the switching noise of a switching power supply, to show the efficiency of our method.