Wavelet based nonparametric regression approach for de-noising and modeling of transient switching noise measurements

In high performance systems, with millions of gates switching at each clock cycle and multiple modules operating at different frequencies, the level of simultaneous switching noise (SSN) is difficult to estimate deterministically; necessitating a measurement based approach. The principal difficulty in pursuing a measurement based approach for measuring the switching noise voltage is the noisy character of the time domain measurement and the partial (magnitude only) information in the frequency domain measurement. Accurate signal measurement is essential for correct extraction of the resonances in the signal, on which in turn depends the accuracy of the modeling results. This paper proposes a wavelet based nonparametric regression approach for processing the transient data such that it can be productively used in accurately extracting the resonances in the signal and using them for modeling purposes.