Verifying Functional Specifications by Regression Techniques on Lissajous Test Signatures

In this paper, a low-cost method to verify functional specifications of analog VLSI circuits is proposed. The method is based on the analysis of Lissajous signatures combined with regression techniques. In order to obtain Lissajous signatures, the observation space is partitioned into zones using hyperplanes, and a set of integer values used as the digital signature of the circuit is generated by Lissajous curve zone crossings. A predictor function obtained by nonlinear regression techniques predicts the functional specification parameters of the circuit under consideration. The viability of this methodology is analyzed and applied to verify the center frequency f(0) of a bandpass biquad filter. Experimental measurements show an accurate prediction of the center frequency of the designed filter.