Performance bound analysis of analog circuits considering process variations

In this paper, we propose a new performance bound analysis of analog circuits considering process variations. We model the variations of component values as intervals measured from tested chip and manufacture processes. The new method applies a graph-based symbolic analysis and affine interval arithmetic to derive the variational transfer functions of analog circuits (linearized) with variational coefficients in forms of intervals. Then the frequency response bounds (maximum and minimum) are obtained by performing analysis of a finite number of transfer functions given by the Kharitonov's polynomial functions. We show that symbolic de-cancellation is critical for the affine interval analysis. The response bound given by the Kharitonov's functions are conservative given the correlations among coefficient intervals in transfer functions. Experimental results demonstrate the effectiveness of the proposed compared to the Monte Carlo method.