Multirate data sampling for dynamic power management and the subsequent reconstruction of data

Methods are presented for multirate data sampling to enable dynamic power management in a low power microsystem and the subsequent reconstruction of data from non-equispaced samples. The sampling system is based on a derivation of the Adams-Bashforth and Adams-Moulton predictor - corrector methods for initial value ordinary differential equations, modified to operate with ordinates rather than derivative values. It is necessary to store 2n + 1 past values of y, where y = g(t) and n is the order of the Adams methods, the sampling method being demonstrated using third order methods. The data reconstruction algorithm is based on a cubic Lagrange interpolation polynomial.