Spectral techniques for multiple valued logic circuits

Canonical representation of multiple valued logic (MVL) functions in any polarity k, k in (0, 1,. . .,p/sup n/ -1), where p is the radix and n denotes the number of variables in a function, was previously presented. The coefficients in a canonical representation are called the spectral coefficients. It is shown that for some MVL functions realizing them as sum of products may not be economical, especially if very few minterms can be combined. Such functions can be efficiently realized as mod-p sum of products in a polarity which provides fewer coefficients. Realization of MVL functions as mod-p sum of products is done using a set of gates which are functionally complete. Implementation of these gates is shown both in I/sup 2/L and CCD technologies. The computation complexity for estimating all the coefficients in the canonical representation is presented.<>