Increasing Implementability of beta-driven Threshold Checkers

Implementation of checkers on the basis of threshold functions is problematic due to low stability of component threshold schemes. For increasing the implementability, the paper investigates two methods of improving stability of threshold based checkers. One of the methods is majorization of the threshold schemes. The second method is so-called calibration of the mentioned schemes. Both of the methods are described and evaluated from the point of expected implementability. The basic difficulty of the checker’s synthesis is the fact that minterms corresponding to the checker’s codewords are unordered. Universal checkers do not check whether a word belongs to the complete set of codewords, but may check whether a word has some specific property, which distinguishes a codeword from a non-codeword. Usually, a number of “1” (or “0”) positions in the word is checked to this end. In the Berger code, this number is set in control positions of the code; in the m-out-of-n code, this number is defined by the code itself. A considerable portion of a universal checker is occupied by a device computing the number of “1” positions in a binary word. It has been shown by Lala [1] that digital adders, if used for the above purpose, result in a considerable overhead. Threshold elements using analog circuits for computing the “1” positions may effectively reduce the overhead [2]. However, there are some considerable difficulties on the way of using such implementations. As long as -driven threshold element [3] is an analog system in its main part, correctness of its functioning is apparently associated with the precision of manufacturing and stability of the parameters’ maintenance during the operation process. This assertion is purified by the definition of implementability, that determines the limitations on the class of threshold functions; therefore correct (faultless) work can be guaranteed with the given technological limitations and operating conditions. It was shown in paper [3], that the implementability can be