We consider a residue number system using n pairwise relatively prime moduli m1,⋯,mnto represent any integer X in the range M/ 2≤X>M/2, when M = ∏mi. The moduli miare chosen to be of the 2-1 type, in order that the residue arithmetic can be implemented by means of binary registers and binary logic. Further, for each residue number X, a magnitude index Pxis maintained for all arithmetic operations. We investigate the properties of such a system and derive the addition, subtraction, multiplication, sign determination, and overflow detection algorithms. The proposed organization is found to improve the operation times for sign detection and overflow detection operations, while rendering multiplication to be a difficult operation.

Except for some elementary definitions and fundamentals, the theory of AN code is by and large the theory of binary (radix = 2) arithmetic codes. It is often believed (erroneously) that this theory can be readily generalized to any nonbinary radix. The very fundamental theorems of Brown and Peterson on single-error-correcting codes have been derived for the binary case only. Whereas a generalized version of Brown's theorem can be stated and proved relatively easily (as shown here), the one for Peterson's theorem is not forthcoming. However, we have succeeded in deriving a theorem for the ternary case (radix = 3) somewhat along the lines of the Peterson's theorem as follows. LetM_3 (A, d)denote the smallest positive integer such that the arithmetic weight ofA M_3 (A, d)in ternary representation is less thand. Also leyA = 2pfor some odd primep. Then 3 is a primitive element ofGF(p)if and only if \begin{equation} M_3 (A, 3)=(3^{(p-1)/2} + 1)/A. \end{equation}