Abstract
Summary form only given. The authors have generalized a previous result in which noise scaling was used to model certain classes of noisy channels. It was shown that orthogonal codes achieve the minimum resource per information bit (MRPIB) for a BSC if the noise is not severe, where the resource is the scaling parameter. They have now shown that orthogonal codes achieve the R 0 -version of MRPIB for two classes of binary input very noisy channels (VNCs) identified by E. E. Majani. They have proved that triply orthogonal codes, in which the codewords are rows of orthogonal arrays of strength three, achieve MRPIB for Class-I binary input VNCs, and orthogonal codes will do if the channels are also symmetric A counterexample has shown that the above does not hold for Class-II VNCs. The results have been generalized to Class-I symmetric VNCs with more than two inputs, and a construction of a sequence of generalized triply orthogonal codes has been found.