A.A. Makhmudov - Head of Scientific and Technical Laboratory, FRPC «RNIIRS»
A.E. Fedorovtsev - Programming Engineer 2 сat., FRPC «RNIIRS»

In existing methods of research of structure of an error-correcting code features of statistical characteristics of coded information sequence traditionally aren't considered. Are limited, as a rule, to the requirement of sufficient «variety» of code words in basic data that it was possible to create system of the independent linear equations which solution will allow to find test ratios of a code. However chances when it is possible to calculate some (not all) test ratios of a correcting code which can be nevertheless used for creation of the noiseproof coder and the decoder. Such situation arises, for example, in case of cascade coding with linear binary block codes with an internal LDPC code when test symbols of an external code bring in a signal a structural regularity and number of linearly independent undistorted code words on any data set there is less number of information symbols of an internal code. The parity-check matrix of a LDPC code restored on such basic data doesn't possess property of a sparseness and for this reason is of little use for creation of the effective iterative (probabilistic) decoder. Excluding in the restored matrix all checks containing linearly dependent information symbols, it is possible to lead a parity-check matrix of a studied LDPC code to the sparse look, possibility of use of the probabilistic decoder thereby is provided. Thus the question of changes of characteristics of the LDPC decoder owing to violation of structure of a code remains open.