The method of the estimation of the probability of errors in the channel, based on recurrence property of convolution codes. The method can be used in computation of the likelihood ratio (LLR) of the code symbols at the demodulator output in communication systems that use both directly convolution code and turbo codes and trellis codes modulations.