The purpose of the paper is to develop a coding method which gives a reasonably good quality of speech by a relatively simple procedure. Easy but effective algorithem for the separation of pulsive and random components of residual signal is described, and the coding method by which difference signals between the original speech wave and the synthesized one excited only by pulsive components are CADM coded after a low pass filtering. The best result we obtained is more than 17 dB in S/N and less than 16 k/bps in bit-rate.