Forward error correction for an atmospheric noise channel

Models for predicting the performance of error correcting codes on an atmospheric noise channel are presented and verified. Two Markov chains are used to model the memory of the atmospheric noise channel. The transition probabilities for these chains are derived from atmospheric noise error processes which were recorded at 306 kHz. The models are used to estimate the probability of codeword error, and these estimates are compared to codeword error rates which are obtained directly from the recorded error processes. The comparisons are made for the Golay code with a variety of bit interleaving depths, and a Reed-Solomon code with a variety of symbol interleaving depths. Both Markov channel models predict the actual performance of the codes with much greater accuracy than the binary symmetric channel. Results for binary convolutional codes are presented