Relation between bit and symbol error probabilities for frequency shift keying with limiter-discriminator detection

In M-ary, Gray coded frequency shift keying with limiter discriminator detection the bit error probability is usually approximated by the symbol error probability divided by the number of bits in a symbol. This approximation is known to be excellent for phase shift keying with large energy-to-noise ratios and Gaussian channel. In other cases this may be questionable. The authors compute the exact values of bit error probability as a function of energy-to-noise-ratio per bit for Gaussian, Rayleigh and Rician channels. They also compute the relative error involved in this approximation