Soft Decision Metrics for Turbo-coded FH M-FSK Ad Hoc Packet Radio Networks

Abstract—This paper addresses turbo-coded non-coherent FH M-FSK ad hoc networks with a Poisson distribution of interferers where multiple access interference can be modeled as symmetric α-stable (SαS) noise and α is inversely proportional to the path loss exponent. The Bayesian Gaussian metric does not perform well in non-Gaussian (α≠2) noise environments and therefore an optimum metric for Cauchy (α=1) noise and a generalized likelihood ratio (GLR) Gaussian metric requiring less side information (amplitude, dispersion) are presented. The robustness of the metrics is evaluated in different SαS noise environments and for mismatched values of the interference dispersion and channel amplitude in an interference-dominated network with no fading or independent Rayleigh fading. Both the Cauchy and GLR Gaussian metric exhibit significant performance gain over the Bayesian Gaussian metric, while the GLR Gaussian metric does so without the knowledge of the dispersion or amplitude. The Cauchy metric is more sensitive to the knowledge of the amplitude than the dispersion, but generally maintains better performance than the GLR Gaussian metric for a wide range of mismatched values of these parameters. Additionally, in an environment consisting of non-negligible Gaussian thermal noise along with multiple access interference, increasing the thermal noise level degrades the performance of the GLR Gaussian and Cauchy metric while for the observed levels both maintain better performance than the Bayesian Gaussian metric. Keywords-soft decision metrics, symmetric α-stable noise, noncoherent detection, ad hoc networks I