Adaptive Demodulation in Impulse Noise Channels.

Electromagnetic interference causes considerable performance degradation on wireless communication systems. The performance degradation is largely affected by the characteristics of the interference. Impulse interference commonly occurs in industrial and automotive environments, where many closely located systems and electronic equipment coexist. In this work, we propose two novel algorithms for adaptive demodulation in impulse noise. The proposed methods compute appropriate log-likelihood ratios based on interference classification and estimation of Middleton’s Class A noise model as well as the symmetric $\alpha$ -Stable (S $\alpha$ S) model. The analytical complexity and intractability of the S $\alpha$ S probability density function (PDF) makes real-time applications unfeasible. Therefore, the proposed algorithms are developed using two different techniques for approximating the S $\alpha$ S PDF. The approximations are achieved through interpolation via i) approximate closed-form expressions of the PDF for a specific parameter set and ii) via numerically pre-calculated look-up tables of the PDF for arbitrary parameter choices. The proposed adaptive techniques are shown to provide a performance gain in terms of bit error rate (BER) of up to 20 dB energy per bit to noise power ratio (Eb/N0), in comparison to traditional, non-adaptive, methods for varying levels of impulse interference. The performance loss of approximating the PDF in comparison to exact calculations amount to less than 8 dB Eb/N0, which is small in relation to the performance gain achieved in comparison to traditional methods. However, the complexity in terms of simulation execution time of the proposed approximate algorithms is reduced by up to a factor 250, compared to methods based on exact computation. The algorithms are validated using regular low density parity check (LDPC) codes, in addition to quasi-cyclic (QC)-LDPC and turbo codes. The proposed adaptive demodulation techniques can be used to improve decoding and demodulation performance in many real-life situations where non-Gaussian interference commonly occurs. [ABSTRACT FROM AUTHOR]