Practical Aspects of Reduced Complexity LDPC Decoding using Forced Convergence

Sammanfattning: This thesis is an extension to the previous work done at Department of EIT in investigating the complexity reduction of LDPC decoding using the stochastic approximation method to improve the energy efficiency of energy-limited applica- tions, such as in mobile phones. The research carried out in this thesis investigates the complexity and conver- gence results of adaptive performance control algorithm (PCAA) that changes the forced convergence threshold in OMS LDPC FC decoding algorithm. The simula- tions are performed on IEEE802.11n OFDM communication system for Additive White Gaussian Noise (AWGN), indoor Rayleigh time varying frequency flat and Rayleigh frequency selective fading channels. Performance results are obtained by running various combinations of LDPC decoding parameters and results are documented for parameter profiles that meet the specified BLER target criteria. In order to visualize the results, simulations are performed by implementing OFDM, PCAA and relevant channel models in MATLAB c and LDPC forced convergence decoder in C. By doing that it was possible to shorten the simulation time as LDPC decoder takes longer in MATLAB c compared to C. Performance metrics such BLER against E b /N 0 , complexity and convergence for various mod- ulation and coding (MCSs) schemes are produced via this setup. The aim of the thesis is accomplished and results obtained indicate signifi- cant convergence time reduction compared to previous methodology [1]. It is also concluded that optimum complexity and convergence trade-off is obtained when Δθ d is set to Block Error Rate target (BLER t ) in AWGN channel set-up. By applying this optimum configuration to both time varying frequency flat fading and frequency selective fading channel encouraging results are observed. PCAA tracks the channel changes in both fading cases while meeting specified BLER t and with significant complexity savings relative to non-forced convergence LDPC decoding.