Research On Detection And Decoding Technology In Non-Ideal Non-Gaussian Channels

The researches on modulation, demodulation, channel coding and decoding technology in tradition-al communication systems are mainly based on the the assumption that the channel noise is additive white Gaussian distributed. However, the assumption doesn’t hold in a large number of practical communication systems, which can be shown mainly in three aspects, non-Gaussian, non-additive and non-linear. This paper studies the detection and decoding algorithms under non-ideal Gaussian channel.After the introduction to the research background and non-ideal Gaussian channel integrated concepts, the second chapter of this paper introduces the decoding algorithms of two near Shannon code, Turbo codes and LDPC codes as the theoretical basis of the paper. In this part, we simulate the MAP and the simplified Max-Log-MAP decoding algorithm for the Turbo codes, the standard BP and the simplified Min-Sum de-coding algorithm for the LDPC codes. The simulation results show that For Turbo codes, the complexity of the Max-Log-MAP algorithm is lower compared to the MAP algorithm and the performance degradation is small when the coefficients are carefully chosen. For LDPC codes, while the performance of the Min-Sum algorithm suffers some losses compared to the standard BP algorithm, the decoding complexity is quite low which is suitable for the hardware implementation.Taking the SaS channel as a example, the third chapter focuses on the non-Gaussian channel. Consid-ering that the pdf of the SαS noise has no closed analytical expression, we set up the SaS channel model, and then propose three method to calculate the LLR out of the channel, the optimal LLR calculation scheme, the clipping method and the improved PWL2n+1 scheme. At last, we simulation the three detection method under the SaS channel model. The simulation results explicit that the proposed two fold fitting LLR calcu-lation methods used in systems with decoding module both have significant performance gain compared to the tradition detection method under the Gaussian assumption, which can resist to the non-Gaussian noise effectively.Taking the atmospheric turbulence channel as an example, the fourth chapter focuses on the non-additive channel. Firstly, we set up the general Gamma-Gamma distributed model and propose a mean estimated method to estimate the channel optical intensity I. Then, in order to compensate the performance loss due to the non-additive noise, we propose an Turbo-OSD scheme to improve the traditional Turbo iterative decoding algorithm. Further more, we also propose two improved Turbo-OSD scheme, additive LLR based Turbo-OSD algorithm and CRC-aided Turbo-OSD algorithm. At last, we simulate the FER for those proposed iterative decoding algorithms under Gamma-Gamma distributed channel with different parameters. The simulation results show that in the slowly changing Gamma-Gamma fading channel, the average estimation method to the signal light intensity is practical and the decoding performance of the proposed iterative algorithms is good enough in the fading channel.Taking the visible light transmission channel as an example, the fifth chapter focuses on the non-linear channel. Considering the nonlinear characteristics of the LED and the high PAPR characteristics of the DCO- OFDM system, we analysis the reason of the nonlinear characteristics of the LED in detail and set up an symmetrical clipping model. To reduce the PAPR of OFDM symbols, we adapt an tone injection algorithm based on convex optimization algorithm expanding the constellation. Then based on the expended constel-lation, we propose an improved soft detection method which supply the LLR to the Turbo/LDPC decoding module. At last we simulate the decoding performance of Turbo and LDPC codes under the symmetrical clipping model in the DCO-OFDM system. Simulation results certify the effectiveness of the TI algorithm to reduce the PAPR in the DCO-OFDM system. Besides, after our modification for the detection scheme, Turbo and LDPC codes can resist the nonlinear effect of the LED and save system resources with lower DC bias.