Investigation of mean least squares on probabilistic shaping direct detection optical OFDM systems

Probabilistic shaping (PS) techniques have been widely used in optical fiber communication systems due to their ability to significantly improve transmission efficiency and spectrum utilization. This paper generates probabilistic shaped 16QAM (PS-16QAM) signals based on the direct detection optical OFDM (DDO-OFDM) system using the block-bit shaping technique for constellation shaping of uniformly distributed 16QAM signals. In the process, considering the impact of inter carrier interference (ICI) and inter symbol interference (ISI) in OFDM signals on the signal quality at the receiver, comb pilot signals are inserted into PS-16QAM signals in the time domain and improve the accuracy of channel vector estimation by the mean least squares (LS) algorithm. We used optical simulation software VPItransmissionMaker10.1 to build the PS-DDO-OFDM system, and used MATLAB for digital signal processing (DSP). The simulation results after a 40-km standard single mode fiber show that the optical signal-to-noise ratio (OSNR) sensitivity of PS-16QAM with mean LS algorithm is improved by 5.6/8.2/10.2 dB, respectively, compared with uniform 16QAM with mean LS algorithm at entropy values of 3.78/3.31/2.82 bits/symbol, and in the PS-16QAM based system, the mean LS algorithm improves the OSNR sensitivity by 5.5/5.5/5.1 dB compared to the common LS algorithm for entropy values of 3.78/3.31/2.82 bits/symbol, respectively. In addition, the error vector magnitude (EVM) values at the receiver side and the convergence analysis of the constellation diagram further validate the significant advantages of the mean LS algorithm in PS-16QAM systems.