Encoding and decoding of serially concatenated codes for data storage systems

Abstract

In this project, we consider the application of serially concatenated codes (SCC) for data storage systems. The SCC typically consists of an outer code and an inner code. The outer code of the SCC can be a high-rate convolutional code, or a low-density parity-check (LDPC) code; while the inner code normally takes the form of the accumulate code or a high-rate recursive convolutional code (RSC). The SCC has the advantage of achieving flexible code rates, low encoding complexities and various decoding schedules, which makes it a promising channel coding scheme for data storage systems. Considering that the conventional LDPC codes have some limitations for applications to high-rate magnetic recording channels, such as high encoding complexities and the lack of sufficient flexibilities in terms of code lengths or code rates, we consider extended irregular-accumulate (eIRA) codes in this project which are flexible in code parameters and feature linear-time encoding complexities.

Furthermore, we study the encoding and decoding of the eIRA code and investigate the effect of decoding schedules on the bit-error-rate (BER) performance over the Additive White Gaussian Noise (AWGN) channels. The simulation results show that the eIRA code performs comparably to LDPC codes of similar code lengths and code rates over these two channels. Besides, the column weight of the eIRA code affects BER performance in the error floor region. A low-rate code is shown to have better performance in the waterfall region; while a high-rate code shows much lower error floor but it is at the expense of a slightly worse-off waterfall performance.

Finally, we consider the application of eIRA codes to magnetic recording channels and simulate the performance of a turbo equalizer consisting of a BCJR channel detector and the eIRA decoder with various detection/decoding schedules. Simulation results showed that the eIRA code can also provide performance similar to a structured LDPC code over the ME2PR4 channel, while achieving much lower encoding complexities.