Towards Fully Optimized BICM Transmissions

Doctoral thesis, 2010

Bit-interleaved coded modulation (BICM) was introduced in 1992 as a new coded modulation (CM) scheme particularly well suited for fading channels and is nowadays considered the de facto CM scheme in wireless standards (HSPA, IEEE 802.11a/g/n, DVB-T2/S2/C2, etc.). A typical BICM configuration is based on a convolutional code, a single bit-level interleaver (S-interleavers), and an equally spaced quadrature amplitude modulation (QAM) input alphabet. In this thesis, we investigate the design of these three building blocks and show that traditional configurations are suboptimal. The main contribution of this thesis is to formally study the design of BICM systems in fading and nonfading channels and to propose new BICM designs. First, the use of multiple interleavers (M-interleavers) in BICM and BICM with iterative decoding (BICM-ID) is formally studied. For BICM, equally spaced QAM input alphabets labeled by the binary reflected Gray code are analyzed. Analytical bounds on the bit error rate (BER) are developed and based on these bounds, the optimum interleaver design is presented. It is shown that in both BICM and BICM-ID, M-interleavers outperform S-interleavers. Furthermore, the asymptotical optimality of BICM-ID with M-interleavers over BICM-ID with S-interleavers is proven. It is also shown that the use of M-interleaver redefines the optimality of the codes used in BICM and BICM-ID. The selection of these optimal codes is also studied in this thesis. A new BICM scheme based on nonequally spaced (hierarchical) QAM input alphabets, a bit-level multiplexer, and M-interleavers is also proposed in this thesis. It is shown that this new scheme outperforms previous BICM designs in fading and nonfading channels. Analytical bounds on the BER are developed and used to optimize the design of the system. The results show that, compared to traditional BICM designs, gains of a few decibels can be obtained. Finally, we study the use of BICM when the interleaver takes a trivial form (BICM-T), i.e., when it does not interleave the bits at all. An analytical model based on a new type of distance spectrum for convolutional codes is developed. This model is used to explain why BICM-T offers considerable gains compared to previous BICM configurations. It is also shown that properly designed BICM-T systems can be asymptotically as good as TCM.