Multi-Layer Interleave-Devision Multiple Access
Final Report Abstract
Concerning the evolution of wireless communications, a key topic is to pursue high spectral efficiency together with high power efficiency. This is motivated by the increasing demand of high-rate data services, like multimedia data in third and forth generation mobile communication systems. Various advanced transmission techniques have been developed to solve these challenges. In this project, focus is on interleave-division multiplexing (IDM). IDM can be used as a multiplexing scheme, but even more importantly as a power and bandwidth efficient coded modulation scheme. In IDM interleaving is employed at chip level. Different layers are separated by different interleavers. Encoded layers are superimposed after being weighted by an amplitude factor and rotated in phase. The amplitudes and phases should be jointly optimized to achieve a good performance. The maximum mutual information obtained by IDM using different power and phase allocation schemes has been investigated in this work. Particle swarm optimization (PSO) is used to find a set of amplitudes and phases which maximize the mutual information in the signal-to-noise ratio (SNR) region of interest. This algorithm can also be extended to jointly optimize the mutual information and the peak-to-average power ratio (PAPR), which is critical with respect to the efficiency of the power amplifier. Besides, the influence of power and phase allocation to the performance of multi-layer detection has been investigated. We observed that the limitation of bandwidth efficiency is not due to the power and phase allocation scheme, but due to the suboptimal Gaussian detector, which assumes that the inter-layer interference is Gaussian distributed. An extrinsic information transfer (EXIT) chart analysis has been used to determine the maximum mutual information which can be supported by different detectors. It is found that the loss of mutual information by using a Gaussian detector grows with increasing SNR. The a posteriori probability (APP) detector has a good performance but its complexity growly exponentially with the number of layers. To compromise between performance and complexity, we derived a novel reliability-based hybrid detector, which allocates the computational power according to the reliability of the decoded bits. Compared to conventional QAM modulation, the constellation diagram of IDM is typically non-uniformly spaced. Due to superposition, most of the transmitted symbols concentrate in the low-energy region of the complex plane, and some of them may overlap. For a large number of layers, the superimposed signal is approximately Gaussian distributed. We compare the mutual information of IDM with that of conventional QAM modulation. It is well known that there is always a gap between the mutual information of QAM and the Shannon limit, which is termed the ultimate shaping gain. Conventionally, complicated shaping algorithms need to be implemented to eliminate this gap. IDM inherently provides a shaping gain. No active shaping is necessary. In the most interesting SNR region, IDM shows a higher potential than QAM. In practical systems, the oscillator instabilities significantly degrade the system performance. The performance of IDM over phase-uncertain channels has been investigated. It is found that IDM is less sensitive to oscillator phase noise than QAM. A graph-based phase noise compensation algorithm has been derived. The results show that near perfect performance can be achieved. Furthermore, the combination of IDM and orthogonal frequency-division multiple access (OFDMA) has also been investigated in this work. For each subcarrier, the number of binary layers to be superimposed can be determined by a bit-loading algorithm according to the channel condition. The capacity of each subchannel can be approached closer by using IDM than using conventional BICM. IDM also provides a universal coded modulation scheme to all the subcarriers, which reduces the amount of side information to be transmitted. As mentioned before, the superposition nature of IDM provides an important difference compared to conventional modulation schemes. As a consequence, the mapping is typically non-bijective. Therefore, this poses challenges with respect to the design of the channel code, whose characteristic should be matched to that of the multi-layer detector. Powerful channel codes like Turbo code or low-density parity check (LDPC) code have a quasi-flat transfer characteristics in their EXIT charts, which typically cross the transfer characteristic of the multi-layer detector after several iterations. Therefore, channel codes suitable for IDM need to be studied. Possible applications are manifold, since IDM can replace any digital modulation scheme. One possible application is 5G wireless radio.
Publications
- “Helical interleaver set design for interleave-division multiplexing and related techniques,” IEEE Communications Letters, vol. 12, pp. 843-845, Nov. 2008
D. Hao and P. A. Hoeher
- “Multilayer APP detection for IDM,” Electronics Letters, vol. 46, no. 1, pp. 96-97, Jan. 2010
M. Noemm, T. Wo, and P. A. Hoeher
- “Superposition mapping with adaptive bit loading for BICM-OFDM systems”, in Proc. International Symposium on Turbo Codes & Iterative Information Processing, Brest, France, Sep. 2010
Z. Shi, T. Wo and P. A. Hoeher
- “Superposition modulation with reliability-based hybrid detection”, in Proc. International Symposium on Turbo Codes & Iterative Information Processing, Brest, France, Sep. 2010
D. Hao and P. A. Hoeher
- Kapitel 13.9 “Kombinierte Modulation und Kanalcodierung”, in Grundlagen der digitalen Informationsübertragung, Wiesbaden: Vieweg + Teubner, 2011
Peter A. Höher
