Practical Issues of Multiuser Detection

2.5.2.1 Computational Complexity

Much of the research on multiuser detection is motivated by the complexity issues of optimum receivers proposed by Verdu [48]. The linear MUD (decorrelator, MMSE) approach to CDMA offered performance comparable to optimum receivers (exponential in complexity) with linear increase in the complexity. The saving in complexity though sounds exciting, linear multiuser re- ceivers are still far too complex than simple matched filter receivers. This is due to the requirement of the inversion of cross correlation matrix for the detection of each user at every symbol period which usually requires K2 operations. The use of interference cancellation receivers does not require the matrix inversion operations and, hence, are attractive in terms of computational com- plexity. Below, a review of most critical problems faced by IC receivers and different approaches to address these issues as proposed by various researchers is provided. Finally, another view on the problems and novel approaches proposed by the author are also described.

2.5.2.2 Estimation and Cancellation of MAI

The IC receivers though are less complex than linear and optimum receivers, their performances are plagued by two practical phenomenons. They are the nearfar user conditions and imperfect channel parameter estimation; which often lead to the problem of serious error propagation and in some cases the performance can be even worse than MF receivers. The central cause for these problems is the very unreliable power estimation of conventional matched filter frontend. It is now

well known that the MF introduce bias into the output decision statistic signals which increases linearly with increase in number of users [47]. Therefore, much of the research on IC receivers have assumed either equal power AWGN environments for PIC [51], [52] or perfect knowledge of received power of users signals in the case of SIC [47].

There exists different approaches in the literature [52], [53], [50] to tackle the problem of un- reliable MAI estimate generation due to the use of conventional matched filter within IC receivers. A popular approach is the partial cancellation by Divsalar et.al. in [52] that shows much improved performance compared with conventional PIC receivers. The intuition behind partial cancellation is backed up by Buehrer et. al. in [60] where they also provide some theoretical analysis of the bias problem in high system loading. Since the partial cancellation approach is based on some trial and error search of cancellation factors, an adaptive weighting of interference estimates for cancellation using a least mean square (LMS) algorithm for multistage PIC is proposed by Xue et. al. in [61]. The scheme has shown to improve the performance of PIC considerably. Recently, another approach of PIC in Rayleigh fading channels for SINR maximization by exploiting the cross correlation of users’ sequences is proposed by Tikiya et. al. in [62]. The scheme is shown to be much improved performance compared with conventional PIC at much higher computational complexity.

Similarly, the partial and limited number of cancellation approaches have been applied to SIC. Another approach of using multistage detection with soft decision at early stage is proposed by Hui and Letaief in [63]. This approach is later refined by Zha [64] which uses combination of hard and soft decision at each stage based on the magnitude of output signal and amplitude averaging [53] to improve the performance of a multistage SIC. Following another course of research on problem of SIC in practical environments using improved power control has been investigated thoroughly by Andrews et. al. in his doctoral dissertation [65]. His approach to solve the problem of imperfect estimation and cancellation is motivated by the fact that if one can learn the static of channel estimation error of the user’s signal at the receiver, that information can be incorporated within the power control algorithm to overcome the capacity penalty due the imperfect cancellation. The optimum power vector is found to be achievable in different channel setting that equalizes the SINR for all users using an iterative algorithm [65]. A comprehensive survey of techniques addressing the issues of imperfect MAI estimation and cancellation can be exhaustive and, hence, only the most significant techniques have been highlighted in this discussion. Based on the brief survey on issues of aforementioned techniques, next a brief discussion of novel blind adaptive approach to address the problem of imperfect MAI estimation and cancellation is carried out.

prove the user capacity of CDMA due to their lower complexity and potentials to overcome nearfar channel conditions. Many variants of IC receivers in the literature are mainly investigated assum- ing a more favorable assumptions such as perfect power control, ideal channel estimation, etc. Practical wireless systems are characterized by rapid changes in users’ channel amplitudes and phases, also with non-ideal distribution of their power due to their varied locations. The main problem for IC receivers is therefore accuracy in the estimation and cancellation of MAI. To im- prove the detection and MAI estimation, a novel blind adaptive approach [16, 17] is proposed in this thesis and applied for both SIC and PIC receivers. The technique uniquely exploits the con- stant modulus property [66, 58] of users’ transmitted data to acquire their channel estimates during the despreading process and makes use of the despreader weights to perform the MAI cancella- tion. The Chapter three describes these new techniques in greater details and their performance analyses.

In this section, it has been appreciated that capacity and performance of non-orthogonal CDMA that is limited by MAI than that of any other effects and MUD can be used to achieve a part of the capacity. The information theoretic capacity of the CDMA, however, lies far above what has been achieved in practice. The traditional approach to multiple access which treats each user as inter- ferer, often require much higher complexity receivers. Another approach of collaborative coding multiple access with error control capabilities is introduced in [14], where it is shown that with simple decoding technique, the scheme achieves much higher sum rate than traditional FDMA, TDMA and CDMA schemes. Also, it is presented in [45] that collaborative coding can be em- bedded within downlink of CDMA to support an overloaded system with much improved perfor- mance compared with other overloaded schemes [44], [26], using very simple and low complexity receivers. We have also noted that, the MUD is an essential technique to improve the performance of CDMA. To achieve high capacity in fading wireless environments, some form of diversity is also needed. Therefore to design an effective system, one has to consider all the aspects users’ channels and enabling techniques such as MUD and diversity. In the sequel, a brief discussion is carried out on what can be done to further improve the performance of existing multiuser systems in practical fading and interference limited wireless environments.

2.5.3 Discussion on New Approaches to Improve the Performance of Multiple Ac-