Overloaded Multiple Access Channels

The causes for the user capacity limitation of OWMA and Non-orthogonal CDMA have been identified in the previous subsection. All these schemes are focused towards ensuring maximum orthogonality between users signals during demodulation process. If the information theoretic limits of MAC are closely looked at, it can be realized that the orthogonal schemes are clearly suboptimal compared with more efficient schemes that pools the available resources among the users. It is shown in [2] that superposition coding of users signals followed by successive interfer-

ence cancellation at the central receiver achieves all the rate corners of the capacity regions of the Gaussian multiple access channels.

The recent work on this interesting topic of communication through an overloaded multiple access channels are now briefly reviewed. While the term MAC has been used, the downlink multiuser transmission channels or equivalently Broadcast Channels (BC) [30] are also the part of our study in this thesis. It will be interesting to mention about a very important idea that the MAC and BC are actually related intimately and information theory has been used to support the duality concept of MAC and BC and their capacity regions [22]. Since the objective of this thesis is to propose and analyse practical schemes for supporting an overloaded MAC and BC with low error rate performance, no detailed study is made on information theoretic aspects of the schemes. The new schemes are influenced by the principles of collaborative coding and decoding [12, 41, 42],[14] and in particular applied for downlink and uplink of an overloaded CDMA. 2.4.3.1 Review of Channel Overloading Schemes in CDMA

The most comprehensive research on overloaded CDMA is conducted by Vanhaverbeke in his doctoral dissertation [43]. Where it is shown that user capacity of CDMA can be increased sub- stantially by effective system design both at transmitters and receivers. More specifically, he inves- tigates the use of scheme using two sets of orthogonal sequences, also called OCDMA/OCDMA . Popular examples of such schemes include the well-known Quasi orthogonal CDMA [44], Ran- dom OCDMA/OCDMA [26], Improved OCDMA/OCDMA [40]. The common idea behind all these schemes is that, the first set of complete N orthogonal sequences are assigned to first N out of K users. The remaining U = K − N users, where U << N , are assigned the same set of orthogonal sequences with set specific mask and overlaid on top of the N sequences.

The premise behind their detection schemes is that the received signals of the users within set with complete sequences (set1) are easier to demodulate since interference received from U (set2) users are low enough to make reliable estimates of set1 users [43]. To detect users’ signals within the set2, the estimates of set1 users’ data can be used. An iterative multistage receiver with decision feedback detection is employed to successively refine the estimates of users signals for both sets. The scheme has been shown to perform very well, giving near single user performance with just three iterations [26]. Similarly, a hybrid multiple access scheme consisting of OCDMA and TDMA has also been proposed in [25] to increase the user capacity of CDMA. As it can be seen that the schemes described are interference limited and therefore they rely on complex inter- ference canceling receivers to overcome the effects of MAI arising due to the loss of orthogonality in overloaded channels.

of exploiting the state of reduced interference to group of N users in the larger set {1, 2, .., N } < K of the system. Since the signals of U users within smaller set {N + 1, N + 2, .., K} are also spread over N degree of freedom, the effective interference of these users on the first set of users will only be σ2 = U/N . Therefore, as long as U << N , the users within the first set will have good SINR and therefore demodulation and estimation of their signals will be very reliable. The obtained estimates can now be incorporated within the next stage of detection to demodulate and estimate users within the second set. One very important point to note is that this process of iterative multistage detection technique comes with higher computational complexity and thus can not be employed in the downlink for mobile users’ receivers. Therefore, it is concluded that while the performance of such schemes are attractive, they can only be applied at significant increase in the complexity. Now the same problem is viewed from a different angle, and the notion of ambiguity and unique decodability is used to address the problem of overloaded channels and to avoid the excessive interference under such channel conditions.

The interference in the above techniques arises from the fact that when the system is over- loaded, the orthogonality is not maintained. In such condition the introduction of other users’ randomly distributed signals on the desired signals may become so severe that the detector may end up with wrong decision of user signal. Lets’s consider a simple example of two users trans- mitting simultaneously their BPSK modulated signals over the common channel to achieve sum rate Rsum = 2. In absence of noise the output of the combined signals can take three values only.

Since there are four possibilities of users’ transmitted signals and the two of them give exactly the same output i.e. addition of 0 and 1 and 1 and 0 both give output of 1. Therefore there is a detection ambiguity when users transmit these bits. Other two possibilities give values of 2 and 0, respectively and can be decoded correctly. The conventional approach to overloaded CDMA tries to avoid the situation of detection ambiguity by employing interference estimate generation and cancellation. This ensures that the remaining signal will most probably contain either 0 or 1 signal values with high probability so that decision error is minimized. A very simple way to resolve the ambiguity problem is to employ collaborative coding or uniquely decodable codes [14], [45] and described earlier in subsection in 2.4.2.2. Also, the multiple access scheme based on such T-user collaborative coding, CCMA has shown to achieve sum rate much higher compared to that of OWMA with Csum = 1[14], [46]. This idea of resolving the ambiguity can be employed within

synchronous CDMA to increase the user capacity. As can be noted, the number of users can be T times the available sequences.

The techniques of multiuser detection and interference cancellation though more complex than conventional matched filter receivers, have received much attention in multiuser CDMA research

community due to their potential to achieve high capacity [47]. This thesis makes a number of new contributions in this subject area. For example, new techniques are proposed that overcome many practical issues limiting the performance of the interference cancellation subclass of multiuser receivers. The detailed descriptions of techniques and their performances and complexities are presented in Chapter 3. However, to provide a an overview of basic principle of different multiuser receiver techniques and a literature review of this subject area, in the next subsection, simplified descriptions of a broad class of multiuser detection techniques and their limitations is provided. Main motivations and ideas behind the new techniques proposed in this thesis are also highlighted.