Name
Title | ||
|---|---|---|
Ergodic Capacity of Block-Fading Gaussian Broadcast and Multi-access Channels for Single-User-Selection and Constant-Power |
Abstract | ||
|---|---|---|
We consider the ergodic capacity region of block-fading Gaussian multiuser channels with channel-state information at both the transmitters and the receivers. We assume a single constraint on the total long-term average power used for both broadcast and multi- access channels. In addition to the optimal solution known from the literature, we provide analytic expressions - some of which are novel - to characterize the boundary surface of the capacity region under auxiliary constraints which include single-user-selection per block, constant total transmit-power per block and the combination of both. We also provide optimal resource allocation schemes to achieve the capacity limits for each case under consideration. Moreover, we provide numerical examples to compare the cases. As an illustrative example, we analyze the two-user case, although the results carry over to the M -user case. I. I NTRODUCTION Fading channels (both time and frequency selective) can be mod- eled as a family of parallel Gaussian channels: this is calle d a block- fading channel (1). Each of the parallel Gaussian channel blocks corresponds to a fading state. In general, the capacity of bl ock-fading multiuser channels with channel-state-information (CSI) at both the transmitter(s) and the receiver(s) can be achieved by (i) op timal power allocation over the channel blocks and (ii) optimal resource (rate and power) allocation over the users in each of the channel blocks. This is applicable to both the broadcast channel (BC) (one-to-many multiuser channel) (2), and the multi-access channel (MAC) (many- to-one multiuser channel) (3). From a practical communications engineering perspective, the optimal solutions are in most cases difficult if not impracti cal to implement. Thus, sub-optimal solutions which have close-to-optimum performance and, at the same time, lend themselves to an easy implementation are favorable. The optimal power allocation scheme over (block-)fading Gaussian broadcast and multi-access channels is given by the water-fi lling approach: more power is allocated when the channel is better and, depending on the desired operating point on the capacity region's boundary surface, some users are assigned higher average power to meet their rate demands. As a consequence of this power allocation policy, the total and individual transmission powers will vary hugely. This will cause problems when, e.g., the transmitter (i.e. the base station in the broadcast case) has maximum power constraints in order not to cause too much interference in adjacent cells. Furthermore, adaptive power control requires additional computational complexity to maintain the average power constraint, and variable transmission power is also likely to require more expensive radio-frequency circuitr y. The optimal resource allocation over a (flat-faded) channel block involves applying the optimal channel-access scheme, which is code division multiple access (in MAC) or superposition coding (in BC) with successive interference cancellation (SIC) at the rec eivers. Furthermore, the number of users scheduled in a channel block varies depending on the channel conditions. Superposition coding with SIC at the receivers can hardly be implemented in practice, because of (i) the complexity involved, (ii) the necessity to inform al l users about the order in which successive cancellation has to be conducted including the coding schemes used (signaling overhead), and (iii) different blocksizes used for encoding of different users: cancellation of a user's signal is only possible when the whole codeword 1 for this user has been received, although the user to be detected - due to delay constraints - may well have a much shorter (although still long) channel coding blocksize. As this user would have to wait for decoding until the "interfering" user's much longer codewo rd has been received, delay constraints are likely to be violated. |
Year | Venue | Field |
|---|---|---|
2009 | EUSIPCO | Topology,Broadcasting,Expression (mathematics),Fading,Computer science,Integrated optics,Ergodic theory,Computer network,Communication channel,Gaussian,Resource allocation |
DocType | Citations | PageRank |
Conference | 1 | 0.42 |
References | Authors | |
8 | 3 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Mohammad Shaqfeh | 1 | 126 | 15.03 |
| Norbert Goertz | 2 | 316 | 28.94 |
| J. Thompson | 3 | 3922 | 267.43 |