Asymptotic performance of linear receivers in MIMO fading channels
IEEE Transactions on Information Theory
The RF-chain limited MIMO system: part I: optimum diversity-multiplexing tradeoff
IEEE Transactions on Wireless Communications
| Hi-index | 0.07 |
The V-BLAST (vertical Bell Labs layered space-time) architecture involves independent coding/decoding per antenna (layer) with equal rate and power per antenna and a fixed order of nulling/canceling decoding but is known to suffer from poor performance; for example, in a multi-input multi-output (MIMO) Rayleigh fading channel with Mt transmit and Mr receive antennas (Mr ges Mt), the diversity-multiplexing gain (D-M) tradeoff is just (Mr-Mt+1)(1-r/Mt) for risin[0, Mt]. There are two remedies available, namely, (i) channel-dependent ordered decoding at the receiver and (ii) allocation of rates and powers across the transmit antennas. However, the former doesn't improve the D-M tradeoff curve and while the latter does (with maximum diversity gain Mr and maximum multiplexing gain Mt), its tradeoff curve is still significantly inferior compared to the D-M tradeoff curve of the optimum (unconstrained) MIMO architecture. In this two-part paper, it is shown that a dramatically better D-M tradeoff and error (e.g. outage) probability can be obtained if the two remedies, i.e., ordered BLAST decoding and rate/power allocation, are judiciously combined. Indeed, a framework is developed for jointly designing channel-dependent ordered decoding at the receiver and decoding order-dependent rate/power allocation at the transmitter. The framework encompasses a large class of new spatial multiplexing architectures (SMAs). In this part, an upper bound to the D-M tradeoff for this class is obtained and found to be quite close to the optimal D-M tradeoff of the MIMO channel. Two special SMAs are proposed corresponding to two different decoding orderings. One is called the norm ordering rate tailored SMA (NRT-SMA), and the other is called the greedy ordering rate tailored SMA (GRT-SMA). The latter is shown to have the D-M tradeoff equal to the upper bound and is hence optimal in D- -M tradeoff among the proposed class of SMAs. Compared to the traditional V-BLAST architecture, the only added complexity of these SMAs is due to (i) feeding back the decoding order information (leslog2(Mt!) bits) per channel realization, and (ii) applying rate and power allocation according to a pre-computed lookup table.