Tommy Guess

Optimum decision feedback multiuser equalization with successive decoding achieves the total capacity of the Gaussian multiple-access channel

The complex Gaussian multiple-access channel (GMAC) Y_=AX_+N_ is considered. The transmitters send information independently with a power constraint so that X_ has a product distribution with E[|X/sub k/|/sup 2/]/spl les/P/sub k/. It is known that multiuser codes exist that will achieve any rate-tuple in the capacity region of the GMAC provided that an optimum joint decoder is used. However, little progress has been made in multiuser coding, and moreover, optimum joint decoding would be too complex. We restrict the decoder to be of a successive decoding type with equalization. Such a decoder is parameterized by feedforward and feedback equalization vectors. The optimum successive decoder (OSD) is obtained by maximizing the mutual information for each user over those vectors. The key result of this paper is that the OSD achieves the total capacity of the GMAC at any vertex of the capacity region. With the OSD, each transmitter can use a single-user code independently of the other users. The complexity of equalization is also only linear in the number of users. For the conventional GMAC, Y=/spl Sigma//sub i=1//sup M/ X/sub i/+N, where A is a row vector with unit elements, the OSD is degenerate and involves no equalization. It reduces to the well-known successive decoder for that channel as described by Wynter (1974) and Cover (1975). Furthermore, for the particular case of the uncoded channel, the OSD reduces to a new decision feedback multiuser detector. This detector is optimum in the sense that it maximizes signal-to-interference ratio for each user.

An information-theoretic framework for deriving canonical decision-feedback receivers in Gaussian channels

A framework is presented that allows a number of known results relating feedback equalization, linear prediction, and mutual information to be easily understood. A lossless, additive decomposition of mutual information in a general class of Gaussian channels is introduced and shown to produce an information-preserving canonical decision-feedback receiver. The approach is applied to intersymbol interference (ISI) channels to derive the well-known minimum mean-square error (MMSE) decision-feedback equalizer (DFE). When applied to the synchronous code-division multiple-access (CDMA) channel, the result is the MMSE (or signal-to-interference ratio (SIR) maximizing) decision-feedback detector, which is shown to achieve the channel sum-capacity at the vertices of the capacity region. Finally, in the case of the asynchronous CDMA channel we are able to give new connections between information theory, decision-feedback receivers, and structured factorizations of multivariate spectra.

Bandwidth-efficient multiple access (BEMA): a new strategy based on signal design under quality-of-service constraints for successive-decoding-type multiuser receivers

This paper considers the design of signature waveforms for successive-decoding-type multiuser receivers (including the optimum successive decoder (OSD)) in a correlated-waveform multiple-access channel. The problem is to obtain signature waveforms that require as little bandwidth as possible while allowing the receiver to meet a given set of quality-of-service (QoS) objectives. The QoS objectives are specified for each user in terms of capacity, or equivalently, the signal-to-interference ratio. A (generally unachievable) lower bound is obtained on the minimum bandwidth required to achieve these QoS constraints. Moreover, a simple algorithm is proposed for obtaining signal sets that meet the QoS constraints when used with the OSD, and which, while not optimal, require a bandwidth that can be very close to the minimum required bandwidth. It is also shown that such signal sets allow for a significantly more efficient use of bandwidth than do orthogonal signals used in time- or frequency-division multiple access (TDMA/FDMA). Based on our signal design approach, we propose a new multiple-access strategy that we refer to as bandwidth-efficient multiple access (BEMA). While BEMA is more bandwidth efficient than TDMA or FDMA, it retains their desirable feature of needing only single-user coding (and decoding) for each user.

Error exponents for maximum-likelihood and successive decoders for the Gaussian CDMA channel

Random-coding error exponents are derived for the Gaussian code-division multiple-access (CDMA) channel for the maximum-likelihood and optimum successive decoders. Error exponents not only specify the capacity region of the channel, which is known, but also give lower bounds on the rate of exponential decay of the average probability of error as a function of the block length of random codes. A comparison of the two decoders in terms of their error exponents is included.

The outage capacity of BLAST for MIMO channels

This paper is concerned with multiple-input multiple-output (MIMO) channel that experience quasi-static flat fading and Gaussian noise. The information-theoretic quantity of outage capacity is precisely defined as it relates to the successive-decoding approach known as the BLAST (or Bell labs layered space-time) architecture. It is shown that finding the outage capacity of a BLAST system is equivalent to the solution of a K-parameter optimization problem, where K is the numbers of transmit antennas. This optimization is solved and the resulting outage capacity of BLAST is found to sustain a loss relative to the unconstrained (i.e., optimal) system, though it still provides for very high-rate data transmission.

Achieving vertices of the capacity region of the synchronous Gaussian correlated-waveform multiple-access channel with decision-feedback receivers

We consider users operating synchronously on a correlated-waveform multiple-access (CWMA) channel with additive Gaussian noise. CWMA is a multiple-access scheme where the users are assigned signature waveforms; it subsumes both orthogonal signaling (e.g., TDMA) and identical-waveform signaling (as in the conventional Gaussian multiple-access channel (C-GMAC)) as special cases. We obtain a decision-feedback receiver (DFR) for the CWMA channel which, in spite of its lower complexity than the optimal receiver, achieves the vertices of the CWMA channels capacity region, implying in addition that the users can choose their single-user codes independently of each other. This result generalizes the well known successive decoder of the C-GMAC.

A new successively decodable coding technique for intersymbol-interference channels

For the Gaussian channel with intersymbol-interference (ISI), it is known that there is no loss in channel capacity if the receiver is an ideal minimum mean-squared error (MMSE) decision-feedback equalizer (DFE) with error-free feedback. However, combining the DFE with channel coding is problematic. Transmitter precoding and reduced-state sequence estimation are two common approaches. This paper introduces a new successively decodable coding technique that effectively combines channel coding with decision-feedback that is housed in the receiver.

Signature sequence and training design for overloaded CDMA systems

The focus of this paper is on training and signature sequence design in an overloaded synchronous CDMA system. The channel fading gains are assumed to be unknown and are estimated using training sequences. We derive a maximum-likelihood (ML) estimator and design sequences to minimize the mean-squared error (MSE) of the estimate. Two design scenarios are considered. One case assumes that the spreading sequences are fixed due to existing system constraints and optimal training sequences are designed using an iterative algorithm in order to minimize the MSE of the estimate. Performance of the iterative algorithm is examined using Welch-bound equality (WBE) sequences as the pre-designed spreading sequences. In the other scenario, spreading sequences and training sequences are designed jointly to minimize the MSE of the estimate. The jointly designed training and spreading sequences achieve optimum performance in the sense of minimizing the MSE. It is observed that when WBE sequences are used as spreading sequences the performance of the iterative algorithm is close to optimum, or even optimum, in certain situations

Asymptotical analysis of the outage capacity of rate-tailored BLAST

The paper considers the multiple-input multiple-output (MIMO) channel with quasi-static Rayleigh fading. Lower bounds are derived for the outage capacity of a recently introduced approach to horizontally encoded BLAST (Bell Labs layered space time) in which the rates of the layers are optimally designed. The outage capacity of this scheme, rate-tailored BLAST (RT-BLAST), is shown to behave in the same manner as the true outage capacity of the MIMO channel (e.g., using a maximum-likelihood, ML, receiver) in the high signal-to-noise ratio (SNR) regime. That is, the asymptotical growth rates of the ML and RT-BLAST outage capacities as a function of SNR are equal to each other.

A comparison of bandwidth-efficient multiple access to other signal designs for correlated waveform multiple-access communications

There have been several papers in the literature that deal with the design of signature waveforms for use by the transmitters in uplink, single-cell, multiple-access communications. In particular, we consider the approach introduced by Guess and Varanasi (1996, 1997), where the signature waveforms are specifically designed for the centralized multiuser receiver at the base so that each transmitter can be guaranteed a preassigned quality-of-service (QoS) requirement in terms of the received signal-to-interference ratio (SIR). The resulting strategy is called bandwidth-efficient multiple access (BEMA). When all users employ pulse amplitude modulation (PAM) and a common signaling rate, the key question in BEMA is how the waveforms must be designed to occupy as little bandwidth as possible and still meet the QoS objectives. For a strict measure of bandwidth, and for a given set of received powers, this question was addressed by the authors for the maximum SIR decision-feedback (MSIR-DF) receiver of Varanasi and Guess (1998). A similar question was addressed by Viswanath, Anantharam and Tse (see ibid., vol.45, p.1968-1983, Sept. 1999) where, for a sum constraint on the received powers, optimal signature signals and transmit powers were obtained for the linear MSIR receiver (without decision feedback). A somewhat different but related non-QoS approach proposes the design of signature signals that maximize the total capacity of the multiple-access channel under a spreading-gain constraint. This article undertakes a comparison of the minimum bandwidth required (to achieve the QoS requirements) for the signals designed for the MSIR-DF receiver, for the linear MSIR receiver, and for sum-capacity maximization as shown by Viswanath and Anantharam (see ibid., vol. 45, p.1984-1991, Sept. 1999). We show that the bandwidth required for multiuser receivers with decision feedback can be significantly less than that required for linear receivers or for sum-capacity maximization when the MSIR-DF receiver is used.