## Optimum Linear Joint Transmit-Receive Processing for MIMO Channels with QoS Constraints (2003)

Venue: | IEEE Trans. Signal Process |

Citations: | 24 - 3 self |

### BibTeX

@ARTICLE{Palomar03optimumlinear,

author = {Daniel Perez Palomar and Miguel Angel Lagunas and John M. Cioffi},

title = {Optimum Linear Joint Transmit-Receive Processing for MIMO Channels with QoS Constraints},

journal = {IEEE Trans. Signal Process},

year = {2003},

volume = {52},

pages = {1179--1197}

}

### OpenURL

### Abstract

This paper considers vector communication through a multi-input multi-output (MIMO) channel with a set of Quality of Service (QoS) requirements for the simultaneously established substreams. Linear transmit-receive processing (also termed linear precoder at the transmitter and linear equalizer at the receiver) is designed to satisfy the QoS constraints with minimum transmitted power. Although the original problem is a complicated nonconvex problem with matrix-valued variables, with the aid of majorization theory, we reformulate it as a simple convex optimization problem with scalar variables. We then propose a practical and e#cient multi-level water-filling algorithm to optimally solve the problem for the general case of different QoS requirements. The optimal transmit-receive processing is shown to diagonalize the channel matrix only after a very specific pre-rotation of the data symbols. For situations in which the resulting transmit power is too large, we give the precise way to relax the QoS constraints in order to reduce the required power based on a perturbation analysis. We also propose a robust design under channel estimation errors which has a great interest in real implementations.

### Citations

8603 |
Elements of information theory
- Cover, Thomas
- 1991
(Show Context)
Citation Context ...y-achieving solution dictates that the channel matrix has to be diagonalized and that the power at the transmitter has to be allocated following a water-filling distribution on the channel eigenmodes =-=[24, 3, 25]-=-. In theory, this solution has the implication that an ideal Gaussian code should be used on each channel eigenmode according to its allocated power [24]. In practice, however, Gaussian codes are subs... |

3703 |
Convex optimization
- Boyd, Vandenberghe
- 2004
(Show Context)
Citation Context ...ered case, the curve is parameterized with respect to , P ( ), and in [15], the curve is parameterized with respect to P , (P )). where and the ’s are the dual variables or Lagrange multipliers [36], =-=[37]-=-. The water-filling solution is easily found from the sufficient and necessary Karush–Kuhn–Tucker (KKT) optimality conditions (the problem satisfies the Slater’s condition and therefore strong duality... |

1838 | Capacity of multi-antenna Gaussian channels
- Telatar
- 1999
(Show Context)
Citation Context ... the receiver have recently attracted a significant interest because they provide an important increase in capacity over single-input single-output (SISO) channels under some uncorrelation conditions =-=[5, 6]-=-. The transmitter and the receiver may or may not have channel state information (CSI), although for slowly varying channels it is generally assumed CSI at the receiver (CSIR). Many publications have ... |

1548 | On the limits of wireless communications in a fading environment when using multiple antennas
- Foschini, Gans
- 1998
(Show Context)
Citation Context ... the receiver have recently attracted a significant interest because they provide an important increase in capacity over single-input single-output (SISO) channels under some uncorrelation conditions =-=[5, 6]-=-. The transmitter and the receiver may or may not have channel state information (CSI), although for slowly varying channels it is generally assumed CSI at the receiver (CSIR). Many publications have ... |

1285 | A simple transmit diversity technique for wireless communications
- Alamouti
- 1998
(Show Context)
Citation Context ...ly varying channels it is generally assumed CSI at the receiver (CSIR). Many publications have dealt with the case of no CSI at the transmitter (CSIT) such as the popular space-time coding techniques =-=[7, 8, 9]-=-. Another great block of research has been devoted to the situation with CSIT [2, 3, 4, 10]. We focus on the latter and, in particular, when linear processing is utilized for the sake of complexity. I... |

1249 | Space-time codes for high data rate wireless communications: Performance criterion and code construction
- Tarokh, Seshadri, et al.
- 1998
(Show Context)
Citation Context ...ly varying channels it is generally assumed CSI at the receiver (CSIR). Many publications have dealt with the case of no CSI at the transmitter (CSIT) such as the popular space-time coding techniques =-=[7, 8, 9]-=-. Another great block of research has been devoted to the situation with CSIT [2, 3, 4, 10]. We focus on the latter and, in particular, when linear processing is utilized for the sake of complexity. I... |

1023 | Multiuser Detection - Verdú - 1998 |

951 |
Layered space-time architecture for wireless communication in a fading environment when using multi-element antennas
- Foschini
- 1996
(Show Context)
Citation Context ...ly varying channels it is generally assumed CSI at the receiver (CSIR). Many publications have dealt with the case of no CSI at the transmitter (CSIT) such as the popular space-time coding techniques =-=[7, 8, 9]-=-. Another great block of research has been devoted to the situation with CSIT [2, 3, 4, 10]. We focus on the latter and, in particular, when linear processing is utilized for the sake of complexity. I... |

804 |
Fundamentals of statistical signal processing: estimation theory
- Kay
(Show Context)
Citation Context ... ai, its minimum value can be found by setting its gradient to zero. The solution is given by A = � HBB H H H �−1 + Rn HB. (8) Expression (8) is the linear minimum MSE (LMMSE) filter or Wiener f=-=ilter [33] and minim-=-izes simultaneously all the diagonal elements of the MSE matrix E as can be seen by “completing the squares” as follows (from (7)): E =(A− (HBB H H H + Rn) −1 HB) H (HBB H H H + Rn)(A − (HBB... |

788 | Optimization by Vector Space Methods
- Luenberger
- 1969
(Show Context)
Citation Context ... that the original nonconvex problem has been reformulated as a simple convex problem, we know that a global optimal solution can be obtained in practice by using, for example, interior-point methods =-=[36, 37]. -=-Nevertheless, the particular convex problem (19) obtained in Theorem 1 can be optimally solved by a simple water-filling algorithm (Algorithm 1) as stated in the following proposition (by setting ˜ρ... |

728 |
Inequalities: theory of majorization and its applications
- Marshall, Olkin
- 1979
(Show Context)
Citation Context ...e convex problems with scalar variables that can be optimally solved. In this section we introduce the basic notion of majorization and state some basic results that will be needed in the sequel (see =-=[29] for a c-=-omplete reference of the subject). Majorization makes precise the vague notion that the components of a vector x are “less spread out” or “more nearly equal” than are the components of a vecto... |

282 | Spatio–temporal coding for wireless communications
- Raleigh, Cioffi
- 1998
(Show Context)
Citation Context ...I) 1999FI 00588, 2001SGR 00268; and Samsung Advanced Institute of Technology. 0s1 Introduction Communications over multi-input multi-output (MIMO) channels have recently gained considerable attention =-=[1, 2, 3, 4]-=-. They arise in many different scenarios such as when a bundle of twisted pair copper wires in digital subscriber lines (DSL) is treated as a whole [1], when multiple antennas are used at both sides o... |

128 | Redundant filterbank precoders and equalizers part I: Unification and optimal designs
- Scaglione, Giannakis, et al.
- 1988
(Show Context)
Citation Context ...I) 1999FI 00588, 2001SGR 00268; and Samsung Advanced Institute of Technology. 0s1 Introduction Communications over multi-input multi-output (MIMO) channels have recently gained considerable attention =-=[1, 2, 3, 4]-=-. They arise in many different scenarios such as when a bundle of twisted pair copper wires in digital subscriber lines (DSL) is treated as a whole [1], when multiple antennas are used at both sides o... |

128 | Joint tx-rx beamforming design for multicarrier mimo channels: A unified framework for convex optimization
- Palomar, Cioffi, et al.
(Show Context)
Citation Context ...l to interference-plus-noise ratio (SINR) with a zero-forcing (ZF) constraint [4], the minimization of the determinant of the MSE matrix [13], and the minimization of the average bit error rate (BER) =-=[14, 15]-=-. A general unifying framework was developed in [15] to treat all such criteria by classifying the variety of objective functions into Schur-concave and Schur-convex functions. In [16], some criteria ... |

116 | Transmit beamforming and power control for cellular wireless systems
- Farrokhi, Liu, et al.
- 1998
(Show Context)
Citation Context ... mainly for multiuser scenarios in which multiple distributed users coexist and joint multiuser processing cannot be assumed at one side of the link since the users are geographically distributed. In =-=[17]-=-, a multiuser scenario with a multielement base station was considered and optimal beamvectors and power allocation were obtained for the uplink and downlink. In [18, 19], the same problem was solved ... |

110 | Optimal designs for space-time linear precoders and decoders - Scaglione, Stoica, et al. |

109 | Optimal sequences and sum capacity of synchronous CDMA systems - Viswanath, Anantharam - 1984 |

107 |
Robust adaptive beamforming
- Cox, Zeskind, et al.
- 1987
(Show Context)
Citation Context ...Design under Channel Estimation Errors Hitherto, perfect CSI has been assumed at both sides of the link. A common problem in practical communication systems arises from channel estimation errors (see =-=[38]-=- and references therein). Whereas a sufficiently accurate CSI at the receiver (CSIR) can be assumed in many cases, CSI at the transmitter (CSIT) will be far from perfect in any realistic situation (se... |

99 | Generalized linear precoder and decoder design for MIMO channels using the weighted MMSE criterion
- Sampath, Stoica, et al.
- 2001
(Show Context)
Citation Context ...ns have dealt with the case of no CSI at the transmitter (CSIT) such as the popular space-time coding techniques [7, 8, 9]. Another great block of research has been devoted to the situation with CSIT =-=[2, 3, 4, 10]-=-. We focus on the latter and, in particular, when linear processing is utilized for the sake of complexity. In many situations, the transmitter is constrained on its average transmit power to limit th... |

79 |
Joint transmitter and receiver optimization for multiple-input multiple-output systems with decision-feedback
- Yang, Roy
- 1994
(Show Context)
Citation Context ...I) 1999FI 00588, 2001SGR 00268; and Samsung Advanced Institute of Technology. 0s1 Introduction Communications over multi-input multi-output (MIMO) channels have recently gained considerable attention =-=[1, 2, 3, 4]-=-. They arise in many different scenarios such as when a bundle of twisted pair copper wires in digital subscriber lines (DSL) is treated as a whole [1], when multiple antennas are used at both sides o... |

79 | Optimal sequences, power control, and user capacity of synchronous CDMA systems with lin169 MMSE multiuser receivers
- Viswanath, Anantharam, et al.
- 1999
(Show Context)
Citation Context ...signal model arising from a singleantenna multiuser multicarrier CDMA system was treated in [21], in which a suboptimal (due to the 1snonconvexity of the problem) gradient-type algorithm was used. In =-=[22]-=-, a single-antenna multiuser CDMA system was characterized in terms of user capacity by optimally designing the CDMA codes of the users. The approach in this paper is similar to the aforementioned exa... |

76 |
Vectored transmission for digital subscriber line systems
- Ginis, Cioffi
- 2002
(Show Context)
Citation Context ...constellations simultaneously through the MIMO channel. In light on these observations, an interesting way to design the uncoded part of a communication system is based on the gap approximation (e.g. =-=[26]-=-) which basically gives the optimal bit distribution (following a water-filling allocation similar to that of the capacity-achieving solution) under the assumption that practical constellations such a... |

61 |
Synthesis of Minimum Roundoff Noise FixedPoint Digital Fiters
- Mullis, Roberts
- 1976
(Show Context)
Citation Context ...le structure plays a key role for practical implementation. Interestingly, an iterative approach to construct a matrix with equal diagonal elements and with a given set of eigenvalues was obtained in =-=[32] based a-=-lso on a sequence of rotations. Lemma 4 [29, 5.A.9.a] For any x, y ∈ IR n satisfying y ≻w x, there exists a vector u such that 4 Optimum Receive Processing u ≤ x and y ≻ u. In this section, we... |

60 |
Broadband radio access networks BRAN:Hiperlan type 2 technical specifications part 1-physical layer.”, ETSI
- ETSI
- 1999
(Show Context)
Citation Context ...h to transmit only over those eigenmodes with a sufficiently high gain), i.e., an equal-rate transmission. Examples of this pragmatic and simple solution are found in the European standard HIPERLAN/2 =-=[27]-=- and in the US standard IEEE 802.11 [28] for Wireless Local Area Networks (WLAN). In any case, once the constellations to be used at each of the substreams are known, the system can be further optimiz... |

58 |
Introduction to Convex Optimization with Engineering Applications
- Boyd, Vandenberghe
- 2003
(Show Context)
Citation Context ...to the constrained convex problem is L = ˇL� k=1 zk + ˇL� µk k=1 ⎛ ⎝ ˇL� i=k 1 − 1+ziλi ˇL� i=k ⎞ ˜ρi⎠ − ˇL� k=1 γk zk, (35) where the µk’s and the γk’s are the d=-=ual variables or Lagrange multipliers [36, 37]. -=-The water-filling solution is easily found from the sufficient and necessary Karush-Kuhn-Tucker (KKT) optimality conditions (the problem satisfies the Slater’s condition and therefore strong duality... |

55 |
11: Wireless LAN Medium Access Control (MAC
- IEEE
- 2003
(Show Context)
Citation Context ... with a sufficiently high gain), i.e., an equal-rate transmission. Examples of this pragmatic and simple solution are found in the European standard HIPERLAN/2 [27] and in the US standard IEEE 802.11 =-=[28]-=- for Wireless Local Area Networks (WLAN). In any case, once the constellations to be used at each of the substreams are known, the system can be further optimized such that each established substream ... |

41 | Filterbank transceivers optimizing information rate in block transmissions over dispersive channels
- Scaglione, Giannakis, et al.
- 1999
(Show Context)
Citation Context ...y-achieving solution dictates that the channel matrix has to be diagonalized and that the power at the transmitter has to be allocated following a water-filling distribution on the channel eigenmodes =-=[24, 3, 25]-=-. In theory, this solution has the implication that an ideal Gaussian code should be used on each channel eigenmode according to its allocated power [24]. In practice, however, Gaussian codes are subs... |

39 |
Joint transmitter-receiver optimization for multi-input multi-output systems with decision feedback
- Yang, Roy
- 1994
(Show Context)
Citation Context ... (MSE) matrix [11, 12, 2, 4, 10], the maximization of the signal to interference-plus-noise ratio (SINR) with a zero-forcing (ZF) constraint [4], the minimization of the determinant of the MSE matrix =-=[13]-=-, and the minimization of the average bit error rate (BER) [14, 15]. A general unifying framework was developed in [15] to treat all such criteria by classifying the variety of objective functions int... |

37 |
Minimum BER block precoder for zero-forcing equalization
- Ding, Davidson, et al.
- 2003
(Show Context)
Citation Context ...l to interference-plus-noise ratio (SINR) with a zero-forcing (ZF) constraint [4], the minimization of the determinant of the MSE matrix [13], and the minimization of the average bit error rate (BER) =-=[14, 15]-=-. A general unifying framework was developed in [15] to treat all such criteria by classifying the variety of objective functions into Schur-concave and Schur-convex functions. In [16], some criteria ... |

36 |
Digital transmission over cross-coupled linear channels,” Bell Syst
- Salz
- 1985
(Show Context)
Citation Context ...ral different criteria have been utilized as a means to measure the quality of the link and to design the system such as the minimization of the (weighted) trace of the mean-square error (MSE) matrix =-=[11, 12, 2, 4, 10]-=-, the maximization of the signal to interference-plus-noise ratio (SINR) with a zero-forcing (ZF) constraint [4], the minimization of the determinant of the MSE matrix [13], and the minimization of th... |

35 |
Optimal linear coding for vector channels
- Lee, Petersen
- 1976
(Show Context)
Citation Context ...ral different criteria have been utilized as a means to measure the quality of the link and to design the system such as the minimization of the (weighted) trace of the mean-square error (MSE) matrix =-=[11, 12, 2, 4, 10]-=-, the maximization of the signal to interference-plus-noise ratio (SINR) with a zero-forcing (ZF) constraint [4], the minimization of the determinant of the MSE matrix [13], and the minimization of th... |

31 | Joint transmitter receiver diversity for efficient space division multiple access
- Chang, Tassiulas, et al.
- 2002
(Show Context)
Citation Context ...heory, which allows the introduction of additional constraints (for example to control the dynamic range of the transmitted signal or to increase the robustness against channel estimation errors). In =-=[20]-=-, the problem was generalized to the case of having multiple antennas at both sides of the link, although a global optimal solution was not found due to the nonconvexity of the problem (a suboptimal i... |

26 | Optimal downlink beamforming using semidefinite optimization,” presented at the 37th Annu
- Bengtsson, Ottersten
- 1999
(Show Context)
Citation Context ...re geographically distributed. In [17], a multiuser scenario with a multielement base station was considered and optimal beamvectors and power allocation were obtained for the uplink and downlink. In =-=[18, 19]-=-, the same problem was solved under the powerful framework of convex optimization theory which allows the introduction of additional constraints (for example to control the dynamic range of the transm... |

17 | Transmitter and Receiver Optimization in Multicarrier CDMA Systems
- Lok, Wong
- 2000
(Show Context)
Citation Context ... due to the nonconvexity of the problem (a suboptimal iterative optimization approach was taken). A similar signal model arising from a singleantenna multiuser multicarrier CDMA system was treated in =-=[21]-=-, in which a suboptimal (due to the 1snonconvexity of the problem) gradient-type algorithm was used. In [22], a single-antenna multiuser CDMA system was characterized in terms of user capacity by opti... |

14 | Multichannel signal processing for data communications in the presence of crosstalk
- Honig, Steiglitz, et al.
- 1990
(Show Context)
Citation Context |

12 |
A unified framework for communications through MIMO channels
- Palomar
(Show Context)
Citation Context ...1) a H i Rni ai where Rni � HBBHHH + Rn − HbibH i HH is the interference-plus-noise covariance matrix seen by the ith substream. It turns out that the A given in (8) also maximizes each of the SIN=-=R’s [15, 34] (-=-although an arbitrary scaling factor on each ai could be included) and the resulting SINR is SINRi = b H i H H R −1 ni Hbi. (12) Using the matrix inversion lemma, 6 the ith diagonal element of (10) ... |

7 |
Channel models for Hiperlan/2 in different indoor scenarios
- ETSI
- 1998
(Show Context)
Citation Context ... on the multicarrier modulation OFDM (64 carriers are used in the simulations). The frequency selectivity of the channel is modeled using the power delay profile type C for HIPERLAN/2 as specified in =-=[39]-=-, which corresponds to a typical large open space indoor environment for NLOS13 conditions with 150ns average r.m.s. delay spread and 1050ns maximum delay (the sampling period is 50ns [27]). The spati... |

3 |
MIMO channel characterisation,” Deliverable D2 V1.1 of IST-1999-11729 METRA project (available at http://www.ist-metra.org
- Schumacher, Kermoal, et al.
- 2001
(Show Context)
Citation Context ...ions with 150ns average r.m.s. delay spread and 1050ns maximum delay (the sampling period is 50ns [27]). The spatial correlation of the MIMO channel is modeled according to the Nokia model defined in =-=[40]-=- (for the uplink) as specified by the correlation matrices of the envelope of the channel fading at 11 In a practical implementation, care has to be taken so that the problem does not become unfeasibl... |

2 | Optimum joint transmitreceive linear processing for vectored DSL transmission with QoS requirements
- Palomar, Lagunas, et al.
(Show Context)
Citation Context ...the problem. Manuscript received May 10, 2002; revised Feb. 24, 2003. Part of the work was presented at the 36th Asilomar Conference on Signals, Systems & Computers, Pacific Grove, CA, Nov. 3-6, 2002 =-=[23]-=- . This work was partially supported by the Spanish Government (CICYT) TIC2000-1025, TIC2001-2356, TIC2002-04594, FIT-070000-2000-649 (MEDEA + A105 UniLAN); the Catalan Government (DURSI) 1999FI 00588... |

2 |
and suboptimal transmit beamforming,” in Handbook of Antennas in
- “Optimal
- 2002
(Show Context)
Citation Context ...tion were obtained for the uplink and downlink under QoS constraints in terms of 1053-587X/04$20.00 © 2004 IEEE1180 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 52, NO. 5, MAY 2004 SINR. In [18] and =-=[19]-=-, the same problem was solved under the powerful framework of convex optimization theory, which allows the introduction of additional constraints (for example to control the dynamic range of the trans... |

1 |
and suboptimal transmit beamforming
- “Optimal
- 2001
(Show Context)
Citation Context ...re geographically distributed. In [17], a multiuser scenario with a multielement base station was considered and optimal beamvectors and power allocation were obtained for the uplink and downlink. In =-=[18, 19]-=-, the same problem was solved under the powerful framework of convex optimization theory which allows the introduction of additional constraints (for example to control the dynamic range of the transm... |

1 |
Joint trasnmitter receiver diversity for efficient space division multiaccess
- Chang, Tassiulas, et al.
- 2002
(Show Context)
Citation Context ...theory which allows the introduction of additional constraints (for example to control the dynamic range of the transmitted signal or to increase the robustness against channel estimation errors). In =-=[20]-=-, the problem was generalized to the case of having multiple antennas at both sides of the link, although a global optimal solution was not found due to the nonconvexity of the problem (a suboptimal i... |

1 |
MIMO channel characterization,”, Deliverable D2 V1.1 of IST-1999-11 729 METRA Projecthttp://www.ist-metra.org, Feb. 2001. et al.: OPTIMUM LINEAR JOINT TRANSMIT-RECEIVE PROCESSING 1197 Daniel Pérez Palomar (S’99–M’03) was born in
- Schumacher, Kermoal, et al.
- 1975
(Show Context)
Citation Context ... a 150-ns average r.m.s. delay spread, and a 1050-ns maximum delay (the sampling period is 50 ns [27]). The spatial correlation of the MIMO channel was modeled according to the Nokia model defined in =-=[40]-=- (for the uplink) as specified by the correlation matrices of the envelope of the channel fading at the transmit and receive side. It provides a large open indoor environment with two floors, which co... |