Beruflich Dokumente
Kultur Dokumente
fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS 1
AbstractIn this paper, we consider massive multiple-input MIMO algorithms with multiple receive antennas have been
multiple-output (MIMO) systems for both downlink and uplink proposed [8], [9]. These systems, however, assume that the
scenarios, where three radio units (RUs) connected via one network supports a maximum of three users through a rela-
digital unit (DU) support multiple user equipments (UEs) at the
cell-boundary through the same radio resource, i.e., the same tively small number of transmit antennas.1
time-frequency slot. For downlink transmitter options, the study Massive MIMO systems in multi-cell environments have
considers zero-forcing (ZF) and maximum ratio transmission also been studied in [5], [7], [10], [11]. Multi-cell massive
(MRT), while for uplink receiver options it considers ZF and MIMO is prone to some critical issues. These include pi-
maximum ratio combining (MRC). For the sum rate of each lot contamination, which becomes, in time division duplex
of these, we derive simple closed-form formulas. In the simple
but practically relevant case where uniform power is allocated (TDD) systems, the main capacity-limiting factor, especially
to all downlink data streams, we observe that, for the downlink, when MRT is used. Joint spatial division and multiplexing
vector normalization is better for ZF while matrix normalization is proposed in [12] to employ frequency division duplex
is better for MRT. For a given antenna and user configuration, we (FDD) systems. The authors in [13] proposed a pilot alignment
also derive analytically the signal-to-noise-ratio (SNR) level below algorithm for a cognitive massive MIMO system. The authors
which MRC should be used instead of ZF. Numerical simulations
confirm our analytical results. in [7] investigated downlink performance with MRT and ZF
precoder for a massive MIMO system.2 In [5], the authors
Index TermsMassive MIMO, cell-boundary users, ergodic studied uplink performance with MRC, ZF, and a minimum
achievable rate, matched filter, zero-forcing, normalization, pre-
coding, and combining filter. mean square error (MMSE) filters for massive MIMO. It was
shown that transmit energy can be conserved by the power-
law 1/M with perfect channel state information (CSI)
scaling
I. I NTRODUCTION and 1/ M with imperfect CSI at the base station (BS),
Multiple-input multiple-output (MIMO) wireless communi- where M represents the number of BS antennas. In [11],
cation techniques have evolved from single-user MIMO (SU- the authors showed theoretically and numerically the impact
MIMO) to multi-user MIMO (MU-MIMO) systems [1]. To of pilot contamination and proposed a multi-cell MMSE-
approach the capacity of the MIMO broadcast channel, [2], based precoding algorithm to reduce both intra- and inter-cell
[3] proposed simple zero-forcing (ZF)-based linear algorithms, interference. In [11], MRT precoding was used; the inter-user
where the transmitter and the receivers are equipped with interference is eventually eliminated once the transmitter has
multiple antennas. The authors in [4] intensively investigated a large enough number of antennas.
the optimality of the linear matched type-combining filter and The assumption of an infinite number of antennas at the
they assumed an infinite number of antennas at the receiver. BS for a finite number of users somehow trivializes many
[4] proved that a simple linear beamforming (coordinated problems (e.g., under this limit MRC/MRT have the same
beamforming in the paper) asymptotically approaches the sum performance as ZF). A more meaningful system scaling is
capacity achieved by dirty paper coding (DPC). considered in [10], [15], [16], where the number of antennas
Recently, massive MIMO (a.k.a. large-scale MIMO) has per BS and the number of users both go to infinity with a
been proposed to further maximize network capacity and to fixed ratio. In this case, the results of the infinite number
conserve energy [5], [6], [7]. The authors in [6] proposed of BS antennas per user can be recovered by letting this
massive MIMO systems that use simple linear algorithms ratio become large. This more refined analysis, however,
such as maximum ratio combining (MRC) for the uplink illuminates all the system performance regimes. For example,
and maximum ratio transmission (MRT) for the downlink. in [16] the massive MIMO regime is defined as the regime
To further increase sum rate performances, several network where pilot contamination dominates with respect to multi-
user interference; it is observed that this regime occurs only
This research was supported by the MSIP (Ministry of Science, ICT when the number of BS antennas per user is impractically
and Future Planning), Korea, under the IT Consilience Creative Program large. These conclusions are also reached, independently and
(IITP-2015-R0346-15-1008) supervised by the IITP (Institute for Informa-
tion & Communications Technology Promotion) and ICT R&D program of in parallel, in [10]. In particular, [10] considers a multi-cell
MSIP/IITP.
1 Note that more than three users can be supported if there is a common
Y.-G. Lim and C.-B. Chae are with the School of Integrated Technology,
Yonsei University, Korea. Email: {yglim, cbchae}@yonsei.ac.kr. message, i.e., for a clustered broadcast channel.
G. Caire is with the School of Electrical Engineering and Computer Science, 2 In [14], the authors investigated the performance of MRT and ZF in large-
Technical University of Berlin, Germany. Email: caire@tu-berlin.de. scale antenna systems, but paid little attention to normalization techniques.
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
2 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
architecture formed by small clusters of cooperating BSs. The deterministic and the large number of users is mathe-
authors proposed a system where the users are partitioned matically significant in analyzing the performance of
into homogeneous classes and the downlink MIMO precod- massive MIMO systems. We investigate whether the
ing scheme is optimized for each class. Then, a scheduler signal-to-interference-plus-noise-ratio (SINR) term
optimally allocates the time-frequency transmission resource in the ergodic sum rate is able to approximate near
across the different user classes, yielding an inherent multi- deterministic such as the form of the expectation
mode multi-cell massive MIMO system. One of the main when M goes to infinity in the low or high SNR
findings of [10] is that it is convenient to serve users at the regime. From this approximation of the SINR term,
cell center in a single-cell mode, while users at the cell edge we derive simple approximations for the ergodic
should be served by small cooperative clusters formed by the achievable sum rate of ZF and MRT/MRC at the
closest three neighboring cell/sectors. low and high SNR regimes considering two simple
This paper, motivated by the results in [10], focuses on the power normalization methods at the downlink.
cell-edge user performance-the system bottleneck for both the The derived approximations are accurate and far
uplink and the downlink. The massive MIMO system under simpler to evaluate than the asymptotic expressions
consideration consists of multiple radio units (RUs) connected given in [10], [15], [16], which were obtained
to one another by optical fibers, and further connected to through asymptotic random matrix theory [17] and
a centralized digital unit (DU), as illustrated in Figs. 1(a) usually given in terms of the solution of multiple
and 2. Through the optical fibers, each RU can share data mes- coupled fixed-point equations.3 Thanks to their sim-
sages and channel state information. With respect to all three plicity, the proposed approximations are suitable to
neighboring BSs forming a cluster, the cell-edge users have analyze the low and high SNR regimes. From the
symmetric and spatially isotropic channel statistics. Hence, derived approximations, we investigate a suitable
the system is equivalent to, as shown in Fig. 1(b), a single-cell normalization method of MRT and transceiver mode
massive MIMO system with cell-edge users that are located in selection algorithms. In their asymptotic expressions,
the low signal-to-noise-ratio (SNR) regime (for the high SNR the authors in [7], [14], [16] did not consider normal-
regime, we provide some results to provide a detailed analysis, ization methods; thus the expressions were unable
such as for determining which normalization method is better to classify which normalization method was better.
for MRT). As in the prior work referenced above, we consider Numerical results demonstrate the tightness of our
the performance of linear precoding/filtering schemes such as analysis.4
ZF and MRT for the downlink and ZF and MRC for the uplink. Downlink precoding normalization methods: We
In particular, we consider two possible normalizations of the compare matrix and vector normalization for down-
precoding filters for the downlink, referred to as vector and, link precoding, in the simple and practical case of
respectively matrix normalization. The main contributions of uniform power allocation over all downlink streams.
this paper are as follows: It is well known that these normalizations, with opti-
mal (waterfilling) power allocation and ZF precoding
Tighter ergodic achievable sum rate of ZF and yield identical results [20]. However, with practical
MRT/MRC: Tighter ergodic achievable sum rate of suboptimal power allocation and in the finite antenna
ZF and MRT/MRC: We provide a methodology of regime these normalizations are generally not equiv-
simple but quite accurate approximations while con- alent. Most prior work on multi-user MIMO paid
sidering the property of near deterministc, which is scant attention to this issue. For example, the authors
defined in Section III-E. This approximation method in [21], considered matrix normalization and those
is valid for massive MIMO systems at low/high in [22] considered vector normalization in [22] with
SNR. The effective channel matrix tends to an neither making any mention of why different nor-
identity matrix when MRT/MRC are assumed with malizations were used; [21] is part I of their papers
perfect CSI in massive MIMO systems. This means dealing with linear precoding (ZF/MMSE), and [22]
that the random channels become near deterministic is part II focusing on non-linear precoding (vector
due to the property of the law of large numbers. If perturbation). Thus, the comparison between vector
the number of user is large, the sum of the inter- and matrix normalization remains an open problem.
ference powers cannot become near deterministic, To solve this issue, we find a suitable normalization
which means the sum of those still have randomness method for MRT precoding by using the proposed
and do not converge to zero. The researchers in [5]
concluded that with perfect CSI at the BS and a 3 Since the exact expressions in [18], [19] for the achievable ergodic sum
large M , the performance of a MU-MIMO system rate of MRC and ZF filters in a SU-MIMO system was given in terms of the
with a transmit power per user scaling with M is sum of a gamma function, it was hard to give insights for the performance
of massive MIMO system, such as the relationship among the sum rate,
equal to the performance of a single-input-single- the number of antennas at the BS, and the number of users, compared to
output system. Assuming a large number of users, asymptotic expressions.
4 In fact, the achievable sum rate can be slightly enhanced in a low SNR
the conclusion in [5] does not hold because the sum
regime by using regularized ZF (a.k.a. MMSE). This strategy, however,
of the interference powers does not converge to zero. requires knowing all users noise variances at the transmitter, which requires
From this aspect, considering the condition of near additional feedback. Thus, in this paper we focus on ZF and/or MRT/MRC.
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
Y.-G. LIM et al.: PERFORMANCE ANALYSIS OF MASSIVE MIMO FOR CELL-BOUNDARY USERS 3
II. S YSTEM M ODEL : M ASSIVE MIMO where, Pu and xk denote the transmit power per each user and
the transmit symbol of the k-th user at uplink, respectively.
Consider a massive MIMO system as shown in Figs. 1
and 2. One DU (cloud BS) controls three RUs and K users.
Each RU is connected with one another by optical fibers. A. Downlink
Fig. 1(a) shows that the cloud BS provides a massive MIMO Eq. (1) contains the desired signal, interference, and noise
environment to cell-edge users under the assumption that RU terms. To eliminate the interference term, we use the following
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
4 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
Radio Unit 1
DL y2 These results are mathematically motivated from some random
h2 User 2
UL x2 matrix theorems, shown in Section III.
s 1) Asymptotic downlink sum rate: In Section IV, we derive
DL: F
Radio Unit 2
r UL: W
DL yk the simple and tight approximations of the ergodic achievable
hk User k
xk sum rate of ZF and MRT. We also find that ZF with vector
UL
Radio Unit 3
normalization is better than ZF with matrix normalization
DL yK while MRT with matrix normalization is better than MRT
hK User K
with vector normalization. The proposed ergodic achievable
UL xK
sum rates of ZF with vector normalization at low SNR and
MRT with matrix normalization at low SNR are given by
Pt (M K + 1)
RZFvec , DLL = K log2 1 + ,
Fig. 2: Block diagram of multi-RU massive MIMO system. K
One DU consists of three RUs connected by optical fibers. Pt (M + 1)
RMRTmat , DLL K log2 1 + .
This model is regarded as a single-RU massive MIMO system. Pt (K 1) + K
2) Asymptotic uplink sum rate: We also evaluate the ap-
precoding: proximations of the ergodic achievable sum rate of ZF and
MRC in the uplink case. The proposed ergodic achievable sum
ZF : F = H (H
H H )1 = [ff 1 f 2 f k f K ], rates of ZF at low SNR and MRC at low SNR are given by
MRT : F = H = [ff 1 f 2 f k f K ]
Pu M
RMRC, ULL K log2 1 + ,
where F is a precoding matrix consisting of each column Pu (K 1) + 1
vector f k . RZF, ULL K log2 {1 + Pu (M K + 1)} .
To satisfy the power constraint, we consider two meth-
ods, i.e., vector/matrix normalizations. The normalized trans- 3) Transceiver mode selection algorithm: In Section VI, we
mit beamforming vectors (columns of a precoding ma- propose two transceiver mode selection algorithms from i) a
with vector/matrix normalizations are given as g k =
trix) transmit power and ii) the number of active users perspectives.
f k /( K||ff k ||) and g k = f k /||FF ||F , respectively. Note We explain how much transmit power and/or the number of
that vector normalization imposes equal power per downlink active users are needed for ZF to provide a better sum rate
stream, while matrix normalization yields streams with differ- than MRT/MRC. The thresholds are given in Lemmas 6-10.
ent power. In this paper, to simplify, we do not consider a
power optimization that could yield a complexity problem in III. M ATHEMATICAL M OTIVATIONS AND P RELIMINARIES
very large array antenna systems. In this section, we introduce some mathematical motiva-
1) ZF/MRT with vector normalization: The received signal tions and preliminaries to evaluate an asymptotic analysis for
at the k-th UE can be expressed as follows: network massive MIMO, which will be used in Sections IV
K and V.
p fk X p f`
yk = Pth Tk sk + Pth Tk s` + nk .
K||ff k || `=1,`6=k
K||ff ` || A. Achievable Rate Bound
(3) In this paper, to maximize the achievable sum rate of
2) ZF/MRT with matrix normalization: Similarly, we can downlink/uplink systems, we evaluate the closed forms of each
rewrite the received signal with matrix normalization as such: systems performance. The achievable rates are bounded as
K
follows:
p fk X p f` ( )
yk = Pth Tk Pth Tk
sk + s` + nk . (4) 1 S
||F
F ||F ||F
F ||F log 2 1 + E log 2 1 +
E I+N
`=1,`6=k
S
I +N
B. Uplink S
log2 1 + E
Similar to the downlink system, to eliminate the interference I +N
term, and to maximize the SNR in (2), we use the following by using Jensens Inequality of convex and concave functions
combining filter at the RUs: where S, I, and N represent signal power, interference power,
1 and noise power, respectively. Note that we use these bounds
ZF : W = (H H H ) H = [w w 1 w 2 w k w K ],
only for ZF cases and show that our results, based on the
MRC : W = H = [w w1 w 2 w k w K ] bounds, are more accurate than prior work [5], [7], [16].
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
Y.-G. LIM et al.: PERFORMANCE ANALYSIS OF MASSIVE MIMO FOR CELL-BOUNDARY USERS 5
B. Expectation and Variance of Random Vectors F. Ergodic Achievable Sum Rate of Massive MIMO Systems
Lemma 1: Let h k and h ` (k 6= `) be M 1 vectors at the low/high SNR regime
whose elements are independent identically distributed (i.i.d.) Lemma 4: Let Xv and vi be a norm of a random
complex Gaussian random variables with zero mean and unit vector v (M 1) and the i-th entry of v , respectively,
variance.
n o
i.e., Xv = v12 + v22 + + vM 2
. Since E X1v =
1) E[||hhk ||2 ] = M , n o n o
hk ||2 ] = M . 1 1 1
Var[||h E v2 +v2 ++v2 = E M (v2 +v2 ++v2 )/M M E v2
hkh ` ] = 0,
2) E[h
1 2 M 1 2 M n o { i}
1 1 1 1
Var[hhkh ` ] = M . and E{X = E{v12 +v22 ++vM
, E con-
v} 2
} M E{vi2 } Xv
3) E[||hhk ||4 ] = M 2 + M , 1
verges to E{Xv } as M goes to infinity.
Var[||hhk ||4 ] = 4M 3 + 10M 2 + 6M . In this Lemma, we assume that the desired signal and the
h 2
4) E[|h k ` | ] = M ,
h interference plus noise terms are norms of a random vector
Var[|hhkh ` |2 ] = M 2 + 2M . (M 1), and some of those terms in the low/high SNR
Proof: See Appendix A. regime are near deterministic as satisfying the condition of
(5); thus we assume at least one of the those terms has the
C. Effective Channel
same property of Xv . From the assumption, if S E {S} from
Lemma 2: In massive MIMO systems, the transmit energy (5) with a low/high SNR assumption
o(e.g., Pt Var {Xnv } o 0,
can be conserved by power-scaling law 1/M with perfect CSI.
n
S 1
1 where S = Pt Xv ), then E I+N E {S} E I+N .
1) lim E[ M hk ||2 ] = 1,
||h n o
M 1 E{S}
1
lim Var[ M hk ||2 ] = 0.
||h Also, E {S} E I+N E{I+N } as M goes to infinity
M
n o
1
2) lim E[ M h kh ` ] = 0, using E X1v E{X 1
v}
. Similarly, for I + N converges to
M n o
1 S E{S}
lim Var[ M h kh ` ] = 0. E {I + N } cases, we also obtain E I+N E{I+N } . So
M
Generally, a transceiver uses MRT or MRC in massive we could obtain the following approximation of SINR when
MIMO, which means that the effective channel of the desired M goes to infinity in the low or high SNR regime:
signal becomes one and the interference signal becomes zero
as the number of antennas (M ) goes to infinity, as illustrated: S E(S)
E . (6)
1 a.s.
I +N E(I + N )
H H I K , as M .
M
From (6), the lower bound of the ergodic sum rate is the
D. Signal and Interference Power same as the upper bound of the ergodic sum rate. Thus we
Lemma 3: In a similar way, the expectation and the variance could also get the following approximation of the ergodic sum
of the signal and the interference power are given by rate when M goes to infinity in the low or high SNR regime:
1) lim E[ M12 ||hhk ||4 ] = 1,
M S E(S)
lim Var[ M12 ||h
hk ||4 ] = 0. E log2 1 + log2 1 + .
M I +N E(I + N )
hkh ` |2 ] = 0,
2) lim E[ M12 |h
M
hkh ` |2 ] = 0.
lim Var[ M12 |h
M
Note that the terms M1 1
hk ||4 and M
||h hk h` |2 do not converge
|h
to M and zero, respectively, as M goes to infinity since their G. Arithmetic-Geometric Inequality
variance does not go to zero, which means that they still have
Lemma 5: Let b1 , b2 , ..., bK be random variables. We can
randomness.
obtain the following inequality through Arithmetic-geometric
E. Chebyshevs Inequality Inequality defined in [23]:
2
Let X be a random variable with variance X , c and be K !
1
X 1 1
scalars, and Y = c X be a random variable with variance log2 1 + K log2 1 + PK .
Y2 = c12 X
2
, respectively. If c is not very large in comparison Kbk k=1 bk
k=1
with E{X}, but c2 >> X 2
, then Y = 1c X is near determin-
istic (nearly all random variables close to the mean), i.e., Proof:
2 2
P [|Y E{Y }| > ] Y2 = 2Xc2 . For fixed > 0, K !
1 X 1 1
X2 log2 1 + log2 1 + 1 PK
if 0, then Y E{Y } (5) K Kbk k=1 Kbk
k=1 K
c2 !
K
with high probability. We use (5) to approximate a SINR term X 1 1
log2 1 + K log2 1 + PK .
in ergodic sum rate expressions. Kbk
k=1 k=1 bk
As an example, to analyze the rate in the high SNR regime,
hk ||2 ] = P12 M converges to zero, then P1t ||h
if Var[ P1t ||h hk ||2
t
1 2 1
converges to Pt E{|h h k | } = Pt M . This is a simple application of Jensens Inequality.
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
6 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
IV. A SYMPTOTIC D OWNLINK S UM R ATE FOR By using Lemma 5, (9) can further be expressed as
C ELL - BOUNDARY U SERS ( !)
1 2
In this section, we derive the achievable rate bounds, and K log2 1 + E Pt
||F
F ||F
show which normalization method is suitable for ZF- and
MRT-type precoding at the downlink. We assume that the cell- 1
= K log2 1 + E Pt
boundary users are in the low SNR regime (Pt Var {Xv } 0) H H )1 )
tr((H
( !)
and that the cell center users are in the high SNR regime 1
( P1t Var {Xv } 0); we also assume that the desired power = K log2 1 + E Pt PK
k=1 ||ff k ||2
term or the interference power term becomes near determin-
istic in the low/high SNR regime. Based on our analytical 1
K log2 1 + E Pt .
results, we will also show which precoding technique is K||ff k ||2
desired for cell-boundary users. So the second upper bound (without an expectation form) of
matrix normalization in the ZF case can be given by
A. Ergodic Performance U2 Pt (M K + 1)
RZFmat , DL = K log2 1 + . (10)
1) Achievable rate bounds for ZF precoding: K
i) Lower bounds: The lower bound of the ergodic sum rate 2) Achievable rate for MRT precoding:
for the ZF precoding is well known, as follows [7]: i) Vector normalization-low SNR regime: From (3), we can
derive the ergodic achievable sum rate of vector normalization
L L Pt (M K)
RZFvec , DL = RZFmat , DL = K log2 1 + in low SNR as follows:
K
Pt M
RMRTvec , DLL K log2 1 + . (11)
using the property of Wishart matrices [17]. Pt (K 1) + K
ii) Vector normalization-upper bound: From (3), we can Proof: See Appendix B1.
derive the SINR of the upper bound of vector normalization ii) Vector normalization-high SNR regime: Similarly, we
in the ZF case, as given by can get the ergodic achievable sum rate of vector normalization
T f k 2 in high SNR as follows:
S
P t h
h k K||ff k ||
E =E Pt (M + 1)
I +N Pt PK T f ` 2
RMRTvec , DLH K log2 1 + . (12)
hk K||ff || + 1 Pt (K 1) + K
`=1,`6=k h `
Proof: See Appendix B2.
Pt
=E iii) Matrix normalization-low/high SNR regime: From (4),
K||ff k ||2
(a) Pt (M K + 1)
we can evaluate the ergodic achievable sum rate of matrix nor-
= (7) malization in low/high SNR by using the following formation:
K
Pt (M + 1)
n (a)o results from the diversity order of ZF
where RMRTmat , DLL/H K log2 1 + . (13)
Pt (K 1) + K
E ||ff 1k ||2 = M K + 1 , as shown in [24]. From
(7), the upper bound of vector normalization in the ZF case Proof: See Appendix B3 and B4.
can be represented as
B. Comparison between Vector and Matrix Normalizations
Pt (M K + 1)
RU ZFvec , DL = K log 2 1 + . 1) Performance comparison of ZF: To find which normal-
K H 1
ization technique is better in ZF, we let bk = [(H H P)t ]kk in
iii) Matrix normalization-upper bound: From (4), the SINR Lemma 5 directly.
of the upper bound of matrix normalization in the ZF case can K
be expressed as
X Pt
log2 1 +
K[(H H H )1 ]kk
T f k 2
k=1
S
h
Pt h k ||FF ||F
!
E =E Pt
I +N T f ` 2
K log2 1 + PK
Pt PK H H )1 ]kk
hk ||FF ||F + 1 k=1 [(H
`=1,`6=k h
(
1 2
) RZFvec , DL RZFmat , DL (14)
= E Pt . (8) K
||F
F ||F H H )1 ]kk = ||ff k ||2 and H H )1 ]kk =
P
where [(H k=1 [(H
PK 2 2
k=1 k|f
f k || = ||F
F || . From (14), since the desired power
From (8), the first upper bound (with an expectation form) of and the interference power are one and zero respectively, the
matrix normalization in the ZF case can be represented as power normalization per user only affects the performance of
ZF. From the perspective of Jesens Inequality, the sum rate
( !)
1 2
U1
RZFmat , DL = K log2 1 + E Pt . (9) with the different power allocation per user is the upper bound
||F
F ||F
of the sum rate with the same power allocation per user at the
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
Y.-G. LIM et al.: PERFORMANCE ANALYSIS OF MASSIVE MIMO FOR CELL-BOUNDARY USERS 7
instant channel and arbitrary SNR. We can conclude that, in V. A SYMPTOTIC U PLINK S UM R ATE FOR
the ZF case, vector normalization is always better than matrix C ELL -B OUNDARY U SERS
normalization. We have focused on a downlink scenario with a sum power
2) Performance comparison of MRT: From (11)-(13), a constraint. In this section, we investigate an uplink case, where
comparison of the ergodic achievable sum rate is given by each user has its own power constraint. From (2), the ergodic
achievable sum rate for the uplink, RUL , is
RMRTmat , DLL ' RMRTvec , DLL (15) "K ( )#
RMRTmat , DLH RMRTvec , DLH (16)
X w Tk h k |2
Pu |w
RUL = E log2 1 + PK .
k=1 Pu `=1,`6=k |w w Tk h ` |2 + ||w
w k ||2
at low and high SNR, respectively. It is well known that (18)
approach methods that maximize the desired signal are better
than those that mitigate the interference power at low SNR. From (18), we can derive the ergodic achievable uplink sum
The effective desired channel gain can be maximized with rate of MRC, RMRC, UL , as follows:
MRT. The desired power per user scales down in proportion RMRC, UL
to its power (1/||ff k ||2 ) in the vector normalization while the "K ( )#
X hk ||4
Pu ||h
desired power per user scales down in the same proportion =E log2 1 + PK .
over all users (1/||F F ||2F ) in the matrix normalization. This k=1 Pu `=1,`6=k hkh ` |2 + ||h
|h hk ||2
means that the gains of better channels with vector normaliza- (19)
tion tend to scale down more than the gains of better channels We approximate the ergodic achievable sum rate of MRC as
with matrix normalization. This yields the difference of the follows:
desired power term of (15) and (16) at low SNR. Therefore, i) High SNR regime:
we confirm that, for MRT precoding, matrix normalization
Pu (M + 1)
is always better than vector normalization at low SNR. We RMRC, ULH K log2 1 + . (20)
Pu (K 1) + 1
conclude, however, that there is marginal performance gap
between vector normalization and matrix normalization at high ii) Low SNR regime:
SNR. Pu M
RMRC, ULL K log2 1 + . (21)
Pu (K 1) + 1
Proof: See Appendix C
C. Ergodic Achievable Sum Rate for Cell-boundary Users
Similar to ZF precoding, the ergodic sum rate for ZF for
with the Best Normalization Method
uplink at low SNR, RZF, ULL , is
As explained in Section IV-B, we conclude that the suitable
Pu
normalization methods are vector normalization for ZF and RZF, ULL RU ZF, UL = K log 2 1 + E
H H )1 ]k,k
[(H
matrix normalization for MRT. We assume that the transmit
power (Pt ) is small for cell-boundary users (low SNR regime). = K log2 {1 + Pu (M K + 1)} .
Using the property of ZF precoding, the ergodic achievable VI. T RANSCEIVER M ODE S ELECTION
sum rate of ZF is represented as A. Algorithm
S In this section, we propose two transceiver mode selection
E log2 1 + algorithms from i) transmit power and ii) the number of
I +N
( !) active users perspectives. To provide a mathematically simple
hTk g k |2
Pt |h
= E log2 1 + PK solution, we first introduce Lemma 6 and Lemma 7 that use a
Pt `=1,`6=k |h hTk g ` |2 + 1 power threshold as follows:
hTk g k |2 Lemma 6: The power threshold to select a better precoder
= E log2 1 + Pt |h
(b)
for downlink is given by
hTk g k |2 )
log2 1 + Pt E(|h (17) K2
Pth, DL = . (22)
where (b) results from (5) with Pt Var |h
T 2
hk g k | 0. Eq. (17) (K 1)(M K + 1)
indicates that the achievable sum rate of ZF precoding can If the RUs have more transmit power than the power threshold
approach its upper bounds at low SNR by (5). Thus, the Pth, DL , the ZF precoder provides a better sum rate perfor-
ergodic achievable sum rate of ZF with vector normalization mance.
at low SNR is given by Proof: To derive (22) for cell-boundary users, we use
the low SNR approximation for ZF and MRT. By letting
Pt (M K + 1) RZFvec , DLL RMRTmat , DLL , we can get (22) as follows:
RZFvec , DLL K log2 1 + .
K
RZFvec , DLL RMRTmat , DLL 0
From (13), we find the ergodic achievable sum rate of matrix Pt (M K + 1) Pt (M + 1)
normalization in low SNR: 0
K Pt (K 1) + K
K2
Pt (M + 1)
RMRTmat , DLL K log2 1 + . Pt Pth, DL = .
Pt (K 1) + K (K 1)(M K + 1)
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
8 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
'LIIHUHQFH
If each UE has larger transmit power per user than Pth, UL , the
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
Y.-G. LIM et al.: PERFORMANCE ANALYSIS OF MASSIVE MIMO FOR CELL-BOUNDARY USERS 9
6LPXODWLRQYHFWRU 6LPXODWLRQ
6LPXODWLRQPDWUL[ 3URSRVHGDQDO\VLV
3URSRVHGDQDO\VLVYHFWRU /RZHUERXQG>@
3URSRVHGDQDO\VLVPDWUL[ &ORVHGIRUP>@
6XP5DWHESV+]
6XP5DWHESV+]
1XPEHURI8VHUV 1XPEHURI8VHUV
(a) Achievable rate of ZF at low SNR. (a) Achievable rate of MRC at high SNR.
6LPXODWLRQ
3URSRVHGDQDO\VLV
/RZHUERXQG>@
&ORVHGIRUP>@
6XP5DWHESV+]
6XP5DWHESV+]
6LPXODWLRQYHFWRU
6LPXODWLRQPDWUL[
3URSRVHGDQDO\VLVYHFWRU
3URSRVHGDQDO\VLVPDWUL[
1XPEHURI8VHUV 1XPEHURI8VHUV
(b) Achievable rate of MRT at low SNR. (b) Achievable rate of MRC at low SNR.
Fig. 4: Achievable rate vs. the number of cell-boundary users, Fig. 5: Achievable rate vs. the number of cell-boundary users,
where M = 24, K = [1, 24], and total SNR = 13.8 dB. where M = 24, K = [1, 24], and user SNR = (a) 13.8 dB, (b)
13.8 dB.
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
10 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
6XP5DWHESV+]
to infinity. This differs from the result in [5] (RMRC, UL = 3WK
M log2 (1+Pu, sum )) because the authors in [5] did not consider
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
Y.-G. LIM et al.: PERFORMANCE ANALYSIS OF MASSIVE MIMO FOR CELL-BOUNDARY USERS 11
. FURVV
057
6XP5DWHESV+]
6XP5DWHESV+]
=)
057PDWUL[QRUPDOL]DWLRQ
=)PDWUL[QRUPDOL]DWLRQ
057PDWUL[QRUPDOL]DWLRQ 057YHFWRUQRUPDOL]DWLRQ
=)YHFWRUQRUPDOL]DWLRQ =)YHFWRUQRUPDOL]DWLRQ
1XPEHURI8VHUV 1XPEHURI8VHUV
05&
=)
057
. FURVV
6XP5DWHESV+]
6XP5DWHESV+]
=)
057PDWUL[QRUPDOL]DWLRQ
=)PDWUL[QRUPDOL]DWLRQ
057YHFWRUQRUPDOL]DWLRQ
=)YHFWRUQRUPDOL]DWLRQ
1XPEHURI8VHUV 1XPEHURI8VHUV
Fig. 7: Achievable rate vs. the number of cell-boundary users, Fig. 8: Achievable rate vs. the number of cell-boundary users,
where M = 24, K = [1, 24], (a) Pt = Pcross, DL = 7.6 dB, where M = 24, K = [1, 24], and total SNR = (a) 0 dB, (b)
(b) Pu = Pcross, UL = 13.6 dB. 5 dB.
[5]. The legend, Simulation, indicates the ergodic achievable In Fig. 6(a), we illustrate the achievable sum rate of MRT
uplink sum rate of MRC (19) while the proposed analysis in precoding and ZF precoding at downlink. Fig. 6(b) illustrates
Figs. 5(a) and 5(b) indicates the approximation shown in (20) the achievable sum rate of MRC and ZF at uplink (Pth =
and (21), respectively.
n Note that the lower
o bound in [5] is given 6.2 dB and 7 dB are calculated by Lemmas 6 and 7 with
Pu (M 1) M = 24 and K = 20 at downlink and uplink, respectively).
by RL MRC = K log 2 1 + Pu (K1)+1 and the closed form in
n o Note that a cross point of MRT (or MRC) curve and ZF
[16] is given by RMRC = K log2 1 + PPu K+1 uM
. In addition, curve is the power threshold. It is well known that MRT/MRC
the sum rate approaches to M when Mn = K at high o SNR, performs better than ZF at low SNR but a borderline between
Pu, sum
and the sum rate approaches to M log2 1 + Pu, sum +1 when MRT/MRC region and ZF region has been unknown and
M = K at low SNR. This shows that the results of (27) and numerically provided [18]. We can easily find this borderline
(28) could hold even M is finite. with the proposed simple equation. Figs. 7(a) and 7(b) show
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
12 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
Precoding normalization technique
ZF Vector normalization Matrix normalization
MRT Matrix normalization Vector normalization
6XP5DWHESV+]
Pth Kcross
K2 Pt (M +1)
Downlink (K1)(M K+1) 1+Pt
1 1
Uplink M K+1
M +1 Pu
6LPXODWLRQ TABLE III: Desired technique in network massive MIMO
3URSRVHGDQDO\VLV
/RZHUERXQG>@>@
systems. ZF (if K Kcross ) and MRT/MRC (if K Kcross ).
&ORVHGIRUP>@
Kcross Precoding technique
1XPEHURI$QWHQQDV Cell-center Large Zero-forcing
Cell-boundary Small MRT/MRC
(a) Achievable rate of MRT at low SNR.
6LPXODWLRQ Section VI, in the low SNR regime, using MRT precoding is
3URSRVHGDQDO\VLV
/RZHUERXQG>@ generally better when the number of active users is larger than
&ORVHGIRUP>@ Kcross, DL . Also, we realize through Figs. 8(a) and 8(b) that as
SNR decreases, Kcross, DL shifts to the left. This means that
the cloud BS could determine a precoding by Kcross, DL at low
6XP5DWHESV+]
SNR.
The performance of MRT/MRC increases as K increases
while the performance of ZF decreases as K increases. This
is because the ergodic sum rate of ZF at M = K goes to a
very small constant at downlink, as shown in (26). Similarly,
at uplink, the ergodic sum rate of ZF at M = K could also
close to a very small value since its lower bound is zero.
Figs. 9(a) and 9(b) show that the achievable sum rate
of MRT/MRC at low SNR. To predict the tightness of our
analysis for infinite M , we compare the simulation result of
our analysis with that of the closed form in [16] and the lower
1XPEHURI$QWHQQDV bound in [7] and [14] in the downlink scenario with a large
number of antennas at the BS (up to M = 100). Similarly,
(b) Achievable rate of MRC at low SNR. we compare the simulation result of our analysis with that
Fig. 9: Achievable rate vs. the number of antennas at the BS, of the closed form in [16] and the lower bound in [5] in the
where K = 10, M = [10, 100], and total/user SNR = -20 dB. uplink scenario. Both numerical results of MRT/MRC show
that, when M goes to infinity, our analyses are quite tighter
than the work given in [5], [7], [14], [16]. Note that the closed
form in [16] is very tight when M = 40 but as M increases not
that MRT/MRC is always better than ZF at very low SNR as tight as our analysis in the low SNR regime and the uplink
regardless of K. Used for simulations were Pt = Pcross, DL = scenario. The computation complexity is the same as the lower
7.6 dB at downlink and Pu = Pcross, UL = 13.6 dB at bounds in [5], [7], [14] but the accuracy is better than the
uplink, and M = 24. This result verifies Lemma 8. We closed form in [16]. Therefore, we conclude that the proposed
summarize also our conclusions in Tables I, II, and III. ergodic sum rate is tighter and simpler than the previous work.
In Fig. 8(a), we also compare the achievable sum rates of ZF
precoding with MRT precoding when the total transmit SNR
VIII. C ONCLUSIONS
is 0 dB. It shows that ZF with vector normalization is better
while MRT with matrix normalization is better for achieving In this paper, we proposed massive MIMO systems support-
higher sum rates. Fig. 8(b) illustrates the achievable sum rates ing multiple cell-boundary UEs. For precoding designs, we
of ZF- and MRT-precoding with 5 dB transmit SNR. The first derived the achievable sum rate bounds of zero-forcing
result is similar to that found in Fig. 8(a). As mentioned in (ZF) and the approximation of the ergodic achievable sum
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
Y.-G. LIM et al.: PERFORMANCE ANALYSIS OF MASSIVE MIMO FOR CELL-BOUNDARY USERS 13
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
14 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
where (a) results from (32).5 Approximation (b) and equality C. Proof of Ergodic Sum Rate of MRC
(c) can be obtained by Lemmas 4 and 1, respectively. 1) Proof of (20):
2) Proof of (12):
RMRC, ULH
" ( )#
4
Pu ||hhk ||4
Pt ||||hKh
K hk ||
X
hk ||2
= E K log2 1 + PK
RMRTvec , DLH = E log2 1 + PK h
|h h ` |2
Pu `=1,`6=k |h hkh ` |2 + ||hhk ||2
k=1
P t `=1,`6=k || Kh
k
h` ||2
+ 1 " ( )#
hk ||4
||h
||h
Pt K||h
hk || 4
= E K log2 1 + PK
`=1,`6=k |hhkh ` |2 + P1u ||hhk ||2
hk ||2
= KE log2 1 + PK h
|h k ` |
h 2
" ( )#
Pt `=1,`6=k K||hh` ||2 + 1
||hhk ||4
" ( hk ||4
)# E K log2 1 + PK
(d) Pt ||hKM `=1,`6=k |hhkh ` |2 + P1u M
KE log2 1 + PK
hkh ` |
|h 2
Pt `=1,`6=k KM +1 Pu (M + 1)
K log 2 1 + . (36)
(e)
Pt (M + 1)
Pu (K 1) + 1
K log2 1 + using (32), (33), Lemmas 1 and 4.
Pt (K 1) + K
2) Proof of (21):
where (d) results from (33) and (e) can be obtained by RMRC, ULL
Lemmas 1 and 4. " ( )#
Pu ||hhk ||4
3) Proof of (13): in the low SNR regime = E K log2 1 + PK
Pu `=1,`6=k |h hkh ` |2 + ||hhk ||2
RMRTmat , DLL P u ||h
h k ||2
4
= E K log2 1 +
Pt ||h k ||
h |h
hkh ` | 2
K P K
X
||H
H ||F2
P u `=1,`6=k ||h hk || 2 + 1
= E log2 1 + PK h
|h 2
kh ` |
k=1
Pt `=1,`6=k ||H H ||F2 + 1 Pu M
" ( )# E K log2 1 +
K |h
h h |2
hk ||4
Pt ||h Pu `=1,`6=k ||hhkk ||` 2 + 1
P
= KE log2 1 + PK
Pt `=1,`6=k |h hkh ` |2 + ||H H ||2F
Pu M
" ( )# K log2 1 +
(f ) Pt (M 2 + M ) Pu (K 1) + 1
KE log2 1 + PK
Pt `=1,`6=k M + ||H H ||2F using the same methods as in (36).
(g) Pt (M + 1) R EFERENCES
K log2 1 +
Pt (K 1) + K
[1] D. Gesbert, M. Kountouris, R. W. Heath, Jr., C.-B. Chae, and T. Salzer,
Shifting the MIMO paradigm: From single user to multiuser commu-
where (f ) can be obtained by using (34) directly. (g) can also nications, IEEE Sig. Proc. Mag., vol. 24, no. 5, pp. 3646, Oct. 2007.
be derived by Lemma 4 and E(||H H ||2F ) = M K. [2] Q. Spencer, A. L. Swindlehurst, and M. Haardt, Zero-forcing methods
for downlink spatial multiplexing in multiuser MIMO channels, IEEE
4) Proof of (13): in the high SNR regime Trans. Sig. Proc., vol. 52, pp. 462471, Feb. 2004.
[3] C.-B. Chae, D. Mazzarese, N. Jindal, and R. W. Heath, Jr., Coordinated
beamforming with limited feedback in the MIMO broadcast channel,
RMRTmat , DLH IEEE Jour. Select. Areas in Comm., vol. 26, no. 8, pp. 15051515, Oct.
hk ||4
2008.
X K Pt ||h H ||2F
||H
[4] C.-B. Chae and R. W. Heath, Jr., On the optimality of linear multiuser
= E log2 1 + PK |h
hkh ` | 2
MIMO beamforming for a two-user two-input multiple-output broadcast
k=1
P t `=1,`6=k ||H H ||2F
+ 1 system, IEEE Sig. Proc. Lett., vol. 16, no. 2, pp. 117120, Feb. 2009.
" ( )# [5] H. Q. Ngo, E. G. Larsson, and T. L. Marzetta, Energy and spectral
||hhk ||4 efficiency of very large multiuser MIMO systems, IEEE Trans. Comm.,
= KE log2 1 + PK vol. 61, no. 4, pp. 14361449, April 2012.
hkh ` |2 + P1t ||H
`=1,`6=k |h H ||2F [6] T. L. Marzetta, Noncooperative cellular wireless with unlimited num-
" ( )# bers of base station antennas, IEEE Trans. Wireless Comm., vol. 9,
(h) ||hhk ||4 no. 11, pp. 35903600, Nov. 2010.
KE log2 1 + PK [7] F. Rusek, D. Persson, B. K. Lau, E. G. Larsson, T. L. Mazetta, O. Edfors,
hk h` |2 + P1t M K
`=1,`6=k |h and F. Tufvesson, Scaling up MIMO: Opportunities and challenges with
(i)
Pt (M + 1)
large arrays, IEEE Sig. Proc. Mag., vol. 30, no. 1, pp. 4060, Jan. 2013.
K log2 1 + [8] C.-B. Chae, S. Kim, and R. W. Heath, Jr., Network coordinated beam-
Pt (K 1) + K forming for cell-boundary users: Linear and non-linear approaches,
IEEE Jour. Select. Topics in Sig. Proc., vol. 3, no. 6, pp. 10941105,
Dec. 2009.
where (h) results from (35) while (i) can be obtained by [9] C.-B. Chae, I. Hwang, R. W. Heath, Jr., and V. Tarokh, Interference
Lemmas 1 and 4 as well. aware-coordinated beamforming in a multi-cell system, IEEE Trans.
Wireless Comm., vol. 11, no. 10, pp. 36923703, Oct. 2012.
[10] H. Huh, G. Caire, H. C. Papadopoulos, and S. A. Ramprashad, Achiev-
5 Since Var[||h hk ||2 ](= M ) < Var|hhk h ` |2 ](= M 2 + 2M ) from Lemma 1, ing massive MIMO spectral efficiency with a not-so-large number of
hk ||2 ] fastly converges to zero than Var[Pt |h
Var[Pt ||h hk h ` |2 ] at low SNR. We antennas, IEEE Trans. Wireless Comm., vol. 11, no. 9, pp. 32263239,
use this property for the rest of the proofs. Sep. 2012.
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TWC.2015.2460751, IEEE Transactions on Wireless Communications
Y.-G. LIM et al.: PERFORMANCE ANALYSIS OF MASSIVE MIMO FOR CELL-BOUNDARY USERS 15
[11] J. Jose, A. Ashikhmin, T. L. Marzetta, and S. Vishwanath, Pilot Chan-Byoung Chae (S06 - M09 - SM12) is
contamination and precoding in multi-cell TDD systems, IEEE Trans. an Assistant Professor in the School of Integrated
Wireless Comm., vol. 10, no. 8, pp. 26402651, Aug. 2011. Technology, College of Engineering, Yonsei Univer-
[12] J. Nam, J.-Y. Ahn, A. Adhikary, and G. Caire, Joint spatial division sity, Korea. He was a Member of Technical Staff
and multiplexing: Realizing massive MIMO gains with limited channel PLACE (Research Scientist) at Bell Laboratories, Alcatel-
state information, in Proc. of Conf. on Info. Scien. and Systems, March PHOTO Lucent, Murray Hill, NJ, USA from 2009 to 2011.
2012, pp. 16. HERE Before joining Bell Laboratories, he was with the
[13] M. Filippou, D. Gesbert., and H. Yin, Decontaminating pilots in School of Engineering and Applied Sciences at
cognitive massive MIMO networks, in Proc. of Int. Symp. on Wireless Harvard University, Cambridge, MA, USA as a Post-
Comm. Systems, Aug. 2012, pp. 816820. Doctoral Research Fellow. He received the Ph.D.
[14] H. Yang and T. L. Marzetta, Performance of conjugate and zero-forcing degree in Electrical and Computer Engineering from
beamforming in large-scale antenna systems, IEEE Jour. Select. Areas The University of Texas (UT), Austin, TX, USA in 2008, where he was a
in Comm., vol. 31, no. 2, pp. 172179, 2013. member of the Wireless Networking and Communications Group (WNCG).
[15] H. Huh, A. M. Tulino, and G. Caire, Network MIMO with linear Prior to joining UT, he was a Research Engineer at the Telecommunications
zero-forcing beamforming: Large system analysis, impact of channel R&D Center, Samsung Electronics, Suwon, Korea, from 2001 to 2005.
estimation, and reduced-complexity scheduling, IEEE Trans. Info. Th., While having worked at Samsung, he participated in the IEEE 802.16e
vol. 58, no. 5, pp. 29112934, May 2012. standardization, where he made several contributions and filed a number of
[16] J. Hoydis, S. ten Brink, and M. Debbah, Massive MIMO in the UL/DL related patents from 2004 to 2005. His current research interests include
of cellular networks: How many antennas do we need? IEEE Jour. capacity analysis and interference management in energy-efficient wireless
Select. Areas in Comm., vol. 31, no. 2, pp. 160171, Feb. 2013. mobile networks and nano (molecular) communications. He has served/serves
[17] A. M. Tulino and S. Verdu, Random matrix theory and wireless as an Editor for the IEEE T RANS . ON W IRELESS C OMM ., IEEE T RANS . ON
communications, Foundations and Trends in Comm. and Info. Th., S MART G RID and IEEE/KICS J OUR . C OMM . N ETS. He is a Guest Editor
vol. 1, no. 1, 2004. for the IEEE J OUR . S EL . A REAS IN C OMM . (special issue on molecular,
[18] R. H. Y. Louie, M. R. McKay, and I. B. Collings, Maximum sum- biological, and multi-scale comm.). He is an IEEE Senior Member.
rate of MIMO multiuser scheduling with linear receivers, IEEE Trans. Dr. Chae was the recipient/co-recipient of the IEEE INFOCOM Best
Comm., vol. 57, no. 11, pp. 35003510, Nov. 2009. Demo Award (2015), the IEIE/IEEE Joint Award for Young IT Engineer
[19] C.-J. Chen and L.-C. Wang, Performance analysis of scheduling in of the Year (2014), the Haedong Young Scholar Award (2013), the IEEE
multiuser MIMO systems with zero-forcing receivers, IEEE Jour. Signal Processing Magazine Best Paper Award (2013), the IEEE ComSoc AP
Select. Areas in Comm., vol. 25, no. 7, pp. 14351445, Sep. 2007. Outstanding Young Researcher Award (2012), the IEEE VTS Dan. E. Noble
[20] G. Caire, N. Jindal, M. Kobayashi, and N. Ravindran, Multiuser MIMO Fellowship Award (2008), the Gold Prize (1st) in the 14th/19th Humantech
achievable rates with downlink training and channel state feedback, Paper Contests, and the KSEA-KUSCO scholarship (2007). He also received
IEEE Trans. Info. Th., vol. 56, no. 6, pp. 28452866, June 2010. the Korea Government Fellowship (KOSEF) during his Ph.D. studies.
[21] C. Peel, B. Hochwald, and A. Swindlehurst, A vector-perturbation
technique for near-capacity multiantenna multiuser communication-part
I: channel inversion and regularization, IEEE Trans. Comm., vol. 53,
no. 1, pp. 195202, Jan. 2005.
[22] B. Hochwald, C. Peel, and A. Swindlehurst, A vector-perturbation
technique for near-capacity multiantenna multiuser communication-part
II: perturbation, IEEE Trans. Comm., vol. 53, no. 3, pp. 537544, March
2005.
[23] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and
Products. Academic Press, 2007. Guiseppe Caire (S92 - M94 - SM03 - F05) was
[24] K. K. Wong and Z. Pan, Array gain and diversity order of multiuser born in Torino, Italy, in 1965. He received the B.Sc.
MISO antenna systems, Int. J. Wireless Inf. Networks, vol. 2008, no. 15, in Electrical Engineering from Politecnico di Torino
pp. 8289, May 2008. (Italy), in 1990, the M.Sc. in Electrical Engineering
[25] S. Sesia, I. Toufik, and M. Baker, LTE, The UMTS Long Term Evolution: PLACE from Princeton University in 1992 and the Ph.D.
From Theory to Practice. Wiley, 2012. PHOTO from Politecnico di Torino in 1994. He was a re-
HERE cipient of the AEI G.Someda Scholarship in 1991,
has been with the European Space Agency (ESTEC,
Noordwijk, The Netherlands) from May 1994 to
February 1995, was a recipient of the COTRAO
Scholarship in 1996 and of a CNR Scholarship
in 1997. He has been visiting Princeton University in Summer 1997 and
Sydney University in Summer 2000. He has been Assistant Professor in
Telecommunications at the Politecnico di Torino, Associate Professor at
the University of Parma, Italy, Professor with the Department of Mobile
Communications at the Eurecom Institute, Sophia-Antipolis, France, and a
professor of Electrical Engineering with the Viterbi School of Engineering,
University of Southern California, Los Angeles, CA. He is now a professor
with the School of Electrical Engineering and Computer Science, Technical
University of Berlin, Germany. He served as Associate Editor for the IEEE
TRANSACTIONS ON COMMUNICATIONS in 1998-2001 and as Associate
Editor for the IEEE TRANSACTIONS ON INFORMATION THEORY in
2001-2003. He received the Jack Neubauer Best System Paper Award from the
IEEE Vehicular Technology Society in 2003, and the IEEE Communications
Society & Information Theory Society Joint Paper Award in 2004 and in 2011.
Giuseppe Caire is Fellow of IEEE since 2005. He has served in the Board of
Yeon-Geun Lim (S12) received his B.S. degree in Governors of the IEEE Information Theory Society from 2004 to 2007, and as
Information and Communications Engineering from President of the IEEE Information Theory Society in 2011. His main research
Sungkyunkwan University, Korea in 2011. He is now interests are in the field of communications theory, information theory, channel
with the School of Integrated Technology, Yonsei and source coding with particular focus on wireless communications.
PLACE University, Korea and is working toward the Ph.D.
PHOTO degree.
HERE His research interest includes massive MIMO and
interference management techniques for smart small
cell networks.
1536-1276 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.