ANALYSIS OF THE PIL OT CONTAMINATION EFFECT IN VERY L ARGE MULTICEL L
MULTIUSER MIMO SYSTEMS FOR PHYSICAL CHANNEL MODEL S
Hien Quoe Ngo * Thomas L. Marzettat
* Department of Electrical Engineering (lSY), Linkoping University, 581 83 Linkoping, Sweden
t Bell Laboratories, Alcatel-Lucent, Murray Hill, NJ 07974, USA
ABSTRACT
We consider multicell multiuser MIMO systems with a very large
number of antennas at the base station. We assume that the channel
is estimated by using uplink training sequences, and we consider a
physical channel model where the angular domain is separated into
a finite number of directions. We analyze the so-called pilot con­
tamination effect discovered in previous work, and show that this
effect persists under the finite-dimensional channel model that we
consider. We further derive closed-form bounds on the achievable
rate of uplink data transmission with maximum-ratio combining, for
a finite and an infinite number of base station antennas.
Index Terms- Pilot contamination, very large MIMO systems.
1. INTRODUCTION
Multiuser multiple-input multiple-output (MU-MIMO) systems can
offer a spatial multiplexing gain without the requirement of multi­
ple antennas at the users [1]. Most studies have assumed that the
base station has some channel state information (CSI). The problem
of not having an a priori CSI at the base station has been considered
in [2,3], assuming that the channel estimation is done by using uplink
pilots. This requires time-division duplex (TDD) operation. Refer­
ences [2, 3] considered a single-cell setting. This is only reasonable
when the pilot sequences available in each cell are orthogonal to
those in other cells. However, in practical cellular networks, chan­
nel coherence times are not long enough to allow for orthogonal­
ity between the pilots in different cells. Therefore, non-orthogonal
training sequences must be utilized and hence, the multicell setting
should be considered.
In the multicell scenario with non-orthogonal pilots in different
cells, channel estimates obtained in a given cell will be impaired by
pilots transmitted by users in other cells. This effect, called "pilot
contamination" has been analyzed in [4]. Recently, [5] considered
the multicell MU-MIMO system with very large antenna arrays at
the base station and showed that when the number of antennas in­
creases without bound, uncorrelated noise and fast fading vanish and
the pilot contamination effect dictates the ultimate limit on the sys­
tem performance.
Most the studies referred to above assume that the channels
are independent [2-4] or the channel vectors for different users
are asymptotically orthogonal [5]. However, in reality, the MIMO
channel is generally correlated because the antennas are not suffi­
ciently well separated or the propagation environment does not offer
This work was supported in part by ELLIIT and the Swedish Research
Council (VR). E. G. Larsson is a Royal Swedish Academy of Sciences (KVA)
Research Fellow supported by a grant from Knut and Alice Wallenberg Foun­
dation.
978-1-4577-0539-7/11/$26.00 ©2011 IEEE
Erik G. Larsson*
rich enough scattering. In this paper, we investigate the multicell
MU-MIMO with large antenna arrays assuming a physical channel
model. More precisely, we consider a finite-dimensional channel
model in which the angular domain is partitioned into a large, but
finite number of directions which is small relative to the number of
base station antennas. The channels are estimated by using uplink
training (assuming TDD operation, as in previous work). For such
channels, the number of parameters to be estimated is fixed regard­
less of the number of antennas. We show that the pilot contamination
effect persists under the finite-dimensional channel model. Further­
more, we derive a closed-form lower bound on the achievable rate
of the uplink transmission, assuming maximum-ratio combining at
the base station. This bound is valid for a large but finite number of
antennas.
2. SYSTEM MODEL
Consider L cells, where each cell contains one base station equipped
with M antennas and K single-antenna users. Assume that the L
base stations share the same frequency band. We consider uplink
transmission, where the lth base station receives signals from all
users in all cells. See Fig. 1. Then, the M x 1 received vector at
the lth base station is given by
L
Yl = 5u2:=TilZi +nl (1)
i=l
where Til represents the M x K channel matrix between the lth base
station and the K users in the ith cell, i.e., [Tillm k is the channel
coefficient between the mth antenna of the lth base station and the
kth user in the ith cell; .,;p;;Zi is the K x 1 transmitted vector of
K users in the ith cell (the average power used by each user is
and nl contains M x 1 additive white Gaussian noise (AW GN). We
Pu);
assume that the elements of nl are Gaussian distributed with zero
mean and unit variance.
2.1. Physical Channel Model
Here we introduce the finite-dimensional channel model that is used
throughout the paper. The angular domain is divided into a large but
finite number of directions P. P is fixed regardless of the number
of base station antennas (P < M). Each direction, corresponding to
the angle (/>k, CPk E [-7r /2, 7r /2]' k = 1, . . . , P, is associated with an
M x 1 array steering vector a (CPk) which is given by
.)p [e-jfIC¢kl,e-j!2(¢kl,...,e-jfM(¢klf
a (CPk) = (2)
where Ji (cp) is some function of cp. The channel vector from kth user
in the ith cell to the lth base station is then a linear combination of the
3464 ICASSP2011
 Base station 6 user
(T 2 K), which satisfies cJ>tcJ> =
Pp =
IK, where
the received pilot matrix at the lth base station is
Tpu. From (6),

i=l D:rcJ>T +Nl

Yp,l = ffpA LHil (7)

1·· 1
• <

3.1. Minimum Mean-Square Error (MMSE) Estimation

/-th cell L-th cell

6

cJ>1 Yp,lYp,IcJ>* cJ>*Y. p,IcJ>* p,IYcJ>1p,IcJ>1,

We assume that the base station uses MMSE estimation. The re­
ceived pilot matrix can be represented by and
Fig. 1. Uplink transmission in multicell MU-MIMO systems. where is the orthogonal complement of Since Y only

steering vectors as follows: 2:=:'= 1 gilkma (rPm) , gilkm

where is the
propagation coefficient from the kth user of the ith cell to the lth base
includes the noise part,
Hl . p,l Yp,IcJ>*.
tion of Let Y �
is a sufficient statistic for the estima­
We have

rPm[gil).k ...gilGilkP f [gill '" gilKl

station, associated with the physical direction m (direction of arrival
Let � be a P x K matrix with g
� il k Y- p,l = ffpA '"i=l HilDil1/2 + WI
D
L

(8)
1 Gil
that contains the path gains from the kth user in
the ith cell to the lth base station. The elements of are assumed
to be independent. Then, the M x K channel matrix between the lth
WI NlcJ>* NM,K (0,1 M,1K). HlHll l
where � � Since
dent columns, we can estimate each column of
has indepen­
independently.
base station and K users in the ith cell is
p,ln
Let Y be the nth column of Y p,l.
Then
Til =AGil

[
a (rP ) ... a (rPp)l
(3)
- ,ln = ffpAhl nf311/n2 + ffpA '"hilnf3il1/n2 + Win
Yp L

D
where � A 1 is a full rank M x P matrix.
Gil il #
(9)

The propagation channel

gilkm
models independent fast fading,
geometric attenuation, and log-normal shadow fading. Its elements wherehiln WinZln �A hiln Hil1/2 + WI,
and are the nth columns of and respec-
are given by
•
tlvely. Denote by
MMSE estimate of hUn
D.
=

is given by
i l 13iln Win·
D
",
L
# Then the

gilkm =hilkm�, = 1, 2, m . . . ,P (4)

h, Un =f311/n2 ffpAt ( p f3Un t + ) -1 Yp- ,ln RZln
p AA
where hilkm
is a fast fading coefficient assumed to be zero mean and
have unit variance, and v7Jilk models the path loss and shadowing
(10)

which are assumed to be independent of the direction m and to be where RZln

we obtain
= {Zlnzin}.
lE By using the matrix inversion lemma,

value of f3ilk
constant and known a priori. This assumption is reasonable since the
changes very slowly with time. Then, we have
1 2 t - 1t
Gil =HilDil1/2 (5)
hUn = f31 � ffp (P pA A t f3 iln + p ) A p,ln
I Y (\\)

where Hil
is the P x K matrix of fast fading coefficients between the
[ Hil l km = h il k m, The kth diagonal element of p
M;p 2:=f iln
pAtA 2:=f=l f3iln in (1\) equals
K users in the ith cell and the lth base station, i.e.,
Dil [ Dillkk =f3ilk.
is a K x K diagonal matrix whose diagonal elements are
=l f3 . Since the uplink is typically interference-limited
T 2:=i=l f3iln
and Mp '
» l. Therefore,
L
given by Therefore, (I) can be written as [4], hUn can be approximated as

L L

Yl = J]J;A LGil i=l xi +nl = J]J;A LHi i=l lD:rXi +nl (6) (12)

Thus, the MMSE estimate of Hu is

3. CHANNEL ESTIMATION H'U - Pp -1/2 (AtA ) -l Aty- p,lDI-1 DI1/2 (\3)

Channel estimation is performed by using training sequences re­

ceived on the uplink. A part of the coherence interval is used for
Dl 2:=f=l
where � Dil. Then, the estimate of the physical channel
matrix between the lth base station and the K users in lth cell is
the uplink training. All users in all cells simultaneously transmit pi­ given by
lot sequences of length T symbols. The assumption on synchronized
transmission represents the worst case from the pilot contamination -
T, u =A Hl' lDI1/2 =Pp -1/2 IIAYp,IDI-1 Dll (14)
point of view, but it makes no fundamental difference to assume un­
synchronized transmission [5]. We assume that the same set of pilot
sequences is used in all L cells. Therefore, the pilot sequences used
Nm,n I (M, E, lit)
em n em m enlItTn.
denotes a matrix-variate complex Gaussian distribu­
M E x IZi E,
in the lth cell can be represented by a T x K matrix �cJ>1 = �cJ> tion with the mean matrix
where
and the variance matrix
IZi denotes the Kronecker product, E E x and lit E x

3465
where IIA � A (AtA) At -1
is the orthogonal projection onto A. 4.2. Analysis of the Pilot Contamination Effect
We can see that since post-multiplication of Y
just multiplication with the pseudoinverse (� � =IK), Y is the
t
p,l
with �. means
p,l To gain further insight into the pilot contamination effect, we derive

conventional least-squares channel estimate. The channel estimator

that we derived thus performs conventional channel estimation and
lower bounds on the achievable rate for a finite and an infinite num­
ber of base station antennas (the analysis requires that > P). To M
then projects the estimate onto the physical (beamspace) model for
the array.
obtain these lower bounds we use the techniques of [3, 4, 6]. We as­

TIn vPP
sume a uniform array at the base station, and the elements of
are Li.d Gaussian radom variables. Let be the nth element of the
Hil
4. UPLINK DATA TRANSMISSION
from (15), we have
rl. Tln
received K x 1 vector DenoteIn· �
Ef-l (Jan r
(Jlln
Then

We consider the uplink transmission represented by (1). The base

station uses its channel estimate obtained by uplink training as in
(14) to perform maximum-ratio combining.
Tln=(y'P;Atgiln +Wlj � (y'Pu"AtGjlXj +nv A (21)

4.1. Maximum Ratio Combining

Equation (21) can be rewritten as
L
The lth base station processes its received signal by mUltiplying it by
the conjugate-transpose of the channel estimate. From (6), (8) and
Tzn =a'fnxl + LafnXj
Nl + Zln· (22)

(14), we have
af n
where �L�=lg!lntA tAGjl + y'Pu"winAGjl,
JpPPu and

rl = ilYlt =Pp -1/2 DllDl1 (y'P;A tGil + Wl)t Zln ( y'P;A Li=l giln +Wln) II l
�
=
L
A n . We have

X (y'Pu"A tGjlXj + nl).

IIA (15)
en
is the nth column of the K x K identity matrix. By adding
(23)

As M ---+ 00, the products of uncorrelated quantities can be

where
{a'fn} a'fn
and subtracting 18: from in (22), and using (23), we obtain
ignored [5]. Then (15) becomes L

1 L L Tln = {a'fn} Xl + (a'fn {a'fn}) Xl + LNlafnXj + Zln

18: 18:
DllD
-

1
rl l ( �G !l) AtA ( GjlXj)
y'Pu" M ---+
M f; (16)
= {gilnAtAgl n} Xln + (a'fn {a'fn}) Xl
JpPPu18: - 18:
We can see that for an unlimited number of antennas, the effect of L

Nl + Zln. (24)
+ LafnXj
uncorrelated noise disappears. In particular, the pilot contamination
effect, which is due to the interference from users in other cells,
persists under the finite-dimensional channel model.
We now consider the special case of a uniform array. Then the ( ) l g (1 + )
Let C x � o 2 x . The nth user in the lth cell can achieve

]
response vector is given by an uplink rate of at least
T
1_ -j2",fsin¢k , ... , e _j2",(M:;:1)dsin¢k
a(,I.'f'k)=_
..;p'
1
[ e (17)
(25)
where d is the antenna spacing, and A is the wavelength. For k i=- I,

Proposition 1 The achievable uplink rate of the nth user in the lth

page, where � (Jjl L;::=l !3ln L�=l (Jiln,

cell in (25) is lower bounded by (26), shown at the top of the next
(Jjlk, � and
eCA)r",CA)
f (A) = L L Xm,n (A) + 1) A1m)
For k =I, kat (CPk) a (CPl) =-j,. Therefore, m=l n=l n (n (27)

1
AtA M-+oo P1 I p. where A � diag(A1,A2, ...,AP), with Ak, k = 1, . . . ,P, is the
M ---+ (19)
{} (A) is the number of distinct eigenval­
AtA,
Tm(1)
kth eigenvalue of
ues, A , ... , ACeCA)) are the distinct eigenvalues in decreasing or­
Substitution of (19) into (16) yields
der, (A) is the multiplicity ofACm) and Xm,n
(A) is the ( m, n)th

y'Pu"Mrl .=:+oo DllDl1 (ti=lG!l) (tj=lGjlXj).

� M characteristic coefficient of A which is defined in [7J.
_pI (20)
Corollary 1 For an unlimited number of base station antennas, the
Gil
Since the elements of are independent, the above result reveals
that the performance of the system under the finite-dimensional
lower bound on the achievable uplink rate of the nth user in the lth
cell becomes
channel model with P angular bins and with an unlimited number
of base station antennas is the same as the performance under an (28)
uncorrelated channel model with P antennas.

3466
 (26)

12.0 16.0

11.0
14.0 --M=20
10.0 -M=50

--...--...'" --...
9.0 � 12.0 ---M=100

:r: :r:
�
N N

R,OO 11
�
8.0 =

fl fl
10.0
7.0
e e
� 6.0
� 8.0

�
'" �
S and 10 dB S
5.0

::> Pu = -10, -5, 0, ::>

6.0
<Zl 4.0 <Zl

3.0 4.0

2.0
2.0
1.0

0.0
2'::
0
-� 40=
-
--:':
6 0----:
8':::0
- -1:-'0-:CO
-:1�20::
- -1 :40�
-� 16:-:
-0
-:- 1-'c8-=

Number of Base Station Antennas (M)

0
-
- 00
:C2"c:---:C
22
'-:c 0-
:2
- �4-:
-
0
::2�
- 60:-:
-
2'-:::-'
80-30 0
0.0
0.0
L-
---'��_�_�_�_�_�__"__"__
0.1 0.2 0.3 0.4 0.5 0.6

Cross Gain (a)

0.7 0.8 0.9
____'
1.0

Fig. 2. Lower bound on the sum-rate of uplink transmission as a Fig. 3. Lower bound on the sum-rate of uplink transmission versus
function of the number of base station antennas M. the cross gain.
6. CONCLUDING REMARKS
Remark 1 If P also goes to infinity, the lower bound on the achiev­

able rate becomes Rr:: = c (L ! � )

J#-l
fl
value for the rate obtained in [5 J. This is due to the fact that when P
is large, things that were random before become deterministic and
hence, the lower bound approaches the exact value.
2
Jln
This equals the exact
This paper has analyzed the pilot contamination effect in multicell
MU-MIMO systems with very large antenna arrays for a physical
channel model with a finite number of scattering centers visible from
the base station. We showed that the pilot contamination effect dis­
covered in [5] persists under the finite-dimensional channel model.
We also derived a closed-form lower bound on the achievable uplink
rate for finite and infinite number of base station antennas.

7. REFERENCES
5. NUMERICAL RESULTS [I] O. Gesbert, M. Kountouris, R. W. Heath Jr., C.-B. Chae, and
T. Salzer, "Shifting the MIMO paradigm," IEEE Sig. Proc.
We consider a system with 4 cells, and K = 10 users per cell. The Mag. , vol. 24, pp. 36-46, Sept. 2007.
training sequence length is T = K, and the number of physical di­ [2] T. L. Marzetta, "How much training is required for multiuser
rections is P = 20. We consider a uniform array at the base station MIMO," in Proc. of Asilomar Conference on Signals, Systems
with * = 0.3 and (Pk = -7r /2 + (k - 1) 7r / P. We further assume and Computers, Oct. 2006.
[3] 1. Jose, A. Ashikhmin, P. Whitning, and S. Vishwanath,
that all direct gains are equal to I and all cross gains are equal to a,
i.e, f311k = 1, and f3jlk = a, Vj =f=. l, k = 1,... , K. The lower bound
"Scheduling and pre-conditioning in multi-user MIMO TOO
on the sum-rate of lth cell is defined as Rl = 'L;;;=1 Rln.
systems," in Proc. IEEE Int. Conj. Communications Workshops
Figure 2 shows the lower bound on the sum-rate of the uplink (ICCW'08), May 2008.
transmission in the lth cell versus the number of antennas M, at
[4] J. Jose, A. Ashikhmin, T. L. Marzetta, and S. Vishwanath, "Pilot
a = 0.1 and for different average transmit powers per user pu =
contamination problem in multi-cell TOO systems," in Proc. of
-10, -5,0, and 10 dB. We see that at low SNR (since the noise
IEEE ISIT, June 2009.
power is unity, the SNR is equal to Pu), using a large number of base
station antennas significantly improves the achievable rate. Further­ [5] T. L. Marzetta, "Noncooperative cellular wireless with unlim­
more, we can see that as M ---+ 00, the sum-rates approach RI. ited numbers of base station antennas," IEEE Trans. Wireless
This is the asymptotic value of sum-rate with an unlimited number Commun. , vol. 9, pp. 3590-3600, Nov. 2010.
of base station antennas and it is independent of the SNR (cf. (28». [6] B. Hassibi and B. M. Hochwald, "How much training is needed
Figure 3 depicts the lower bound on the sum-rate of uplink trans­ in multiple-antenna wireless links?" IEEE Trans. In! Theory,
mission versus the cross gain at an SNR of Pu = 0 dB for different vol. 49, pp. 951-963, Apr. 2003.
M = 20, 50 and 100. We can see that the effect of pilot contamina­ [7] H. Shin and M. Z. Win, "MIMO diversity in the presence of
tion can be very significant if the value of the cross gain is close to double scattering," IEEE Trans. In! Theory, vol. 54, pp. 2976-
the value of the direct gain, regardless of M. 2996, Jul. 2008.

3467