Chapter 3

MU-MIMO BC Capacity
Signal Model
Let K users with single antennas are served by a base station with M T antennas. So the total
K

power average power constraint would be

Pk =P
k

. We assume that the BS has full CSI

and the channel is a flat fading channel, and the signals travel over different vector channels.
Under these assumptions the MIMO BC channel can be modelled as [1]Y =Hx+ n
Where Y=[y1,y2,y3.yk]T is a K 1 vector where each element represents the received
signal by a MS. x=[x1,x2,x3.xk]T is a MT 1 vector where each element represents
the transmitted signal from the BS towards all MS. H is the channel matrix with the
dimension of MR MT where MR is the total number of receiver antenna. In this case MR=K.
Also n=[n1,n2,n3 . nk]T is a K 1 vector that represents Gaussian noise.
It should be noted here if each receivers have multiple receive antennas then the dimensions
of the above equation will be changed accordingly.

Capacity calculation:
The achievable capacity region for MIMO broadcast channels are still an open problem [1]. A
set of achievable region was published first by Caire and Shamai [2] by using Dirty Paper
coding concepts by Costa [3]. A duality between the achievable rate region of BC channels
and multiple access capacity region was published by Vishwanath et Al [4]. From the work of
Yu and Cioffi [5] the achievable rate region of MIMO BC was mentioned as [1]
C BC =

min

max log 2

R nn >0,[ Rnn ] k ,k =N o Tr ( R xx) =P

det ( H R xx H H + R nn)
det ( Rnn )

However, this capacity region is achievable through DPC which in non-linear and very
complex to implement in practical systems. So in practical systems, a less optimal yet easy to
implement linear precoding schemes are used. In this thesis, we used Zero Forcing Precoding
and MMSE precoding techniques to evaluate capacity.

Linear Precoding Techniques:

Let us go back to our system description, X is the transmitted signal vector from the base
station. If we consider the actual data stream for the user K as S k, then X =WS , where
W=[w1,w2,w3..wk]T is the MT MR precoding vector applied by the base station. If we
consider single antenna users then the sum rate capacity can be written as [6]-

C BC = log ( 1+ k )
k=1

Where k is the SINR for user k. It can be calculated as2

H(k) w k|
|
k =
2
N o + | H(k) bi|
i k

W is in general a pre-coder matrix that can be designed. As mentioned, in this work we

investigated the ZF and MMSE precoding techniques.
Zero Forcing Precoding [1]:
In Zero forcing precoding, the goal is to transmit the user signal to the desired user and
steering nulls to the other users. In this case the ith column of the zero forcing precoding
matrix WZF,i is given by

Bibliography
[1 A. Paulraj, R. Nabar and D. Gore, Introduction to Space-time Wireless
] Communications, Cambridge University Press, 2008.
[2 G. Caire and S. Shamai, "On Achivable Rates in a Multi-Antenna Broadcast
] Downlink," 2000.
[3 M. Costa, "Writing On Dirty Paper," IEEE. Trans. Inform. Theory, Vols. IT-29, pp.
] 439-441, 1983.
[4 S. Vishwanath, N. Jindal and A. Goldsmith, "Duality, Achievable Rates, and
] Sum-Rate Capacity of Gaussian MIMO Broadcast Channels," IEEE International
Conference on Communications, vol. 49, pp. 2658-2668, 2003.
[5 W. Yu and J. Cioffi, "Sum Capacity of Gaussian vector broadcast channels,"
] IEEE Trans. Inform. Theory, vol. 50, no. 9, pp. 1875-1892, 2004.
[6 B. Lau, M. Jensen, J. Medbo and J. Furuskog, "Single and Multi-User
] Cooperative MIMO in a Measured Urban Macrocellular Environment," IEEE
Transactions on Antennas and Propagation, vol. 60, no. 2, pp. 624-632, 2012.