1
|P a g e
Master Thesis Electrical Engineering Emphasis on Telecommunications Thesis no: MEE10:78 September 2010.
User Scheduling Algorithm for MU-MIMO System with limited feedback Sudhir Kumar Burra (861019-4717) Reddy Prasad Reddy Yendrapalli (860808-1553)
Under the esteemed guidance of Prof. Abbas Mohammed
Blekinge Institute of Technology September 2010
Department of Electrical Engineering School of Engineering Blekinge Institute of Technology SE-37179 Karlskrona Sweden.
2
|P a g e
ACKNOWLEDGEMENTS
We express our sincere gratitude to our guide P r o f . A B B A S M O H A M M E D for his guidance and constant encouragement throughout the project work. We are very grateful to him for providing us his valuable time through various sessions to discuss the issues related to our thesis work which enabled us to take this thesis to fruitful completion. Special thanks to Mr. MICHAEL ASMAN, Program Manager for DDP in SWEDEN, for his valuable suggestions and guidance during the entire course work. We would like to thank Mr. GURUDUTT KUMAR VELPULA, International Coordinator for DDP, for providing us the opportunity to study in BTH, SWEDEN and also we would like to thank Dr. MADHAVI LATHA, Coordinator for DDP, JNTU. We would like to convey our heartful thanks to all the Professors of BTH and JNTU for their immense help and moral support in completing our course work successfully. We express our sincere thanks to all my friends at BTH, who supported us during our stay in SWEDEN and made it really enjoyable and memo rable. We are very grateful to our parents and our sisters for their support and constant encouragement.
Sudhir Kumar Burra Reddy Prasad Reddy Yendrapalli.
2
|P a g e
ACKNOWLEDGEMENTS
We express our sincere gratitude to our guide P r o f . A B B A S M O H A M M E D for his guidance and constant encouragement throughout the project work. We are very grateful to him for providing us his valuable time through various sessions to discuss the issues related to our thesis work which enabled us to take this thesis to fruitful completion. Special thanks to Mr. MICHAEL ASMAN, Program Manager for DDP in SWEDEN, for his valuable suggestions and guidance during the entire course work. We would like to thank Mr. GURUDUTT KUMAR VELPULA, International Coordinator for DDP, for providing us the opportunity to study in BTH, SWEDEN and also we would like to thank Dr. MADHAVI LATHA, Coordinator for DDP, JNTU. We would like to convey our heartful thanks to all the Professors of BTH and JNTU for their immense help and moral support in completing our course work successfully. We express our sincere thanks to all my friends at BTH, who supported us during our stay in SWEDEN and made it really enjoyable and memo rable. We are very grateful to our parents and our sisters for their support and constant encouragement.
Sudhir Kumar Burra Reddy Prasad Reddy Yendrapalli.
3
|P a g e
ABSTRACT
In conventional cellular systems, each base station (BS) transmits signals intended for a single user in a particular resource allocation. As bandwidth is a scarce resource, effective utilization of the available bandwidth in the system is essential in modern wireless systems especially for applications such as video streaming and voice over internet protocol (VoIP) which demands high data rate. Fortunately since the users feedback the channel state information to the network, there is an opportunity for the BS to schedule more than one users data in a single resource allocation by designing precoders which beamform the data to the intended user. This technique which is called multi-user multiple-input and multiple-output (MU-MIMO) is adopted in the evolving radio interface technologies. For properly utilizing the feedback information, scheduling algorithms are designed which selects pairs of users which would maximize system capacity. In this thesis we describe MU-MIMO technique with codebook based precoding that has been proposed for the IEEE 802.16m mobile broadband standard. A multi-user proportional fair (PF) scheduling algorithm is designed to improve both sum capacity and fairness among users.
Key words: MU-MIMO, Bandwidth, Limited feedback, Data rate, Precoder.
4
|P a g e
GLOSSARY
AWGN
Additive White Gaussian Noise
BS
Base Station
CCI
Co-Channel Interference
CQI
Channel Quality Indicator
CL
Closed Loop
DR
DataRegion
FDD
Frequency Division Duplex
LOS
Line Of Sight
LTE
Long Term Evolution
LTE-A
Long Term Evolution Advanced
MIMO
Multiple Input Multiple Output
MMSE
Minimum Mean Square Error
MRC
Maximum Ratio Combining
MS
Mobile Station
MU-MIMO
Multi User MIMO
MUI
Multilingual User Interface
OFDM
Orthogonal Frequency Division Multiplexing
OL
Open Loop
PF
Proportional Fair
PMI
Precoder Matrix Information
QoS
Quality of Service
5
|P a g e
RF
Radio Frequency
SFBC
Space Frequency Block Coding
SINR
Signal to Interference plus Noise Ratio
SISO
Single Input Single Output
SNR
Signal to Noise Ratio
STBC
Space Time Block Coding
UE
User Environment
UT
User Terminal
VoIP
Voice over Internet Protocol
ZF
Zero Forcing
6
|P a g e
CONTENTS
Ack nowledgements.
....................... .............. 2
Abstract. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1
2
3
Intr o du ction
9
1.1
Motivation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10
1.2
Objectives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3
Back ground . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . .
1.4
Outline. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
11
Theory of MIMO System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13 2.1
MIMO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13
2.2
MIMO System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3
MIMO Precoding
15
2.3.1
Optimal Unitary or SVD precoding. . . . . . . . . . . . . . .16
2.3.2
Codebook Based Precoding . . . . . . . . . . . .
16
2.3.3
Zero Forcing Precoding
16
2.3.4
Dirty Paper Precoding . . . . . . . . . . . . . . . . . . . . . . . . .16
Literature Survey
17
3.1
17
Multi User MIMO . . . . . . . . . . . . . . . . . . 3.1.1
Spatial Diversity Gain . . . . . . . . . . . . . . 17
3.1.2
Spatial Multiplexing Gain. . . . . . . . . . . .
17
7|
Page
3.1.3
Multi User Diversity Gain.
. . . . .
18
3.2 Challenges of MU-MIMO. . . . . . . . . . . . . . . . . . 19 3.2.1
Interference . . . . . . . . . . . . . . .
19
3.2.2
Post-processing. . . . . . . . . . . .
19
3.2.3
Pre-processing/Precoding. . . . . . . . . . . . . . . . . . . . . . .20
3.2.4
CQI Modelling . . . . . . . . . . . . . .
3.2.5
Scheduling. . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
20
3.3 Precoding Techniques of MU-MIMO . . . . . . . . . . . . 21 3.3.1 Zero Forcing Beamforming and Unitary Precoding . . . .21 3.3.2 Channel Inversion Method and Diagnolization Method.21 3.4 System Model. . . . . . . . . . . . .
3.5
4
22
3.4.1
Uncorrelated Channel Model . . . . . . . . . . .
23
3.4.2
Correlated Channel Model. . . . . . . . . . . .
24
Proportional Fair Scheduler . . . . . . . . . . . . . . .
24
Mathematical Procedure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..25 4.1
4.2
MU-MIMO Capacity Formulation. . . . . . . . . . . . . . 25 4.1.1
MMSE Receiver . . . . . . . . . . . . . . . . . . . . 26
4.1.2
CQI Calculation. . . . . . . . . . . . . . . . . . . . . . . 27
Procedure
29
5 Results and Discussion…………………………………………… 31
6 Conclusions and Future work.. …………………………………… . 36
7
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 .7
8|
Page
List of Figures
2.1 MIMO system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.2 MIMO pre and post processing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .15 3.1 MU-MIMO system setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 4.1 Procedure for MU-MIMO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 5.1 Sum capacity of MU-MIMO for 4×2 with uncorrelated channel. . . . . . . . . . 33 5.2 Sum capacity of MU-MIMO for 4×1with uncorrelated channel. . . . . . . . . . .33 5.3 Sum capacity of MU-MIMO for 4×2 with correlated channel. . . . . . . . . . . . 34 5.4 Sum capacity of MU-MIMO for 4×1 with correlated channel. . . . . . . . . . .. 34
9|Page
CHAPTER 1 1. INTRODUCTION
As the demand for high data rate applications like video and audio streaming, VoIP, video conferencing are increasing, future wireless systems should be able to provide high speed broad band services for mobile users with sufficient quality of service (QoS) support. As the bandwidth and power are scarce or limited resources, techniques which lead to efficient utilization of these resources are quite necessary in the next generation wireless systems. At the same time the wireless channel creates a challenging environment because of variety of channel impairments. Thus, future wireless systems are to be designed taking all these factors into consideration.
For scenarios with a large number of users to be served in one cell, high capacity gains can be achieved by transmitting independent data streams to different users sharing the same time-frequency resources. This technique is referred to as multi-user multiple-input multiple-output (MU-MIMO) [1]. It is one of the techniques which can be used in cellular systems to increase spectral efficiency.
In MU-MIMO operation two or more user environment’s (UE) share the same timefrequency resources. Several parallel data streams are transmitted simultaneously, one for each UE. It is assumed that the UE feeds back a quantized version of the observed channel, so that base station (BS) can schedule in MU-MIMO mode terminals with good channel separation.
Long term evolution (LTE) and its successor LTE-Advanced (LTE-A) are some of next generation wireless systems, which use advanced features like MIMO, link adaptation, orthogonal frequency division multiplexing (OFDM) and many other techniques to help in achieving high spectral efficiencies.
10 | P a g e
1.1 MOTIVATION
MIMO systems which employ multiple antennas at transmitter and receiver is a very useful technique in wireless environments to combat the effects of fading and to use the radio channel efficiently by transmitting multiple streams to a user in the same resource allocation, thus achieving diversity gain and multiplexing gain, which is the first step in achieving high spectral efficiencies. Transmit diversity (space frequency block coding) is one such scheme which sends multiple copies of the same data to a user which makes use of multiple replicas of transmitted data to combat fading. This is a very useful scheme in fast fading environments. Also the fact that each user experiences different channel conditions and quantized channel state information is available at the BS, which can be utilized to achieve additional gain by jointly precoding users which makes them orthogonal to each other. Thus allowing the BS to schedule more than one user in a resource allocation. This leads to a situation where efficient user scheduling and pairing algorithms are required, which uses the feedback information intelligently at BS to maximize the capacity achievable in the system.
11 | P a g e
1.2 OBJECTIVE The main goal is to study and evaluate the performance of MU-MIMO systems using feedback sent by the users.
Create a cellular environment with single or multiple BSs with multiple users with multiple antennas spread randomly positioned inside the cells.
First the BS will transmit reference symbols which are known to the users.
Calculate signal to interference plus noise ratio (SINR) and sum capacity.
Plot the graph of sum capacity vs signal-to-noise ratio (SNR).
1.3 Background MU-MIMO has generated considerable interest recently. The main idea emerging from this research is that multiple users can be simultaneously multiplexed to take simultaneous advantage of multi-user and spatial diversity. The optimal BS interference cancellation strategy is the so called dirty paper coding (DPC) [6], but it is not directly practical. More realistic linear multi-user precoding techniques have been developed. BS does not have knowledge of the channel unless it is feedback by the mobile station (MS) such as in the case of Frequency Division Duplex (FDD) systems. It is huge amount for MS to feedback the complete channel, hence quest for f inite feedback systems arises. One of the solutions to this problem is codebook based precoding technique which is a linear precoding technique. Here, a codebook which contains a set of precoding matrices known to the BS and MS is used.
This thesis describes MU-MIMO technique with codebook based precoding that has been proposed for the IEEE 802.16m mobile broadband standard. A multi-user PF scheduling algorithm is designed to improve both sum capacity and fairness among users.
12 | P a g e
1.4 Outline Chapter 2 This chapter will give a detailed description of MIMO systems. Chapter 3 This chapter will give a detailed description of MU-MIMO systems. Chapter 4 This chapter will discuss the actual formulation used to calculate channel quality indicator (CQI) and capacity of users. Chapter 5 This chapter includes the MATLAB results and discussion. Chapter 6 This chapter presents the conclusions.
13 | P a g e
CHAPTER 2
2. Theory of MIMO System
2.1 MIMO The use of multiple antennas allows independent channels to be created in space and it is possible to achieve spatial diversity, which can be created without any additional bandwidth and transmit power . In addition to providing spatial diversity, antenna arrays can be used to focus energy (beamforming) or create multiple parallel channels for carrying unique data streams (spatial multiplexing). When multiple antennas are used at both the transmitter and the receiver, it is commonly referred as MIMO system. These systems can be used to:
Increase the system reliability (decrease the b it or packet error rate).
Increase the achievable data rate and hence system capacity.
Increase the coverage area.
Decrease the required transmits power.
However, these four desirable attributes usually compete with one another. For example, an increase in data rate will often require an increase in either the error rate or transmit power.
14 | P a g e
2.2 MIMO System Figure 2.1 shows MIMO system where there are M(>1) antennas at the BS and N (>1) antennas at the MS.
BS
1
1
2
2
.
H
.
M
MS N
Figure 2.1: MIMO System The wireless channel matrix H can be expressed as
h h H h N
11
h12
21
h22
1
hN 2
h M hNM NXM h1 M 2
(2.1)
where hij is the channel gain from ith receive antenna to the j th transmit antenna. In case of MIMO systems along with diversity, spatial multiplexing can also be exploited which refers to breaking the incoming high rate data stream into M independent data streams. Assuming that the streams can be successfully decoded, the nominal spectral efficiency is thus increased by a factor of M. This is certainly exciting wh ic h implies that adding antenna elements can greatly increase the viability of the high data rates desired for wireless broadband access. The MS has to estimate M×1 transmit vector from N×1 receive vector. In order to adjust the number of streams, some sort of pre-processing also called precoding is done before actual transmission, which can be thought as a kind of beamforming. More insights about MIMO can be found in reference [2]. MIMO systems can be classified as:
15 | P a g e
Single-user or Multi-user
When a single-user is scheduled in a dataregion it is referred as single-user MIMO (SUMIMO). If more than one user is scheduled in a dataregion then it becomes MU-MIMO. These two are differ in terms of precoding and scheduling. The number of streams allocated to each user is configurable.
Open loop or Closed loop
When the precoders are fixed to subbands and chosen from a codebook which is known to BS and MS is referred as Open loop MIMO (OL-MIMO). If the precoders are formed by the scheduler based on the preferred matrix index (PMI) feedback from each of the MSs, then it is called Closed loop MIMO (CL-MIMO). From the above classification, there are four possible MIMO configurations: (1) OL-SU-MIMO (2) OL-MU-MIMO (3) CL-SU-MIMO (4) CL-MU-MIMO.
2.3 MIMO precoding Precoding is done mainly to map K transmitted symbols to M transmitting antennas. In case of SU-MIMO these K symbols belong to single-user, whereas in MU-MIMO they are intended for K different users (assuming single stream per user). There are various ways of precoding, some of them are discussed here in brief. x
1
1
1
2
2
.
.
x 1
x
2
. .
Pre Processing
.
Post processing
.
x 2
. .
x K
M
N
Figure 2.2: MIMO pre and post processing.
x k
16 | P a g e
2.3.1 Optimal Unitary or SVD precoding In case of SU-MIMO, when the channel is known at the transmitter, the best precoder to use is the matrix formed by the right singular vectors of the channel, which has proven to achieve the channel capacity of MIMO systems, at the cost of feeding back signaling of the channel state information from MS to the BS.
2.3.2 Codebook based precoding Here a codebook which contains a set of precoders is known at the transmitter and receiver. In case of OL-MU-MIMO, precoders are fixed to all the subbands and the user needs to feedback which precoding vector in the precoder is to be used to precode the data. Detail description on how code books are designed is found in references [7] and [8].
2.3.3 Zero Forcing (ZF) precoding For SU-MIMO the precoder is just the pseudo inverse of the channel, which can completely cancel out the inter stream interference and reproduce the data vector transmitted with additive noise. In case of MU-MIMO the precoder has to cancel out multi-user interference, so a block diagonalization method is proposed in reference [9] which is used to find the precoder under some constraints.
2.3.4 Dirty Paper Coding (DPC) All the techniques discussed previously were linear techniques, but DPC is a non-linear coding technique that pre-cancels known interference without power penalty. Once the transmitter is assumed to know the interference signal regardless of channel state information knowledge at the receiver. This category includes Costa Precoding, Tomlinson-Harashima Precoding and the Vector Perturbation Technique as discussed in references [4] and [5].
17 | P a g e
CHAPTER 3
3. Literature Survey of MU-MIMO 3.1 MU-MIMO In MU-MIMO more than one user can be served in the same bandwidth using appropriate precoders at BS. This technique is just like SU-MIMO where one or more streams transmitted at a time using multiple antennas belonging to the same user. In MU-MIMO each stream could belong to a different user i.e., instead of stream multiplexing, MU-MIMO does user multiplexing. For scenarios where large number of users is to be served in one cell or to serve a limited number of users with increased throughput, MU-MIMO can be used.
The three gains that are useful in increasing the performance of MU-MIMO systems are defined as follows [11].
3.1.1 Spatial diversity gain This is the technique for improving communication quality by transmitting and receiving with multiple antennas. Each pair of transmit and receive antennas provides a signal path by sending signals that carry the same information through different paths. Hence multiple independently faded replicas of the data symbol can be obtained and more reliable reception is achieved.
3.1.2 Spatial multiplexing gain This is the performance improvement derived from using multiple antennas to transmit multiple signal flows through space in parallel. For a MIMO system with Nt transmitting antennas and Nr receiving antennas, the maximum achievable spatial multiplexing gain is minimum of Nt and Nr .
18 | P a g e
3.1.3 Multi-user diversity gain The improvement in system throughput derived from using a scheduler wh ich exploits the disparities fading and interference characteristics between users. The first two (spatial diversity gain, spatial multiplexing gain) can be typically achieved using precoders at the transmitter side by using the feedback information sent by UE and using multiple antennas. But the latter can be achieved by using proper scheduling techniques. MU-MIMO offers additional degrees of freedom when compared to SU-MIMO since multiple users are multiplexed in the same physical channel. This can be achieved by pairing users whose precoders are orthogonal to each other in a dataregion and then precoding them appropriately so that each user sees only its own information. As UE feedback quantized channel information, the users will not be perfectly orthogonal to each other so some remnant inter user interference will be seen by each of the users who are paired. This can be minimized by using a minimum mean square error (MMSE) receiver at UE to minimize the effect of multilingual user interference (MUI) on capacity [17]. The main advantages that lead to MIMO paradigm shift to MU-MIMO from SU-MIMO communications are 1. MU-MIMO schemes allow for direct gain in multiple-access capacity (proportional to number of transmit antennas) because of multiplexing of data of several users in the same radio channel. 2. MU-MIMO schemes are more immune to loss of channel rank because of line of sight (LOS) conditions or antenna correlation, which is a major problem that causes performance degradation in SU-MIMO communications.
19 | P a g e
3.2 Challenges of MU-MIMO MU-MIMO has tremendous benefits which are achieved by overcoming some challenges. Multiple users using the same resources at the same time would lead to several issues that need to be considered, some of them are mentioned here.
3.2.1 Interference When multiple users are using the same resources at the same time, there would be severe interference between their signals. Each user should be capable of decoding his respective stream by reducing the interference due to other stream. This can be achieved by careful pre-processing at the transmitter and post-processing at the receiver.
3.2.2 Post-processing In single-user transmission, MIMO could be used for spatial multiplexing, where multiple symbols are transmitted to the same user. For example, c onsi der a 2×2 single-user system, in which the received vector can be represented as y
=
Hx
+
n
(3.1)
where the transmitted 2× 1 vector x represents 2 symbols that are transmitted simultaneously to a particular user, thus doubling the user throughput. In order to decode the 2 symbols from the received 2×1 vector y, a simple approach would be to build a linear receiver that diagonalises the system, i.e., multiply the received vector y by H−1. This decouples the system and we get back the two transmitted symbols.
In the MU-MIMO case, the effective received vector y, is a concatenation of the symbols received by geographically separated users, and post processing must be done in such a way to reduce the interference from the other user. Several receiver configurations such as MMSE, maximum ratio combining (MRC) and ZF are possible but MMSE receiver is shown to reduce the interference effectively.
20 | P a g e
3.2.3 Pre-processing/precoding Because of this limitation on the interference cancellation that can be done at the MS, good precoders need to be designed, such that we beamform efficiently towards the two users. However this would require good knowledge of the channels to both users at the BS, which requires heavy amounts of feedback. So we would need to come up with the best possible precoders to use at the BS, with a limitation on the feedback rate.
3.2.4 Channel Quality Indicator (CQI) modeling CQI is a feedback by the user in frame (n), for the allocation of modulation and coding schemes in frame (n+1). CQI modeling is to be done so that the user experiences a good throughput. In the single input single output (SISO) case, CQI is a function of the channel to a particular user, which (for low Doppler ’s shift) does not fluctuate much between adjacent frames. But in case of MU-MIMO, in addition to being a function of the channel to the user, CQI is also a function of the precoder used at the BS. Hence, better the precoding is lesser is the interference and higher CQI will be.
3.2.5 Scheduling When we have a number of users contending for same resource, throughputs can be increased by scheduling those users who experience a good channel. This increase in system performance merely because of scheduling the best-set of users at any point of time is known as multi-user diversity. However, maximizing system throughput must not come as a result of cell-edge users (who face poor channel conditions) never being scheduled. System performance must be maximized and at the same time a certain amount of fairness must be ensured among the users in the system. A multi-user scheduler that meets these demands needs to be implemented.
21 | P a g e
3.3 Precoding techniques of MU-MIMO 3.3.1 Zero Forcing Beamforming (ZFBF) and Unitary precoding ZFBF and unitary precoding are two useful precoding techniques for MU-MIMO in limited feedback environments. ZF precoding is a potential precoder design for MU-MIMO. The main benefit of ZF scheme is that the interference is pre-cancelled at the transmitter side. It implies that eNodeB has most of the computational complexity in designing the p recoder and each terminal needs only information regarding its own data streams for reception. However the quantized channel information has to be precise, so that the multi-user interference becomes sufficiently low in order to get gains from this scheme. The ZF precoder can be designed using the moore-penrose psuedo inverse as given below (assuming “u” users are paired together)
W T eq (eq eq )
where eq is the equivalent channel feedback and
(3.2) W T
is the precoder used.
3.3.2 Channel Inversion Method and Diagnolization method (BD) Channel inversion method is one of the linear precoding MU-MIMO techniques which is simple and has capacity limit. When spatial correlation is increased, the multi-user channel capacity decreases rapidly. BD can perfectly cancel co-channel interference (CCI), but has antenna constraint at the BS and MS. The computation burden for system is very heavy when the number of users is very large. Both channel inversion method and BD are based on the feedback of the MIMO channel matrix, so the feedback is very large. More information about these techniques can be found in [13][14][15].
22 | P a g e
3.4 System Model Let us consider a MIMO system with M transmitting antennas at the BS and N receiving antennas at each MS as shown in figure 3.1. There is a codebook known a prior to both the BS and all MSs. If U users are waiting to be scheduled at the BS, the scheduler will determine K (
x 1
User 1
User 1 data
x 1
1
User 2 data
x 2
2
SCHEDULER
PRECODER
.
.
.
User U data
x k
M
x 2
User 2
. .
. .
x k
User K
Feedback: CQI, Precoder Vector index.
Figure 3.1: MU-MIMO System setup.
The OFDM technique has become one of the most promising techniques for next generation wireless communication systems. Since OFDM technique can deal frequency selective fading as flat fading, so in this thesis, we model the MIMO channel as the flat fading MIMO channel [3].
23 | P a g e
3.4.1 Uncorrelated Channel Model The received signal vector at the k th MS is given by
yk =Hk Wx+n k
k = 1,2,.....,U
(3.3)
where
x1 x 2 x and xk KX 1
hk11 h k12 hk 1 M k h 21 h k 22 h k 2 M H k k k k h N 1 h N 2 h NM NXM
•
xk is the complex symbol transmitted for k th user.
•
Hk ∈ C N ×M is the N×M wireless channel matrix from the k th MS to BS and hij ∼C N (0,
1) which represents the channel impulse response coupling the j th antenna at the BS to the ith antenna at the MS and its amplitude obeys independent and identical Rayleigh-distribution. •
W= v1 v2 . . . vk Ml K W∈C is a precoder chosen from the codebook C which contains set of
unitary precoders and vk represents precoding vector used to precode k th user data, where k is called stream indicator or precoding vector index. •
nk ∼ C N (0, No I N ) is noise vector at the k th MS.
3.4.2 Correlated Channel model The antennas at the BS are magnitude correlated i.e., each antenna at the BS sees same channel gain to all receiver antennas of k th user. Now the channel in the previous subsection can be modified as
h1 (e j e j 2 e j 3 e j 4 H k j j 2 j 3 j 4 h2 (e e e e k
k
k
k
k
k
k
k
)
)
(3.4)
24 | P a g e
Here
θk ∼ u [−π / 3, +π / 3]
for k th user, hi is Rayleigh distributed random variable representing
the channel gain between ith receive antenna and any of the antennas at BS.
3.5 Proportional Fair (PF) Scheduler The PF scheduler is designed to take advantage of multi-user diversity while maintaining comparable long-term throughput for all users. Let R k (t) denote the instantaneous data rate that user k can achieve at time t, and let Tk (t) be the average throughput for user k up to time slot t. The PF scheduler selects the user denoted as k ∗ with the highest R k (t)/ Tk (t) for transmission. In the long term, this is equivalent to selecting the user with the highest instantaneous rate relative to its average throughput. The average throughput Tk (t) for all users is then updated according to Tk (t 1) (1 = (1
1
tc 1
t c
)Tk (t )
1
t c
Rk (t )
kk *
)Tk (t )
PF Metric =
*
k=k
Rk (t ) Tk (t )
(3.5)
Thus consistently underserved users receive scheduling priority, which promotes fairness. The parameter tc controls the latency of the system. If tc is large, the latency increases, with the benefit of higher sum throughput. If tc is small, the latency decreases, since the average throughput values change more quickly, at the expense of some throughput. There are other schedulers like Round Robin scheduler (RR scheduler) which schedules users one after another without any priority and greedy scheduler, which schedule the users based on their instantaneous rates by ignoring the average throughput and sacrificing the fairness.
25 | P a g e
CHAPTER 4
4. Mathematical Procedure In this chapter we introduce formulation used to calculate CQI and capacity of users.
4.1 MU-MIMO Capacity Formulation Referring to the system model in figure 3.1 the received signal vector for 1st user can be expressed as
y1 H1Wx n1
(4.1)
Now the transmitted symbol for the 1st user xi is to be estimated from this received vector. For traditional MIMO detection a linear receiver is used to detect the transmit data. ZF, MMSE and MRC detection criterions are commonly employed. In order to obtain good performance, we consider a linear MMSE receiver equation discussed earlier and can be rewritten as
x1 x 2 y1 H1 v1v 2 vk n1 xk y1 H1v1x1
K
H vx 1 i
i
n1
(4.2)
i 1,i 1
The first term represents the desired signal, the second term is inter stream interference caused by scheduling more than one user and the third term is complex Additive White Gaussian Noise (AWGN) at the receiver. Let the effective channel after precoding be expressed as
˜ 1 = H1W, where H ˜ 1 is N H
×
K precoded channel matrix.
26 | P a g e
4.1.1 MMSE
Receiver
MMSE receiver will try to reduce the MSE between the desired and estimated symbols. The linear 1 × N MMSE receiver b1 to decode 1st user data can be expressed as
b1 min {|x1 x1 | } min {|x1 y1 | } p 2
2
b
b
1
*R
(4.3)
where •
R is N
× N
•
p is N
×
auto-correlation matrix of received vector y1 .
1 cross-correlation vector between the desired symbol x1 and received
vector y1 .
Assuming noise and data are i.i.d and uncorrelated random vectors. The total power constraint P is divided equally among K users, expressions for R and p is calculated as follows
R {y1 y1 } *
1 { XX }1* {n1n1* } *
P * 1 1 N0 I N K
=
(4.4)
p { y1 x1 } *
1v1{x1 x1*} P 1 v1 K
=
P h1 K
=
(4.5)
27 | P a g e
˜1 where h
=
H1 v1
=
˜ 1 ]1 [H
=
[H1 W ]1 and
“
∗ ”
indicates conjugate transpose operation. The
final expression for MMSE filter is written as
KN 0 * b1 h 1 1 I N P
1
* 1
(4.6)
From the matrix inversion term, it is evident that MMSE receiver will try to reduce inter stream interference but cannot remove it completely. Here the criterion is not to make the inter stream interference zero but to minimize the MSE.
4.1.2 CQI
Calculation
Let us calculate the C QI of the 1st user. The estimated symbol at the k th MS can be written as
K
b H v x b n
x1 b1 y1 b1H1v1x1
1
i 1,i 1
1 i
i
1 1
(4.7)
where bl is MMSE receiver vector. Assuming the total power P is divided equally among K users, expressions for signal (S), interference (I), and noise powers (N) of 1st user is calculated as follows
S {(b1H1v1x1 )(b1H1v1x1 )*}
=|b1H1v1 |2 E{x1x1*} P 2 |b1H1v1 | K
=
I
(4.8)
K
{(b H v x )(b H v x ) } *
1
1 i i
1
1 i i
i 1,i 1
K
* 1 1
= E{x x }
2 |b1H1vi |
i 1,i 1
P K 2 = |b1H1vi | K i 1,i 1
(4.9)
28 | P a g e
*
N = E{(b1n1 )(b1n1 ) } *
*
= b1 E{b1n1 }b1
b1 2 N 0 S
CQI 1
(4.10)
I +N
P 2 |b1H1v1 | K P K 2 2 |b1H1vi | b1 N 0 K i1,i 1
(4.11)
Here “ . ”indicates norm of the vector. Since noise is assumed to be Gaussian distributed, Capacity of 1st user C1 can be calculated from Shannon’s channel capacity theorem as
C1
=
log2 (1 + C QI1)
(4.12) After calculating all individual capacities, the sum capacity can be calculated by adding individual capacities. .
C sum
K
C i
i 1
(4.13)
29 | P a g e
4.2 Procedure This section explains how abstraction is performed at the BS and MS. 1. In case of MU-MIMO each user has to feedback the following information to the BS. (a) CQI for all the subbands. (b) Stream Indicator or Precoding vector index for all the subbands. The above information can be calculated based on the desired and interfering channels to each user. Note that the feedback sent by the MS during frame number (n) is used to schedule users in frame number (n+1). Hence the feedback path is indicated as frame (n+1). 2. Now it is the job of PF Scheduler to select which set of K users need to be scheduled out of U users in a particular subband based on the above feedback. (a) First the PF Scheduler will calculate the PF metrics for all users. (b) It will try to find set of K users who prefer different stream indicators. (c) If there are more than one set of users then it will select those set of users who has maximum sum PF metric. (d) If there is no such set of users then scheduler will randomly force the users to use different stream indicator so that pairing can be done. 3. After scheduling a dataregion the average throughput of all the users are updated. Then the scheduler will schedule the users for next dataregion. This way scheduler will schedule users to all dataregions one by one. When there are large numbers of users contending for servicethen pairing is not a problem, since there is very high probability that at least K users will choose different stream indicators. 4. After scheduling, the SINR is calculated at the MS. 5. The SINR calculated in the previous step is used to calculate performance metrics like sum capacity and throughput.
30 | P a g e
BS
MS
frame (n) Scheduler
frame (n) Desired and CQI and precoding vector calculation
interfering channel information
frame (n+1)
Feedback from MS: 1) CQI 2) PVI
frame(n)
SINR calculation
frame (n+1)
Performance Metrics: 1) sum capacity 2) Throughput
Figure 4.1 Procedure for MU-MIMO.
frame (n)
31 | P a g e
CHAPTER 5
5. Results and Discussion 5.1 Matlab Simulation Results In this section we will discuss how the capacity will effect when multiple users are paired in a dataregion. Here in total we have 12 subbands and 20 users competing for resources.
System Parameters
Parameters
Values
No. of antennas at BS station
2/4
No. of antennas at MS station
2/4
Frames transmitted
100
Channel
uncorrelated/correlated
Subbands
12
Iterations
100
SNR (dB)
0:1:50
Users contending for resource
20
Users served
[234]
Channel subbands
repetition
across 1
32 | P a g e
Figure 5.1 represents the sum capacity of MU-MIMO for 4 ×2 w i t h u n c o r r e l a t e d channel. In this figure we clearly observe that the sum capacity of 2 users is higher when compared to 3 users and 4 users. It is well understood that as the number of users increases, the inter stream interference will increase hence the CQI of each user will decrease and also the capacity per user decreases. Since the number of antennas at the mobile station are only two, the receiver can suppress only one interferer effectively resulting in higher sum capacity in case of 2 users when compare to sum capacity of 3 users and 4.
Figure 5.2 represents the sum capacity of MU-MIMO for 4×1 with uncorrelated channel. In this figure, we can clearly notice that the sum capacity of 2 users is higher when compared to that of 3 and 4 users. The main difference between the two graphs is that, in figure 5.1 the sum capacity for 2 users increases linearly, where as in case of figure 5.2 the sum capacity increases exponentially.
Figure 5.3 and figure 5.4 represents the sum capacity of MU-MIMO for 4×2 and 4×1 with correlated channel. Here also, as the number of users increases the sum capacity decreases. In the case of single receiving antenna (i.e. M=1) the sum capacity of users is smaller when compared to that of two receiving antenna at the MS. For both correlated and uncorrelated cases, as the number of users increases, the sum c apacity decreases.
33 | P a g e
SUM CAPACITY of MU-MIMO for 4Tx X 2Rx,UnCorrelated channel z 35 H / c e s / s 30 l o b m y 25 s n i d n 20 a b b u s r 15 e p Y T I 10 C A P A 5 C M U S 0
2-Users 3-Users 4-Users
0
5
10
15
20
25 SNR in dB
30
35
40
45
50
Figure 5.1Sum capacity of MU-MIMO for 4×2 with uncorrelated channel
z H / 12 c e s / s l o b 10 m y s n i d 8 n a b b u s r 6 e p Y T I C A 4 P A C M U 2 S 0
SUM CAPACITY of MU-MIMO for 4Tx X 1Rx,UnCorrelated Channel 2-Users 3-Users 4-Users
5
10
15
20
25 SNR in dB
30
35
40
45
Figure 5.2 Sum capacity of MU-MIMO for 4×1 with uncorrelated channel
50
34 | P a g e
z H35 / c e s / s 30 l o b m y s 25 n i d n 20 a b b u s 15 r e p Y 10 T I C A P 5 A C M U 0 S 0
SUM CAPACITY of MU-MIMO for 4Tx X 2Rx,Correlated channel 2-Users 3-Users 4-Users
5
10
15
20
25 SNR in dB
30
35
40
45
50
45
50
Figure 5.3 Sum capacity of MU-MIMO for 4×2 with correlated channel
z H / 12 c e s / s l o b 10 m y s n i d 8 n a b b u s r 6 e p Y T I C A 4 P A C M U 2 S 0
SUM CAPACITY of MU-MIMO for 4Tx X 1Rx,Correlated channel 2-Users 3-Users 4-Users
5
10
15
20
25 SNR in dB
30
35
40
Figure 5.4 Sum capacity of MU-MIMO for 4×2 with correlated channel
35 | P a g e
From figures 5.1 to 5.4, we can clearly observe that
As the number of users increases the sum capacity decreases.
The optimum number of users that can be scheduled in a dataregion to achieve maximum throughput is equal to minimum of the number of antennas at the BS and MS.
The sum capacity increases linearly with S N R (dB) when number of users paired is not greater than minimum of the number of antennas at the BS and MS i.e., when K min {M, N} then sum capacity increases linearly, otherwise it saturates at some point.
36 | P a g e
CHAPTER 6 CONCLUSIONS
MU-MIMO is a promising technique which allows more than one user that can be served in each subband. An efficient multi-user proportional fair (PF) scheduler algorithm is designed and implemented in MU-MIMO technique with code book based precoding that has been proposed for the IEEE 802.16m mobile broadband standard. Optimum number of users can be scheduled in a dataregion to achieve maximum sum capacity, which is equal to minimum number of antennas at the base station and the mobile station. The sum capacity increases linearly with SNR (db) when the number of users paired is not greater than the minimum of the number of antennas at the BS and MS i.e. when K min {M, N} then sum capacity increases linearly, otherwise it saturates at some other point. From the results discussed in the previous chapter, multiple users can be paired in a dataregion resulting in higher sum capacity. In addition, we observe that as the number of user’s increases, the sum capacity decreases.
37 | P a g e
REFERENCES
[1] Riberio C. B, Hugl K, Lampinen M, “Performance of Linear Multi-User MIMO Precoding in LTE system,” Third International symposium on Wireless pervasive Computing (ISWPC 2008), pp. 410-414, 2008.
[ 2 ] C a t r e u x S , G r e e n s t e i n L . J , E r c e g V , Some results and insights on the performance “
gains of MIMO systems ,
”
IEEE Journal on selected areas in communications, vol. 21, June
2003. [3] Fang Shu, Li Lihua, Cui Qimei and Zhang Ping, Non-Unitary Codebook Based Precoding “
Scheme for Muiti-User MIMO with Limited Feedback , IEEE Wireless Communication and ”
Networking conference, pp. 678-682, 2008. [4] C. Peel, B. Hochwald, A. Swindlehurst,
Vector-Perturbation Technique for near-capacity
“
muitiantenna muitiuser communication- Part I: Channel Inversion and Reguiarization, IEEE ”
Transactions on Communications, vol. 53, January 2005. [5] C. Peel, B. Hochwald, A. Swindlehurst,
Vector- Perturbation Technique for near-capacity
“
muitiantenna muitiuser communication- Part II: Perturbation,” IEEE Transactions on Communications, vol. 53, January 2005. [6] M. Costa, “Writing on dirty paper,” IEEE Transactions on Information Theory,
vol.
29, no. 3,
pp. 439-441, May 1983. [7] Stankovic V and Haardt M, Generaiized Design of Multi-User MIMO Precoding “
Matrices
, ”
IEEE Transactions on communications, vol. 7, no. 3, March 2008. [8] Fang Shu, Wu Gang, Xiao Yue, and Li Shao-qian, Multi-User MIMO Linear Precoding “
with Grassmannian Codebook, International Conference on Communications and Mobile ”
Computing , pp. 250-255, January 09.