multiple antenna apwcs12

Multiple-antenna Techniques in LTE-Advanced and Beyond

Multiple-antenna Techniques in LTE-Advancedand Beyond

Byonghyo ShimKorea University

IEEE VTS APWCS12 Presentation

August 24, 2012


Table of contents

Wireless Cellular Communications

Multiple Antennas in WCDMA/HSPA Systems

Multiple Antennas in LTE Systems (SU/MU-MIMO)

Multiple Antennas in LTE-Advanced Systems (CoMP/Relay)

Multiple Antenna technique in the future



Wireless CommunicationsI Without wire : reachable anywhereI Various topologies (one to one, one to many, many to one,

etc.)I Interference problemI Fading



Wireless Systems - Antennas

I Transform wire propagated waves into space propagatedwaves and vice versa.

I Means to achieve degree of freedom of wireless systems(spatial, temporal, spatio−temporal, etc.)



Benefits of Multiple Antennas in Wireless Communications

I Provide more resolvable signal paths : diversity gain



Benefits of Multiple Antennas in Wireless CommunicationsI Supports more data streams : multiplexing gain

0 5 10 15 20 25 300

5

10

15

20

25

30

35

Es/N0 (dB)

Capacity (

bits/c

hannel use)

SISO

2x2 MIMO

4x4 MIMO



Benefits of Multiple Antennas in Wireless CommunicationsI Higher chance to select good quality mobile : opportunistic

gain

I When the number of users in a cell is large, almost surelythere exists a user whose channel strength is much larger thanthe average. By allocating system resources to that user,multiuser diversity is fully capitalized



Benefits of Multiple Antennas in Wireless Communications

I Allow collaborative communication : cooperative gain



Multiple Antennas in 3G Wireless Communications(WCDMA, HSDPA, HSPA)

System model: yn = hn ∗ xn + vn (xn is transmitted chip)

I Temporal diversity : rake receiverI Tx diversity

I Alamouti coding: ML-detection with only linear processing[y1y2

]=

[h1 h2h∗2 −h∗1

] [x1x2

]+

[w1

w∗2

]I Rx diversity

I Equalization using CIR with best power hin(i = arg max{|h1n|, |h2n|})

I Joint equalization x̂n = E [xnyHn ]R−1ynyny

I All of these approaches are suboptimal!




Complicated system model: y = HCWGd + v (C: scramblingmatrix, W: ovsf matrix, G: gain matrix)

I Latest release of HSPA system adopts 2x2 MIMO system

I Hard to implement

I Frequency domain processing (may use circular approximationof H)

I SINR after equalization SINR = PIinter+Iintra+N0

= PαP+N0




I SINR after equalization SINR = PIinter+Iintra+N0

= PαP+N0

(saturation in performance P →∞)I Solution for interference is required to improve spectral

efficiency

0 5 10 15 20 25 30 350

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

SNR (dB)

Capacity (

bps/H

z)




I Type3i receiver : interference aware LMMSE receiver

I IC algorithm: interference cancellation and equalizationalgorithm (e.g., Qualcomm Q-ICE)

I When multiple antenna techniques are combined, systemmodel becomes very complicated



LTE - OFDM based SystemI Cyclic prefix + frequency domain symbol processing → free

from intracell interferencey = H̃x + vYi = λiXi + Vi

I scalar processing per subcarrierI New start from the simple scalar model



LTE - New Framework to incorporate Multiple Antennas

I Yi = λiXi + Vi → Allow per subcarrier processing withoutintracell interference

I Provide nice and clean framework for building up multipleantenna techniques

I Multiple Tx/Rx antennas can be integrated per subcarrier →Singleuser MIMO

I Basestation with multiple transmit antenna is transmittingsignal to multiple receiver with one or more antenna →Multiuser MIMO

I Cooperation among multiple basestations is possible → CoMP



Multiple Antenna Techniques in LTE Systems

Brief historyI Rel. 8-9 (LTE System)

I Receive diversity (RxD)I Transmit diversity (TxD)I Cyclic delay diversity (CDD)I SU-MIMOI MU-MIMO (primitive form)

I Rel. 10 and beyond (LTE-Advanced)I High-dimensional SU-MIMO (e.g., 8× 8)I Elaborated MU-MIMOI Coordinated Multipoint (CoMP) transmission and receptionI Relay (primitive form)



SU-MIMO in LTE Systems

I The SU-MIMO scheme is applied to the physical downlinkshared channel (PDSCH)

I Support both closed-loop (codebook based) and open-loopoperation




I Open-loop: when reliable PMI feedback is unavailable at theeNB.

I Closed-loop: UE needs to feedback the channel qualityindicator (CQI), precoding matrix indicator (PMI), and rankindicator (RI).




I Channel quality indicator (CQI)I The CQI index is the largest index guaranteeing BLER < 0.1.




I Precoding matrix indicator (PMI)I Indicates the preferred codebook element of the resource block.

I Rank indicator (RI) : the number of layers supportedI Open loop MIMO : the RI decides transmit diversity (RI = 1)

or open-loop spatial multiplexing (RI > 1).I Single RI represents whole bandwidth.




I Layer= 1 and layer= 2.



Limitations of SU-MIMO

I Capacity of SU-MIMO is scaled by min{M,N} at high SNR.However...

I The number of receive antennas N at the battery poweredmobile devices is smaller than M (limit the capacity gain).

I SU-MIMO systems are susceptible to the ill-behavior ofpropagation channels (LOS, correlations among antennaelements).

I As an alternative option, MU-MIMO has received muchattention



MU-MIMO System

I Single subcarrier supports multiple distinct receivers havingone or more antennas.

I Research focus is more on Tx.



Scaling Law of the MU-MIMOI Capacity of MU-MIMO is scaled by min{M,Nnuser}

limSNR→∞Csum

log SNR = min{M,Nnuser}I In contrast to SU-MIMO, as long as Nnuser is large, capacity

can be scaled by M

0 5 10 15 20 25 300

5

10

15

20

25

30

35

40

SNR(dB)

Sum

Rate

(bps/H

z)

4Tx,1Rx,4User

2Tx,1Rx,2User

3Tx,1Rx,3User



MU-MIMO in LTE - In Reality

I Finite feedback: Tradeoff between feedback accuracy anduplink resource consumption → Codebook based precoding(PMI feedback)

I Suboptimal user scheduling: user selection at the basestation(eNB) is imperfect due to the inaccurate CQI. Also, CQI inthe mobile cannot be accurate due to the lack of knowledgeon precoding information of other users

I Pilot issue (common pilot or dedicated pilot?)



MU-MIMO in LTE Systems

Reference signal (pilot) for measuring parameters: common pilotand dedicated pilot

I Common pilot (CRS): yp = hsp + v (pilot) and yd = hpsd + v(data)

I Common for all subcarriers; the mobile needs to obtain theprecoding matrix information via the control signaling (onlycodebook based precoding)

I Dedicated pilot (DM-RS) : yp = hpsp + v (pilot) andyd = hpsd + v (data)

I Specific to the user; transparent operation is possibleI Non-codebook based precoding (e.g., ZF, MMSE) is possible



Spectral Efficiency for Various Cellular Systems

WCDMA (R99)

HSPA (Rel. 5)

HSPA (Rel. 6)

HSPA (Rel. 7)

WiMAX (Rel. 1)

LTE (Rel. 8)

LTE-A (Rel. 10)

Ave

rage sp

ectra

l efficie

ncy

(bps/H

z/cell)

1.0

2.0

3.0

?

Beyond LTE-A

I Solution for the interference (cell-edge scenario) is veryimportant issue to improve spectral efficiency.



Coordinated MultiPoint (CoMP) Transmission/Reception

I CDMA systems has scrambling code but no cell-edgeprotection for the OFDM systems → susceptible to theintercell interference (y = hs + hI sI + v).

I The premise of CoMP is to share information to reduceintercell interference or harness it to improve network capacity.



CoMP schemes in the downlinkI Mobile (UE) may be able to receive signals from multiple cell

sites by joint processing.I Information is transmitted from a single eNB in coordinated

scheduling/beamformingI In both schemes, eNBs are connected via backhaul to

exchange scheduling and beamforming information.



Joint Processing

I In joint processing, data is transmitted from different point ata time and then coherently combined at the terminal.

I Joint processing achieves rate gain but requires more systemresources.

log2

(1 +

Pt‖∑L

i=1H(i)k w(i)‖2

σ2

)

> log2

(1 +

Pt‖H(k)k w(k)‖2

σ2 + Pt‖∑L

i=1,i 6=k H(i)k w(i)‖2

)



Coordinated scheduling/beamforming

I Received signal model at user k is

yk = H(k)k w(k)√pksk +

∑Li=1,i 6=k H

(i)k w(i)√pi si + nk

I Various beamforming methods (e.g., maximizing SINR/SLNR,ZF, MMSE) exist.



I Beamforming generation methods

- Maximizing SINR

w(k)opt = arg max

w(k)

Pt‖H(k)k w(k)‖2

σ2 + ‖∑L

i=1,i 6=k H(i)k w(i)‖2

- Maximizing SLNR

w(k)opt = arg max

w(k)

Pt‖H(k)k w(k)‖2

σ2 + ‖∑L

i=1,i 6=k H(i)k w(k)‖2

- ZF

w(k) = HH(HHH

)−1, H = [H1H2 · · ·HL] .

- MMSE

w(k) = H(k)k

H(

L∑i=1

H(i)k H

(i)k

H+σ2

Pt)−1



Relay Technique

I Relay increases throughput and extends coverage of mobilecommunication systems.

I AF, DF, CF, etc.



AF Relay Technique

I Amplify and forward (AF) techniqueI Amplifies relay received signals and transmits.

I Already used in 2G and 3G systems.I To improve coverage in urban areas (e.g., hot spot, subway),

indoor environments.I Low cost, short processing delay but amplifies inter-cell

interference and noise simultaneously.



DF Relay Technique

I Decode and forward (DF) techniqueI Relay decodes the received signals and re-encodes for

transmission.

I Pros and consI Perfect elimination of noise.I Processing delay due to mod/demod and enc/dec.



DF Relay Technique

I Additional processing at relay for DFI User-data regeneration processing.I User-data transmission processing.

I Agreed to standardize specifications in LTE Rel.10.



Interference Channel – Status

I No cell-edge protection for the OFDM systems.

I These days, small cells (pico/femto cells) are prevalent forimproving cell throughput → solution for intercell interferencerequired

I Uncovering fundamental limit of interference channel is stillopen problem (No essential progress from Han andKobayashi’s result)

I Instead of performing cancellation/mitigation/suppression,why don’t we harmonize interferences?



Possible Options for Interference Channel – InterferenceAlignment (IA)

I What is needed for a receiver to distinguish a desired symbolfrom the interference?

Y = H1x1 + H2x2 + · · ·+ HKxK

I Signal space should be orthogonal to the interference space

H1 6∈ span (H2, · · · ,HK )



Interference Alignment (IA) - Illustrative Example

I Signal space orthogonalization can be achieved eithertemporal domain or spatial domain.



Interference Alignment (IA)

I IA on the 3 User Interference Channel with 2 antenna nodes.

I Since only one of the unknowns is desired, IA makes itpossible to recover its value by aligning the two undesiredsymbols along the same dimension at each receiver.

I Problem: inaccurate CSI might deteriorate aligningperformance.



Massive MIMO: large Tx/Rx antennas

I Strong candidate for achieving improvement in spectralefficiencies (study items in Rel. 12 and beyond)

I How can we support dozens of antennas → Antenna, RF,DSP, etc.



In an Ideal Scenario...

I Example of Massive-MIMO MAC (N × K channel)I N : the number of basestation antennasI K : the number of users

y = Hx + n

I By matched filtering (MF), the received signal becomes

1

NHHy =

1

NHHHx +

1

NHHn

N→∞−→ x

I The transmit power (per each antenna) can be made arbitrarily

smallI Full DOF (proportional to the number of users) can be achievedI Uncorrelated interference and noise vanish as N → ∞



Pilot contamination problemI Number of pilot, length of pilot sequences increase with the

number of antennasI If the number of antennas N is larger than the coherence

interval T , then orthogonality can no longer guaranteedI Training with non-orthogonal pilots leads to bad channel

estimate, resuling in severe inter-cell interference



Receiver... What should we do?

I A practical scheme to achieve full multiplexing gain → Iterativedetection and decoding (IDD) algorithm

I Perform symbol detection and channel decoding in an iterativefashion

I Exploit the soft-output from each other as a prior informationI A posteriori probability (APP) is expressed as log-likelihood

ratio (LLR)

LD(bk |y) = lnP(bk = +1|y)

P(bk = −1|y)

= lnP(bk = +1)

P(bk = −1)︸︷︷︸LA

+ ln

∑bk+

P(y|b) exp(12b

Tk LA,k

)∑bk−

P(y|b) exp(12b

Tk LA,k

)︸︷︷︸LE



IDD Receiver... What should we do?I Feedback of prior information gradually improves the

detection/decoding qualityI An approximate low-complexity APP detection required (e.g.,

group list detection, probabilistic tree-pruning detection)



Overall Structure of IDD



Summary of Multiple Antenna Technologies

I Multiple antenna techniques have been one of most successfulstory of wireless communications in the last decade

I Information theoretical concept has been commercialized in ashort period of time

I Fruitful field and still many important problems exist

I Future keywords include M2M, ad-hoc, massive MIMO, · · ·

multiple antenna apwcs12

Documents