MIMO and Transmit Diversity for SC-FDMA

Vision Innovation Speed Performance

MIMO and Transmit Diversity for SC-FDMA

Donald Grieco

InterDigital, Inc.

Melville, NY

March 13, 2009

© 2009 InterDigital, Inc. All rights reserved.

2 © 2009 InterDigital, Inc. All rights reserved.

Outline of presentation

• Benefits of using MIMO and transmit diversity

for uplink transmission

• General description of MIMO/transmit diversity

for SC-FDMA

• Transmit diversity techniques

• MIMO techniques

• MIMO receivers

• MIMO performance

• SC-FDMA vs. OFDMA MIMO comparisons






for SC-FDMA


• MIMO techniques

• MIMO receivers



Benefits of using MIMO for uplink transmission

• Improved spectrum efficiency for the uplink

• Take maximum advantage of a multiple antenna

solution for the terminal

• Improved bit rate and robustness at the cell edge

• Reduced inter-cell and intra-cell interference due to

beamforming

• Improvement in system capacity even when

considering additional feedback signaling overhead

• Reduced average transmit power requirements at the

terminal by operating at a lower received SNR.

• Use of transmit diversity for control information in the uplink

4







for SC-FDMA


• MIMO techniques

• MIMO receivers



Single-codeword MIMO beamformer block diagram

• Single codeword requires only one HARQ process and less signaling overhead

• Channel state info, fed back from base station, based on channel sounding from

terminal

• Spatial transformer can implement beamforming or precoding

• Pilots are beamformed/precoded by time (or frequency) multiplexing with data


Channel Encoder

Puncture

Frequency Interleaver


Constellation

Mapping

Constellation

Mapping

iFFT

iFFT CP

CP

Data

Pilots

Pilots

Spatial

Parser

Mux

Mux

Channel State

Information

Reverse Link for Transmitter Beamforming

-

FFT

FFT

Control

Control

(Spatial Multiplexing

BeamformingSpatial

Spreading)

2-D

Sp

atia

l Tra

nsf

orm

Su

b-C

arr

ier

Ma

pp

ing

Dual-codeword MIMO beamformer block diagram

• Dual codewords shown requires two HARQ processes [1]

• Can use simple CRC-based SIC receiver to improve performance

• Precoder can implement spatial multiplexing codebook or transmit diversity

• Codebook index, fed back from base station, based on channel sounding from terminal

• Pilots, shown not precoded, can be used for both channel estimation and channel

sounding to select precoder


Channel Encoder

Rate

Matching



Constellation

Mapping

Constellation

Mapping

Data

-

FFT

FFT

Spatial Multiplexing

BeamformingSpatial

Spreading)

Pre

cod

er

Su

b-C

arr

ier

Ma

pp

ing

Mux

Mux

Pilots

Pilots

iFFTVariable

size

CP

iFFTVariable

size

CP

Channel Encoder

Rate

Matching

DeMux

Precoder

GeneratorCodebook

index






for SC-FDMA


• MIMO techniques

• MIMO receivers



Transmit diversity techniques

• Transmit antenna selection/switching (TAS)

• Precoding vector switching (PVS)

• Space-time block code (STBC)

• Space-frequency block code (SFBC)

• Frequency switch transmit diversity (FSTD )

• Cyclic-delay diversity (CDD)


Summary of uplink 2-TxD schemes


Scheme Pros Cons

Slot-based

TAS/PVS

•Low PAPR

•Transparent to the receiver

•Low diversity gain

CDD •Preserves single-carrier (SC)

property

•Requires one set of precoded

pilots

•Poor performance in correlated

channels

•No uncoded diversity

STBC •Preserves SC property

•Uncoded diversity

•Needs even number of symbols

•Requires two sets of precoded pilots

SFBC •Uncoded diversity •Breaks SC property


FSTD •Preserves SC property

(with two DFTs)

•No uncoded diversity


Refs [3], [4]






for SC-FDMA


• MIMO techniques

• MIMO receivers



SU-MIMO techniques

• Spatial multiplexing

• Eigen beamforming

• Codebook precoding

• Layer permutation


Spatial multiplexing

• Separate streams, either single or multiple

codewords, are sent on separate antennas

– No precoder feedback overhead

– Each antenna can be rate controlled

• Selective per-antenna rate control (S-PARC) [17]


Eigen beamforming

• For eigen-beamforming the channel matrix is decomposed using a

single-value decomposition (SVD) or equivalent operation as

• The 2-D transform for spatial multiplexing, beamforming, etc. can be

expressed as

where the matrix T is a generalized transform matrix.

• In the case when transmit eigen-beamforming is used, the transform

matrix T is chosen to be a beamforming matrix V which is obtained

from the SVD operation above, i.e., T = V.

• Requires feedback of either channel matrix H or beamformer matrix T


HUDVH

Tsx

Codebook precoding

• Precoding for spatial multiplexing is defined by

• The values of the precoding matrix W(i) are selected among the

precoder elements in the codebook configured in the BS and the

terminal.

• The codebook index is fed back to the terminal


)(

)(

)(

)(

)(

)1(

)0(

)1(

)0(

ix

ix

iW

iy

iy

P

• The precoder takes as input a block of vectors x(i) from the

layer mapping and generates a block of vectors y(i) to be

mapped onto resources on each of the antenna ports, where

represents the signal for antenna port p. )()( iy p

LTE downlink codebook for 2 antenna ports [5]

• Codewords have constant modulus property

• Number of layers depends on rank of channel


Codebook index

Number of layers

1 2

0

1

1

2

1

10

01

2

1

1

1

1

2

1

11

11

2

1

2

j

1

2

1

jj

11

2

1

3

j

1

2

1 -

Downlink codebooks may be used for uplink (TBD)

Layer permutation

• With layer permutation, also called large-delay

CDD, the data streams are distributed evenly across

all the layers, resulting in equal SNRs at the receiver

• For LD-CDD in LTE, precoding for spatial

multiplexing is defined by [5]

where W(i) is the precoding matrix, and matrix U

and diagonal matrix D(i) supporting cyclic delay

diversity are shown in the following slide.


)(

)(

)()(

)(

)(

)1(

)0(

)1(

)0(

ix

ix

UiDiW

iy

iy

P

Large-delay cyclic delay diversity


Number of layers

U )(iD

2

221

11

2

1je

220

01ije

3

3834

3432

1

1

111

3

1

jj

jj

ee

ee

34

32

00

00

001

ij

ij

e

e

4

41841246

4124844

464442

1

1

1

1111

2

1

jjj

jjj

jjj

eee

eee

eee

46

44

42

000

000

000

0001

ij

ij

ij

e

e

e






for SC-FDMA


• MIMO techniques

• MIMO receivers



Types of MIMO receivers

• Linear Minimum Mean-Square Error (LMMSE)

• LMMSE-Successive Interference Canceller (SIC)

• Turbo-SIC

• Maximum Likelihood Detector (MLD) – List sphere decoder [13]

– QRD-ML [14]


LMMSE receiver for dual codewords


Channel

Estimation

FFT

Spatial

Deparser

FEC

Data Demodulation

Data Demodulation

De-Interleaver

De-Interleaver

Data

Sub-Carrier

Demapping

IFFT

Remove CP

FFTRemove CP

IFFT

MMSE

FEC Data

MIMO detection using an LMMSE receiver can be expressed as

1)~~

(~ vv

H

ss

H

ss RHRHHRR

where R is the receive processing matrix, ssR and vvR are the signal and

interference correlation matrices and H~

is the effective channel matrix which

includes the effect of the precoding matrix on the estimated channel response.






for SC-FDMA


• MIMO techniques

• MIMO receivers



SC-FDMA MIMO performance

• Comparison with SIMO and TAS [8]

• Comparison with transmit diversity [8]

• Single vs. dual codewords [9]


Simulation parameters


Carrier frequency 2.0 GHz

Symbol rate 4.096 million symbols/sec

Transmission bandwidth 5 MHz

TTI length 0.5 ms (2048 symbols)

Number of data blocks per TTI 6

Number of data symbols per TTI 1536

FFT block size 256

Cyclic Prefix (CP) length 7.8125 μsec (32 samples)

Channel model Typical Urban (TU6)

Antenna configurations 1x2 (SIMO), 1x2 (TAS), 2 x 2 (MIMO)

Fading correlation between transmit/receive

antennas = 0

Moving speed 3 km/h

Data modulation QPSK and 16QAM

Channel coding Turbo code with R = 1/2, 1/3 and

soft-decision decoding

Equalizer LMMSE

Feedback error None

Channel Estimation Perfect channel estimation

Comparison of MIMO with SIMO and TAS

• The figure below compares throughputs for TxBF without AMC compared with a

single transmit antenna or transmit antenna switching with AMC [18]

– At 5 Mbps TxBF using coding rate 1/3 exhibits about 4.5 dB advantage over TAS and about

5 dB over a SIMO

– At 8 Mbps TxBF using a coding rate of ½ exhibits about 4 dB advantage over TAS and

about 5 dB over SIMO.


-2 0 2 4 6 8 10 12 14 16 18 0

1

2

3

4

5

6

7

8

9

10

SNR in dB

TU Channel, rate 1/2 and 1/3

16QAM QPSK,1/2, BF

16QAM QPSK,1/3, BF

SIMO 1x2

TAS 1x2

Throuput in Mbps

Comparison of TxBF with AMC vs. transmit diversity


gain (dB) TxBF to SFBC vs. throughput Throughput TxBF vs SFBC (16QAM, ½) TxBF vs SFBC (16QAM, 1/3)

1 Mbps 5.5 4.2

2 Mbps 4.5 3.6

3 Mbps 2.9 1.4

4 Mbps 1.9 0.9

TxBF achieves much higher throughput, 9.2 Mbps, than SFBC 1/2 and 1/3 which

achieve maximum data rate 6.1 and 4.1 Mbps [8]

Comparison of single and dual codewords for UL MIMO


-5 0 5 10 15 20 25 30 0

2

4

6

8

10

12

14

SNR dB

Throughput (MBPS)

Throughput comparison of Single codeword and double codeword, SCME-C Channel

Dual CW 16QAM r1/2 - QPSK r1/2




Single CW 16QAM - QPSK r1/2




• In the case of DCW, with TxBF the upper eigenmode has higher SNR than the total system

SNR. Thus at low SNR that codeword contributes some successful transmissions while the

lower codeword generally does not.

• In the case of SCW, the upper stream protects the lower stream because the coding covers

both streams. This results in an overall lower BLER for SCW at higher SNRs [9].






for SC-FDMA


• MIMO techniques

• MIMO receivers



Evaluation of UL transmission schemes for LTE-A

• Multiple access schemes:

– SC-FDMA: As in LTE, contiguous RBs are allocated for each UE.

– Clustered DFT-S-FDMA (2,3,4,8 equal sized clusters): DFT precoding output

is mapped to multiple clustered RBs [11].

– OFDMA: As in LTE DL, per RB (or sub-band) based frequency dependent

scheduling is considered.

• MIMO techniques:

– Assuming 2 transmit antennas at the UE, we consider Rank-1 and rank-2

precoding with the E-UTRA downlink precoding codebook [5].

– In SC-FDMA, wideband precoding is used such that a selected precoding

vector/matrix is used for all the allocated RBs. In clustered DFT-S-FDMA, one

precoding matrix/vector is used per cluster, while in OFDMA two options are

considered, precoding per RB and per 3 RB [10] .


Clustered DFT-S-OFDMA with chunk specific filter


• Clustered transmission in the frequency domain is realized using non-contiguous

sub-carrier mapping [11]

– Unlike in SC-FDMA, the DFT precoded data is mapped to multiple subcarrier clusters each

having contiguous sub-carrier mapping in frequency

D F T

Su b - car - rie

M a p - p ing

Filter & Cyclic Extens . Filter & Cyclic Extens .

Filter & Cyclic Extens .

IFFT

CP Insertion Modulator

Channel

Coding

Code Block

segmentation

20 MHz chunk

System simulation assumptions for uplink throughput


Parameter Assumption

Transmission bandwidth 20 MHz

Total number of RBs allocated 24

Subband (cluster) size for DFT-S-FDMA 24, 12, 8, 6, 3 RBs

Subband size for OFDMA 1, 3 RBs

MCS As per [15]

Antenna configurations 2 x 2

Multiple access (MA) scheme SC-FDMA, Clustered DFT-S-FDMA, OFDMA

MIMO configuration LTE downlink precoding: Rank-1 and 2

Channel type SCM, PA, VA

Resource allocation Best M clusters from partition system BW

Receiver LMMSE

Ref. 10

Throughput performance with non-power limited geometry

• OFMDA outperforms Clustered DFT-S-OFDMA and SC-FDMA, due to its

inherent advantage [16].

• The performance of Clustered DFT-S-OFDMA improves as the number of

clusters increases, for a given total number of RBs (equivalently, the cluster

size gets smaller). This is due to more frequency-scheduling flexibility [10].

32


Rank-2 MIMO (VA channel)

-2 0 2 4 6 8 10 12 140

0.5

1

1.5

2

2.5

3

3.5

4

4.5

Es/No (dB)

Norm

alized t

hro

ugput

(bits/s

ec/s

ubcarr

ier)

OFDMA

Clustered DFT: 8

Clustered DFT: 4

Clustered DFT: 2

SC-FDMA

Rank-1 MIMO (SCM channel)

-2 0 2 4 6 8 10 12 140

0.5

1

1.5

2

2.5

3

3.5

Es/No (dB)

Norm

alized t

hro

ugput

(bits/s

ec/s

ubcarr

ier)

OFDMA

Clustered DFT: 8

Clustered DFT: 4

Clustered DFT: 2

SC-FDMA

Throughput performance with power limited geometry

• In Rank-1 precoding MIMO, OFDMA performs worse than other MA schemes in most

of the channel types/conditions under consideration.

• For Rank-2 MIMO, OFDMA performs slightly better than other MA schemes in SCM

and VA channels but performs much worse in the PA channel which is less frequency

selective.

• In both Rank-1 and Rank-2 MIMO options, Clustered DFT-S-OFDMA can provide

better performance than SC-FDMA [10].

33


Throughput gain relative to SC-FDMA, using Rank-1 precoding under various channels

2x2 SCM 2x2 PA 2x2 VA

MA

Es/No

(dB) -2 6 14 -2 6 14 -2 6 14

OFDMA (1 RB) -4% 5% 8% -36% -15% -11% -4% -3% -2%

OFDMA (3 RBs) -11% -1% 2% -36% -16% -11% -6% -5% -3%

Clustered DFT:

Max. 8 clusters 11% 11% 9% 3% 2% 1% 18% 9% 9%

Clustered DFT:

Max. 4 clusters 7% 7% 6% 3% 2% 1% 18% 9% 9%

Clustered DFT:

Max. 3 clusters 6% 5% 5% 3% 2% 1% 14% 8% 7%

Clustered DFT:

Max. 2 clusters 3% 4% 4% 2% 2% 1% 8% 4% 4%

Throughput gain relative to SC-FDMA, using Rank-2 precoding under various channels

2x2 SCM 2x2 PA 2x2 VA

MA

Es/No

(dB) -2 6 14 -2 6 14 -2 6 14

OFDMA 1 RB 6% 17% 22% -13% -13% -4.4% 3.3% 14% 24%

OFDMA 3 RBs 3% 12% 15% -13% -13% -4.5% 3.0% 10% 21%

Clustered

DFT:

Max. 8 clusters

9% 17% 18% 4.9% 4.9% 5.4% 6.9% 20% 25%

Clustered

DFT:

Max. 4 clusters

5% 13% 12% 4.9% 4.9% 5.3% 6.1% 18% 22%

Clustered

DFT:

Max. 3 clusters

4% 9% 9% 4.9% 4.7% 5.0% 4.5% 15% 17%

Clustered

DFT:

Max. 2 clusters

3% 6% 6% 4.4% 3.8% 4.2% 1.9% 8.4% 9.7%

Conclusions on UL transmission schemes for LTE-A

• In non-power limited geometry OFMDA outperforms Clustered DFT-

S-OFDMA and SC-FDMA due to its inherent performance advantage

– Greatest gains are in more frequency-selective channels.

• In power limited geometry where the power is backed off by the CM

increase, OFDMA performs worse than other MA schemes in many

of the considered channel types/conditions.

– The inherent performance advantage of OFDMA cannot make up for the SNR

loss due to backoff at cell edges.

• Clustered DFT-S-OFDMA provides better performance than SC-

FDMA, even when the UE maximum power must be backed off by

the CM increase

– The benefit is a function of the maximum number of clusters and the frequency

selectivity of the channel.


SC-FDMA vs. OFDMA MIMO comparisons [12]


R1-083732 / 2008-09-23

• With 2x4 the performance of SC-FDMA with Turbo SIC is equal to that of OFDMA

• MLD outperforms SIC only in the highest MCS: 64QAM 8/9

• With 2x2 OFDMA performs a bit better

•However, the required SNR is extremely high making this scenario impractical [12]

SCM-C 2X2

0

1

2

3

4

5

6

7

8

9

10

0 3 6 9 12 15 18 21 24 27 30 33

SNR [dB]

Sp

ec

tra

l E

ffic

ien

cy

[B

it/H

z]

OFDMA MLD 2x2

OFDMA SIC 2x2

SC-FDMA SIC 2x2

SCM-C 2X4

0

1

2

3

4

5

6

7

8

9

10

0 3 6 9 12 15 18 21 24SNR [dB]

Sp

ec

tra

l E

ffic

ien

cy

[B

it/H

z]

OFDMA MLD 2x4

OFDMA SIC 2x4

SC-FDMA SIC 2x4

SNR this high is very

hard to achieve in

practice

Agreements reached on uplink scheme for LTE-A

• Uplink shared channel (PUSCH) transmission,

both MIMO and non-MIMO, uses DFT

precoding (SC-FDMA)

• On top of LTE Rel-8 operation:

– Control-data decoupling (simultaneous PUCCH and

PUSCH transmission) supported in addition to TDM

type multiplexing

– For multiple component carriers, non-contiguous

data transmission with single DFT per component

carrier


Summary of presentation

• SC-FDMA used as uplink air interface for 3GPP LTE

– Transmit Antenna Selection is the only transmit diversity

scheme supported

– Uplink MIMO, although shown to offer higher throughput,

was not included in Rel 8 due to lack of time

• SC-FDMA continues as uplink air interface for 3GPP

LTE-Advanced

– MIMO will be incorporated (2x2 and 4x4)

– Transmit diversity will be added


Backup slides


Transmit antenna or Precoding vector switching

• With TAS the transmit antenna can be selected

by the base station (BS) by channel sounding

with each antenna

• Alternatively, the transmit antenna can be

randomly or periodically switched between

symbols or slots

• With PVS virtual antennas obtained by unitary

precoding can be switched or chosen randomly


Alamouti STBC • A well-known space-time block code (STBC) is the Alamouti scheme, which ensures a diversity order equal

to 2 and a low-complexity optimal decoding if the channel remains unchanged during the STBC

transmission time.

• Two symbols s1 and s2 are sent during two transmission periods on two transmit antennas: s1 and –s2* on

antenna 1 and s2 and s1* on antenna 2.

• When combined with SC-FDMA, the Alamouti scheme can be applied on two symbols transmitted on the

same sub-carrier of two successive SC-FDMA symbols as depicted below [2]


DFTIFFT

f1

f2f3f4

SC

-FD

MA

sym

bol 1

Alamouti

STBC

Encoded

data

f1

f2f3f4

f1

f2f3f4

f1

f2f3f4

SC-FDMA

symbols to be

transmitted on

antenna 1

IFFT

f1

f2f3f4

f1

f2f3f4

SC-FDMA

symbols to be

transmitted on

antenna 2

s'1

s'2

s'3

s'4

s1

s2

s3

s4

SC

-FD

MA

sym

bol 2

SC

-FD

MA

sym

bol 1

on

ante

nna

1

-s'1*

-s'2*

-s'3*

-s'4*

s1

s2

s3

s4

SC

-FD

MA

sym

bol 2

on

ante

nna

1

s'1

s'2

s'3

s'4

s1*

s2*

s3*

s4*

SC

-FD

MA

sym

bol 1

on

ante

nna

2

SC

-FD

MA

sym

bol 2

on

ante

nna

2

Alamouti SFBC • The Alamouti scheme can also be applied on two symbols transmitted on two adjacent sub-carriers of the

same SC-FDMA symbol. The Alamouti scheme can be modified as depicted below [2] in order to guarantee

the same cubic metric (CM) as STBC on both transmit antennas.

• In practice, the N sub-carriers are divided into two groups of N/2 sub-carriers and within each group the

Alamouti scheme is applied on the first sub-carrier and the last sub-carrier, on the second sub-carrier and

the next to last sub-carrier, …, on N/4th sub-carrier and the (N/4+1)th sub-carrier.


DFT IFFT

SC-FDMA

symbol

Alamouti

SFBC

with low

CM

Encoded

data

s1

-s4*

s3

-s2*

f1

f2f3f4

f1

f2f3f4 SC-FDMA

symbol to be

transmitted on

antenna 1

s2

s3*

s4

s1*

SC-FDMA

symbol to be

transmitted on

antenna 2

SC-FDMA

symbol

s5

s7

f5

f6f7f8

f5

f6f7f8

s6

s8

IFFT

f1

f2f3f4

f1

f2f3f4

f5

f6f7f8

f5

f6f7f8

-s8*

-s6*

s7*

s5*

s1

s3

f1

f2f3f4

f1

f2f3f4

s2

s4

s5

s7

f5

f6f7f8

f5

f6f7f8

s6

s8

Frequency switch transmit diversity

• The FSTD scheme can be easily applied by mapping symbols to be

• transmitted on the first antenna on odd sub-carriers and symbols to be transmitted

on the second antenna on even sub-carriers as depicted below [2]


DFTIFFT

s1

0

s2

0

f1

f2f3f4

SC-FDMA

symbol

Encoded

data

f1

f2f3f4

SC-FDMA

symbol to be

transmitted on

antenna 1

IFFT

SC-FDMA

symbol to be

transmitted on

antenna 2

DFT

0

s'1

0

s'2

f1

f2f3f4

f1

f2f3f4

S/P

Cyclic delay diversity

• The CDD scheme can be easily applied by cyclically shifting the

signal transmitted on the second antenna with respect to the one

transmitted on the first antenna as depicted below [2]

• Either small-delay or large-delay CDD can be used


DFTIFFT

s1

s2

s3

s4

f1

f2f3f4

SC-FDMA

symbol

Encoded

data

f1

f2f3f4

SC-FDMA

symbol to be

transmitted on

antenna 1P/S

SC-FDMA

symbol to be

transmitted on

antenna 2

Cyclic

delay

Performance of transmit diversity schemes [4] • SD CDD shows similar performance as the 1x2 baseline in uncorrelated channels. With small number of RBs (2RBs) allocated, the

coded diversity advantage of SD CDD is limited.

• Among other TxD schemes, since LD CDD and FSTD do not have uncoded diversity benefit, the BLER performance gap between

them and STBC becomes wider as we increase the coding rate (left plots r=2/3; right plots r=1/3).

• STBC Alts 1, 3 and 4 show similar BLER performance; at BLER with 16-QAM rate 2/3, they are better than STBC Alt2 and LD CDD by

about 1 dB, and better than SD CDD and FSTD by more than 3.0 dB.

• SD CDD shows extreme behaviors in the positively correlated ( = 0.7) channels.

• LD CDD shows worse BLER than STBC. The performance gap becomes larger as we increase the coding rate, as LD CDD relies only

on coded diversity.

• FSTD shows similar performance as base 1x2 scheme, as the two Tx antenna channels are highly correlated.


r=1/3 r=2/3

-2 0 2 4 6 8 10 12 14 16 10

-3

10 -2

10 -1

10 0

SNR

BL

ER

UL 2TxD with 2 RBs in TU6 UnCorrelated, extended CP: 16QAM

Base 1x2

STBC Alt 1: Rep

STBC Alt 2: A1

STBC Alt 3: top-down

STBC Alt 4: even-odd

FSTD

SD CDD

LD CDD

r=1/3 r=2/3

-2 0 2 4 6 8 10 12 14 16 10

-3

10 -2

10 -1

10 0

SNR at each receive antenna

BL

ER

UL 2-TxD with 2 RBs in TU6 Correlated, extended CP: 16QAM

Base 1x2

STBC Alt 1: Rep

STBC Alt 2: A1

STBC Alt 3: top-down

STBC Alt 4: even-odd

FSTD

SD CDD

LD CDD

LTE downlink codebook for 4 antenna ports [5]

• The quantity denotes the matrix defined by the columns given by the set {s}

from the expression

where I is the 4x4 identity matrix and the vector is given by the table.


Codebook index

nu Number of layers

1 2 3 4

0 Tu 11110 }1{

0W 2}14{

0W 3}124{

0W 2}1234{

0W

1 Tjju 111 }1{

1W 2}12{

1W 3}123{

1W 2}1234{

1W

2 Tu 11112 }1{

2W 2}12{

2W 3}123{

2W 2}3214{

2W

3 Tjju 113 }1{

3W 2}12{

3W 3}123{

3W 2}3214{

3W

4 Tjjju 2)1(2)1(14 }1{

4W 2}14{

4W 3}124{

4W 2}1234{

4W

5 Tjjju 2)1(2)1(15 }1{

5W 2}14{

5W 3}124{

5W 2}1234{

5W

6 Tjjju 2)1(2)1(16 }1{

6W 2}13{

6W 3}134{

6W 2}1324{

6W

7 Tjjju 2)1(2)1(17 }1{

7W 2}13{

7W 3}134{

7W 2}1324{

7W

8 Tu 11118 }1{

8W 2}12{

8W 3}124{

8W 2}1234{

8W

9 Tjju 119 }1{

9W 2}14{

9W 3}134{

9W 2}1234{

9W

10 Tu 111110 }1{

10W 2}13{

10W 3}123{

10W 2}1324{

10W

11 Tjju 1111 }1{

11W 2}13{

11W 3}134{

11W 2}1324{

11W

12 Tu 111112 }1{

12W 2}12{

12W 3}123{

12W 2}1234{

12W

13 Tu 111113 }1{

13W 2}13{

13W 3}123{

13W 2}1324{

13W

14 Tu 111114 }1{

14W 2}13{

14W 3}123{

14W 2}3214{

14W

15 Tu 111115 }1{

15W 2}12{

15W 3}123{

15W 2}1234{

15W

}{snW

n

Hn

Hnnn uuuuIW 2

nu

LMMSE-SIC receiver for dual codewords


• The basic idea of an LMMSE-SIC detector is to

detect and decode multiple data layers codeword by

codeword.

• To achieve the best performance, the codeword with

the highest strength is detected first.

• Once a codeword is successfully decoded, it is

reconstructed and subtracted from the received

signal.

• An LMMSE with reduced size is then applied to the

resulting received signal to obtain the symbol

estimation of the second codeword.

High level block diagram of LMMSE-SIC detector


MMSE

Sorting

Decoder

CSI: H

Modulation & Coding Scheme

Received Signal

Noise Power

Signal Reconstruction

-Σ

Bu

ffer

Detected data for

codeword 1

Detected data for

codeword 2

MMSE Decoder

From Sorting

Turbo-SIC receiver

• Iterative SIC receiver utilizing soft decision metric

• Turbo equalizer performs multi-stream equalization

by utilizing an iterative loop between decoding and

equalization [6]


Receive antenna

MMSE

Equalizer

Soft

Modulator

Soft

Modulator

Turbo

decoder

Turbo

decoder

Rate

matching

Rate

dematching

Rate

dematching

Rate

matching

Layer

Demapper

Soft

Demodulator

Soft

Demodulator

Interference

Canceller

Interference

Canceller

M-DFT

M-DFT

M-IDFT

M-IDFT

Layer

mapper

Layer

mapper

RE

Demapper

RE

Demapper

N-FFT

N-FFT

soft values

soft values

hard values

hard values

Maximum Likelihood Detector

• MLD compares the received signal with the full set of

possible symbols in search for most likely transmitted

signal

– In OFDMA, candidate symbol set is limited to a single sub-carrier

– MLD has lower complexity in OFDMA than in SC-FDMA

• Computational complexity increases exponentially with

increasing modulation order and/or MIMO rank

• Complexity can be reduced by limiting the candidate

symbol set, but at the cost of degraded performance

– List sphere decoder [13]

– QRD-ML [14]

49


Maximum Likelihood Detector

MLD receiver block diagram


N-FFT

Turbo

Decoder

Turbo

Decoder

N-FFT

RE Demapper

RE Demapper

QRD

ML

Receiver

Layer

Demapper

Rate

Dematching

Rate

Dematching

receive antennas soft values

hard values

hard values

Computational complexity


• The computational complexity of the above two

receivers is dependent on the number of transmit

antennas, the number of receive antennas, modulation

order, the size of constellation alphabet and the code

rate, as well as the number of survival path retained at

the m-th stage for QRD-ML receiver [7].

• The computational complexities of a QRD-ML receiver

mainly focus on the QR decomposition, Euclidian

distance estimation, sorting and LLR computation.

• The computational complexities of Turbo-SIC receiver

mainly focus on interference cancellation, soft

modulation/demodulation and the soft decoding.

References 1. R1-061481 “User Throughput and Spectrum Efficiency for E-UTRA”, InterDigital Communications, 3GPP TSG RAN WG1 #45, May 8-12, 2006

2. R1-09739 “Comparison of uplink transmit diversity schemes for LTE-Advanced”, Mitsubishi Electric, 3GPP TSG RAN WG1 #56, Feb. 9-13, 2008

3. R1-090690 “Views on TxD schemes for PUCCH”, Panasonic, 3GPP TSG RAN WG1 #56, Feb. 9-13, 2008

4. R1-090614 “Discussions on UL 2Tx Transmit Diversity Schemes in LTE-A”, Samsung, 3GPP TSG RAN WG1 #56, Feb. 9-13, 2008

5. 3GPP TS 36.211 Physical Channels and Modulation (Release 8), 3rd Generation Partnership Project

6. R1-083732 “Comparison between SC-FDMA and OFDMA for LTE-Advanced Uplink”, Nokia Siemens Networks, 3GPP TSG RAN WG1 #54bis,

September 29- October 3, 2008

7. R1-090125 “Performance comparison of UL MIMO in OFDMA vs. SC-FDMA”, Huawei, 3GPP TSG RAN WG1 #55bis, Jan 12 – 16, 2009

8. R1-061082 “Uplink MIMO SC-FDMA with Adaptive Modulation and Coding “, InterDigital Communications, 3GPP TSG RAN WG1 #44bis, March 27

– 31, 2006

9. R1-062159 “Comparison of SCW and MCW for UL MIMO Using Precoding”, InterDigital Communications, 3GPP TSG RAN WG1 #44, August 28 –

September 1, 2006

10. R1-083515 “Throughput evaluation of UL Transmission Schemes for LTE-A”, InterDigital Communications, 3GPP TSG RAN WG1 #54bis, Sept. 29

– Oct. 3, 2008

11. R1-082609 “Uplink Multiple Access for LTE-Advanced”, Nokia Siemens Network, Nokia, 3GPP TSG RAN WG1 #53b, June, 2008

12. R1-090232 “Comparing performance, complexity and latency of SC-FDMA SIC and OFDM MLD”, Nokia Siemens Network, Nokia, 3GPP TSG RAN

WG1 #55bis, January 12 – 16, 2009

13. M. O. Damen, H. E. Gamal, and G. Caire, “On maximum-likelihood detection and the search for the closest lattice point,“ IEEE Trans. Inform.

Theory, vol. 49, no. 10, pp. 2389{2402, Oct. 2003.

14. J. Yue, K. J. Kim, J. D. Gibson and R. A. Iltis, “Channel estimation and data detection for MIMO-OFDM systems”, IEEE Global Telecommunications

Conference, vol. 22, no. 1, pp. 581 - 585, Dec 2003.

15. 3GPP TS 36.213: Physical layer procedures (Release 8), 3rd Generation Partnership Project

16. T. Shi, et al, “Capacity of single carrier systems with frequency-domain equalization,” IEEE 6th CAS Symposium on Emerging Technologies:

Mobile and Wireless Comm., 2004

17. S.J. Grant, K.J. Molnar, and L. Krasny, “System-level performance gains of selective per-antenna-rate-control (S-PARC) ”, IEEE VTC Spring 2005

18. R1-051398, “Transmit Antenna Selection Techniques for Uplink E-UTRA”, Institute for Infocomm Research (I2R), Mitsubishi Electric, NTT

DoCoMo.


Documents

MIMO and Transmit Diversity for SC-FDMA