Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Broadband wireless communications techniques
2012/4/12 FA/Tohoku University 1
OUTLINE Wireless Evolution Challenge for Future Wireless Gigabit Wireless Concluding Remarks
Fumiyuki Adachi Wireless Signal Processing & Networking (WSP&N) Lab.
Dept. of Electrical and Communications Engineering,
Tohoku University, JapanE-mail: [email protected]
http://www.mobile.ecei.tohoku.ac.jp/
Lecture note #00
Matsushima
Tohoku U. Aobayama-campus
Wireless EvolutionTo 3G
From 2G to 3G systems Then, from 3G to 3.9G (LTE)
systems
2012/4/12 FA/Tohoku University 2
F. Adachi, “Wireless past and future - evolving mobilecommunications systems,” IEICE Trans. Fundamentals, vol.E84-A, pp. 55-60,Jan. 2001.
Wireless Evolution To 3G In early 1980’s, communications systems changed from
fixed “point-to-point” to wireless “anytime, anywhere”communication.
Cellular systems have evolved from narrowband network ofaround 10kbps to wideband networks of around 10Mbps.
Now on the way to broadband networks of 100Mps (LTE).
2012/4/12 FA/Tohoku University 4We are here
Gigabitwireless
1980 1990 2000 Year
2G~64kbps
1G~2.4kbps
NarrowbandEra
Ser
vice
typ
eVo
ice
Mul
timed
ia
2010
3G~2Mbps
4G100M~1Gbps
W-CDMACDMA2000TD-SCDMA
HSDPA(W-CDMA)
~14Mbps0GVoice only
point-to-point
50~100Mbps
WidebandEra
BroadbandEra
3G LTE
2020
3.5G 3.9G
Cellular Networks Taking A Role of Fixed Networks No. of cellular+PHS phone
users exceeded that of fixedanalog telephone circuits inMarch 2000 in Japan Cellular: 51.141M PHS: 5.708M Fixed phone: 55.446M
This clearly shows that peoplewant to communicate withpeople, not with places
Cellular networks will take arole of the conventional fixednetworks, soon, worldwide
2012/4/12 FA/Tohoku University 5
Year0
10
20
30
40
50
60
70
1985 1990 1995 2000 2005
Mobile
Fixed
61.22M@1997
No.
of
user
s (M
)
There was a big technical leap from 2G to 3Gsystems.
2012/4/12 8FA/Tohoku University
1GAnalog(FDMA)
2GDigital(TDMA)
Voice+data
Improved frequency utilizationNarrowband
3G/3.5GDigital(CDMA)
Increased peak rateIncreased throughput
Increased no. of channels
Broadband
Voice+Data
Big leap
~64kbps ~2Mbps~14Mbps
~2.4kbps
Evolution of Wireless Access Techniques Wireless access technique used in cellular systems has been
changed from FDMA to DS-CDMA. FDMA was used in 1G, TDMA in 2G, DS-CDMA in 3G OFDMA and SC-FDMA are used for downlink and uplink, respectively,
in 3.9G(LTE)
2012/4/12 FA/Tohoku University 9
Time
FDMAf3f2f1
Freq
uenc
y
1G
TDMA
Time
1 3 1 2 3
2G Single-carrier
3.9G(LTE)
OFDMA,SC-FDMA
Frequency-domain processing
2
Time-domain processing
DS-CDMA3G Freq
uenc
y
Time
# 2# 3
Spreading code#1
Freq
uenc
y
20ms
Ex. 6.25kHz
Ex.
5MH
z
25kHz
3G Systems Using W-CDMA Data transfer rates in 2G systems are too slow for
downloading rich information distributed in the Internet. 3G cellular systems are designed to offer cellular users a
significantly higher data-rate services using W-CDMAtechnology (3.84cps/5MHz bandwidth). Indoor: 2Mbps Pedestrian: 384kbps Vehicular: 144kbps
2012/4/12 FA/Tohoku University 11
F. Adachi, M. Sawahashi and H. Suda, “WidebandDS-CDMA for next generation mobilecommunications systems,” IEEE Commun. Mag., vol.36, pp. 56-69, Sept. 1998.
Indoors~2Mbps
Mobile~144kbps
IMT2000Network
Pedestrian~384kbps
2012/4/12 FA/Tohoku University 29
Problem of 3G DS-CDMA Wireless Access
1G, 2G and 3G used FDMA, TDMA, and DS-CDMA, respectively All of them are based on time-domain signal
processing What signal processing should be used for
4G? Recently, multicarrier wireless techniques and
frequency-domain signal processing have beenattracting much attention
OFDMA, SC-FDMA, MC-CDMA, or DS-CDMA in4G?
F. Adachi, D. Garg, S. Takaoka, and K. Takeda, “BroadbandCDMA techniques,” IEEE Wireless Commun. Mag., Vol. 12, No.2, pp. 8-18, April 2005.
Single-Carrier DS-CDMA Transmit signal is spread over a broad bandwidth and
suffers from severe spectrum distortion due to severefrequency-selectivity of the channel.
2012/4/12 FA/Tohoku University 31
(a) Transmitter
Datamodulation
Data
Time-domainspreading
Channelcoding &
interleaving
c(t)
Chipshaping
FrequencyfcCarrier frequency
Bandwidth (1+)/Tc
(b) Power spectrum
Channel Model
2012/4/12 FA/Tohoku University 32
Transmitted radio waves are reflected or diffracted by somelarge buildings, creating resolvable paths having time delaysof multiple of (signal bandwidth)-1.
Each resolvable path is the sum of irresolvable pathscreated by local scatterers surrounding a mobile.
The path gain hl(t) varies in time according to themovement of mobile terminal since resolvable paths areadded constructively at one time and destructively atanother time.
Localscatterers
Large obstacles
Transmitter
ReceiverReflection/diffraction
c-4
Channel Impulse Response Multipath channel can be viewed as a time varying linear
filter having impulse response h(, t). Many impulses are received with different time delays when
one impulse is transmitted from a transmitter at time t. Frequency-selective fading channel is produced.
2012/4/12 FA/Tohoku University 33
Time varyingFIRfilter
Transmit signals(t)
Received signalr(t)
1
0
)()(),(L
lll thth
time
Transmit pulse
Am
plitu
de
Wirelesschannel
0
Impu
lse r
esp
onse
Inverse ofthe signal bandwidth, 1/W
Time delay0
Observed pulses (each pulse consisting of many unresolvable pulses)
0 1 2
Rake Receiver for DS Receivers of present 3G systems use time-domain rake
combining, which is a channel matched filter. Rake combining can improve the BER performance if the
channel frequency-selectivity is not too strong (or thenumber L of resolvable paths is not too large).
2012/4/12 FA/Tohoku University 34
F. Adachi, M. Sawahashi and H. Suda, “Wideband DS-CDMAfor next generation mobile communications systems,” IEEECommun. Mag., vol. 36, pp. 56-69, Sept. 1998.
c*(t) Recovered data
De-interleaving& channeldecoding
Datademodulation
*0h
Time-domain despreading
Coherent rake combining(a matched filter to channel)
0
L-1
*1Lh
Integrate& dump
l=1l=0
Delay Impu
lse
resp
onse
l=2
L=3Equalizermatched topropagationchannel
Limitation of Time-domain Rake Number of resolvable paths increases as the transmission
rate or the signal bandwidth increases. This increases the inter-path interference (IPI), thereby
degrading the BER performance. Furthermore, finite number of Rake fingers can collect only
a fraction of total power, reducing the power efficiency.
2012/4/12 FA/Tohoku University 35
Time delay
Impu
lse
resp
onse
0
Wideband (~5MHz)
Broadband (25~100MHz)
BER Performance w/ Rake Combining As the number of resolvable
paths increases, the channel frequency-selectivity gets stronger.
The achievable BER performance of 3G SC(DS)-CDMA with rake combining degrades significantly due to strong IPI. For a heavily loaded channel,
even with L=2, a high BER floor appears if the code-multiplexing order is high.
On the other hand, MC-CDMA with MMSE-FDE provides much better performance. Performance improves as L
increases.
2012/4/12 FA/Tohoku University 36
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
0 5 10 15 20 25 30
Average received E b /N 0 (dB)
Ave
rage
BER
DS-CDMA withrake combining
MC-CDMA withMMSE equalization
N c =256SF =16, C =16
16
L=2
Uniform delay profile
L=2
16
SC-CDMA for downlink w/full code-multiplexing
QPSK
SC-CDMA with rake combining
Multi-carrier vs
Single-carrier
In 3G systems, DS-CDMA (or single-carrierCDMA) is adopted for both uplink anddownlink since it is a very flexible multi-access technique.
Which will be an optimal wireless accesstechnique in a severe frequency-selectivechannel, single-carrier CDMA or multicarrierCDMA or OFDMA for 4G wireless systems?
2012/4/12 FA/Tohoku University 39
F. Adachi, D. Garg, S. Takaoka, and K. Takeda,“Broadband CDMA techniques,” IEEE WirelessCommun. Mag., Vol. 12, No. 2, pp. 8-18, April 2005
MC-CDMA MC-CDMA is a combination of OFDM and CDMA.
A simple one-tap FDE Very robust transmission against frequency-selective channel
A DS-CDMA chip block is transformed by IFFT into the frequency-domainsignal block (MC-CDMA).
2012/4/12 FA/Tohoku University 54
(a) Transmitter
(b) Power spectrum FrequencyfcCarrier frequency
Bandwidth 1/Tc
Transmittingdata Channel
coding &interleaving
Datamodulation +CP
#0
c(t)
Time-domainspreading
IFFTS/P
Conversion to freq.-domain spread signal
#Nc-1
Frequency-domain equalization (FDE) is used to exploit thefrequency selectivity of the channel.
FDE based on the minimum mean square error (MMSE)criterion can provide the best downlink performance. MMSE weight is the one which minimizes the mean square error (MSE)
between the transmit subcarrier component and the received distortedcomponent.
2012/4/12 FA/Tohoku University 59
(c) Receiver
-CPDe-interleaving
& channeldecoding
Datademodulation
W(0)
W(k)Recovered
data
Frequency-domainequalization
FFT
#0
#Nc-1
P/S
c*(t)
Time-domaindespreading
W(Nc-1)
Integrate& dump
Comparison of MMSE-, MRC-, and ZF- FDE MMSE provides the
best BERperformance. No. of subcarriers
Nc=256, spreadingfactor SF=256
Rayleigh fadingchannel with no.paths L=2 and 16,ideal channelestimation
Single user case(U=1)
2012/4/12 FA/Tohoku University 70
MC-CDMA with U=1
10-5
10-4
10-3
10-2
10-1
100
0 5 10 15 20 25 30
AWGNMRCZFMMSE
Ave
rage
BER
Average received Eb/N0 (dB)
L=1L=16L=2
MMSE,MRCL=2
L=16
ZF
QPSK
DS-CDMA with Rake vs MC-CDMA with FDE As the number L of resolvable
paths increases, the channel frequency-selectivity gets stronger.
Achievable BER performance of 3G SC(DS)-CDMA with rake combining degrades significantly due to strong IPI. For a heavily loaded channel,
even with L=2, a high BER floor appears if the code-multiplexing order is high.
On the other hand, MC-CDMA with MMSE-FDE provides much better performance. Performance improves as L
increases.
2012/4/12 FA/Tohoku University 72
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
0 5 10 15 20 25 30
Average received E b /N 0 (dB)
Ave
rage
BER
DS-CDMA withrake combining
MC-CDMA withMMSE equalization
N c =256SF =16, C =16
16
L=2
Uniform delay profile
L=2
16
SC-CDMA for downlink w/full code-multiplexing
QPSK SC-CDMA with rake combining
Frequency-domain Equalization (FDE) for DS-
CDMA
One-tap FDE can replace rake combining to pprovide much improved performance
2012/4/12 FA/Tohoku University 73
F. Adachi, D. Garg, S. Takaoka, and K. Takeda,“Broadband CDMA techniques,” IEEE WirelessCommun. Mag., Vol. 12, No. 2, pp. 8-18, April 2005F. Adachi, Kazuki Takeda, and H. Tomeba,“Introduction of Frequency-Domain Signal Processingto Broadband Single-Carrier Transmissions in aWireless Channel,” IEICE Trans. Commun., Vol. E92-B,No.09, pp.2789-2808, Sep. 2009.
FA/Tohoku University 74
Can Single-Carrier CDMA (DS-CDMA) Survive in The Future? Can DS-CDMA (single-carrier CDMA) survive in the future?
This is an interesting question. My answer is “YES”. Single-carrier technique with spreading is believed to still
remain an important technique although single-carriertechnique without spreading will be used for uplink accessin 3.9G (LTE).
To take advantage of the frequency-selectivity, MMSE-FDEcan be applied to improve the BER performance of SC-CDMA, similar to MC-CDMA.
2012/4/12 FA/Tohoku University 74
FDE Improves SC-CDMA Performance Coherent Rake combining can be replaced by one-tap FDE
Block transmission of Nc chips Insertion of cyclic prefix (CP) at the transmitter FFT/IFFT at the receiver
2012/4/12 FA/Tohoku University 75
(b) Receiver
Receiveddata
・・・
Dat
ade
mod
.
Removalof CP IF
FTFFT
c*(t )
Time-domaindespreading
W(0)
W(k)
W(Nc-1)・・・
FDE
Integrate& dump
(a) Transmitter
Inse
rtio
n of
CPTransmit
data
Dat
a m
od.
Time-domain spreading
c(t )
DataCP
Copy
Nc chipsNg chips
*D.Falconer, S. Ariyavisitakul, A,Benyamin-Seeyar and B. Eidson,“Frequency Domain Equalizationfor Single-Carrier BroadbandWireless Systems,” IEEECommunications Magazine, Vol.40, No. 4, pp. 58-66, April 2002.
*F. Adachi, D. Garg, S. Takaoka,and K. Takeda, “BroadbandCDMA techniques,” IEEEWireless Commun. Mag., Vol. 12,No. 2, pp. 8-18, April 2005.
Similarity Between DS- and MC-CDMAWith FDE, both DS- and MC-CDMA can take
advantage of the channel frequency-selectivityand have a similar improved BER performance.
Their uplink transmitter/receiver structures aresimilar.
2012/4/12 FA/Tohoku University 78
cu(t)
Scrambling
cscr(t)Transmit data
Dat
a m
od.
Inse
rtio
n of
CP
DS
MC
Nc-P
oint
IF
FT
Transmitter
Spreading
Descrambling
cscr(t)
Received
data
Despreading
cu(t)**
Dat
a de
mod
.
Nc-P
oint
FFT
Rem
oval
of
CP
FDE MC
DS
Receiver Nc-P
oint
IF
FT
Performance Comparison Among SC- and MC-CDMA and OFDM BER performance (single
user) can be significantly improved compared to the coherent rake receiver. Better BER performance than
OFDM even for full code-multiplexing (no. of users, U, is equal to SF=256).
However, there is still a big performance gap from the theoretical lower bound. This is due to residual ISI
after MMSE-FDE. Introduction of ISI
cancellation technique can reduce the performance gap.
2012/4/12 FA/Tohoku University 89
# F. Adachi, D. Garg, S. Takaoka, and K. Takeda,“Broadband CDMA techniques,” IEEE Wireless Commun.Mag., Vol. 12, No. 2, pp. 8-18, April 2005
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
0 5 10 15 20 25
Average received E b /N 0 (dB)
Ave
rage
BER
DS-FDE (MMSE)DS-rakeMC-FDE (MMSE)OFDM
QPSK data modulationRayleigh fadingL=16, uniformSF=256, C =256
SC-FDE (MMSE)SC-rake
Uncoded
Uncoded single user
SC-rake
OFDM
SC-,MC-FDE
Significant improvement
Lower bound
Wireless Evolutionto
3.9G (LTE)
FDMA was used in 1G, TDMA in 2G, DS-CDMA in3G
OFDMA and SC-FDMA are used for downlink anduplink, respectively, in 3.9G(LTE)
2012/4/12 FA/Tohoku University 90
Before 4G, There Will Be 3.9G3G systems will continue to evolve to meet the
demands of (internet-related) broadband wirelessservices and substantially strengthen its downlinkdata rate capability. High-speed downlink packet access (HSDPA) of
~14Mbps/5MHz, called 3.5G systems, started in Japanin 2006.
Even 3.5G of ~14Mbps will sooner or later becomeinsufficient.
A 3.9G (3G LTE) will appear to provide broadband dataservices of 50~100Mbps/20MHz using 3G bands.
2012/4/12 FA/Tohoku University 91
DS-CDMA(DL/UL)(~14Mbps)
OFDMA(DL),SC-FDMA(UL)(50~100Mbps)
3.5G (HSDPA) 3.9G (LTE)
3G band will be used No available bandwidth of around 100MHz (a hot matter
of WRC-07) Present 3G bandwidth (1.25~20MHz) will be used to
provide much faster rate data services Target: 100Mbps for downlink, 50Mbps for uplink
Difference from 3G wireless technology 3~3.5G:DS-CDMA (DL/UL) 3.9G (LTE): OFDMA(DL)/SC-FDMA(UL) Better transmission performance than 3G and good
commonality with WLAN
2012/4/12 FA/Tohoku University 92 2012/4/12 FA/Tohoku University 93
Evolution of Wireless Access Technique Cellular and WLAN (incl. WiMAX) are both using
OFDM technique Frequency-domain signal processing to exploit the
channel frequency-selectivity Standardization world
FDMA TDMA DS-CDMA SC-FDMAOFDMA
DS-CDMA OFDMA802.11b 802.11a, 11n
Cellular
WLAN
1G 2G 3G 3.9G (LTE)
WiMAX802.16
ITU world
IEEE world
Evolution of Wireless Access Techniques Wireless access technique used in cellular systems has been
changed from FDMA to DS-CDMA. FDMA was used in 1G, TDMA in 2G, DS-CDMA in 3G OFDMA and SC-FDMA are used for downlink and uplink, respectively,
in 3.9G(LTE)
2012/4/12 FA/Tohoku University 94
Time
FDMAf3f2f1
Freq
uenc
y
1G
TDMA
Time
1 3 1 2 3
2G Single-carrier
3.9G(LTE)
OFDMA,SC-FDMA
Frequency-domain processing
2
Time-domain processing
DS-CDMA3G Freq
uenc
yTime
# 2# 3
Spreading code#1
Freq
uenc
y
20ms
Ex. 6.25kHz
Ex.
5MH
z
25kHz
Uplink and Downlink Access Techniques Are Different For a long time, the same access technique has been
adopted for the uplink and downlink. Different multi-access techniques between downlink and
uplink in LTE (this is the first time in the history) Downlink: MC-based multi-access is more appropriate since
high PAPR is not a big problem at the base station. Uplink: as the data rates increases, the PAPR problem
becomes more serious. SC-based multi-access is moreappropriate because of its lower PAPR.
2012/4/12 FA/Tohoku University 95
Base station(DL)Higher peak power amp.(UP) Complex MMSE-FDE
Mobile terminal(UL) Lower peak power amp.(DL) Simple ZF-FDE
Wireless Access of 3.9G Different multi-access techniques between downlink and
uplink (this is the first time in the history) Downlink: OFDMA, ~100Mbps Uplink: SC-FDMA, ~50Mbps
Scheduling for packet access Multiuser diversity in wireless channel Hybrid ARQ using incremental redundancy (IR) strategy Non real time data services
2012/4/12 FA/Tohoku University 97
Downlink UplinkBandwidth(MHz) 1.4/3/5/10/15/20IFFT/FFT block size
128/256/512/1024/1536/2048
Multi-access OFDMA SC-FDMAScheduling Multi-user diversity gainARQ Turbo-coded IR-HARQ
Downlink and uplink access schemes are different
OFDMA Downlink Resource allocation: one or more resource blocks of 1msec
and 12 subcarriers (180kHz) each are allocated accordingto each user’s channel condition (scheduling) to obtainmultiuser diversity gain
Proportional fairness (PF)* scheduling can maximize thethroughput while keeping fairness among users
2012/4/12 FA/Tohoku University 98
Freq.
Freq.
User A User B User C
Cha
nnel
gai
n
12 subcarriers(180kHz)
* A. Jalali, R. Padovani, and R. Pankaj, “Data throughput of CDMA-HDR a high efficiency-high data rate personal communication wireless system,” Proc. IEEE VTC 2000-Spring, vol. 3, pp. 1854 -1858, Tokyo, 15-18May 2000.
S/P
P/S
IFFT
User A
+CPS/PUser B
Map
ping OFDMA
signal
Frequency-domain scheduling
Frequency and time -domain scheduling
2012/4/12 FA/Tohoku University 99
A
A
A
A
A
B
B
B
B
B
B
B
B A
12 subcarriers(180kHz)
BS
User A
User B1ms
Tim
e
Frequency
Why Single-carrier (SC)Transmission for Uplink? SC has lower PAPR than
OFDM. No ISI at different symbol
positions. SC is suitable for the uplink
transmission. Less expensive power
amplifier is required.
2012/4/12 FA/Tohoku University 100
-4
-3
-2
-1
0
1
2
3
4
55 57 59 61 63 65
α= 0.00
-4
-3
-2
-1
0
1
2
3
4
55 57 59 61 63 65
α= 1.00
-4
-3
-2
-1
0
1
2
3
4
55 57 59 61 63 65
α= 0.50
SC
-4
-3
-2
-1
0
1
2
3
4
55 57 59 61 63 65
OFDM
OFDMA
256 subcarriers
)2sin()(Im)2cos()(Re
)2exp()(Reveformcarrier wa Modulated
tftstftstfts
cc
c
SC signal has lessPAPR than OFDM.
Reason for this isbecause time-domainSC signal is a Nyquistfiltered signal whichhas the sameamplitude as theoriginal data symbol(ISI free) at every Tssecond (where Ts is thesymbol length).
2012/4/12 FA/Tohoku University 101PAPR(dB)
CCD
F
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
2 4 6 8 10 12
α=0.00
α=0.50
α=1.00
OFDM
OFDM
SC
SC-FDMA Uplink Uplink MAI
The MAI can be avoided by combining SC-FDE with either FDMA orCDMA while alleviating the complexity problem of MUD (remainingproblem is other-cell MAI).
SC-FDMA is a combination of SC-FDE and FDMA A block of data symbols to be transmitted is transformed into
frequency-domain signal which is then mapped onto a different set ofsubcarriers of OFDM.
To preserve the low PAPR property of SC-FDMA signals, equidistancespectrum mapping (either localized or equidistantly distributed) isused.
2012/4/12 FA/Tohoku University 104
# R. Dinis, D. Falconer, C. T. Lam, and M. Sabbaghian, Proc.GlobeCom2004, vol.6, pp. 3808-3812, Dallas, TX, USA, 29Nov.-3 Dec. 2004
# K. Takeda and F. Adachi, Proc. IEEE VTC2005-Fall, Dallas,U.S.A., 26-28 Sept. 2005.
# M. Schnell, I. Broeck, and U. Sorger, ETT, Vol. 10, No.4,pp. 417-427, Jul.-Aug. 1999.
Transmitfiltering
M-pointDFT
Nc-pointIFFT
Coded data
0 M-1
・・・
1 2
User u
f
SC-FDMAsignal+CPMapping
Same as downlink OFDMA
Distributed FDMA Each user’s frequency components are mapped equal
distantly over entire signal bandwidth. Large frequency diversity gain can be obtained.
Localized FDMA Each user’s frequency components are mapped over
consecutive subcarriers. Lower frequency diversity gain, but larger multiuser
diversity gain can be obtained by scheduling.
Distributed FDMA vs Localized FDMA
2012/4/12 106FA/Tohoku University
Distributed
Localized
User #0#1#2#3
f
fUser #0 User #1 User #2 User #3
SC-FDMA vs OFDMA A different subcarrier group is assigned to a
different user. SC-FDMA vs OFDMA
In SC-FDMA, M-symbol block is transformed into thefrequency-domain signal of M components, which is thenmapped onto a subcarrier group assigned to that user.
In OFDMA, M symbols in a block are mapped onto asubcarrier group assigned to that user.
2012/4/12 FA/Tohoku University 107
Transmit data
Data m
od.
+CP
SC
OFDMN
c-Point
IFFT
Transmitter
M -Point D
FT
マッ
ピン
グ
Freq.
0d 1d 1Md
)0(D
)1(D)1( MD
Freq.
OFDMA
0 1 k 2cN 1cN
0 1 k 2cN 1cN
SC-FDMA
2012/4/12 FA/Tohoku University 108
Frequency Diversity vs Multi-user Diversity Uplink SC-FDMA
Localized FDMA: multi-user diversity Distributed FDMA:frequency diversity
Multi-user diversity(freq./time-domain scheduling)
Frequency diversity
Distributed FDMA
Whole bandwidth
userA
userC
userB
userD
Localized FDMA
Whole bandwidth
Cha
nnel
gai
n
f
Different users’ spectra are mapped based on FDMA so that they are not overlapped to avoid the MAI.
At a BS, each user’s spectrum is extracted after FFT and then, joint FDE & diversity combining is carried out.
M-point D
FT
Mapping
Nc -point IFFT
Transmit data
Data m
od.
+CP
Joint FD
E/diversity com
bining
Joint FD
E/diversity com
bining –CP –C
P
Nc -point FFT
De-m
apping
M-point ID
FTM
-point ID
FT
User #U-1
Data
demod.
Data
demod.
Recovered dataUser #U-1
User #0
User #0
Antenna #0
#Nr–1
User transmitter
BS receiver
2012/4/12 109FA/Tohoku University
BER performance whenU=16 users aremultiplexed per block.M=16 subcarriers are
assigned to each user,which is equal to OFDMAsystem.
The distributed SC-FDMAobtains larger frequencydiversity gain and canachieve better BERperformance than thelocalized SC-FDMA.
2012/4/12 FA/Tohoku University 110
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
0 5 10 15 20
Ave
rage
BER
Average received Eb/N0 (dB)
L=16-path uniformpower delay profile
SC-FDMA
QPSKNc=256 Ng=32U=16
Distributed
Localized2
4
Nr=1
3
10
1
0
2
1
0
/|)(|
)()(
)()()(ˆ
NEkH
kHkW
kRkWkR
s
N
mm
mMMSEm
N
mm
MMSEm
r
r
Wireless Evolution To
4G (LTE-Advanced) Cellular vs Wireless LAN
2012/4/12 FA/Tohoku University 116
4G VisionBroadband services: data rates for mobile
services may be up to 100 Mbps and those fornomadic services may be up to 1Gbps.
Gigabit wireless: an important technology for therealization of 4G systems.
2012/4/12 FA/Tohoku University 117
High(60~250km/h)
Low(Pede-strianspeed)
IMT2000 EnhancedIMT2000
New MobileAccess
New nomadic/LocalArea wireless access
Enhancement
Digital broadcast system
Inter-connection
1 10 100 1000Peak useful data rate (Mb/s)
New capabilities of systems beyond IMT-2000
Mobility
ITU-R WP8F (Ottawa, June 2002) :illustration of Capabilities of IMT2000 and Systems Beyond
Beyond IMT-2000
Wireless Evolution In early 1980’s, communications systems changed from
fixed “point-to-point” to wireless “anytime, anywhere”communication.
Cellular systems have evolved from narrowband network ofaround 10kbps to wideband networks of around 10Mbps.
Now on the way to broadband networks of 1Gbps (4G).
2012/4/12 FA/Tohoku University 118We are here
Gigabitwireless
1980 1990 2000 Year
2G~64kbps
1G~2.4kbps
NarrowbandEra
Ser
vice
typ
eVo
ice
Mul
timed
ia
2010
3G~2Mbps
4G100M~1Gbps
W-CDMACDMA2000TD-SCDMA
HSDPA(W-CDMA)
~14Mbps0GVoice only
point-to-point
3G LTE
50~100Mbps
WidebandEra
BroadbandEra
2020
There was a big technical leap from 2G to 3Gsystems.
In 4G systems, much higher throughput than 3Gsystems is demanded.
2012/4/12 119FA/Tohoku University
1GAnalog(FDMA)
2GDigital(TDMA)
Voice+data
Narrowband
3GDigital(CDMA)
4GDigital
(OFDMA?)
Increased peak rateIncreased throughput
Increased no. of channels
Broadband
Voice+Data
Big leap
3G LTEDigital(OFDMA
/SC-FDMA)
Big leap
2012/4/12 FA/Tohoku University 120
What Is a Killer Application in 4G? It is quite difficult to predict which services will
become popular in the coming 10 years However, it is no doubt that Internet-related
services will dominate in 4G Another promising service in 4G is visual
communication Our natural communication way is to speak/hear and
look. Earlier generations of communication networks provided
voice services (speak/hear) only. Visual communication everywhere will be one of
important services.
Wireless Voice to Wireless Video
2012/4/12 FA/Tohoku University 121
1G Voice
2G Voice Text & data (e-mailing, data access)
3G Video Picture&video
4GVideocommunications Video data
3.5G3.9G
Non-Real timeservice
Real time communications
Voice
Wireless Everywhere How to offer both cellular and nomadic users a
broad range of wireless multimedia services,everywhere?
Almost impossible to build a new, single superwireless system to meet all demands for a broadrange of wireless services
A good solution may be a seamless and secureintegration of various wireless networks, thoseare optimized to each environment
2012/4/12 FA/Tohoku University 122
Evolution of Wireless Access In 3G, FDMA was used in 1G, TDMA in 2G, DS-CDMA. In 3.9G(LTE), OFDMA and SC-FDMA will be used for
downlink and uplink, respectively. In 4G, what wireless access technique is used?
2012/4/12 FA/Tohoku University 123
1G
TDMA2G
Time
Single-carrier
3.9G(LTE)
OFDMA,SC-FDMA
Time-domain signal processing
Frequency-domainSignal processing
Time
f3f2f1
Freq
uenc
y 6.25kHzFDMA
Freq
uenc
y
Time
1 3 1 2 32
Ex. 20ms
CDMA3G
Freq
uenc
y
# 2# 3
Spreading code#15M
Hz
?4G
25kHz
ITU allocated the spectrum for 4G systems in Dec.2007. 450~470MHz (20MHz), 790~806MHz (16MHz),
2.3~2.4GMHz (100MHz), 3.4G~3.6GHz Global use(200MHz)
2012/4/12 FA/Tohoku University 124
DS-CDMA(DL/UL)(~14Mbps)
OFDMA(DL),SC-FDMA(UL)(50~100Mbps)
OFDMA, SC-FDMAMC-CDMA, or SC-CDMA?
(100M~1Gbps)
3.5G 3.9G
4GWRC’07
New band for 4G
3G band
2012/4/12 FA/Tohoku University 125
Cellular vs WLAN Until 3G,
Cellular systems WLANCellular and WLAN systems have been advancing
independently Will this continue?
Cellular systems
WLAN
Everywhere Wide coverage area Demand: increasing
data rate
High speed, but hotspot coverage
Demand: extension of coverage area
2012/4/12 FA/Tohoku University 126
Convergence of Cellular and Wireless LAN Will Happen? Cellular Wireless LAN
Cellular
DS-CDMA
WLAN
OFDM
Next GenerationWirelessDS-CDMA?MC-CDMA?
Single-carrier?OFDM?
+ Broadband data
+ Mobility
Present Year of around 2015?
IP-based Wireless Access Network Wireless access network may become closer to
present wireless LAN but with nationwide mobilitymanagement.
2012/4/12 FA/Tohoku University 127
Basestation
WNC
IP-basedcore network
LR
WNC: Wireless Network ControlLR: Location Register
IP-basedwireless access
network
2012/4/12 FA/Tohoku University 128
Heterogeneous Networks Cellular provides wide area coverage, but WLAN
provides hot spot coverage WLAN has a power problem Perhaps, hot spot coverage can only be possible due to
power problem Cellular world
Counterpart of WLAN is 3.9G/4G However, the development of 3.9G will take 5 years.
WLAN world WLAN is quite difficult to cover the whole nation WiMAX has appeared, but needs time to become a
popular system
4G wireless systems may not be based on a singlestandard
A global wireless system that consists of manydedicated wireless systems inter-connected bybroadband Internet technology
2012/4/12 FA/Tohoku University 129
Broadcasting
WLAN
Cellular
Broadcasting
WLAN
Advancement
Advancement
Advancement
2G, 3G cellular
3.5G, 3.9G cellular
WiMAX4G (Gigabit)Cellular
Technical Goal Of Wireless The available bandwidth and power are limited The ultimate goal of wireless technology is to
provide extremely high rate data services uniformly over an areawith as much low energy and narrow bandwidthas possible
2012/4/12 FA/Tohoku University 130
1GAnalog(FDMA)
2GDigital(TDMA)
3G/3.5GDigital(CDMA)
Big leap
~64kbps ~2Mbps~14Mbps
~2.4kbps
4G/5GDigital
Ultimate goal
Extremely high datarate with as much
low energyand
narrow bandwidthas possible
Wireless Network for Cloud Computing
2012/4/12 FA/Tohoku University 131
Cloud Computing Network
Wireless Access Network
User terminals (ultrahigh memory capacity, high speed data communication)
Big wireless pipe (>1Gbps)Low delay (~10ms)
A variety of data services through Internet
ITU allocated the spectrum for 4G systems in Dec. 2007. 450~470MHz (20MHz) 790~806MHz (16MHz) 2.3~2.4GMHz (100MHz) 3.4G~3.6GHz for Global use (200MHz)
200MHz bandwidth in the global frequency band must beshared by several operators (at least 2) and reusedeverywhere.
The single frequency reuse is possible with either SC orOFDM.
However, probably, an effective frequency reuse factor canbe around 25% an effective bandwidth/BS is only around12.5MHz/link under severe co-channel interference.
1Gbps/12.5MHz/BS is equivalent to 80bps/Hz/BS!!
2012/4/12 FA/Tohoku University 132
LTE-Advanced May Not Be Sufficient Peak data rate:1Gbps/BS
Total bit rate per one BS There may be many users accessing the same BS Therefore, data rate per user is less than 1Gbps
1Gbps/BS is not enough in densely populatedarea Active user density: one user/(50x50)m2
1km cell 1,256 users/BS 0.8Mbps/user
Ultimate goal >1 Gbps/BS wireless access (>10Mbps/user) Extremely low transmit power (least interference to
other users) Uniform transmission quality over a service area
2012/4/12 FA/Tohoku University 133
How To Achieve ~1Gbps with Limited Bandwidth and Power? Future wireless networks may require >1Gbps/BS
capability. LTE advanced networks (4G) are expected to
provide broadband packet data services of up to1Gbps. However, available bandwidth is limited. In December 2007, ITU allocated 3.4~3.6GHz band for
4G services. Only 200MHz is available for global use. This must be shared by at least 2 operators and by the
up/down links. Although one-cell reuse of 100MHz ispossible, effective bandwidth which can be used at eachBS is only around 12.5MHz/link.
Development of advanced wireless techniquesthat achieve a spectrum efficiency of>80bps/Hz/BS with least transmit power isdemanded.
2012/4/12 FA/Tohoku University 134
Important Technical Issues Severe channel Advanced equalization
Limited bandwidthMIMO multiplexing
Limited transmit power Distributed antenna network (DAN)
2012/4/12 FA/Tohoku University 135
Limited Bandwidth Problem
2012/4/12 137FA/Tohoku University
0
0.1
0.2
0 5 10 15 20 25 30 35 40
(dB)
S1
)6(1
)1(log3
5.3
212
4 7N=3
×S/3
F3F2
F1
F3F1
F3F4
F1F2
F3F1
F4F5
F6F1
F2F7
F3
F6F7
F1F1
F1
N=7 N=4
N=3N=1
Resources (bandwidth, power) are limited Spectrum efficiency is most important
Frequency reuse is indispensible Cluster size of 4~7 may maximize the cellular Therefore, each BS can only utilize a fraction of the
system bandwidth
How To Achieve ~1Gbps with Limited Bandwidth? Use of higher level of modulation has a limitation
1024QAM provides only 10bps/Hz. In addition, the achievable BER performance severely
degrades.
2012/4/12 FA/Tohoku University 139
16QAM(4bps/Hz)
4QAM(2bps/Hz) 1024QAM
(10bps/Hz)
MIMO Multiplexing May Be A Savior Independent data streams are transmitted simultaneously
from transmit antennas using the same carrier frequency. SDM is to increase achievable data rate within the limited
bandwidth, i.e., the channel capacity in bps/Hz.
2012/4/12 FA/Tohoku University 141
Coding S/P Signal]detection Decod.
Multipathchannel
Nt antennas Nr antennas
Channel Model
2012/4/12 FA/Tohoku University 143
Transmitted radio waves are reflected or diffracted by somelarge buildings, creating resolvable paths having time delaysof multiple of (signal bandwidth)-1.
Each resolvable path is the sum of irresolvable pathscreated by local scatterers surrounding a mobile.
The path gain hl(t) varies in time according to themovement of mobile terminal since resolvable paths areadded constructively at one time and destructively atanother time.
Localscatterers
Large obstacles
Transmitter
ReceiverReflection/diffraction
d-4
Channel Impulse Response Multipath channel can be viewed as a time varying linear
filter having impulse response h(, t). Many impulses are received with different time delays when
one impulse is transmitted from a transmitter at time t. Frequency-selective fading channel is produced.
2012/4/12 FA/Tohoku University 144
Time varyingFIRfilter
Transmit signals(t)
Received signalr(t)
1
0
)()(),(L
lll thth
time
Transmit pulse
Am
plitu
de
Wirelesschannel
0
Impu
lse r
esp
onse
Inverse ofthe signal bandwidth, 1/W
Time delay0
Observed pulses (each pulse consisting of many unresolvable pulses)
0 1 2
0.01
0.1
1
10
0 1 2 3 4 5 6 7 8 9 10
0.01
0.1
1
10
0 10 20 30 40 50 60 70 80 90 100
Frequency (MHz)
Frequency (MHz)
Cha
nnel
gai
nC
hann
el g
ain
Frequency-selective Channel The transfer function
H(f, t) of broadband channel at time t is not constant and varies over the signalbandwidth.
Challenge is totransmit broadbanddata with high qualityover such a severefrequency-selectivechannel.
2012/4/12 FA/Tohoku University 145
L=16Uniform power delay profilel-th path time delay=100l + [-50,50)ns
FA/Tohoku University 147
Transmit Power Problem Peak power is in proportion to “transmission rate” x “fc2.6
[Hata-formula]” where fc is the carrier frequency. Assume that the required transmit power for 8kbps@2GHz is 1Watt for
a communication range of 1,000m. The required peak transmission power for [email protected] needs to be
increased by 1Gbps/8kbps x (3.5GHz/2GHz)2.6 = 535,561 times, that is,536kWatt.
Obviously, this cannot be allowed.
To keep the 1W power, the communication range should bereduced by 43 times(i.e., 1,000m 23m). Femto cellularnetwork.
FA/Tohoku University 147
# M. Hata, “Empirical formula for propagation loss in land mobile radio services”, IEEE Trans. Veh. Technol., VT-29, pp. 317-325, 1980. 2012/4/12
8kbps@[email protected]
1000m23m
For broadband communications, communicationrange shrinks significantly because of thetransmit power limitation.
Fundamental change is necessary in wirelessaccess network.
2012/4/12 FA/Tohoku University 148
Radio control station
Network
Base station
Network
Base station
Advances in SC-FDE
2012/4/12 FA/Tohoku University 149
F. Adachi, Kazuki Takeda, and H. Tomeba, “Frequency-Domain Equalization for Broadband Single-CarrierMultiple Access,” IEICE Trans. Commun., Vol.E92-B,No. 05, pp. 1441-1456, May 2009.F. Adachi, Kazuki Takeda, and H. Tomeba,“Introduction of Frequency-Domain Signal Processingto Broadband Single-Carrier Transmissions in aWireless Channel,” IEICE Trans. Commun., Vol. E92-B,No.09, pp.2789-2808, Sep. 2009.
Why Single-carrier (SC)Transmission for Uplink? SC signal has lower PAPR
than OFDM. No ISI at different symbol
positions. SC is suitable for the uplink
transmission. Less expensive power
amplifier is required.
2012/4/12 FA/Tohoku University 150
-4
-3
-2
-1
0
1
2
3
4
55 57 59 61 63 65
α= 0.00
-4
-3
-2
-1
0
1
2
3
4
55 57 59 61 63 65
α= 1.00
-4
-3
-2
-1
0
1
2
3
4
55 57 59 61 63 65
α= 0.50
SC
-4
-3
-2
-1
0
1
2
3
4
55 57 59 61 63 65
OFDM
OFDMA
256 subcarriers
)2sin()(Im)2cos()(Re
)2exp()(Reveformcarrier wa Modulated
tftstftstfts
cc
c
SC signal has lessPAPR than OFDM.
Reason for this isbecause time-domainSC signal is a Nyquistfiltered signal whichhas the sameamplitude as theoriginal data symbol(ISI free) at every Tssecond(where Ts is thesymbol length).
2012/4/12 FA/Tohoku University 151PAPR(dB)
CCD
F
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
2 4 6 8 10 12
α=0.00
α=0.50
α=1.00
OFDM
OFDM
SC
Single-carrier Frequency-domain Equalization (SC-FDE) Simple one-tap FDE
Block transmission of Nc symbols Insertion of cyclic prefix (CP) at the transmitter FFT/IFFT for FDE at the receiver
2012/4/12 FA/Tohoku University 152
(b) Rx
(a) Tx
+CPTransmit
data block
Data m
od.
*D.Falconer, S. Ariyavisitakul, A, Benyamin-Seeyar and B. Eidson, “FrequencyDomain Equalization for Single-Carrier Broadband Wireless Systems,” IEEECommunications Magazine, Vol. 40, No. 4, pp. 58-66, April 2002.
*F. Adachi, D. Garg, S. Takaoka, and K. Takeda, “Broadband CDMA techniques,”IEEE Wireless Commun. Mag., Vol. 12, No. 2, pp. 8-18, April 2005.
Received data block
・・・Data
Dem
od.
Nc -point IFFT
W(0)
W(k)
W(Nc-1)
・・・
FDE
Nc -point FFT
-CP
SC-FDE can take advantage of the channel frequency-selectivity and can improve the BER performance.
Transceiver structure of SC-FDE is similar to OFDMtransceiver. Difference is the location of IFFT.
2012/4/12 FA/Tohoku University 153
CP Nc symbols
Ng symbols
Copy
Received data
Data dem
od.
-CP
FDE
OFDM
SC
ReceiverN
c-Point
IFFT
Nc–Point
FFT
Transmit data
Data m
od.
+CP
SC
OFDM Nc-Point
IFFTTransmitter
Transceiver can be designed based on OFDM For SC, FFT and IFFT are added at transmitter and receiver,
respectively.
2012/4/12 FA/Tohoku University 154
Receiver
Transmit data
Data m
od.
+CP
SC
OFDM
Nc-Point
IFFT
Transmitter
Nc-Point FFT
Received data
Data dem
od.
FDE
OFDM
SC
Nc-Point
IFFT
Nc-Point FFT
-CP
CP Nc symbols
Ng symbols
Copy
2012/4/12 FA/Tohoku University 155
Time-domain signal and noise after FDE
(b) MRC
-5
0
5
0 50 100 150 200 250
Time t
s(t)
-10
0
10
0 50 100 150 200 250
Time t
Re[
(t)]
0.01
0.1
1
10
0 50 100 150 200 250
Subcarrier index k
H(k
)
-5
0
5
0 50 100 150 200 250
Time t
s(t)
-10
0
10
0 50 100 150 200 250
Time t
Re[
(t)]
0.01
0.1
1
10
0 50 100 150 200 250
Subcarrier index k
H(k
)
(c) ZF
-5
0
5
0 50 100 150 200 250
Time t
s(t)
-10
0
10
0 50 100 150 200 250
Time t
Re[
(t)]
0.01
0.1
1
10
0 50 100 150 200 250
Subcarrier index k
H(k
)
(a) MMSE
Equivalent channel
ZFMRCMMSE
Signal
Noise
BER Performance Improvement BER performance of SC can
be significantly improved with MMSE-FDE.
However, there is still a big performance gap from the theoretical lower bound. Due to residual ISI after
MMSE-FDE. Introduction of ISI cancellation
technique can reduce the performance gap.
2012/4/12 FA/Tohoku University 156
F. Adachi, K. Takeda, and H. Tomeba, “Introduction of Frequency-DomainSignal Processing to Broadband Single-Carrier Transmissions in a WirelessChannel,” IEICE Trans. Commun., Vol. E92-B, No.09, pp.2789-2808, Sep.2009.
1.0E-05
1.0E-04
1.0E-03
1.0E-02
1.0E-01
1.0E+00
0 5 10 15 20
Ave
rage
BER
Average received Eb/N0 (dB)
MMSE
ZF
MRC
L=16-pathuniform powerdelay profileQPSKNc=256Ng=32
EGC
Matchedfilter boundLower bound
Performancegap
MMSE ,
)/()()(
MRC ),(EGC |,)(|/)(
ZF ),(/1
)(
10
2
*
*
*
NEkHkH
kHkHkH
kH
kW
s
Performance gap fromthe MF bound increasesfor higher multileveldata modulation. Performance gap from
MF bound at BER=10-3
QPSK: 6.5dB 16QAM: 8.5dB
64QAM: 11.5dB
157
0 5 10 15 20 25 30105
104
103
102
101
10
MF bound
MMSE-FDE
Average received Eb/N0(dB)
Aver
age
BER
QPSK
16QAM
64QAM
Nc=64Ng=16L=16-path
2012/4/12 FA/Tohoku University
Performancegap
Nc-P
oint
FFT
–CP
Nc-P
oint
FFT
–CP
Joint FDE/antenna diversity combining
Antenna #0
#Nr–1
)(0 kW
)(1 kWrN
Nc-P
oint
IFF
T
Dat
a de
mod
.
Recovered data
Joint FDE & Antenna Diversity Combining Transmitter/receiver structure
2012/4/12 158FA/Tohoku University
Transmit data
Dat
a m
od.
+CP
* F. Adachi, K. Takeda, and H. Tomeba, “Frequency-Domain Equalization for Broadband Single-Carrier Multiple Access,” IEICE Trans. Commun., Vol.E92-B, No. 05, pp. 1441-1456, May 2009.
* F. Adachi, K. Takeda, and H. Tomeba, “Introduction of Frequency-Domain Signal Processing to Broadband Single-Carrier Transmissions in a Wireless Channel,” IEICE Trans. Commun., Vol. E92-B, No.09, pp.2789-2808, Sep. 2009.
BER performance Antenna diversity is
powerful to improvethe BER performance.
2012/4/12 FA/Tohoku University 159
MMSE ,/|)(|
)(
ZF,|)(|
)(
)(
)()()(ˆ
10
1
0
2
1
0
2
,
1
0,
NEkH
kH
kH
kH
kW
kRkWkR
s
N
mm
m
N
mm
m
mMMSE
N
mmmMMSE
r
r
r
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
0 5 10 15 20
Ave
rage
BER
Average received Eb/N0 per receive antenna (dB)
L=16-path uniformpower delay profile
QPSKNc=256Ng=32
MMSE
ZF
Nr=1(No diversity)
Nr=2
Nr=4
Nr=3
F. Adachi, Kazuki Takeda, and H. Tomeba, “Introduction ofFrequency-Domain Signal Processing to Broadband Single-Carrier Transmissions in a Wireless Channel,” IEICE Trans.Commun., Vol. E92-B, No.09, pp.2789-2808, Sep. 2009.
Frequency-domain Block Space-time Transmit Diversity(STTD) Alamouti’s STTD for frequency-nonselective channel
Simple 2-antenna space-time transmit diversity (STTD) is aspace-time block coding that does not need transmit channelknowledge and can alleviate the complexity problem of mobileterminal.
Similar diversity gain to MRC receive diversity is obtained butwith 3 dB power penalty.
Transmit diversity can alleviate the complexity problem whenusing receive antenna diversity
2012/4/12 FA/Tohoku University 160
se sose -s*o
so s*eSTT
Den
codi
ng
Antenna #0
A pair of2 symbols
time time
Antenna #1
h0
h1 STT
Dde
codi
ng
re rotime
)ˆ,ˆ( oe ss),( oe ss
S. M. Alamouti, “A simple transmit diversity technique for wireless communications,” IEEE J. Sel. Areas Commun., vol. 16, no. 8, pp.1451-1458, 1998.
Frequency-domain Block STTD STTD encoding is applied to each frequency component of a
pair of two symbol blocks to be transmitted. Equivalent time-domain STTD encoding is derived.
2012/4/12 FA/Tohoku University 161
so(t)t STTD
encoding+ CP
n=0
n=1se(t)
so(t)
se(t) )( tNs co
)( tNs ce
t
tso(t)
se(t) )( tNs co
t
t
A pair of two symbol blockseven odd
Nc symbolseven odd
STTD-encoded symbol blocks
Nc +Ng symbols
CPTransmitantennas
)( tNs ce
Transmitter
Receiver
K. Takeda, T. Itagaki, and F. Adachi, "Application of space-time transmit diversityto single-carrier transmission withfrequency-domain equalization and receiveantenna diversity in a frequency-selectivefading channel," IEE Proc.-Commun., vol.151, No.6, pp. 627-632, Dec. 2004.
Nc-P
oint
FFT
–CP
Nc-P
oint
FFT
–CP
Antenna #0
#Nr–1
Recovered data
BER Performance With Frequency-domain Block STTD
2012/4/12 FA/Tohoku University 162
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
0 5 10 15 20 25 30
Average received E b /N 0 (dB)
Ave
rage
BER
L =16N r =1
STTDno STTD
=0dB=2dB=8dB=dB
with MMSE-STTD (Nt=2, Nr=1)
MMSEfor
,
1|)(||)(|
)(
MRCfor ,)(
for ZF |)(|
)(
)(
1
0
1
0
2,1
2,0
,1/0
,1/0
1
0
1
0
2,
,1/0
,1/0
c
gsN
mmm
m
m
n
N
mmn
m
m
NN
NEkHkH
kH
kH
kH
kH
kW
r
r
Frequency-domain blockMMSE-STTD gives afurther improved BERperformance.
An MMSE-STTD gain of3.2dB at BER=10-4 isobtained when =0 dB.
Cyclic Delay Transmit Diversity (CDTD) Well known transmit
diversity is cyclic delaytransmit diversity (CDTD).
The same signal block istransmitted simultaneouslyfrom Nt transmit antennaswith different cyclic timedelays.
Weak frequency selectivechannel can betransformed into strongerselective channel to obtainlarger frequency diversitygain.
2012/4/12 FA/Tohoku University 163
J. H. Winters, “Diversity gain of transmit diversity inwireless systems with Rayleigh fading,” IEEE Trans.Veh. Technol., Vol. 47, pp.119-123, Feb. 1998.
(a) Transmitter
Datamodulation
Data
n=0
n=Nt-1+CPNt-1
+CP
Dat
ade
mod
. Data・・
・
(b) Receiver
m=0
m=Nr-1
IFFT
FFT
W0(Nc-1)
WNr-1(Nc-1)
W0(0)
W0(Nc-1)
Rem
oval
of
CP
Rem
oval
of
CP
0(t)
Nr-1
(t)
2012/4/12 FA/Tohoku University 164
STTDとDTDの特性比較
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
0 5 10 15 20 25 30
Average received E b /N 0 (dB)
Ave
rage
BER
=0dB=8dB=dB
L =16N r =1
STTDDTDNo transmitdiversity
Nt=2
CDTD
QPSKNc=256
As the channel frequency-selectivity gets stronger (as gets smaller), the BERperformance improves forboth MMSE-STTD and -CDTD.
When =8 dB, the requiredEb/N0 for BER=10-4 is 2.5dBsmaller with MMSE-STTD thanwith MMSE-CDTD. This isbecause the equivalent no. ofreceive antennas is two timeslarger with MMSE-STTD.
When =0 dB (the channelselectivity is strong enough),MMSE-CDTD gain is as smallas 0.4dB, but MMSE-STTDgain is as much as 3.2dB.
Iterative ISI Cancellation Residual ISI replica is generated and subtracted from the
received signal.
MMSE weight
2012/4/12 FA/Tohoku University 166
10
)()(
)1()()()(
)()(
)(
)()()(
)(ˆ)/1(
ˆ
~ˆ/2~ˆ
replica ISI residual theis ~ where
~ˆ~
cNkc
ii
iiiss
i
ii
i
iii
kHNA
ATE
HWH
sFIHM
RWR
M
MRR
# Kazuaki Takeda and F. Adachi, “Throughput of multicode DS-CDMA HARQ using MMSE turbo equalization” IEICE Technical report, RCS2006-167, pp.59-64, Nov. 2006.
# K. Ishihara, K. Takeda, and F. Adachi, “Frequency-domain Multi-stage Soft Interference Cancellation for DS-CDMA Uplink Signal Transmission,” IEICE Trans. Commun., Vol.E90-B, No.5, pp. 1152-1161, May 2007.
.estimatesoft its andblock signal ed transmittebetween therror squared average theis |)(~)(| where
)/(|)(|)()(
bygiven is ,~or error vect theofelement each of
error squared average theminimizes which weight,MMSE The
2)1()1(
10
2)1(
*)(
)()()(
tstsE
NEkHkHkW
A
iis
ii
iii
FsRE
-CP
Nc-
poin
t FF
T
Nc-
poin
t IF
FT
ReceivedSignal block
Soft replicaN
c -point FFT
Residual ISIreplica
MMSEweight
・・
・・
・・
・・
・・
・・
)()( kW i
)(kR
)(~ )1( kS i
)()( kM i
)1(~ is
)(ˆ )( kR i )(~ )( kR i
-+
MMSE-FDE/w residual ISI cancellation can improve the BER performance.
But, the performance improvement is smaller for a higher multilevel modulation. Performance gap from
MF bound @BER=10-4
QPSK: 2dB 16QAM: 5dB 64QAM: 12dB
1702012/4/12 FA/Tohoku University
0 5 10 15 20 25 3010-5
10-4
10-3
10-2
10-1
100
MF bound
MMSE-FDE
Average received Eb/N0(dB)
Aver
age
BER
QPSK
16QAM
64QAM
Nc=64Ng=16L=16-path
Performancegap
K. Takeda and F. Adachi, “Frequency-domain ICICancellation with MMSE Equalization for DS-CDMADownlink,” IEICE Trans. Commun., Vol.E89-B,No.12, pp.3335-3343, Dec. 2006.
Frequency-domain Block Signal Detection BER performance of the broadband single-carrier signal
transmissions significantly degrades due to the stronginter-symbol interference (ISI).
Frequency-domain block signal detection which combinesfrequency-domain equalization (FDE) and maximumlikelihood detection (MLD) is very promising.
2012/4/12 FA/Tohoku University 171
y YFrequency-
domain block signal
detection
-CP
Nc-
poin
t FF
T
Dat
aD
emod
. Received
data
Data
Dat
am
od. d
+CP
timed(0) d(2) d(Nc-1)∙∙∙
Nc-symbol transmit block
Frequency-selective
channel H
NdHFY s
s
TE2
T. Yamamoto, K. Takeda, and F. Adachi, “A Study of Frequency-Domain Signal Detection forSingle-Carrier Transmission,” Proc. IEEE VTC 2009 Fall, 20-23 Sept, 2009, Araska, US.
Equivalent channel
MMSE-FDE Combined With SIC
...
...
......
...
...
Iteration stage0th layer
1th layer
...
Nc-1th layer
Initial stage
...
0th layer
Nc-1th layer
1th layer
no cancellation
cancellation
d(0) d(1) d(2) ・・・ d(Nc-1)
d(0) d(1) d(2) ・・・ d(Nc-1)
d(0) d(1) d(2) ・・・ d(Nc-1)
d(0) d(1) d(2) ・・・ d(Nc-1)
d(0) d(1) d(2) ・・・ d(Nc-1)
d(0) d(1) d(2) ・・・ d(Nc-1)
cancellation
cancellation
cancellationcancellation
cancellation
no cancellation
2012/4/12 FA/Tohoku University
Interference cancellation
One-tap MMSE-FDE detectionReplica
generation
Decision resultY
172
T. Yamamoto, K. Takeda, and F. Adachi, “A Study of Frequency-Domain Signal Detection for Single-Carrier Transmission,” Proc. IEEE VTC 2009 Fall, 20-23 Sept, 2009, Araska, US.
Hard decision iterative SIC
2012/4/12 FA/Tohoku University 173
1.0E-01
1.0E-02
1.0E-03
1.0E-04
1.0E-050 5 10 15 20
QPSKNcNg=16
Theoretical lower bound
Average received Eb/N0 (dB)
Ave
rage
BER
One-tap MMSE-FDE
Hard decisioniterative V-BLAST
i=0i=1i=2i=3
MMSE detection
1.0E-01
1.0E-02
1.0E-03
1.0E-04
1.0E-055 10 15 20 25
16QAMNcNg=16
Theoretical lower bound
Average received Eb/N0 (dB)
Ave
rage
BER
One-tap MMSE-FDE
Hard decisioniterative V-BLAST
i=0i=1i=2i=3
MMSE detection
(b) 16QAM (a) QPSK
Soft decision iterative SIC One iteration is sufficient. More than one iterations do not provide
noticeable additional performance improvement.
2012/4/12 FA/Tohoku University 174
1.0E-01
1.0E-02
1.0E-03
1.0E-04
1.0E-050 5 10 15 20
QPSKNcNg=16
Theoretical lower bound
Average received Eb/N0 (dB)
Ave
rage
BER
One-tap MMSE-FDE
Soft decisioniterative V-BLAST
i=0i=1i=2i=3
MMSE detection
1.0E-01
1.0E-02
1.0E-03
1.0E-04
1.0E-055 10 15 20 25
16QAMNcNg=16
Theoretical lower bound
Average received Eb/N0 (dB)
Ave
rage
BER
One-tap MMSE-FDE
Soft decisioniterative V-BLAST
i=0i=1i=2i=3
MMSE detection
(b) 16QAM (a) QPSK
T. Yamamoto, K. Takeda, and F. Adachi, “Frequency-Domain Block Signal Detection for Single-Carrier Transmission,” IEICE Trans. Commun., Vol.E93-B, No.08, pp. 2104 – 2112, Aug. 2010.
Frequency-domain QRM-MLD Maximum likelihood detection employing QR decomposition
and M-algorithm (called QRM-MLD) can achieve near MLDperformance with much reduced computational complexity.
Cyclic prefix insertedSC transmission (CP-SC)
2012/4/12 FA/Tohoku University 175
Data
Dat
am
odul
atio
n
+CP
Nc-
poin
t D
FT
FDBD
usi
ng
QRM
-MLD
Received
data
Dat
a de
-m
odul
atio
n
d s Y(TA)
QR decomposition M
LD u
sing
M
-alg
orith
m
QH
R
QRFH
Y Decisionvariable
H
F
Data symbols(0)CP(0) Data symbols(1)CP(1)
Ng symbolsNc symbols
DFT block size
QR decomposition is applied to Y to transform detect thesymbols in decreasing order of reliability using MLD.
MLD
2012/4/12 FA/Tohoku University 176
Y QR decomposition
Multiplication of QH
MLD based on M-algorithm
Decision result
QRHF YQY Hˆ
NQ
0
NQRdYQY
H
cNN
N
N
s
s
c
H
s
sH
Nd
dd
R
RRRRR
TE
NY
YY
TE
cc
c
c
)1(
)1()0(
2
)1(ˆ
)1(ˆ)0(ˆ
2ˆ
1,1
1,11,1
1,01,00,0
.
1
0
2
01,1
2
1
1
)1(2)1(ˆ2ˆ )(
and level modulation thedenotes where
)(minargˆ
c
ccc
cNc
N
n
n
ncnNnN
s
sc
s
sN
NX
nNdRTEnNY
TEL
X
L
RdYd
ddd
In case of iterativeSIC&MMSED, as thenumber of iterationsincreases, the BERperformance improves andapproaches that of thetheoretical lower bound.
Much better performancecan be achieved by usingQRM-MLD than by usingiterative SIC&MMSED.
However, a large M isrequired.
2012/4/12 FA/Tohoku University 181
[Iterative SIC&MMSED]T. Yamamoto, K. Takeda, and F. Adachi, “A Study of Frequency-Domain Signal Detection for Single-Carrier Transmission,” Proc. IEEE VTC 2009 Fall, 20-23 Sept, 2009, Araska, USA.[QRM-MLD]T. Yamamoto, Kazuki Takeda, and F. Adachi, "Single-carrier transmission using QRM-MLD with antenna diversity," Proc. The 12th International Symposium on Wireless Personal Multimedia Communications (WPMC2009), Sendai, Japan, 7-10 Sep. 2009.
Average received Eb/N0 (dB)
1.0E-01
1.0E-02
1.0E-03
1.0E-04
1.0E-050 5 10 15 20
Aver
age
BER MMSE-FDE
QRM-MLD
Theoretical lower bound
Iterative SIC&MMSED
QPSKNc=64Ng=16L=16-path uniform
i=1i=0
M=16 M=64M=256
i=2
QPSK
In case of iterativeSIC&MMSED, as thenumber of iterationsincreases, the BERperformance improves andapproaches that of thetheoretical lower bound.
Much better performancecan be achieved by usingQRM-MLD than by usingiterative SIC&MMSED.
However, a large M isrequired.
2012/4/12 FA/Tohoku University 182
Average received Eb/N0 (dB)
1.0E-01
1.0E-02
1.0E-03
1.0E-04
1.0E-055 10 15 20 25
Aver
age
BER MMSE-FDE
QRM-MLD
Theoretical lower bound
Iterative SIC&MMSED
16QAMNc=64Ng=16L=16-path uniform
i=1i=0
M=16 M=64M=256
i=2
16QAM
[Iterative SIC&MMSED]T. Yamamoto, K. Takeda, and F. Adachi, “A Study of Frequency-Domain Signal Detection for Single-Carrier Transmission,” Proc. IEEE VTC 2009 Fall, 20-23 Sept, 2009, Araska, USA.[QRM-MLD]T. Yamamoto, Kazuki Takeda, and F. Adachi, "Single-carrier transmission using QRM-MLD with antenna diversity," Proc. The 12th International Symposium on Wireless Personal Multimedia Communications (WPMC2009), Sendai, Japan, 7-10 Sep. 2009.
Cyclic prefix inserted SC (CP-SC) transmission Last Ng symbols in transmit symbol block are used as CP to
make the received signal block as a circular convolution oftransmit symbol block and channel.
Since last Ng symbols are unknown, a large number ofsurviving symbols in QRM-MLD is requried.
Two types of CP
2012/4/12 FA/Tohoku University 183
Data
Dat
a m
od.
+CP
Nc-
poin
t D
FT
FDBD
usi
ng
QRM
-MLD
Received
data
Dat
a de
mod
.
-CPd Y(CP)
Data symbols(0)CP(0) Data symbols(1)CP(1)
Ng symbolsNc symbols
DFT block size
Training sequence aided SC (TA-SC) Last Ng symbols in DFT block are known and therefore, more
accurate signal detection using QRM-MLD is possible than CP-SC.
However, DFT size is larger than the case of CP-SC, increasingthe complexity.
2012/4/12 FA/Tohoku University 184
Data
Dat
a m
od.
+TS
Nc+
Ng-
poin
t D
FT
FDBD
usi
ng
QRM
-MLD
Received
data
Dat
a de
mod
.
d s Y(TA)
Data symbols(0)TS Data symbols(1)TS
Nc symbols Ng symbols
DFT block size
T. Yamamoto, K. Takeda, and F. Adachi, “Frequency-domain Block Signal Detection with QRM-MLD for Training Sequence-aided Single-carrier Transmission,” EURASIP Journal on Advances in Signal Processing, Vol. 2011, Article ID 575706, Vol. 2011, doi:10.1155/2011/575706.
Frequency-domain QRM-MLD Maximum likelihood detection employing QR decomposition
and M-algorithm (called QRM-MLD) can achieve near MLDperformance with much reduced computational complexity.
Training sequence aided SC(TA-SC)
2012/4/12 FA/Tohoku University 185
Data
Dat
am
odul
atio
n
+TS
Nc+
Ng-
poin
t D
FT
FDBD
usi
ng
QRM
-MLD
Received
data
Dat
a de
-m
odul
atio
n
d s Y(TA)
QR decomposition M
LD u
sing
M
-alg
orith
m
QH
R
QRFH
Y Decisionvariable
H
F
Data symbols(0)TS Data symbols(1)TS
Nc symbols Ng symbols
DFT block size
5 10 15 20 25
16QAMNc=64Ng=16
Average received Eb/N0 (dB)
Aver
age
BER MF
bound
10-1
10-2
10-3
10-4
10-5
M=4M=16M=256
CP-SCTA-SC
Frequency-domain QRM-MLD
MMSE-FDE
When TA-SC is used,the required number ofsurviving paths in theM-algorithm is greatlyreduced whileachieving almost thesame BER performanceas CP-SC.
The overall complexityrequired for SCfrequency-domainQRM-MLD blockdetection in TA-SC isreduced to about 7.4%of that in CP-SC incase of 16QAM.
2012/4/12 186FA/Tohoku University
T. Yamamoto, K. Takeda, and F. Adachi, “Frequency-domain Block Signal Detection with QRM-MLD for Training Sequence-aided Single-carrier Transmission,” EURASIP Journal on Advances in Signal Processing, Vol. 2011, Article ID 575706, Vol. 2011, doi:10.1155/2011/575706.
Overlap FDE The insertion of CP reduces the throughput by a factor of
1/(1+Ng/Nc).
Without CP insertion, the inter-block interference (IBI) isproduced.
2012/4/12 FA/Tohoku University 187
DataCPCopy
Nc symbolsNg symbols
NoiseIBI residualISI residual)()(12ˆ
FDEMMSEafter signaldomain -Time
2
IBI theis )( where
2signal Received
1
0
1
0mod)(
1
0mod)(
n
N
kcc
sn
L
llnnNnlnl
c
sn
nn
L
lNlnl
c
sn
skHkWNT
Es
uusshTE
t
nshTEr
c
cl
c
:
IBI
Block average equivalent channel gain
*Kazuki Takeda, H. Tomeba, K. Takeda, and F. Adachi, “DS-CDMAHARQ with Overlap FDE,” IEICE Trans. Commun., Vol. E90-B,No.11, pp.3189-3169, Nov. 2007.
*K. Takeda, H. Tomeba, and F. Adachi, “Iterative Overlap FDE forDS-CDMA without GI,” Proc. 2006 IEEE 64th Vehicular TechnologyConference (VTC), Montreal, Quebec, Canada, 25-28 Sept. 2006.
Impulse response of circular FDE filter concentrates at the vicinity of t=0. This means that the residual
IBI after FDE is localized only at the both ends of FFT block.
This property can be exploited to mitigate the IBI after FDE.
Overlap FDE
2012/4/12 FA/Tohoku University 188
0
0.1
0.2
0.3
0.4
0.5
-128 -64 0 64 128
t
Inpu
lse re
spon
se a
fter F
DE MMSE-FDE
N c =256L =16E b /N 0=10dB
FFT blockFFT block
time
FFT block
A partial sequence of Msymbols is picked up
*Kazuki Takeda, H. Tomeba, K. Takeda, and F.Adachi, “DS-CDMA HARQ with Overlap FDE,” IEICETrans. Commun., Vol. E90-B, No.11, pp.3189-3169,Nov. 2007.
*K. Takeda, H. Tomeba, and F. Adachi, “IterativeOverlap FDE for DS-CDMA without GI,” Proc. 2006IEEE 64th Vehicular Technology Conference (VTC),Montreal, Quebec, Canada, 25-28 Sept. 2006.
If M128 is used, the residual IBI can be minimized and theBER performance close to MMSE-FDE with CP insertion canbe achieved.
2012/4/12 FA/Tohoku University 189
1.0E-05
1.0E-04
1.0E-03
1.0E-02
1.0E-01
0 5 10 15 20
Ave
rage
BER
Average received Eb/N0 (dB)
L=16-pathuniform power delay profileQPSKNc=256
Conventional MMSE-FDEusing Ng=32-GI
Overlap FDEM = 64
128192224240256
M symbols
FFT block window
0 Nc-1
Joint overlap FDE & ICI cancellation provides good BERperformance close to the lower-bound.
2012/4/12 FA/Tohoku University 190
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
0 5 10 15 20Average received E b /N 0 (dB)
Ave
rage
BER
QPSKN c =256, M =160
L =16SF =U =16
Lowerbound
Overlap FDEonly
Iterativejoint overlap FDE& ICI cancellationi =1
2 3 4
Number of iteration
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
0 5 10 15 20Average received E b /N 0 (dB)
Ave
rage
BER
QPSKN c =256, M =160
L =16SF =1
Lowerbound
Overlap FDEonly
Iterativejoint overlap FDE& ICI cancellation
i =1 2 3 4
Number of iteration
(a) SF=U=1 (b) SF=U=16
Application of Overlap FDE to DS-CDMA Overlap FDE
improves the throughput of SC HARQ.
Since CP insertion is not required, overlap FDE can be applied to the present HSDPA using SC-CDMA. Much higher
throughput can be achieved.
The 3G air interface does not need to be changed at all.
2012/4/12 FA/Tohoku University 191
0
0.5
1
1.5
2
-5 0 5 10 15 20 25 30
Thro
ughp
ut (
bps/
Hz)
Average received Es/N0 (dB)
L=16-pathuniform power delay profileQPSKTurbo coded HARQType II S-P2Nc=256, SF=C=16
Overlap FDE (M=64)
Conventional FDEusing Ng=32-chip GI
Coherent rake combining
Frequency-domain Transmit Filtering Frequency-domain
filtering is a partial spreading of the transmit signal in the frequency domain. Frequency domain filtering
increases the number of frequency components from Nc to (1+)Nc.
Lower PAPR is achieved at the price of wider signal bandwidth.
Larger frequency-diversity gain can be obtained through joint MMSE-FDE & spectrum combining.
2012/4/12 FA/Tohoku University 192
Transmit data
Dat
a m
od.
+CP
Nc-P
oint
FF
T
Transmitter
Freq
uenc
y-do
mai
nFi
lter
2Nc-P
oint
IF
FT
Filtering
Nc frequency components
(1+)Nc frequency components
NcNc/2Nc/2Nc 0
NcNc/2Nc/2Nc 0Frequency index k
Am
plitu
de o
f si
gnal
spec
trum
Am
plitu
de o
f si
gnal
spec
trum
Joint MMSE-FDE & Spectrum Combining Weight that minimizes the restored signal spectrum after
spectrum combining and transmitted signal spectrum
196
1
0
1
1
2
*
1
1
)(
)()(
weightMMSE theis )( where
12/~2/,)()()(ˆ
NENpkH
pNkHpNkW
kW
NNkpNkWpNkRkR
s
pc
cc
ccp
cc
Nc 2Nc0-Nc-2Nc Frequency index k
Am
plitu
de o
f si
gnal
spe
ctru
m
)(ˆ kR
)( kR
Nc frequency components
2012/4/12 FA/Tohoku University
T. Obara, K. Takeda, and F. Adachi, "Joint Frequency-domain Equalization & Spectrum Combining for The Reception of SC Signals in the Presence of Timing Offset," Proc. IEEE VTC2010-Spring, Taipei, Taiwan, 16–19 May 2010.
When transmit filtering is used,the BER performance of SC-FDE is sensitive to the timingoffset between transmitterand receiver. As the roll-off factor of the transmit
filter increases, larger frequencydiversity gain can be achieved.
However, the performance degradesif the timing offset exists.
Timing Offset Problem
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
5 7 9 11 13 15 17 19
= 0.220.51.0
Ideal timing( = 0)
QPSKNc = 256Ng = 32L =16-path uniform
]5.0,5.0[
Aver
age
BER
Average received Eb/N0 (dB)
Conventional MMSE-FDE
Timing offset
* T. Obara, H. Tomeba, Kazuki Takeda, and F. Adachi,“Impact of Timing Offset on DS-CDMA with Overlap FDE,”The 5th IEEE VTS Asia Pacific Wireless CommunicationsSymposium (APWCS2008), Tohoku University, Sendai,Japan, 21-22 Aug., 2008.
2012/4/12 200FA/Tohoku University
Non Zero (ISI)
Timing offset
0 Ts 2Ts 3Ts 4TsTs2Ts3Ts4Tst
Out
put o
f rec
eive
filte
r
BER Performance As the filter roll-off factor increases, the signal bandwidth
widens and hence, larger frequency diversity gain isobtained.
2012/4/12 FA/Tohoku University 201
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
0 10 20
Aver
age
BER
Average received Eb/N0 (dB)
QPSKNc=256, Ng=32L=32-path uniform=0
□ =0△ 0.5× 1
=0
=1
100
10-1
10-2
10-3
10-4
102 4 6 8
QPSK
OFDM=0
=0.25
=0.5
=0.75=1
PAPR0 (dB)
Pr(P
APR
>PA
PR0)
Robust Against Timing Offset Joint MMSE-FDE & spectrum combining is robust against
the receive timing offset. MMSE weight removes the phase rotation produced by timing offset. BER performance is almost insensitive to the timing offset.
202
Symbol rate MMSE-FDE
=0.5
Symbol rate MMSE-FDE
Joint MMSE-FDE & spectrum combining
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
5 10 15 20
Aver
age
BER
Average received Eb/N0 (dB)
QPSKNc = 256Ng = 32L=32-path uniform=0.5
∈[0.5, 0.5]
Freq. diversity gain
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
5 10 15 20
Aver
age
BER
Average received Eb/N0 (dB)
QPSKNc = 256Ng = 32L=32-path uniform=1
∈[0.5, 0.5]
=1
Joint MMSE-FDE & spectrum combining
Freq. diversity gain
2012/4/12 FA/Tohoku University
MIMO Combined With FDE
2012/4/12 FA/Tohoku University 245
Even using FDE, reliable communication over an entire service is quite difficult due to shadowing and distance dependent path losses.
Furthermore, effective bandwidth per each link seems to be around 12.5MHz. How to achieve 1Gbps using the limited bandwidth?
MIMO Multiplexing Next generation (4G) wireless systems are
expected to provide broadband packet dataservices of up to 1Gbps. However, availablebandwidth is limited. In December 2007, ITU allocated 3.4~3.6GHz band for
4G services. Only 200MHz is available for global use. This must be shared by at least 2 operators and by the
up/down links. Although one-cell reuse of 100MHz ispossible, effective bandwidth which can be used at eachBS is only around 12.5MHz/link. 1Gbps/12.5MHz isequivalent to 80bps/Hz/BS!!
Future 5G wireless systems require >1Gbps/BScapability. Development of a highly spectrum efficient wireless
transmission technology of >80bps/Hz/BS is demanded.2012/4/12 FA/Tohoku University 259 FA/Tohoku University 260
Is Super-high Level Modulation Helpful? To achieve 1Gbps/BS using 12.5MHz で bandwidth, the
required spectrum efficiency is1Gbps/12.5MHz=80bps/Hz/BS!
We need to adopt 280QAM. However, this is impractical. BER increases as the modulation level increases.
16QAM(4bps/Hz)
256QAM(8bps/Hz)
1024QAM(10bps/Hz)
260FA/Tohoku University
Signal constellation
2012/4/12
MIMO Multiplexing It is not a practical approach to use super-high level modulation
like 280QAM. Independent data streams are transmitted simultaneously from
transmit antennas using the same carrier frequency. Multiplexing is to increase achievable data rate within the limited
bandwidth, i.e., the channel capacity in bps/Hz.
2012/4/12 FA/Tohoku University 261
Coding S/P Signaldetection Decod.
Multipathchannel
Nt antennas Nr antennas
G. J. Foschini and M. J. Gans, “On limits of wireless communicationsin a fading environment when using multiple antennas,” WirelessPersonal Commun., Vol.6, No. 3, pp.311-335, Mar. 1998.
Single-carrier MIMO Multiplexing Joint MMSE frequency-domain equalization (FDE) and
parallel interference cancellation (PIC) is repeated for de-multiplexing while achieving frequency-diversity gain.
2012/4/12 FA/Tohoku University 263
Frequency-domainiterative PIC
-CP
・・・
-CP
Nr antennasReceived
data
・・・ Nt
Nt antennas
+CP
・・・
+CP
S/P
Data Mod.
P/S
・・・
・・・
LLR cal.
・・・
・・・2D MMSE-FDE
Nt signals
Frequency-domain Time-domain
Nr signals
Nt signals
SISO channel
Nr
Nt
TFFTPIC… …
… …
......
...
...
SIMO channel
Nt signals
FFT
i=0
i>0
Replica
generation
IFFT・・・・・・
… …
… …
...
...
......
MIMO channel
・・・
A. Nakajima, D. Garg and F.Adachi, “Frequency-domainiterative parallel interferencecancellation for multicode DS-CDMA-MIMO multiplexing,” Proc.IEEE VTC’05 Fall, Vol.1, pp. 73-77,Dallas, U.S.A., 26-28 Sept. 2005.A. Nakajima and F. Adachi,“Iterative FDIC using 2D-MMSEFDE for turbo-coded HARQ in SC-MIMO multiplexing,” IEICE Trans.Commun. Vol. E90-B, No.3,pp.693-695, Mar. 2007.
Throughput of SC MIMO Multiplexing MIMO signal detection
using FDE and FDIC provides higher throughput than OFDM.
2012/4/12 FA/Tohoku University 267
FDIC
Replica gene.
Start
LLR comp.
Output
MMSE FDE Weight comp.
# A. Nakajima, D. Garg and F. Adachi, “Frequency-domain iterative parallel interference cancellation formulticode DS-CDMA-MIMO multiplexing,” Proc. IEEEVTC’05 Fall, Vol.1, pp. 73-77, Dallas, U.S.A., 26-28Sept. 2005.
# A. Nakajima and F. Adachi, “Iterative FDIC using 2D-MMSE FDE for turbo-coded HARQ in SC-MIMOmultiplexing,” IEICE Trans. Commun. Vol. E90-B, No.3,pp.693-695, Mar. 2007.
0
1
2
3
4
5
6
7
8
-5 0 5 10 15 20 25
Th
rou
ghp
ut (
bit
/s/H
z)
Average received Es/N0 per receive antenna (dB)
SC (4, 4)MIMO SDM with iterative FDIC(i=4)HARQ type-II S-P2QPSK, Nc=256, Ng=32L=16-path uniform power delay profile
SC w/ FDIC
SC w/o FDIC
OFDMw/ FDIC
OFDM w/o FDIC Single transmit antenna limit
0
1
2
3
4
5
6
7
8
-5 0 5 10 15 20 25 30
Average received E s /N 0 per antenna (dB)
Thro
ughp
ut (b
ps/H
z)
MC-CDMA (4,4)SDM with iterative FDIC, QPSK,SF (=U )=256, N c=256, N g =32, L=16, S-P2, i=4
OFDM
MC DS
OFDM
w/ FDICw/o FDIC
MC, DS
MC, DS
w/o FDIC
w/ FDIC
Throughput of Spread MIMO Multiplexing Spread MIMO multiplexing
w/FDIC provides the throughput superior to non-spread SC. Much better throughput than OFDM
in a high Es/N0 region. Almost the same throughput as
OFDM in a low Es/N0 region.
2012/4/12 FA/Tohoku University 270
FDIC
Replica gene.
Start
LLR comp.
End
2D-MMSE FDE
Multicodedespreading
Multicodespreading
Weight comp.
A. Nakajima, D. Garg and F. Adachi, “Frequency-domain iterative parallelinterference cancellation for multicode DS-CDMA-MIMO multiplexing,”Proc. IEEE VTC’05 Fall, Vol.1, pp. 73-77, Dallas, U.S.A., 26-28 Sept. 2005.A. Nakajima and F. Adachi, “Iterative FDIC using 2D-MMSE FDE for turbo-coded HARQ in SC-MIMO multiplexing,” IEICE Trans. Commun. Vol. E90-B,No.3, pp.693-695, Mar. 2007.
0
1
2
3
4
5
6
7
8
-5 0 5 10 15 20 25
Th
rou
ghp
ut (
bit
/s/H
z)
Average received Es/N0 per receive antenna (dB)
SC (4, 4)MIMO SDM with iterative FDIC (i=4)HARQ type-II S-P2QPSK, Nc=256, Ng=32L=16-path uniform power delay profile
SC w/ FDIC
SC w/o
FDIC
OFDMw/ FDIC
OFDM w/o FDIC
MIMO Multiplexing MIMO multiplexing is powerful.
But, use of too many antennas at a mobile terminal is notpractical
Furthermore, strong co-channel interference exists due tofrequency reuse.
2012/4/12 FA/Tohoku University 272
dB.in loss shadowing theis andexponent losspath theis MT, and BSbetween distance theis where
,10
1, if,min
1log,min
,min11log,min
10/
2
1
0
1
0
2
2
rr
NNNN
NNN
hNNN
NNC
t
rtrt
rrt
N
n
N
nnn
rttrt
r
r
t
ttr
SNRTransmit
capacity Channel
One Possible Solution How to implement many antennas in a hand
portable unit? Wearable antenna on your head?
2012/4/12 FA/Tohoku University 273
Wearable antennaNo space available at a small hand portable unit
MIMO Cannot Solve Power Problem Transmit power is another important issue Broadband radio requires prohibitively large
transmit power
2012/4/12 FA/Tohoku University 275
dB.in loss shadowing theis andexponent losspath theis MT, and BSbetween distance theis where
,10
1, if,min
1log,min
,min11log,min
10/
2
1
0
1
0
2
2
rr
NNNN
NNN
hNNN
NNC
t
rtrt
rrt
N
n
N
nnn
rttrt
r
r
t
ttr
SNRTransmit
capacity ChannelReceived
SNR
FA/Tohoku University 276
Peak power is in proportion to “transmission rate” x “fc2.6
[Hata-formula]” where fc is the carrier frequency. Assume that the required transmit power for 8kbps@2GHz is 1Watt for
a communication range of 1,000m. The required peak transmission power for [email protected] needs to be
increased by 1Gbps/8kbps x (3.5GHz/2GHz)2.6 = 535,561 times, that is,536kWatt. Obviously, this cannot be allowed.
To keep the 1W power, the communication range should be reducedby 43 times(i.e., 1,000m 23m).
A new wireless access network architecture needs to bedeveloped.
FA/Tohoku University 276
# M. Hata, “Empirical formula for propagation loss in land mobile radio services”, IEEE Trans. Veh. Technol., VT-29, pp. 317-325, 1980.
2012/4/12
8kbps@[email protected]
1000m23m
F. Aachi, “Wireless past and future - evolvingmobile communications systems,” IEICE Trans.Fundamentals, vol. E84-A, pp. 55-60,Jan. 2001.E. Kudoh and F. Adachi, “Power and FrequencyEfficient Wireless Multi-hop Virtual CellularConcept,” IEICE Trans. Commun., Vol.E88-B,No.4, pp.1613-1621, Apr. 2005.
Wireless EvolutionTo
5G (Gigabit Wireless)
Voice to Video Cloud Computing Network >1Gbps access capability 3 important technical issues
Limited bandwidth MIMO multiplexing Severe channel Advanced equalization Transmit power Distributed antenna
network (DAN)
2012/4/12 FA/Tohoku University 278
Wireless Network for Cloud Computing
2012/4/12 FA/Tohoku University 279
Cloud Computing Network
Wireless Access Network
User terminals (ultrahigh memory capacity, high speed data communication)
Big wireless pipe (>1Gbps)Low delay (~10ms)
A variety of data services through Internet
LTE-Advanced May Not Be Sufficient Peak data rate:1Gbps/BS
Total bit rate per one BS There may be many users accessing the same BS Therefore, data rate per user is less than 1Gbps
1Gbps/BS is not enough in densely populatedarea Active user density: one user/(50x50)m2
1km cell 1,256 users/BS 0.8Mbps/user
Ultimate goal >1 Gbps/BS wireless access (>10Mbps/user) Extremely low transmit power (least interference to
other users) Uniform transmission quality over a service area
2012/4/12 FA/Tohoku University 280
How To Achieve ~1Gbps with Limited Bandwidth? LTE advanced networks (4G) are expected to
provide broadband packet data services of up to1Gbps. However, available bandwidth is limited. In December 2007, ITU allocated 3.4~3.6GHz band for
4G services. Only 200MHz is available for global use. This must be shared by at least 2 operators and by the
up/down links. Although one-cell reuse of 100MHz is possible, an
effective bandwidth which can be used at each BS isonly around 12.5MHz/link. 1Gbps/12.5MHz is equivalentto 80bps/Hz/BS!!
Future wireless networks may require >1Gbps/BScapability. Development of advanced wireless techniques that
achieve a spectrum efficiency of >80bps/Hz/BS withleast transmit power is demanded.
2012/4/12 FA/Tohoku University 281
Important Technical Issues Severe channel Advanced equalization
Limited bandwidthMIMO multiplexing
Limited transmit power Distributed antenna network (DAN)
2012/4/12 FA/Tohoku University 282
Reduced Communication Range For broadband communications, communication
range shrinks significantly because of thetransmit power limitation.
Fundamental change is necessary in wirelessaccess network.
2012/4/12 FA/Tohoku University 283
Radio control stationCore
Network
Base station
CoreNetwork
Base station
Frequency Efficiency Let the total information bit rate be Ctotal (bps), the total
bandwidth be B (Hz), and the cluster area size be A (km2). Since the information service of bit rate Ctotal is provided in
the area of size A using the total bandwidth B, the cellularefficiency can be defined
2012/4/12 FA/Tohoku University 284
)km(bps/Hz 11 , , since and
s)channel(bpper ratebit n informatioHz)bandwidth( channel
)(km size cell
Let
)km(bps/Hz
2
2
2
SNWC
NCCNWBNSA
CWS
ABC
total
total
E.g. N=7
F4F5
F6F1
F2F7
F3
f
B
2012/4/12 285
)km(bps/Hz 11 2SNW
C
Reducing the cluster size N (e.g., N=7→4→3→1)by reducing the required SIR(a) Error control, antenna diversity(b) Cell sectorization, beam tilting
Reducing the cell size:Macro cell system (cell radius of several km)micro cell system (cell radius of several 100m)pico cell system (cell radius of several 10m)
FA/Tohoku University
Communication range reduces from 1,000m to23m, resulting in a nano-cell network of radius of23m.
Radio resource control based on present cellulararchitecture becomes troublesome.
Radio resource controle Core network
Core netowork
Base station
Nano-cell network
Nano-cell Network
2012/4/12 FA/Tohoku University 286
4G(LTE-Advanced) Is Still Not Sufficient Peak data rate:1Gbps/BS
Total bit rate per one BS There may be many users accessing the same BS Therefore, data rate per user is less than 1Gbps
1Gbps/BS is not enough in densely populatedarea Active user density: one user/(50x50)m2
1km cell 1,256 users/BS 0.8Mbps/user
2012/4/12 FA/Tohoku University 287
Evolution Into 5G5G systems are required to provide much faster
packet data services of a peak rate of >1Gbps.>1Gbps with extremely low transmit power is a
requirement.
2012/4/12 FA/Tohoku University 288
We are here
Broadbandwireless
1980 1990 2000 Year
2G~64kbps
1G~2.4kbps
Ser
vice
typ
eVo
ice
Mul
timed
ia
2010
3G~2Mbps
4G100M~1Gbps
IMT-2000
HSDPA
~14Mbps0GVoice only
point-to-point
3G LTE
50~100Mbps
NarrowbandEra
WidebandEra
BroadbandEra
?Gigabitwireless
5G>1Gbps
2020
5G System Requirement >1 Gbps/BS wireless access
(>10Mbps/user) Extremely low transmit power (least
interference to other users) Uniform transmission quality over a
service area
2012/4/12 FA/Tohoku University 289
Distributed Antenna Network (DAN)
A new wireless access network with >1Gbpswhile achieving extremely low transmit powerand providing uniform transmission qualityover a service area.
This can be achieved by introducing thedistributed antenna concept.
2012/4/12 FA/Tohoku University 290
Important Technical Issues Limited bandwidth Spatial reuse of the same frequency
bps/Hz bps/Hz/BS or bps/Hz/km2
Frequency reuse + MIMO multiplexing Severe channel Advanced equalization
Limited transmit power Distributed antenna network (DAN)
2012/4/12 FA/Tohoku University 291
Frequency Reuse To increase the bandwidth/BS (or bps/Hz/km2),
interference management is essential The same frequency must be reused because of
limited available bandwidth From the spectrum efficiency (bps/Hz/BS) point of view,
the same frequency needs to be reused at locations asclose as possible
Cochannel interference is a limiting factor on frequencyefficiency
Cochannel interference management becomes acrucial issue to realize spectrum efficientbroadband networks
2012/4/12 FA/Tohoku University 292
Reduced Communication Range For broadband communications, communication
range shrinks significantly because of thetransmit power limitation.
Fundamental change is necessary in wirelessaccess network.
2012/4/12 FA/Tohoku University 294
Radio control stationCore
Network
Base station
CoreNetwork
Base station
CoreNetwork
Base station
Coordinated Multi-point Transmission (CoMP) To improve the communication quality for a user
near the cell edge, the coordinated multi-pointtransmission (CoMP) was introduced. Improved SINR (increased capacity) with limited
transmit power This is the first step towards the realization of
distributed antenna network (DAN).
2012/4/12 FA/Tohoku University 295
3. Distributed Antenna Network
Distributed antenna network is designed torealize a nano-cell network with increasedspectrum efficiency and reduced transmit power.
Many antennas belonging to a base station (SPC:signal processing center) are distributed aroundSPC.
Each distributed antenna forms a cell Resource allocation control (frequency, time,
power) for distributed antennas is carried out bySPC.
2012/4/12 FA/Tohoku University 296
2012/4/12 FA/Tohoku University 297
Path lossShadowing
ㇾ Fading
ㇾ Path lossㇾ Shadowingㇾ Fading
Co-located antennas Distributed antennas
Distributed Antenna Network (DAN) Many antennas cooperate and act as cooperative
spatial multiplexing, diversity, or relay The problems can be mitigated which result from distance-
dependent path loss and shadowing loss as well as thefrequency-selective fading
Significant reduction in transmit power is possible
2012/4/12 FA/Tohoku University 298
DANSignal Processing
Center
Optical fiber cable
Distributed antenna
Distributed relay
Spatialmultiplexing/diversity
Relay with network coding
Huge transmit power Low spectrum efficiency Non-uniform quality
Huge transmit power Low spectrum efficiency Non-uniform quality
Very low transmit power High spectrum efficiency Uniform quality
Very low transmit power High spectrum efficiency Uniform quality
2012/4/12 FA/Tohoku University 300
High bit rate services with very low transmit power/uniform quality over a service area Distributed antenna network (DAN)
High bit rate with limited bandwidthdistributed MIMO
multiplexing/diversity/relay/beamforming
Formulation of Personal Cell Center of personal cell is a user Personal cell moves according to user movement
2012/4/12 FA/Tohoku University 301
Huge transmit power Low spectrum efficiency No-nuniform quality
Huge transmit power Low spectrum efficiency No-nuniform quality
Very low transmit power High spectrum efficiency Uniform quality
Very low transmit power High spectrum efficiency Uniform quality
BS is the center of cell
Each user is the center of cell
Uniform Quality Uniform quality over an BS area
2012/4/12 FA/Tohoku University 302
Distance from BS
Thro
ughp
ut
Present Cellular
Uniform qualityover an BS area
Cell edge
Distributed Antenna Network (DAN) Many antennas are spatially distributed around a signal
processing center (SPC), which is a gateway to the network With a high probability, some antennas close to an MT can always be
visible Antennas are connected with a SPC by a means of optical or wireless
links
2012/4/12 FA/Tohoku University 303
SPC
SPC
SPC
SPC
Distributed Antenna Layer to form a user centric cell
Signal Processing Layer
0.01
0.1
1
10
0 1 2 3 4 5 6 7 8 9 10
0.01
0.1
1
10
0 10 20 30 40 50 60 70 80 90 100
Frequency (MHz)
Frequency (MHz)
Cha
nnel
gai
nC
hann
el g
ain
Frequency-selective Channel The transfer function
H(f, t) of broadband channel at time t is not constant and varies over the signalbandwidth.
This channel is called the frequency-selective channel.
Advanced equalizationis necessary.
2012/4/12 FA/Tohoku University 304
L=16Uniform power delay profilel-th path time delay=100l + [-50,50)ns
Distributed MIMO Transmit Diversity With Pre FDE (Downlink)The distributed MIMO diversity increases
significantly the channel capacity while significantly reducing the transmit power.
2012/4/12 FA/Tohoku University 305
W0(k)
Cha
nnel
co
ding
Dat
a m
od.
Opt
. m
od.
Distributed antennas
Tran
smit
bina
ry
data
Opt. fibercable
FFT
FFT
WNt-1(k)
IFFT
IFFT
+CP
+CP
Opt
. de
-m
od.
Opt
. de
-m
od.
Distributed Antenna NetworkMT
BS
F. Adachi, K. Takeda, T. Obara, and T. Yamamoto,“Recent Advances in Single-carrier Frequency-domain Equalization and Distributed AntennaNetwork,” (invited) IEICE Trans. Fundamentals,Vol.E93-A, No.11, pp.2201-2211, Nov. 2010.
SC-MRT transmit diversity(downlink) Arbitrary number of transmit
antennas for single receiveantennas
The use of only around 5antennas near a user providesa sufficient improvement.
2012/4/12 FA/Tohoku University 306
# H. Matsuda, H. Tomeba, and F. Adachi,“Channel capacity of distributed antenna systemusing maximal ratio transmission,” The 5thIEEE VTS Asia Pacific Wireless CommunicationsSymposium (APWCS2008), Tohoku University,21~22 Aug., 2008.
1.0E-03
1.0E-02
1.0E-01
1.0E+00
0 1 2 3 4 5 6 7 8 9 10Channel capacity C (bit/s/Hz)
Prob
[C<a
bsci
ssa]
=3.5, =7.0
Es/N0=10dBL=16-path uniform
t
tt
t
t
MRT diversity
transmit diversity(downlink)
2012/4/12 FA/Tohoku University 307
c
N
k
N
nt
N
nt
s
N
nt
tt
NknW
knHNE
knH
knHknW
c t
t
t
t
t
t
1
0
1
0
2
1
0
2
1
02D
1
0
2
*
),(
constraintpower transmit under the
0 ,
,
1max
,
,,
bygiven weight,MRT-joint WF called weight FDE pre following the
using maximized iscapacity downlink Theantenna. receive single for the used be
can antennas transmit ofnumber arbitrary An
H. Matsuda, K. Takeda, and F. Adachi, “Downlink Transmit Diversity For Broadband Single-carrier Distributed Antenna Network,” to be presented at IEEE VTC 2010 Spring, Taipei, May 2010.
1.0E-02
1.0E-01
1.0E+00
0 1 2 3 4 5 6 7 8 9 10
= 3.5 = 7.0L = 16Es/N0 = 10 dBNc = 256
C (bps/Hz)
Prob
[C<a
bsci
ssa]
Nt = 12
3
5
10
2D-WF1D-WFMRT
WF-MRTWFMRT
Joint WF-MRT
Frequency-domain STBC-JTRD(downlink) Space-Time Block Coded Joint Transmit/Receive Diversity
(STBC-JTRD) is suitable for downlink application since itallows an arbitrary number of transmit antennas. Transmit FDE (channel state information (CSI) is necessary at
TX) to obtain frequency-diversity gain. Only simple addition/subtraction and complex conjugation
operations are required at the receiver.
3082012/4/12 FA/Tohoku University
Terminal
Dat
a m
od.
#0
#Ndan-1
STB
C-JT
RD
en
code
r
IFFTFFT
・・・
・・・
+CP
SPC
Ndan distributedantennas
#1
Dat
a de
mod
.
#0
STB
C-JT
RD
de
code
r
IFFT
FFT
・・・
・・・
CP
#Nmt -1Nmt receiveantennas
Cooperative Diversity (STBC-JTRD) (downlink) An arbitrary number of transmit antennas can be used.
Suitable to downlink transmission. Pre FDE to achieve frequency-diversity gain (channel state
information (CSI) is necessary). Receiver needs simple addition/subtraction and complex
conjugation operations only.
309
N (no. of transmit
antennas)
M (no. of receive
antennas)J Q Coding
rate
Arbitrary
1 1 1 12 2 2 13 3 4 3/44 3 4 3/4
2012/4/12 FA/Tohoku University
# R. Matsukawa, T. Obara, K. Takeda, and F. Adachi, “Single-carrier Distributed Antenna Network Downlink Using JointTransmit/Receive Diversity,” Proc. 12th IEEE InternationalConference on Communication Systems 2010, Singapore, 18-19 November, 2010.
# H. Matsuda, R. Matsukawa, T. Obara, K. Takeda, and F.Adachi, “Channel Capacity of Distributed Antenna NetworkUsing Space-Time Block Coded-Joint Transmit/ReceiveDiversity,” Proc. 12th IEEE International Conference onCommunication Systems 2010, Singapore, 18-19 November,2010.
* H. Tomeba, K. Takeda and F. Adachi, “Space-Time BlockCoded Joint Transmit/Receive Diversity in a Frequency-Nonselective Rayleigh Fading Channel,” IEICE Trans. Commun.,Vol.E89-B, No.8, pp.2189-2195, Aug. 2006.
#H. Tomeba and F. Adachi, “Frequency-domain Space-TimeBlock Coded-Joint Transmit/Receive Diversity for The SingleCarrier Transmission,” Proc.10th IEEE InternationalConference on Communication Systems (ICCS 2006),Singapore, 30 Oct. – Nov. 2006.
J blocks Q blocks
time…
・・・
・・・
・・・
timeWTransmit FDE
Cooperative Diversity (STBC-JTRD) (downlink) M=2 STBC-JTRD encoding/decoding
3102012/4/12 FA/Tohoku University
)()()()(
)()(
)(ˆ
*0,11,0
*1,10,0
1
0
kRkRkRkR
kDkD
kD
)()()(2
)()()()(
)(1,10.1
1,00.0
kkkTE
kRkRkRkR
k
s
s NSH
R
)}(),({ 10 kDkD )(kD
)(kW
)(kSTransmit FDEEncoding
Received signal
Decoding
Transmitsignal
)()()()()()(
)(
where)()()(
)()()()()(
with
)()(
)()(
)()()()(
)(
1,11,10,1
1,01,00,0
1,10,1
1,10,1
1,00,0
*01
*10
kHkHkHkHkHkH
k
kkAk
kDkDkDkDk
kk
kSkS
kSkSkSkS
k
danN
danN
danNdanN
H
jtrdstbc
H
HW
D
DWS
Cooperative Diversity (STBC-JTRD) (downlink) Joint WF-MRT weight that maximizes the channel
capacity
MMSE weight
2012/4/12 FA/Tohoku University 311
MRT WF
21
1
0
1
0
2,
10
2D1
0
1
0
2,
0,)(
)2/(max
)(
1)(
mt danmt danN
m
N
nnm
sN
m
N
nnm
mrtwf
kH
NE
kH
kA
11
0
1
0
1
0
2, )(1)(
N
EkH
MkA s
N
m
N
nnm
mmsemt dan
Cooperative Diversity (STBC-JTRD) (downlink) STBC-JTRD encoding when Nt=Nr=2
312
*
0,0*1
*0,1
*0
*0,11
*0,00
0,1
0,0
hdhdhdhd
ss
*
1,0*1
*1,1
*0
*1,11
*1,00
1,1
1,0
hdhdhdhd
ss
1
021,1
21,0
20,1
20,0
*0,11,0
*1,10,0
1
0ˆˆ
dd
hhhh
rrrr
dd
2012/4/12 FA/Tohoku University
Models of DAN and CN DAN
Ntotal=7 antennas areassumed.
Ndan antennas are selectedfrom Ntotal=7 antennas,based on the local averagereceived signal power (i.e.,the antenna selection isbased on the path loss plusshadowing loss).
CN Ndan antennas are selected
from Ntotal antennas co-located at the BS, basedon the instantaneousreceived signal power.
2012/4/12 FA/Tohoku University 313
Increasing the number of transmit antennas reduces the 10% outage capacity.
The use of 2 receive antennas maximizes the capacity.
4
5
6
7
8
1 2 3 4 5 6 7 8 9 10Ndan
10%
outa
ge c
apac
ity (
bps/
Hz)
Transmit Es/N0 = 10dB=3.5, =7.0, L=16Nc=256
○ Nmt = 1□ 2△ 3◇ 4
Nmt =1
2
3, 4
2012/4/12 FA/Tohoku University 314
Joint MRT-WF weight
H. Matsuda, R. Matsukawa, T. Obara, Kazuki Takeda, and F. Adachi, "Channel Capacity of Distributed Antenna Network Using Space-Time Block Coded-Joint Transmit/Receive Diversity," Proc. 12th IEEE International Conference on Communication Systems (ICCS 2010), Singapore, 17-19, Nov. 2010.
BER Distribution Increasing the number
of transmit antennas reduces the BER outage probability (probability of BER exceeding the allowable BER) significantly.
Also effective is increasing of no. of receive antennas.
2012/4/12 FA/Tohoku University 315
0.001
0.01
0.1
1
1.00E-04 1.00E-03 1.00E-02 1.00E-01
Prob
[BER
>abs
ciss
a]
BER
Transmit Es/N0=5dBL = 16 = 3.5= 7.0
Nt=1
2
3
4
Nr = 12
MMSE weight
R. Matsukawa, T. Obara, K. Takeda, and F. Adachi, "Single-carrier Distributed Antenna Network Downlink Using Joint Transmit/Receive Diversity," Proc. 12th IEEE International Conference on Communication Systems (ICCS2010), Singapore, 17-19, Nov. 2010.
Uplink FD-STTD
2012/4/12 316
Mobile Terminal
Nmt=2
cc
cc
NtNstsNtNsts
mod)()(mod)()(~
*01
*10S
FA/Tohoku University
Dat
a m
od.
STT
Den
codi
ng +CP
+CP
#0
#1s1(t) s0(t)
2 symbol blocks DAN SPC
Estimated symbol blocks
Dat
a de
mod
.
STT
D
deco
ding
with
FD
E
・・・
-CP#0
#Ndan-1
#1
+
s1(t)^ s0(t)
^
FFT
FFT
S.M. Alamouti, “A simple transmit diversity technique for wireless communications,” IEEE Journal on Selected Areas in Communications, Vol. 16, No. 8, pp. 1451–1458, October 1998.
Up and Downlink Comparison The same BER
performancefor up anddownlinks
2012/4/12 FA/Tohoku University 317
0.001
0.01
0.1
1
1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01
Prob
[BER
>abs
ciss
a]
BER
FD-STTD(Uplink , MMSE weight)FD-STBC-JTRD(Downlink, MMSE weight)
Ndan =1
23
4
10-1
10-2
10-3
Transmit Es/N0 = 5dB=3.5, =7.0, L=16Nc=256, Ng=32Nmt=2 *
Throughput DAN can achieve much higher throughput
throughput at the cell edge is about 2.6 (bps/Hz) in DAN while it isabout 1.2 (bps/Hz) in CN.
Type II S-P2 (Incremental Redundancy) using rate-1/3 turbo coding.
2012/4/12 FA/Tohoku University 318
Spatial distribution of throughput when (Ndan, Nmt) = (4, 2) for DAN using 7 distributed antennas, 16QAM, and Es/N0=0 (dB)
Thro
ughp
ut (
bps/
Hz)
16QAMEs/N0=0 (dB)(Ndan,Nmt)=(4, 2)
MMSE weight
Increasing Ndanachieves higherthroughput due toits larger spatialdiversity gain. The use of more
than Ndan=4provides only amarginal increasein the throughput.
By increasing Ndanfrom 1 to 4, thetransmit power canbe reduced byabout 6dB.
2012/4/12 FA/Tohoku University 319
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
-10 -5 0 5 10 15 20Normalized transmit Es/N0 (dB)
DAN=3.5, =7.0, L=16
16QAM, Nc=256, Ng=32Nmt=2
Ndan= 12
4, 7
1%-o
utag
e th
roug
hput
(bps
/Hz)
◇ Ndan = 1□ 2△ 4× 7
Cooperative AF Relay Cooperative relay can
extend the coverage A simple protocol is 2-slot
amplify and forward (AF) .
2012/4/12 FA/Tohoku University 320
Cooperate AF relay
SPCMT
RSi
rMi riB
rMB
1.0
MTRS,SPC RSSPC
1st slot 2nd slot
time
Frame
Cooperative 2-slot AF relay does not always improve the throughput. In a good channel
condition, the direct communication provides higher throughput.
Switching betweendirect and relaying(D/AR switching)provides always higherthroughput. Cooperative AF relay is
used only if it provides larger channel capacity than the direct communication.
2012/4/12 FA/Tohoku University 324
Capacity improvement by D/AR switching
0
2
4
6
8
10
12
14
16
-10 0 10 20 30
Out
age
capa
city
(bps
/Hz)
(dB)
U=1, K=6Nc=128, M=64L=16=7.0(dB), =3.5
■ D/AR switching○ Cooperative AF relaying△ Direct communication
t
50% outage
1% outage
Cooperative AF Relay To better exploit the frequency selectivity of the channel,
spectrum division/adaptive subcarrier allocation (SDASA) isproposed. Each user’s M frequency components after DFT are divided into D
sub-blocks of of M/D components each. They are mapped over Nc/(M/D) resource blocks (Nc subcarriers)
based on the channel conditions of MT-RS-BS and MT-BS.
2012/4/12 FA/Tohoku University 325
An example of SDASA when (M, D, Nc) = (8, 4, 16)
M. Nakada, K. Takeda, and F. Adachi, “Channel capacity of SC-FDMA cooperative AF relay using spectrum division & adaptive subcarrier allocation,” Proc. IC-NIDC 2010, pp.579-583, Sept. 2010.
Frequency
Channel gain
Nc =16
Sub-block
① ② ③ ④ ⑤ ⑥ ⑦ ⑧
Resource block
k
Cooperative AF Relay The additional use of
SDASA obtains the frequency diversity gain.
It can further reduce the required transmit power compared to the D/AR switching without SDASA.
2012/4/12 FA/Tohoku University 326
Additional capacity improvement by SDASA
0
2
4
6
8
10
12
14
16
-10 0 10 20 30
Out
age
capa
city
(bps
/Hz)
(dB)
◆ With SDASA (D=4)■ Without SDASA
U=1, K=6Nc=128, M=64L=16s=7.0(dB), a=3.5D/AR switching
1%
50%
t
327
Cooperative Beamforming Multiple distributed antennas are used to reduce
the other user interference using the same carrierfrequency.
Omuniantenna
Directional beam
Distributed antenna(a group of antennas)
DANSignal Processing
Center
2012/4/12 327FA/Tohoku University
Cooperative Beamforming Multiple distributed antennas are used to reduce
the other user interference using the same carrierfrequency.
Omuniantenna
Directional beam
Distributed antenna(a group of antennas)
DANSignal Processing
Center
2012/4/12 328FA/Tohoku University
Cooperative Beamforming SC-FDAAA acts as an adaptive beamformer as
well as one-tap FDE based on the MMSE criterion Beamforming can suppress MAI while FDE
suppresses ISI.
2012/4/12 FA/Tohoku University 329
-GI
Data m
od.
+GI
-GI
Data D
emod.
Data m
od.
+GI
SPCUser#0
User #U-1
*W. Peng and F. Adachi, “Frequency Domain Adaptive Antenna Array Algorithm for Uplink Transmission,” Proc. The 20th Personal, Indoor and Mobile Radio Communications Symposium 2009 (PIMRC’09), Tokyo, Japan, 13-16 Sept., 2009.
Cooperative Beamforming Using a group of Nr receive antennas, up to Nr
users can be accommodated with satisfactoryperformance even in a strong frequency-selectivechannel.
2012/4/12 FA/Tohoku University 330
0 2 4 6 8 10 12 14 16 18 20
10-4
10-3
10-2
10-1
100
Average received Eb/N0(dB)
Ave
rage
BE
R
U=12
3
4
5
L=16, Nr=4
*W. Peng and F. Adachi, “Frequency Domain Adaptive Antenna Array Algorithm for Uplink Transmission,” Proc. The 20th Personal, Indoor and Mobile Radio Communications Symposium 2009 (PIMRC’09), Tokyo, Japan, 13-16 Sept., 2009.
*W. Peng and F. Adachi, “Single-carrier Frequency Domain Adaptive Antenna Array for Cellular Systems,” Proc. 2010 IEEE 72nd Vehicular Technology Conference (VTC2010-Fall), Ottawa, Canada, 6-9 Sept. 2010.
SC-FDAAA
Two DAN SC-FDAAA schemes The distributed SC-FDAAA
generates the SC-FDAAA weightfor each active cluster ofantennas respectively and thencombines the post SC-FDAAAsignals.
On the other hand, the unifiedSC-FDAAA generates the AAAweight for all active clusters ofantennas.
The unified SC-FDAAA schemeachieves better performancethan the distributed SC-FDAAAscheme.
2012/4/12 FA/Tohoku University 331
Distributed antenna clustersfor SC-FDAAA
Cooperative Beamforming Frequency-domain cooperative beamforming allows the
single frequency-reuse (FRF=1). U=2 case.
2012/4/12 FA/Tohoku University 332
0 2 4 6 8 10 12 14 16 18 20
10-6
10-5
10-4
10-3
10-2
10-1
Transmit SNR(dB)
Ave
rage
BE
R
FRF=1FRF=3FRF=4FRF=7
Uplink TransmissionTwo active clusters of four antennas(D)=2
Unified SC-FDAAA
Distributed SC-FDAAAMulti-cell
F6F1
F2
F3F5F4
F7
FRF=7
F1F1
F1
F1F1F1
F1
FRF=1
U=2 case
Cooperative Multiplexing Distributed antennas can be used for spatial
multiplexing to significantly increase the datarate.
2012/4/12 333FA/Tohoku University
Distributed antenna
DANSPC
Cooperative Multiplexing TS-SC MIMO multiplexing using QRM-MLBD
2012/4/12 FA/Tohoku University 334
Nc+Ng-pointDFT
QRD of equivalent channel matrix
…
Nda
ndi
strib
uted
ant
enna
s
Nc+Ng-point DFT
Y1
YNdan
…
… Mul
tiplic
atio
n of
QH
MLD
usi
ng M
-al
gorit
hmY
De-
mod
ulat
ion
SPC
Info. bits
Dat
am
odul
atio
n +TSd
S/P
…
+TS
d1
dNmt
s1
sNmt
…
Nmt antennas
Mobile terminal
T. Yamamoto, K. Takeda, and F. Adachi, “Training sequence-aided QRM-MLD block signal detection for single-carrier MIMO spatial multiplexing,” Proc. IEEE International Conference on Communications (ICC 2011), Kyoto, Japan, 5-9 June, 2011.
Cooperative Multiplexing Uplink transmission using Nmt transmit antennas
(MT) and Ndan distributed receive antennas (SPC) Antenna distribution model: Nmt=Ndan=2
2012/4/12 FA/Tohoku University 335
1.0
2/3
R1
R2 1.0R
DAN Conventional cellular system
Cooperative Multiplexing Received signal representation at mth antenna
2012/4/12 FA/Tohoku University 336
symbols of vector TS:)]1(),...,0([
antennat th transmi ofblock symbol data : )]1(),...,(),...,0([
with
][
vectornoisedomain -frequency:antenna ddistribute )~1(th and antenna transmit MT
)~1(th between matrix channeldomain -frequency : size ofmatrix DFT :
where
2
antenna ddistribute)~1(th at the vector signal receiveddomain -frequency The
,
)(
1
)(,
gT
gnnn
Tcnnnn
TTn
Tnn
m
dan
mtnm
JJ
N
nmn
NNnm
s
sm
dan
NNuu
nNdtdd
NmmNnn
J×J
TE
Nmm
mtgc
u
d
uds
N
HF
NsFHY
Cooperative Multiplexing Overall received signal representation
2012/4/12 FA/Tohoku University 337
1×)+( size of vector noise overall:
)+(×)+( size ofmatrix channel equivalent:
where
2
2
1×)+( size of signal receiveddomain -frequency Overall
1
11
,1,
,11,11
gcdan
gcmtgcdan
Ns
s
NNNNN
N
s
s
N
gcdan
NNNNNNNNN
TE
TE
NNN
mt
danmtmtdandan
mt
dan
NH
Ns
sH
N
N
s
s
FHFH
FHFH
Y
YY
Cooperative Multiplexing QRM-MLBD consists of three steps
OrderingQR decompositionMLD using M-algorithm.
Ordering
2012/4/12 FA/Tohoku University 338
1× size of symbolth at vector TS: )(
1× size of symbolth at vector data:)(
where
)]1()()0(),1()()0([
vector) theof bottom at the symbols training theall relocate (to orderingafter vector symbol transmit The
danT
danT
Tg
TTTc
TTTorder
Ntt
Ntt
NtNt
u
d
uuuddds
Cooperative Multiplexing QR decomposition
2012/4/12 FA/Tohoku University 339
)+(×)+( size ofmatrix ngular upper tria :
)+(×)+( size ofmatrix unitary :with where
2ˆ
signal receiveddomain -frequency ed transformThe
gcmtgcmt
gcmtgcdan
Horder
s
sH
NNNNNNNNNNNN
TE
RQ
QR=H
NQRsYQY
Cooperative Multiplexing MLD using M algorithm
2012/4/12 FA/Tohoku University 340
stagelast at the metricpath smallest thehavingpath back the by tracingon demodulati Data (c)
metricspath thecomparingby paths surviving as selected are paths best thestage,each In (b)
diagram treein the distanceEuclidean minimum thehavingpath best for the Searching (a)
M
101010101010
10
10
10
CN
DAN
MMSED
QRM-MLBD
M=1M=4M=16
NcNg=16Transmit Es/N0=10dB16-path uniform
Nmt=Ndan=2QPSK
BER
Prob
[BER
>abs
ciss
a]
Cooperative Multiplexing DAN can significantly
improve thetransmission qualitycompared to the caseof co-located antennasat BS (conventionalcellular system).M is the number of
surviving candidate symbols in the Malgorithm of QRM-MLD.
As M increases, the achievable link performance improves but, the complexity for the signal detection reduces.
2012/4/12 FA/Tohoku University 341
DAN achieves higher throughput than CN The cell edge throughput is about 3bps/Hz higher with DAN
than with CN. CN can achieve high throughput only near SPC.
2012/4/12 FA/Tohoku University 342
Thro
ughp
ut (b
ps/H
z)
2.0
3.0
4.0
5.0
6.0
7.0
Es/N0=5dB(b) CN
(a) DAN
Spatial distribution of throughput
02.0
3.0
4.0
16QAMNdan=Nmt=2NcNg=16=3.5, =7dB16-path uniform
Normalized transmit Es/N0 (dB)
10%
out
age t
hrou
ghpu
t (bp
s/Hz)
5 10 15 20 25
5.0
6.0
7.0
DAN
CN
QRM-MLBD(M=16)MMSED
DAN can reduce thenormalized transmitEs/N0 required forachieving the samethroughput as CN. Es/N0 reduction from CN is
as much as about 10dBfor a 10%-outagethroughput of 5bps/Hz.
QRM-MLBD can improvethe throughputcompared to the MMSEdetection. Es/N0 reduction from
MMSED is as much asabout 9dB in DAN for a10%-outage throughput of5bps/Hz.
2012/4/12 FA/Tohoku University 343
Some Concluding Remarks Next generation (4G) and future (5G) wireless networks will
require Giga-bit wireless technology of >1Gbps and>80bps/Hz/BS under severe MAI and co-channelinterference.
Wireless Signal Processing Frequency-domain equalization & MIMO may be indispensible
techniques. But, still insufficient! Reducing transmit power
Distributed antenna network (DAN) with multiplexing, diversityor relay can solve the transmit power problem while increasingthe spectrum efficiency.
Other promising solutions? Lots of interesting and important research topics remain
before the born of next generation wireless systems.
2012/4/12 FA/Tohoku University 364
Please Visit Our Homepage for More Info.
http: //www.mobile.ecei.tohoku.ac.jp
2012/4/12 FA/Tohoku University 365