Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Outline
Comunicazioni Wireless Ubique: Tecnologie Esistenti e Future
Introduction to Wireless Communication Systems.
CISM - Centro Internazionale di Scienze Meccaniche
Udine - May 20, 2004 Part I
Wireless Communications
Satellite Systems
Terrestrial Radio
Systems Free space
Optical Comm.
Introduction
Some reasons for success
– Wireless connection
– Simple and cheap deployment enjoy ubiquitous communications !
– Coverage
– Mobility
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 3 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 4
Research and Technology Drivers Existing and Emerging Wireless Technologies
Existing Technologies
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 5 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 6
Signal Bandwidth < Channel Coherence Band Signal Bandwidth > Channel Coherence Band
Symbol duration ~ Channel Coherence Time Symbol duration < Channel Coherence Time
h(t;τ) +
Time-Variant Flat Fading Frequency Selective Fading
Propagation Phenomena η(t)
NP
Attenuation (path-loss) h(t ;τ ) = α (t )δ (τ ) h(t ;τ ) = ∑ α p (t )δ (τ − τ p )
p =0
Slow Fading (shadow fading) α1
α2
α3
Slow variations of the received power caused by obstructions. α4
xk = x(kT ) = α (kT )ak + ηk τ1 τ2 τ3 τ4
Fast Fading
NP
Fast variations of the received power caused by multipath propagation
xk = x( kT ) = ∑ α p ak − p + η k
in correspondence to movement in the order of the wavelength. p =0
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 7 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 8
Multiple Antenna Systems
MS Smart Antennas
MS
RX
Generate beams with phased arrays to sectorize
coverage.
MS
MS
PART I
Adaptive Antennas
Multiple Antenna Systems MS RX
Receive antenna combining to gain spatial diversity
and cancel co-channel interference.
MS
MS
Space-Time Coding
TX RX
Multiple transmit and receive antennas to increase
capacity.
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 9 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 10
1 1 ⎡ y1 ⎤ ⎡ α 1,1 ... α 1, NT ⎤ ⎡ x1 ⎤ ⎡ n1 ⎤
⎢ ⎥ Es ⎢ ⎥⎢ ⎥ ⎢ ⎥ Es
⎢ ... ⎥ = ... ... ... ⎥ ⎢ ... ⎥ + ⎢ ... ⎥ y= Hx +n
TX RX NT ⎢ N R ,1 NT
⎢y R ⎥
N ⎢α α R T ⎥⎦ ⎢⎣ x NT
N , N ⎥ ⎢n R ⎥
N
NT NR ⎣ ⎦ ⎣ ⎦ ⎣ ⎦
NT
∑α r ,t
x t
+ n r
r = 1,..., N R ⎛ E / N0 ⎞
t =1 CH = log 2 det ⎜ I + S HH † ⎟ bit / s / Hz
AWGN, m=0, σ2=N0
⎝ NT ⎠
transmitted complex signal by antenna t
We assume H to have independent complex Gaussian entries (Rayleigh fading)
channel weight link antenna (t-r)
The Outage Capacity is the distribution of CH
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 11 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 12
Mean Capacity Space-Time Coding
240
30 dB
xk1 yk1
200
25 dB
xk2 yk2
160 bi bi
Capacity (bit/s/Hz)
Space-Time
20 dB ST Encoder S/P
Decoder
120 SNR=15 dB
xkNT ykNR
80 10 dB
5 dB
40
0 dB
To approach the Shannon Capacity we need to design powerful space-time codes:
C < 9.16 bit/s/Hz
0
with NT=NR=1 2 4 6 8 10 12 14 16 18 20 22 24
joint channel coding, modulation, with transmission over multiple antennas.
Number of Antennas (NT=NR )
Ergodic Capacity is used to characterize fast fading channels. Fundamental contribution by Tarokh, Seshadri, and Calderbank (1998 AT&T Labs)
Outage Capacity is used to characterize quasi-static fading channels.
Fundamental contribution by Foschini (1996 Bell Labs):
Capacity increases linearly with the number of TX antennas if NR≥ NT.
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 13 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 14
Diversity Gains and Coding Gains are determined by the rank and determinant of certain
diNT xkNT
matrices constructed from complex codewords. Recall that the transmitted signals overlap, Bit-Symbol
Mapper
therefore, the ST code must have a structure that allows to separate the signals at the receiver.
ST-BICM comprises
L
⎛ 1 ⎞ ⎛ 1 ⎞
L
~ (χ2 )
−L
decreasing
Prob. Error
decreasing
Prob. Error
~ ( SNR )
−L
Flexible approach.
Full diversity codes can be designed for both quasi-static and time-variant fading
channels.
SNR SNR
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 15 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 16
Turbo MIMO Decoding Example of Application to GSM/EDGE Air Interface
yk1 λe(di1)
1 Bit/s/Hz 2 Bit/s/Hz 3 Bit/s/Hz
λa(ci1)
Bit-DeInterlever
yk2 λe(di2)
Joint Soft-In bi
Soft-In Soft-Out
Soft-Out P/S
Decoder
Detection
λe(ci1)
ykNR λe(diNT) Bit-Interlever 1 TX
it=1
1 TX
λa(diNT) λa(di1)
1 TX it=4
260 bits
S/P
it=4
NT
2 TX 1 TX
ykr = y r (kT ) = ∑∑ x t (nT )g CH
t ,r
(kT − nT ) + η (kT ) it=4 it=4
t =1 n
1 TX
it=4
The receiver has to separate the overlapping signals and recover the information bits.
Iterative (turbo) decoding procedure:
MIMO Demapping at the Detector: A Posteriori Probability Calculator for Each Coded Bit.
Single receive antenna – TU channel model – 4/8 PSK with STBI convolutional coding.
Maximum a Posteriori Channel Decoder: Improved Extrinsic Information for the Coded Bits.
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 17 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 18
Remarks
It is fundamental to
PART II
Study and model the MIMO channel
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 19 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 20
Multicarrier Transmission Principles General Multicarrier Architecture
f1 − f1
a1 (lT0 ) y1 ( nT0 ) aˆ1 (lT0 )
Motivation h(.) x x g(.) Equalizer
η (t )
– Simplify the equalization task in wide band frequency selective channels.
a k (lT0 ) RF RF
y k (nT0 ) aˆ k (lT0 )
Principle h(.) x + Channel + x g(.) Equalizer
Mod Dem
fk − fk
– Divide the spectrum in a number of narrow band sub-channels (flat faded).
y M ( nT0 )
a M (lT0 ) aˆ M (lT0 )
h(.) x x g(.) Equalizer
– Allocate transmission power over the good channels (water filling principle).
QAM / PSK symbols fM − fM
|H(f)|
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 21 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 22
DMT - OFDM
Sub-Ch. Data Period: Transmitter
T = MT
0
Tone Spacing: M points IDFT
∆ f = 1 / MT
CS-FMT
Sub-channel symbol period T0=NT=(M+ µ)T
Sub-Ch. Data Period:
T = MT M 2π
( n − µ )( k −1)
x( nT + lT0 ) = ∑ a k (lT0 )e
0
j
n = 0,..., N − 1
Tone Spacing: M
∆f = 1 / MT
k =1
Receiver
NCS-FMT
Sub-Ch. Data Period: Disregard cyclic prefix
T = NT
0
Tone Spacing:
∆f=1 / MT
M points DFT
N/M =K>1
One tap equalizer
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 23 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 24
One Tap Equalization for CP-DMT FMT: Digital Implementation
a1 (lT0 ) z1 (lT0 )
Hypothesis g1(mT0) RF Channel RF g1(-mT0)
NP ≤µ
M
• Static over a DMT symbol M
a (lT0 )
gM(mT0) gM(-mT0)
z (lT0 ) ICI~0
M M M M
x( nT ) y (nT )
Thesis
• The DFT output equals the data symbol weighted by the channel frequency response.
CS-FMT
• The receiver simplifies into a simple one-tap equalizer.
k −1
fk = T0 = MT k = 1,..., M
T0
⎛ NP −j
2π
p ( k −1) ⎞
z k (lT0 ) = DFT {y (lT0 )}( k ) = DFT {A (lT0 )}( k ) DFT {α}( k ) = a k (lT0 ) ⎜ ∑ α p e M ⎟ g k ( mT0 ) = h((k − 1)T + mT0 )
⎝ p =0 ⎠
Prototype pulse
DFT received block Transmitted block is cyclically
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 25 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 26
The sub-channel equalizer has low complexity since the sub-channel impulse
l-th output sample of the k-th RX filter response is short (sub-channel is narrow band).
z k (lT0 ) = a k (lT0 ) g EQ
k
(0) + ∑ a k (lT0 − mT0 ) g EQ
k
(mT0 ) + η (lT0 ) Practical issues (real world !!)
m≠0
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 27 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 28
Example of Sub-Channel Frequency Response Probability [ Achievable Bit Rate > K ]
DMT
0
CS-FMT
-10 1
0
|H(f)| (dB)
-50
-40
0.7
-60
-50 0.6
-70
-60
-80 0.5
0.5 0.505 0.51 0.515 0.52 0.525 0.53 -70
fT
-80 0.4
0.5 0.505 0.51 0.515 0.52 0.525 0.53
fT NCS-FMT
0 0.3
|H(f)| (dB)
-40 0
40 45 50 55 60 65 70 75 80 85 90
N/M = 1.125 + 4 virtual carriers -50 M bit/s
-60
carriers -80
0.5 0.505 0.51 0.515 0.52 0.525 0.53 c Rayleigh exponential with τrms=100 ns
fT ------ DMT
------ CS-FMT ¦ Rayleigh exponential with τrms=40 ns
CP-DMT: CP length = 30 chips + 16 virtual carriers ------ NCS-FMT * Ricean exponential with R=5 dB, τrms=40ns
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 29 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 30
Remarks
FMT is more complex than DMT since it requires filtering and equalization.
Ultra Wide Band Communications
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 31 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 32
UWB Main Characteristics Impulse Modulation
0.5
Signal bandwidth > 500 MHz or Bandwidth / Center-frequency ≥ 0.2
s(t)
0
Most popular schemes are based on impulse modulation with short duration pulses Tg
-0.5
Tf
Simple base band (carrier less) implementation. -1
0 1 2 3 4 t
Good penetration properties.
Convey a bit sequence via a sequence of monocycles (short duration pulses)
Good spatial and temporal resolution.
Co-existence with other radio systems. • Bi-Phase PAM modulation or Time-Hopped modulation.
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 33 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 34
T Tg T Tg
cuI ,m = cuI , m = cuI ,m =
slots: NST
−1 1 −1 1 −1 1 −1 1 −1 1 −1 1
frame: Tf
slots: NST
burst: LTf frame: Tf
burst: LTf
TX signal of user u
TX signal of user u
L −1
su (t ) = ∑ bu , k ∑ g (t − cuTH,l T − lT f − kLT f ) ⇒ = ∑ bu , k vuTH (t − kLT f ) L −1 N S −1
k l =0 k su (t ) = ∑ bu , k ∑ cuO,l ∑c I
u,m g (t − mT − lT f − kLT f ) ⇒ su (t ) = ∑ bu , k vuDS (t − kLT f )
k l =0 m=0 k
signature waveform
± 1 bit sequence monocycle signature waveform
± 1 bit sequence monocycle
hopping codeword of user u and length L
inner codeword user u and length Ns
|G(f)|2 (dB)
|G(f)|2 (dB)
g(t)
g(t)
0 0 -20 -20
-1 -1 -40 -40
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
|G(f)|2 (dB)
|G(f)|2 (dB)
g(t)
g(t)
0 0 -20 -20
We can use time-frequency concentrated pulses as the family of: The FCC specifications are very tight:
π
− ( t / T0 )2 Transmission band 3.1 -10.6 GHz with Spectral Density of -41 dBm/MHz
Derivatives of Gaussian monocycle g (t ) ~ e 2
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 37 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 38
TX y(t) Z0(k) ∞
NU-1 g0MF(t) LTf Z 0 (k ) = ∫ y (t ) g 0MF ( kLT f + ∆t − t ) dt
−∞
RX 0
TX 2 TX 3
NU −1 1
Z 0 (k ) = b0, k ∫ ∑ ∑b ∫
LT f LT f
| v0EQ (t ) |2 dt + u ,k − n vuEQ (t + nLT f + ∆t0 − ∆tu )v0EQ (t )dt + w(k )
0 0
Convolution of the u-th user’s signature waveform with the u-th user channel u =1 n = 0
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 39 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 40
Synthesis of Matched Filter and Rake Reception Cont. ed
UWB channel is highly frequency selective.
We need to estimate the channel tap delays and amplitudes. It can be done with a
We assume to synthesize the channel with a finite number of taps (FIR filter). training approach.
α1
α3 We can implement a mixed Analog-Digital Rake Receiver
α5
τ2 τ4
NP
g uCH (t ) = ∑ α p ,u δ (t − τ p ,u ) α1
τ1 τ 3 τ5 nTf+τ1
α6 p =1 x
α4 α2
α2 nTf+τ2 x
y(t) Z0(k)
The received signature waveform can be synthesized as follows g(-t) +
L −1 N P
αN
vˆ0EQ ,TH (t ) = ∑∑ αˆ 0, p g (t − c0,THl T − lT f − τˆ0, p ) nTf+τN x
l = 0 p =1
L −1 N S −1 NP
vˆ0EQ , DS (t ) = ∑ c0,Ol ∑ c ∑ αˆ I
0, m 0, p g (t − mT − lT f − τˆ0, p ) No Spreading / TH
l =0 m=0 p =1
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 41 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 42
Channel: Exponential delay profile, 10 Rayleigh faded rays (with sign) and with uniform delay distribution in [0 4D].
te
ma
ma
-1 -1 -1 -1
10 10 10 10
sti
sti
lE
lE
ne
ne
A
an
an
DM
Nu=8
Ch
Ch
-C
Nu=8
DS
al
TH
al
Ide
Ide
-2 -2 Nu=8 -2 -2
10 10 10 10
Nu=2
BER
BER
BER
BER
Nu=2
Nu=8
-3 -3 Nu=2 -3 -3
10 10 10 10
Nu=2
Nu=1 Nu=1
Nu=1 Nu=1
-4 -4 -4 -4
10 10 10 10
0 3 6 9 12 15 18 0 3 6 9 12 15 18 0 3 6 9 12 15 18 0 3 6 9 12 15 18
SNR (dB) SNR (dB) SNR (dB) SNR (dB)
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 43 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 44
Cont. ed Remarks
With practical channel estimation we get a tolerable performance loss.
UWB systems allow very high transmission speed (say in the order of Gb/s)
DS-CDMA -- Ns =8 -- L=1 DS-CDMA -- Ns=8 --L=8 Potentially, their implementation is simple, however,
te
ma
sti -1 -1
lE
10 10
ne
Nu=8
A
Ch
DM
al
-C
tic
DS
-2 -2
10 Nu=2 10 Nu=8
BER
BER
Nu=2
Nu=1
-3 -3
10 10
Nu=1
Nu=1
-4 -4
10 10
0 3 6 9 12 15 18 0 3 6 9 12 15 18
SNR (dB) SNR (dB)
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 45 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 46
Conclusions References
Space-Time Coding
Reliable spectral efficient transmission over wireless channels is an Algorithmic
Foschini, Bell Labs Tech. Journal 1996
Challenge.
Tarokh, Seshadri, Calderbank, IEEE Trans. IT, 3-1998 & 3-1999.
The state of the art is represented by
Tonello, IEEE Trans. Comm, 2-2003.
Multiple Antenna Technology
OFDM – FMT Systems
Multi-carrier Modulation
Cherubini, Eleftheriou, Olcer, Cioffi, IEEE Comm. Mag. 5-2000.
Ultra Wide Band Transmission.
Assalini, Pupolin, Tomba, Proc. WPMC 2003.
A powerful air interface allows for Tonello, Bell Labs Tech. Journal 2003.
Higher receiver sensitivity (coverage)
UWB Systems
Robustness to co-channel interference (spectral efficiency)
Win, Scholtz, IEEE Comm. Letters, 1-1998.
Lower transmission power
Durisi, Benedetto, Proc. ICC 2003.
Simpler processing complexity.
Tonello, Rinaldo, Bellin, Proc. ISPLC 2004.
DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 47 DIEGM UNIVERSITÀ DEGLI STUDI DI UDINE CISM COURSE 2004 48