Module-3 : Transmission Lecture-10 (18/5/00)
Marc Moonen
Dept. E.E./ESAT, K.U.Leuven
marc.moonen@esat.kuleuven.ac.be
www.esat.kuleuven.ac.be/sista/~moonen/
Lecture 10 : Smart Antennas -Overview
• Introduction:
Smart Antennas
SDMA (`driver application’)
• SDMA v1.0
Line-of-sight propagation & beamforming DOA estimation and signal reconstruction
• SDMA v2.0
Multi-path propagation
MIMO channel modeling & source separation
• Related Topics
CDMA multi-user detection (see Lecture-9)
MIMO transmission (see Lecture-2)
Introduction : Smart Antennas
• Antenna arrays (hardware) with (software) `beam-forming’
(`beam-steering’), or similar (in multi-path scenario, see below).
• `Antenna diversity’
Introduction : Smart Antennas
• Aim : increase signal-to-interference-and-noise ratio, hence improved performance/increased capacity (e.g. in CDMA systems)
• Antenna arrays mostly considered for base station systems, not (often) for mobile terminals.
• Currently simple systems with switching between antenna signals (=select best signal), fixed directional antennas for sectorization (e.g. GSM), ...
• More advance systems considered for WLANs, for W-CDMA, etc...
• Will consider SDMA as `driver application’
Introduction : SDMA
• `Conventional’ wireless communications (`SISO’, TDMA/FDMA/CDMA)
• What we have in mind is ….
(MIMO transmission, SDMA)
transmitter radio channel
x bits/sec/Hz/km2 receiver
transmitter receiver
transmitter
radio channel 2x bits/sec/Hz/km2
receiver
Introduction : SDMA
• Example : cellular mobile telephony (e.g. GSM)
• Basic network architecture : -country covered by a grid of cells -each cell has a base station
-base station connected to land telephone network and
communicates with mobiles via a radio interface
Introduction : SDMA
• Why cellular ?
Capacity increase by spectrum reuse , pico-cells, etc.
• Capacity increase by multiplexing :
- GSM (900MHz) has 125 frequency channels/cell (FDMA) 8 time slots/channel (TDMA)
In practice, capacity per cell << 8*125 !
- Spatial multiplexing : allows different users in 1 cell to use the same freq./time slot
4
1
P d
Introduction : SDMA
• PS: in GSM neighboring cells cannot use same
frequency bands (intercell interference). Same
frequency band used in each 7th cell.
Introduction : SDMA
• SDMA (`spatial division multiple access’) allows different users in the same cell to use the same frequency
channel/time slot/code, and thereby offers substantial capacity increases when superimposed on a current system!
• SDMA supports multiple directional connections on a
single conventional radio channel through the usage of
antenna arrays and advanced signal processing.Introduction : SDMA
PS: SDMA ~ `dynamic sectorization’
WARNING:
• Major practical impediment is computational complexity (cfr.
linear algebra-type operations at high sampling rates). … Gflops requirement….
• Major challenge for VLSI/ASIC design
• First products probably in WLAN-type applications instead of cellular/mobile
AIM:
• Illustrate (near) future system design concepts….
SDMA v1.0: Beamforming Approach
• Assumptions:
- sources are in the far-field
- line-of-sight (LOS) connections - no multi-path effects
- homogeneous medium/ideal channel characteristics - additive white Gaussian noise (AWGN)
- no inter-symbol interference
SDMA v1.0: Beamforming Approach
• Beamforming (`spatial filtering’):
PS: compare with regular temporal (FIR) filtering
SDMA v1.0: Beamforming Approach
• Data Model:
= *
antenna outputs sources
array gain matrix
time samples for antenna-1 time samples for antenna-2
time samples for source-1
steering vector source-1 steering vector source-2
SDMA v1.0: Beamforming Approach
• Data Model:
`Steering vector’ a(theta)
= array response vector, contains gains and phase shifts for a narrow-band wavefront impinging from direction-of- arrival (DOA) theta (and for a certain carrier frequency)
The collection of `steering vectors’ for all possible angles theta, is referred to as the `array manifold’
Knowledge of `array manifold’ is crucial is beamforming approach
SDMA v1.0: Beamforming Approach
• Array manifold example:
Uniform Linear Array
where f = phase shift =
d angle
)) sin(
. . . .2
exp( c
d
j fc
1
2...
1
)
( f f f
Ma
SDMA v1.0: Beamforming Approach
• Significance of array manifold:
-array manifold is a parametrization of the steering vector as a function of the DOA
-if array manifold is known (by calibration or
physical modeling), `channel modeling’ is reduced to DOA estimation. If the DOA for one particular source is identified, its complete steering vector is known.
SDMA v1.0: Beamforming Approach
• Problem Statement:
Given antenna outputs & array manifold, compute :
-directions-of-arrival (DOA’s) -source signals
= *
antenna outputs sources array gain matrix
?
SDMA v1.0: Beamforming Approach
• Solution (Part-1): DOA estimation
`low-resolution algorithms’ : Fourier-based (e.g. for ULA’s)
`high-resolution algorithms’ :
-MUSIC [Schmidt 1979]: search for DOA such that steering vector optimally matches `column space’ of antenna output matrix
-ESPRIT [Roy et al, 1987]: DOA’s identified as generalized eigenvalues of a matrix `pencil’
= *
antenna outputs sources
array gain matrix
SDMA v1.0: Beamforming Approach
• Solution (Part-2):
Beamforming and signal reconstruction
Given steering vectors of signal-of-interest and interferers, compute
beamformer weights
such that interference
signals are eliminated
(`null steering’)
SDMA v1.0: Beamforming Approach
• Solution (Part-2):
Beamforming and signal reconstruction
compute weight vector w1, w2,…. such that….
= *
antenna outputs sources array gain matrix
* *
w1 w2 :
w1 w2 :
1 0 : 0
SDMA v1.0: Beamforming Approach
• Solution (Part-2):
- Compute weight vector w1, w2,…. that cancels all
interferers, and retains the signal of interest (cfr. supra) - This is `zero-forcing’ solution. With additive noise, a minimum-mean-squared-error solution is preferred.
- Other : Generalized sidelobe canceller, minimum variance distortionless response beamforming, Griffiths-Jim
beamforming : adaptive beamformers, based on knowledge of steering vector of (only) the signal-of-
interest, and where noise environment (incl. interferers)
SDMA v1.0: Beamforming Approach
• Beamforming approach deficiencies :
- not always line-of-sight (LOS) connection - multi-path effects
long/short term fading (e.g. wavelength=30cm @ 900MHz) - inter-symbol-interference
(e.g. symbol ~ 1km @ 270kbits/sec)
• Conclusion:
- array manifold concept no longer useful
- need more sophisticated data models/algorithms
SDMA v2.0: Channel Modeling Approach
• Instead of this…..
• we have this…..
SDMA v2.0: Channel Modeling Approach
• SDMA with multi-path corresponds to multi-user (multiple- input/multiple output) channel equalization problem :
a) identify channel model
b) reconstruct channel inputs from outputs+model
single-user (e.g. GSM) multi-user (SDMA)
SDMA v2.0: Channel Modeling Approach
• Step-1 is a channel modeling, i.e. identify...
• Training sequence based versus `blind algorithms’
(see Lecture 5-6)
L - user signal
:
2 - user signal
1 - user signal
*
CHANNEL
) ( ..
) ( )
(
: :
:
) ( ..
) ( )
(
) ( ..
) ( )
(
M - antenna output
:
2 - antenna output
1 - antenna output
2 1
2 22
21
1 12
11
H z H z H z
z H
z H
z H
z H
z H z
H
ML M
M
L L
SDMA v2.0: Channel Modeling Approach
• Step-2 is equalizer design, i.e. identify…
• Zero-forcing (ISI=MUI=0) versus MMSE (see Lecture 5-6)
• This is combined equalization & source separation
L - user signal
:
2 - user signal
1 - user signal
*
0 ..
0 1
) ( ..
) ( )
(
: :
:
) ( ..
) ( )
(
) ( ..
) ( )
(
* ) ( M
- antenna output
:
2 - antenna output
1 - antenna output
* ) (
) ( ..
) ( )
( )
(
2 1
2 22
21
1 12
11 2
1
H z H z H z
z H z
H z
H
z H z
H z
H z
C z
C
z C z
C z
C z
C
ML M
M
L L M
SDMA v2.0: Channel Modeling Approach
• Step-1 & -2 may be combined : direct (training sequence based) equalizer design (see Lecture-5/6). Only training sequence for user-of-interest needed (not for other users).
• Recursive vs batch processing (Lecture-5/6)
• `Oversampling’ (i.e. having more outputs (antennas) than inputs (users)) is crucial for the existence of zero-forcing solutions (for FIR channels).
• Connections with fractionally spaced equalization theory and filter bank theory.
• Active area of research (blind algorithms based on 2nd
order statistics, finite alphabet properties, etc.).
Related Topics
CDMA multi-user detection algorithms (Lecture-9) MUD algorithms are conceptually similar :
-Spreading viewed as a (transmit) filtering operation and part of the `channel’.
-Nyquist-rate sampling at the receiver is symbol- rate oversampling, which is equivalent to spatial oversampling (multiple antennas).
-etc...
Related Topics
MIMO Transmission (Lecture-2)
- Point-to-point transmission, where both transmitter and sender have antenna array
- additional flexibility for sender (beamforming, …) - with M antennas at both ends, allows for M-fold
channel capacity increase with the same transmit power budget (!)
- example : V-BLAST
Conclusions
• Smart Antennas
- Advantages : improved signal-to-interference-and-noise ratio, increased capacity (CDMA).
- Considered for W-CDMA, ...
• SDMA v1.0
- Beamforming approach
- Conceptually simple, but not applicable in multi-path environment
• SDMA v2.0
- Multi-path/MIMO channel modeling approach - Powerful but complex
• Related Topics
- CDMA multi-user detection, MIMO transmission
Assignment 5.1
`Brain Teaser’ :
• In Lecture-2, we have considered MIMO-transmission from a channel capacity point of view. Look at the conclusions again. One of the conclusions was that `one has to be lucky with the channel characteristic’.
• Think of a similar channel capacity analysis for SDMA.
Does one again have to be `lucky with the channel’ ?
• What would be a most advantageous channel, in terms of
channel capacity ? What would be the obtained channel
capacity ? Is it `what we had in mind’ (cfr slide-5) ?
Assignment 5.2
`Brain Teaser’:
• In Lecture 7-8 we have considered multi-tone transmission, where a (high-rate) bit stream is split up into (low-rate)
parallel bit streams, which are then used to QAM modulate different carriers.
• Now consider these low-rate streams as being different users, accessing the same transmission channel. The carrier modulation may be viewed/compared with a spreading operation a la DS-CDMA.
• Based on this, compare DMT with CDMA and MIMO, both from a capacity and a receiver structure point of view.
Look for similarities and differences.
• In a similar fashion, compare DMT with MIMO transmission