Page 1
© Imperial College LondonPage 1
WSEAS PLENARY LECTURE:
The Challenges of Subspace Techniques and
their Impact on Space-Time Communications.
29th December 2004Dr Athanassios ManikasDeputy Head Comms & Signal ProcessingDepartment of Electrical & Electronic Engineering
Page 2
© Imperial College LondonPage 2
Outline
1. Notation
2. General Problem Formulation
3. Subspace Techniques• Signal-Subspace• Manifolds• Performance Bounds
4. Space-Time Communications
5. Conclusions
Page 3
© Imperial College LondonPage 3
Notation
Page 4
© Imperial College LondonPage 4
cont. - Notation
Page 5
© Imperial College LondonPage 5
cont. - Notation
Origin
bs
pL[ ]
L[ ]
Page 6
© Imperial College LondonPage 6
cont. - Notation
Origin
bs
pL[ ]x
L[ ] x
x
Page 7
© Imperial College LondonPage 7
General Problem Formulation
Condition
AWGN
Page 8
© Imperial College LondonPage 8
n(t)
m (t)1
m (t)2
m (t)M
+
cont. - General Problem Formulation
Page 9
© Imperial College LondonPage 9
cont. - General Problem Formulation
Page 10
© Imperial College LondonPage 10
Page 11
© Imperial College LondonPage 11
Subspace Techniques
Page 12
© Imperial College LondonPage 12
cont. - Subspace Techniques
Page 13
© Imperial College LondonPage 13
The “Signal-Subspace” Concept
Page 14
© Imperial College LondonPage 14
cont. - The “Signal-Subspace” Concept
Page 15
© Imperial College LondonPage 15
cont. - The “Signal-Subspace” Concept
Page 16
© Imperial College LondonPage 16
cont. - The “Signal-Subspace” Concept
Page 17
© Imperial College LondonPage 18
cont. – The “Signal-Subspace” Concept
Page 18
© Imperial College LondonPage 19
The “Manifold” Concept
Page 19
© Imperial College LondonPage 20
cont. – The “Manifold” Concept
Page 20
© Imperial College LondonPage 21
cont. – The “Manifold” Concept
Page 21
© Imperial College LondonPage 22
cont. – The “Manifold” Concept
Page 22
© Imperial College LondonPage 23
cont. – The “Manifold” Concept
Page 23
© Imperial College LondonPage 24
cont. Manifold
Page 24
© Imperial College LondonPage 25
cont. – The “Manifold” Concept
Page 25
© Imperial College LondonPage 26
cont. – The “Manifold” Concept
Page 26
© Imperial College LondonPage 27
cont. – The “Manifold” Concept
Page 27
© Imperial College LondonPage 28
cont. – The “Manifold” Concept
Page 28
© Imperial College LondonPage 29
Performance Bounds
Page 29
© Imperial College LondonPage 30
Space-Time Communications
Page 30
© Imperial College LondonPage 31
cont. - Space-Time Communications
Page 31
© Imperial College LondonPage 32
cont. - Space-Time Communications
• All ‘conventional’ CDMA receivers can be modified to become Space-Time CDMA receivers
• Enhancements would result in considerable performance gains
• Enhancements are not trivial however
Page 32
© Imperial College LondonPage 33
cont. - Space-Time Communications
• Two classes1. ST - Single-User Rx: requires no knowledge beyond
the PN-sequence and the timing of the user it wants to demodulate/receive ('desired' user)
2. ST - Multi-User Rx: requires knowledge of the PN-sequence & the timing of every active user as well as knowledge of the received amplitudes of all users and the noise level
> Can be Optimal or Sub-optimal depending on whether the decision making criteria for symbol detection are fully met (Optimal) or partially met (Sub-optimal)
Page 33
© Imperial College LondonPage 34
Optimum ST-CDMA Receivers
• ST - RAKE Receiver: Optimum Single-User Receiver
• ST - MLSE Receiver: Optimum Multi-User Receiver which must take into account the PN-codes of all other CDMA users in the system. Non-linear and computationally far too complex
> Huge gap in performance and complexity between an optimum single-user and an optimum multi-user receiver
> Decorrelating MU Receiver:This is a typical sub-optimal MU Rx (but much simpler that the optimum MU Rx):
Page 34
© Imperial College LondonPage 35
Space-Time Channels
T xT x
T x
V IV OC h a n n el
T xT xS p a c e-T im e
R xM IV O
C h a n n e l
S p a c e-T im eR x
S p a ce -T im eT x
S p a ce -T im eT x
S p a ce -T im eT x
Page 35
© Imperial College LondonPage 36
+ n(t)
SIVO Channel
of 1st Source
m (t)1
SIVO Channel
of 2nd Source
SIVO Channel
of -th SourceM
VIVO Channel
Space-TimeRx
m (t)2
m (t)M
T xT x
T x
V IV OC h a n n el
T xT xS p a c e-T im e
R x
Page 36
© Imperial College LondonPage 37
cont. Space-Time Channels
i1 i2 iK i
S i1 S i2 S iK i
+
S p a c e-T im eR x
i1 i2-i1 i3-i2S IV O
E tc .s (t)i
Page 37
© Imperial College LondonPage 38
cont. Space-Time Channels
Page 38
© Imperial College LondonPage 39
Page 39
© Imperial College LondonPage 40
Page 40
© Imperial College LondonPage 41
Example – ULA of 3 elements
User1 (Desired) Path 1 Path 2 Path 3 Path 4 Path 5
Path Delay (T c ) 1 9 17 21 27
Path Direction (O ) 50 94 125 141 76Path Coefficient -0.10 + 0.26j -0.01 - 0.24j -0.31 - 0.02j -0.31 - 0.02j 0.42 - 0.35j
User2 (Interference) Path 1 Path 2 Path 3 Path 4 Path 5
Path Delay (T c ) 4 8 17 26 27
Path Direction (O ) 92 35 149 67 61Path Coefficient -0.20 + 0.56j -0.41 - 0.74j -0.39 - 0.92j -0.91 - 0.12j 0.76 - 0.00j
User3 (Interference) Path 1 Path 2 Path 3 Path 4 Path 5
Path Delay (T c ) 2 13 19 25 27
Path Direction (O ) 103 84 80 79 116Path Coefficient -0.15 + 0.27j -0.71 - 0.24j -0.11 - 0.01j -0.21 - 0.05j 0.45 - 0.55j
Page 41
© Imperial College LondonPage 42
Example - Channel Estimator
> Surface and contour plots shows that all 5 path delays and directions are correctly estimated
Page 42
© Imperial College LondonPage 43
Example - Decision Variables
Page 43
© Imperial College LondonPage 44
Example - SNIR Comparisons
0 5 10 15 20 25 30 35 40 45 50-20
-10
0
10
20
30
40
50Average Output SNIR Against Number of Users
Number of Users
Ave
rage
Out
put S
NIR
(in
dB
)Decorr
Proposed
ST-RAKE
Decorrelating Rx (MU) incomp.
0 5 10 15 20 25 30 35 40 45 50-20
-10
0
10
20
30
40
50
0 5 10 15 20 25 30 35 40 45 50-20
-10
0
10
20
30
40
50Average Output SNIR Against Number of Users
Number of Users
Ave
rage
Out
put S
NIR
(in
dB
)Decorr
Proposed
ST-RAKE
Decorrelating Rx (MU) incomp.
Page 44
© Imperial College LondonPage 45
Example – “Near-Far” Resistance
Page 45
© Imperial College LondonPage 46
Example – Subspace Tracking
Page 46
© Imperial College LondonPage 47
Page 47
© Imperial College LondonPage 48
Page 48
© Imperial College LondonPage 49
Conclusions:
ST Comms based on Subspace-Techniques:
– blind
– near-far resistant,
– superresolution capabilities
– The number of multipaths that can be resolved is not constrained by the number of array elements (antennas).
Page 49
© Imperial College LondonPage 50