Upload
mehdimajidi797144
View
14
Download
0
Embed Size (px)
DESCRIPTION
Lattice reduction aided MIMO Precoding
Citation preview
Lattice-Reduction-Aided MIMO Precoding
Robert Fischer
LEHRSTUHL FUR INFORMATIONSUBERTRAGUNGFriedrich–Alexander–Universitat Erlangen–Nurnberg
Outline 1
Introduction
Decision-Feedback Equalization and Precoding
Lattice-Reduction-Aided Precoding
Simulation Results
Summary
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Introduction 2
Situation: multiple–input/multiple–output (MIMO) transmission
y = Ha + n
Problem: interference of data streams transmitted in parallel
=> separation/equalization of signals required
Recently: increasing interest in schemes for
transmitter side signal processing
i.e., “pre-equalization”, “precoding”
Here: precoding for MIMO channels
broadcast scenario — downlink, no receiver side cooperation
schemes based on lattice basis reduction
duality to receiver side schemes
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Introduction (II) 3
Receiver Side Cooperation
Yes NoTra
nsm
itte
rSid
eCooper
ation
Yes
Point-to-Point Point-to-Multipoint(Downlink)
No
Multipoint-to-Point(Uplink)
—
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Decision-Feedback Equalization 4
Terms: decision-feedback equalization (DFE), successive cancellation, V-BLAST
H
a1
a2
a
y
n
a1
h1
y2
h2
y3
a2g2
fT1
fT2
= [h1 · · ·hNT]
aNT
g1
Successive procedure:
subtraction of interference of already detected symbols aκ
filtering by fTk (||f k||2 = 1) for suppression of residual interference
scaling by gk for unit gain signal transfer function
decision of symbol ak
Decision order can be optimized!
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Decision-Feedback Equalization (II) 5
Redrawing the receiver: Matrix DFE
a
H
a1
a2
a
y
n
F G
B − I
P
aNT
Calculation of required matrices: (sorted QR-type decomposition)
H P = F H G−1 B
feedforward matrix F : unitary matrix with rows fTk
scaling matrix G: diagonal matrix with entries gk
feedback matrix B: lower triangular matrix with unit main diagonal
permutation matrix P : representation of decision order
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Precoding 6
Multipoint-to-point transmission: DFE
B − I
F
a
G PH
a1
a2
a
y
n
aNT
Point-to-multipoint transmission: Precoding
P T MOD F Hx
a1y1
B − I a
aa
n1
g1
x
nNR
yNRgNR
aNR
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Precoding (II) 7
Tomlinson-Harashima precoding:
P T MOD F Hx
a1y1
B − I a
aa
n1
g1
x
nNR
yNRgNR
aNR
Calculation of required matrices: (sorted QR-type decomposition)
P T H = G−1
B F H
feedforward matrix F : unitary matrix (short-term power constraint)
scaling matrix G: scaling factors (G = PGP T) (SNRs ∼ 1g2k)
feedback matrix B: lower triangular matrix with unit main diagonal
permutation matrix P : representation of processing order
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Precoding (III) 8
P T F Hx
a1y1
B − I a
aa
n1
d
nNR
yNRaNR
g1
gNR
x
with x = MOD{z} = z + d
P T H = G−1
B F H => HF B−1 = PG−1
= G−1P
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Precoding (III) 8
P T
a1y1
a
aa
n1
d
nNR
yNRaNR
g1
gNR
G−1P
with x = MOD{z} = z + d
P T H = G−1
B F H => HF B−1 = PG−1
= G−1P
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Precoding (III) 8
a1y1
a
a
d
yNRaNR
gNRnNR
g1n1
1−1
1
−1
Precoding = linear pre-equalization+ periodic extension of signal constellation
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Lattice Basis Reduction 9
Complex-valued MIMO channel model:
y = Ha + n
Equivalent real-valued MIMO channel model with doubled dimensionality:[Re{y}Im{y}
]=
[Re{H} − Im{H}Im{H} Re{H}
] [Re{a}Im{a}
]+
[Re{n}Im{n}
]
In shortyr = H rar + nr
Lattice structure of signals: (example NT = 1, 4–ASK)
ar H rar yr = H rar + nr
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Lattice Basis Reduction (II) 10
ML detection:I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Lattice Basis Reduction (II) 10
Linear equalization:I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Lattice Basis Reduction (III) 11
Interpretation as lattice:I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Lattice Basis Reduction (III) 11
Equalization to rectangular grid:I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Lattice Basis Reduction (IV) 12
Procedure:
Lattice basis reduction: (e.g. LLL algorithm)
H r = H red R
with H red: “more suited” description of the latticeR: integer coefficients and det(R) = ±1
Equalization of only H red instead of H r:
yr
nr
R−1
a1
a2
aNT
a/ar
H/H r
ar/aH−1
red
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Lattice Basis Reduction (IV) 12
Procedure:
Lattice basis reduction: (e.g. LLL algorithm)
H r = H red R
with H red: “more suited” description of the latticeR: integer coefficients and det(R) = ±1
Equalization of only H red instead of H r:
yr
nr
R−1F P
B − I
a1
a2
aNT
a/ar
H/H r
ar/a
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Lattice-Reduction-Aided Precoding 13
Multipoint-to-point transmission:
yr
nra1
a2
aNT
a/ar
F
B − I
P R−1
H/H r
ar/a
Point-to-multipoint transmission: [Windpassinger et al. 2004], [Stierstorfer et al. 2004]
MOD Fx
a1y1
B − I
xa
n1
nNR
yNR gNR
aNR
g1
H/H r
a/ar
ar/a
P TR
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Simulation Results 14
Parameters:
NT = 4 transmit antennas at joint transmitter
NR = 4 separated receivers
uncoded transmission;
– all parallel data streams use 4-ary QAM
– constant average transmit power
“flat block-fading channels”
– random channel matrix H (i.i.d. Gaussian entries with unit variance)
– averaging over a large number of channel realizations
optimal processing order
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Simulation Results (II) 15
−5 0 5 10 15 20 2510
−5
10−4
10−3
10−2
10−1
100
10 log10
(Eb,TX/N0
)[dB] −→
BE
R−→
linear pre-equal.precoding (C/R)
vector precoding
LR lin. pre-equal.LR precoding
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding
Summary 16
Precoding for MIMO channels
nonlinear transmitter side preequalization
of special interest in point-to-multipoint transmission (downlink)
schemes dual to that applicable in receiver side processing
=> “uplink-downlink duality”
considerable gains possible by applying the concept of
lattice-reduction-aided equalization
independent, parallel channels with periodic extension of the signal set result
=> channel coding immediately applicable
I
Robert Fischer: Lattice-Reduction-Aided MIMO Precoding