D IGITAL PROCESSING ALGORITHMS AND ARCHITECTURES FOR UWB LOW COST COMMUNICATION SYSTEM S AMI MEKKI Télécom ParisTech (ENST)
July 3rd, 2009
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
O UTLINE
1
I NTRODUCTION AND M OTIVATION
2
E NERGY D ETECTION R ECEIVER
3
P ROBABILISTIC UWB E QUALIZER
4
C HANNEL PARAMETERS E STIMATION
5
A RCHITECTURE F EASIBILITY
6
C ONCLUSIONS
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
2 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
1
I NTRODUCTION AND M OTIVATION
2
E NERGY D ETECTION R ECEIVER
3
P ROBABILISTIC UWB E QUALIZER
4
C HANNEL PARAMETERS E STIMATION
5
A RCHITECTURE F EASIBILITY
6
C ONCLUSIONS
Sami MEKKI
Architecture Feasibility
Digital processing algorithms and architectures for UWB low cost communication system
Conclusions
July 3rd, 2009
3 / 39
Introduction and Motivation
Energy Detection Receiver
I NDOOR UWB
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
APPLICATIONS
Wireless personal area network (WPAN)
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
4 / 39
Introduction and Motivation
Energy Detection Receiver
T HE SPECTRUM
Probabilistic UWB Equalizer
OF THE
UWB
Channel Parameters Estimation
Architecture Feasibility
Conclusions
SIGNALS VERSUS CONVENTIONAL
SIGNALS
PSD
Power Spectrum Density
Conventional radio signals
UWB radio signal
Frequency
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
5 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
D EFINITION UWB D EFINITION UWB systems transmit signals whose bandwidths exceed 20% of their center (f −f ) frequency fc (i.e Hf L > 20%), or have a −10dB bandwidth of more than 500 MHz c
Scholtz monocycle with τ = 0.13ns m
1.5 pS(t)
0 −5 Relative Power (dB)
Signal Amplitude
1
0.5
0
−0.5
−10
Bandwidth
−15 −20 −25 −30
−1
fH
fL
−35
pS(t) FCC Mask
−1.5 −2
−1.5
−1
−0.5
0 Time [s]
0.5
1
1.5
2 −10
x 10
1
2
3
4
5 6 7 8 Frequency (GHz)
9
10
11
12
F IGURE : Scholtz monocycle in time domain and in frequency domain for τ = 0.13 ns.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
6 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
O BJECTIVES
G OALS Design an Impulse Radio (IR) UWB receiver based on low cost analog components. Transmit at high data rate (≥ 100 Mbps) with M -PPM modulation. Define new equalization algorithm with energy detection. Develop new methods for channel parameters estimation based on energy detection. Study the associated architecture for hardware implementation
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
7 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
D IFFICULTIES 1
UWB Channel Models (CM) [IEEE 802.15.3a].
CM1
CM4
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
−0.1
−0.1
−0.2
−0.2
−0.3
−0.3
−0.4
−0.4
−0.5
0
50
100
150 Time (nS)
200
250
−0.5
0
50
100
150
200
250
Time (nS)
F IGURE : Impulse responses for Channel Model CM1 and CM4. Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
8 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
D IFFICULTIES 1
UWB Channel Models (CM) [IEEE 802.15.3a].
2
Use of M -PPM modulation for high data transmission (i.e. transmit at 100 Mbps with a 4-PPM).
Time Symbol Ts Time Slot Tslot
Binary Sequence 00
01
10
11
01
F IGURE : An example of 4−PPM modulation and straight binary mapping.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
8 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
D IFFICULTIES 1
UWB Channel Models (CM) [IEEE 802.15.3a].
2
Use of M -PPM modulation for high data transmission (i.e. transmit at 100 Mbps with a 4-PPM). Transmitted data
Ts= 4.Tslot 1
e(t) 0
0
20
40
60
80
100
120
140
160
180
200
220
Time(nS)
Transmitted symbol: 00 0.3
x
(t) n−2
0
−0.3
0
20
40
60
80
100 120 Time(nS) Transmitted symbol: 10
140
160
180
200
220
0
20
40
60
80
100 120 Time(nS) Transmitted symbol: 00
140
160
180
200
220
0
20
40
60
80
140
160
180
200
220
0.3
x
(t)
n−1
0
−0.3
0.3
xn(t)
0
−0.3
100
120 Time(nS)
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
8 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
D IFFICULTIES 1
UWB Channel Models (CM) [IEEE 802.15.3a].
2
Use of M -PPM modulation for high data transmission (i.e. transmit at 100 Mbps with a 4-PPM). Transmitted data
Ts= 4.Tslot 1
e(t) 0
0
20
40
60
80
100
120
140
160
180
200
220
120
140
160
180
200
220
Time(nS)
0.3
0.2
0.1
0
−0.1
−0.2
−0.3
0
20
40
60
80
100 Time(nS)
(b) Inter-Symbol Interference (ISI) and Inter-Slot Interference (IStI).
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
8 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
D IFFICULTIES 1
UWB Channel Models (CM) [IEEE 802.15.3a].
2
Use of M -PPM modulation for high data transmission (i.e. transmit at 100 Mbps with a 4-PPM). Transmitted data
Ts= 4.Tslot 1
e(t) 0
0
20
40
60
80
100
120
140
160
180
200
220
Time(nS)
The resulting signal 0.3
0.2
0.1
0
−0.1
−0.2
−0.3
0
20
40
60
80
100
120
140
160
180
200
220
Time(nS)
(c) The resulting signal sn (t).
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
8 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
R ECEIVER T YPES 1
Rake Receiver used for DS-UWB (coherent receiver).
Channel impulse response: h(t) = Drawbacks:
PK−1
`=1 a` e
jθ` δ(t − τ
RT
`)
0
a1e−jθ1
c∗(t − τ1 )
RT
A lot of Rake fingers is required to capture energy (up to 50 resolvable dominant paths in NLOS) [Win
0
Σ
Decision
and Scholtz, 1998, 2002] Using a single LOS path signal as template is suboptimal [Win and Scholtz, 1997] High sampling rate is required for channel estimation (14 Ghz [Nikookar and Prasad, 2009]).
⇒ Increase the complexity and cost of the receiver
a2 e−jθ2 ...
c∗(t − τ2 )
RT 0
aK−1e−jθK−1
c∗(t − τK−1)
F IGURE : Rake Receiver for DS-UWB. Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
9 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
R ECEIVER T YPES 1
Rake Receiver used for DS-UWB (coherent receiver).
2
Transmit Reference (T-R) receiver is pilot-based receiver (non-coherent receiver).
Two transmitted pulses are used: 1
reference pulse
2
data pulse
t
Drawbacks:
D
T-R receivers is poor at low signal-to-noise ratios or in the presence of narrow band interference [Dowla al. 2004]
1
⇒ Not Low Cost! Sami MEKKI
D
D
1
Mixer Filter and Amplifier
Half of the transmitted waveform is used as pilot. A mixer and a delay line are necessary.
D
0
Z
D 1
0
Differential data: "1 1 0 1 0" A/D T
Delay D
F IGURE : A DPSK sequence and TR-UWB receiver.
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
9 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
R ECEIVER T YPES 1
Rake Receiver used for DS-UWB (coherent receiver).
2
Transmit Reference (T-R) receiver is pilot-based receiver (non-coherent receiver).
3
Energy Detection (ED) receiver (non-coherent receiver).
Based on energy detection. Typically used for low data rate (Wireless Sensor Networks) Drawbacks:
(.)2
ADC
DSP
de
0
Suffers more from noise due to a square-law device. Dedicated for low data rate transmission.
ZT
F IGURE : Typical ED receiver.
⇒ Low cost receiver (analog components point of view)
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
9 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
R ECEIVER T YPES 1
Rake Receiver used for DS-UWB (coherent receiver).
2
Transmit Reference (T-R) receiver is pilot-based receiver (non-coherent receiver).
3
Energy Detection (ED) receiver (non-coherent receiver).
⇒ ED receiver is the best choice for low cost analog components.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
9 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
1
I NTRODUCTION AND M OTIVATION
2
E NERGY D ETECTION R ECEIVER
3
P ROBABILISTIC UWB E QUALIZER
4
C HANNEL PARAMETERS E STIMATION
5
A RCHITECTURE F EASIBILITY
6
C ONCLUSIONS
Sami MEKKI
Architecture Feasibility
Digital processing algorithms and architectures for UWB low cost communication system
Conclusions
July 3rd, 2009
10 / 39
Introduction and Motivation
Energy Detection Receiver
T RANSMITTER
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
AND RECEIVER DESIGN .
Channel
zn (t) Data
Encoder
Modulator
Pulse Generator
{en−k (t)}
Channel Filter
sn (t)
H
SISO Decoder
Equalizer
En,m
Z
(.)2 Tslot
F IGURE : Block diagram of transmission and reception with an equalizer and SISO decoder.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
11 / 39
Introduction and Motivation
Energy Detection Receiver
T RANSMITTER
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
AND RECEIVER DESIGN .
Channel
zn (t) Data
Encoder
Modulator
Pulse Generator
{en−k (t)}
Channel Filter
sn (t)
H
SISO Decoder
Equalizer
En,m
Z
(.)2 Tslot
F IGURE : Block diagram of transmission and reception with an equalizer and SISO decoder.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
11 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
The output of the channel filter is sn (t) =
Z En,m
Channel Parameters Estimation
PK−1
k=0 xn−k (t) (sum of
nTs +(m)Tslot
=
Architecture Feasibility
Conclusions
K interfered symbols).
(sn (t) + zn (t))2 dt
(1)
nTs +(m−1)Tslot
According to [Urkowitz-67]:
En,m =
2L X
(s`n,m + z`n,m )2
(2)
`=1
where 2L = 2Tslot W (i.e. 2L is the Degree of Freedom (DoF) over Tslot ).
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
12 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
E NERGY DISTRIBUTION From (2) χ` = s`n,m + z`n,m ∼ N (s`n,m , σ2 ). ⇒ The energy En,m follows a Chi2 (χ2 ) distribution with 2L DoF, the pdf is given [Proakis, 2000]: 1
if Bn,m =
` 2 `=1 (sn,m )
P2L
, 0 (non-centrality parameter)
1 En,m p(En,m |Bn,m ) = 2σ2 Bn,m 2
if Bn,m =
` 2 `=1 (sn,m )
P2L
! L−1 2
−
e
(En,m +Bn,m ) 2σ2
p Bn,m En,m IL−1 σ2
(3)
=0 p(En,m |0) =
1
σ2L 2L Γ(L)
EL−1 n,m e
−En,m 2σ2
(4)
where IL−1 (u) and Γ(z) are respectively the (L − 1)th -order modified Bessel function of the first kind and the Gamma function [Abramowitz, 1964].
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
13 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
1
I NTRODUCTION AND M OTIVATION
2
E NERGY D ETECTION R ECEIVER
3
P ROBABILISTIC UWB E QUALIZER
4
C HANNEL PARAMETERS E STIMATION
5
A RCHITECTURE F EASIBILITY
6
C ONCLUSIONS
Sami MEKKI
Architecture Feasibility
Digital processing algorithms and architectures for UWB low cost communication system
Conclusions
July 3rd, 2009
14 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
M OTIVATION Manage the Inter-Symbol Interference (ISI) and intra-Symbol Interference induced by the channel models (defined by IEEE802.15.3a).
P ROPOSAL Develop a new probabilistic equalizer based on ED. 1
The equalizer computes p(En |xn ).
2
This probability (after normalization) is forwarded to the SISO (Soft-Input Soft Output) decoder to compute the extrinsic probability.
3
The equalizer computes again p(En |xn ) taking into account the probabilities provided by the decoder previously. Steps 2 and 3 are iterated until convergence.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
15 / 39
Introduction and Motivation
H OW
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
THE PROBABILISTIC EQUALIZER WORKS !
The probabilistic equalizer takes into account all the possible interference and computes p(En |xn ) as follows
p(En |xn ) =
X
M K−1 Y Y p(En,m |Bn,m ) π(xn−k )
xn−1 ,...,xn−K+1 m=1
(5)
k =1
where π(xn−k ) is the a priori probability of xn−k and Bn,m ∈ B = {βj }. TABLE : Exhaustive Listing of Energy Patterns βj for two consecutive 4-PPM interfering symbols, K = 2.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
16 / 39
Introduction and Motivation
H OW
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
THE PROBABILISTIC EQUALIZER WORKS !
The probabilistic equalizer takes into account all the possible interference and computes p(En |xn ) as follows
p(En |xn ) =
X
M K−1 Y Y p(En,m |Bn,m ) π(xn−k )
xn−1 ,...,xn−K+1 m=1
(5)
k =1
where π(xn−k ) is the a priori probability of xn−k and Bn,m ∈ B = {βj }. TABLE : Exhaustive Listing of Energy Patterns βj for two consecutive 4-PPM interfering symbols, K = 2.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
16 / 39
Introduction and Motivation
H OW
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
THE PROBABILISTIC EQUALIZER WORKS !
The probabilistic equalizer takes into account all the possible interference and computes p(En |xn ) as follows
p(En |xn ) =
M K−1 Y Y p(En,m |Bn,m ) π(xn−k )
X
xn−1 ,...,xn−K+1 m=1
(5)
k =1
where π(xn−k ) is the a priori probability of xn−k and Bn,m ∈ B = {βj }.
TABLE : Relation between the presumed number of interfering symbols K and |B|
Sami MEKKI
K
P
|B|
2 3 4
5 9 13
15 88 424
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
16 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
P ROBABILISTIC E QUALIZER S UMMARY IN P ERFECT CSI
Encoder
Decoded bits {dei}
z(t)
codeword
Data{di}
SISO Decoder
c
Mapper Pulse Generator
p(En|xn )
{pn−k (t)}
Channel Filter H
En,m
Equalizer
s(t)
Z
(.)2 Tslot
π(xk )
Mem
{βj } σ2
χ2 Table p(En,m|Bn,m )
F IGURE : Probabilistic equalizer procedure
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
17 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
S IMULATIONS WITHOUT CHANNEL CODING IN P ERFECT CSI Energetic Equalization with Perfect CSI without Channel Coding at 100Mbps and P=5 100
Bit Error Rate (information)
10-1
10-2
10-3
CM1 with Equalization CM1 without equalization CM2 with Equalization CM2 without equalization CM3 with Equalization CM3 without equalization CM4 with Equalization CM4 without equalization
10-4
10-5
6
8
10
12 Eb/N0 (dB)
14
16
18
20
BER vs Eb /N0 for different channel models IEEE802.15.3a for K = 2 at 100Mbit/s (Tslot = 5ns) using a 4-PPM modulation, W = 3 GHz (2L = 30). Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
18 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
S IMULATIONS WITHOUT CHANNEL CODING IN P ERFECT CSI Energetic Equalization with Perfect CSI without Channel Coding at 100Mbps and P=9 100
Bit Error Rate (information)
10-1
10-2
10-3
10-4
10-5
CM1 with Equalization CM2 with Equalization CM3 with Equalization CM4 with Equalization 6
8
10
12 Eb/N0 (dB)
14
16
18
20
BER vs Eb /N0 for different channel models IEEE802.15.3a for K = 3 at 100Mbit/s (Tslot = 5ns) using a 4-PPM modulation, W = 3 GHz (2L = 30). Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
18 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
S IMULATIONS WITHOUT CHANNEL CODING IN P ERFECT CSI Energetic Equalization with Perfect CSI without Channel Coding at 100Mbps and P=13 100
Bit Error Rate (information)
10-1
10-2
10-3
10-4
10-5
CM1 with Equalization CM2 with Equalization CM3 with Equalization CM4 with Equalization 6
8
10
12 Eb/N0 (dB)
14
16
18
20
BER vs Eb /N0 for different channel models IEEE802.15.3a for K = 4 at 100Mbit/s (Tslot = 5ns) using a 4-PPM modulation, W = 3 GHz (2L = 30). Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
18 / 39
Introduction and Motivation
Energy Detection Receiver
S IMULATIONS WITH
Probabilistic UWB Equalizer
Channel Parameters Estimation
CHANNEL CODING IN
Architecture Feasibility
Conclusions
P ERFECT CSI
P=5, BICM (23,35) at rate 1/2, 10 SISO iterations and Energetic Equalization in Perfect CSI at 100Mbps 100
Bit Error Rate (information)
10-1
10-2
10-3
10-4
10-5
CM1 CM2 CM3 CM4 6
8
10
12 Eb/N0 (dB)
14
16
18
20
BER vs Eb /N0 for different channel models using BICM(23,35) with code rate 1/2 length N=1024 bits, 10 iterations of the SISO decoder for K = 2, data rate 100Mbit/s and a 4-PPM modulation, W = 3 GHz (2L = 30). Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
19 / 39
Introduction and Motivation
Energy Detection Receiver
S IMULATIONS WITH
Probabilistic UWB Equalizer
Channel Parameters Estimation
CHANNEL CODING IN
Architecture Feasibility
Conclusions
P ERFECT CSI
P=9, BICM (23,35) at rate 1/2, 10 SISO iterations and Energetic Equalization in Perfect CSI at 100Mbps 100
Bit Error Rate (information)
10-1
10-2
10-3
10-4
10-5
CM1 CM2 CM3 CM4 6
8
10
12 Eb/N0 (dB)
14
16
18
20
BER vs Eb /N0 for different channel models using BICM(23,35) with code rate 1/2 length N=1024 bits, 10 iterations of the SISO decoder for K = 3, data rate 100Mbit/s and a 4-PPM modulation, W = 3 GHz (2L = 30). Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
19 / 39
Introduction and Motivation
Energy Detection Receiver
S IMULATIONS WITH
Probabilistic UWB Equalizer
Channel Parameters Estimation
CHANNEL CODING IN
Architecture Feasibility
Conclusions
P ERFECT CSI
P=13, BICM (23,35) at rate 1/2, 10 SISO iterations and Energetic Equalization in Perfect CSI at 100Mbps 100
Bit Error Rate (information)
10-1
10-2
10-3
10-4
10-5
CM1 CM2 CM3 CM4 6
8
10
12 Eb/N0 (dB)
14
16
18
20
BER vs Eb /N0 for different channel models using BICM(23,35) with code rate 1/2 length N=1024 bits, 10 iterations of the SISO decoder for K = 4, data rate 100Mbit/s and a 4-PPM modulation, W = 3 GHz (2L = 30). Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
19 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
1
I NTRODUCTION AND M OTIVATION
2
E NERGY D ETECTION R ECEIVER
3
P ROBABILISTIC UWB E QUALIZER
4
C HANNEL PARAMETERS E STIMATION
5
A RCHITECTURE F EASIBILITY
6
C ONCLUSIONS
Sami MEKKI
Architecture Feasibility
Digital processing algorithms and architectures for UWB low cost communication system
Conclusions
July 3rd, 2009
20 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
M OTIVATION Estimate the channel parameters (B and σ2 ) for equalization application.
P ROPOSALS 1
Apply the expectation-maximization (EM) algorithm [Dempster, Laird and Rudin, 1977] to obtain the maximum likelihood of estimates that might not be available otherwise.
2
Transmit special training sequences to estimate the channel parameters.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
21 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
EM A LGORITHM
This algorithm allows to compute maximum-likelihood estimate of parameters when direct access to the full data necessary to make the estimate is impossible. Equalization needs to access to B = {βj } and σ2 .
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
22 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
EM A LGORITHM
This algorithm allows to compute maximum-likelihood estimate of parameters when direct access to the full data necessary to make the estimate is impossible. Equalization needs to access to B = {βj } and σ2 . ⇒ The EM algorithm is ideally suited to problems of this sort.
y is the incomplete data (received data); x is the missing data (transmitted data); (x, y) is the complete data. θ is the parameter to be estimated (the channel parameters (B = {βj }, σ2 )).
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
22 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
A LGORITHM D ESCRIPTION The EM algorithm consists of two primary steps: an expectation step, followed by a maximization step. It proceeds as follows: 1
Choose an initial parameter θ(0)
2
E-step (Expectation step):Estimate unobserved data using θi
Q(θ|θ(i) ) = Ex [log p(x, y|θ)|y, θ(i) ]
(6)
where Q is the auxiliary function. 3
M-step (Maximization step): compute maximum likelihood estimate of parameter θ(i+1) using estimated data.
θ(i+1) = arg max Q(θ|θ(i) )
(7)
θ
The E-step and the M-step are alternated until the parameter estimate has converged (no more change on the estimate).
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
23 / 39
Introduction and Motivation
EM
Energy Detection Receiver
APPLICATION TO
Probabilistic UWB Equalizer
ED
Channel Parameters Estimation
Architecture Feasibility
Conclusions
MODEL
The auxiliary function with ED is Q(θ|θi )
=
N X M XX x
log p(En,m |Bn,m , θ)APP(x)
(8)
n m=1
The update channel parameters are 2 N M q X X (i) (i) f (En,m , 2Lσ2 ) pi (Bn,m = βj ) n=1 m=1 (i+1) βj ≈ N M XX (i) pi (Bn,m = βj )
(9)
n=1 m=1
σ2
(i+1)
=
N M 1 XXX (i) (i) (En,m − βj )pi (Bn,m = βj ) 2LMN n=1 m=1 (i)
(10)
Bj
where f (En,m
Sami MEKKI
En,m − 2Lσ2 (i) 0
(i) , 2Lσ2 ) =
if En,m > 2Lσ2
(i)
if En,m ≤ 2Lσ2
(i)
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
24 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
E QUALIZER AND EM ALGORITHM
Encoder
SISO Decoder
Architecture Feasibility
Conclusions
DESIGN
z(t)
codeword
Data{di}
Decoded bits
Channel Parameters Estimation
c
Mapper Pulse Generator
p(En|xn )
{pn−k (t)}
Channel Filter H
En,m
Equalizer
s(t)
Z
(.)2 Tslot
AP Pi (x) π(x) Mem
EM
θ
i
{βj } σ
2
χ2 Table p(En,m|Bn,m )
F IGURE : Transmitter and receiver design with joint channel parameters estimation and energy equalization.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
25 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
K=2, BICM (23,35) at rate 1/2, 10 SISO iterations and Energy Equalization with 1 iteration of EM at 100Mbps 100
Bit Error Rate (information)
10-1
10-2
10-3
10-4
10-5
CM1-Perfect CSI CM1-EM Algorithm CM2-Perfect CSI CM2-EM Algorithm CM3-Perfect CSI CM3-EM Algorithm CM4-Perfect CSI CM4-EM Algorithm 6
8
10
12 Eb/N0 (dB)
14
16
18
20
BER for different channel models using BICM(23,35) with code rate 1/2, N=1024 with K = 2, 10 SISO decoder iterations with one iteration of the EM algorithm per decoder iteration (10 EM iterations). 10 symbols for EM initialization, EM algorithm versus Perfect CSI at 100 Mbps using 4-PPM modulation. Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
26 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
K=3, BICM (23,35) at rate 1/2, 10 SISO iterations and Energy Equalization with 1 iteration of EM at 100Mbps 100
Bit Error Rate (information)
10-1
10-2
10-3
10-4
10-5
CM1-Perfect CSI CM1-EM Algorithm CM2-Perfect CSI CM2-EM Algorithm CM3-Perfect CSI CM3-EM Algorithm CM4-Perfect CSI CM4-EM Algorithm 6
8
10
12 Eb/N0 (dB)
14
16
18
20
BER for different channel models using BICM(23,35) with code rate 1/2, N=1024 with K = 3, 10 SISO decoder iterations with one iteration of the EM algorithm per decoder iteration (10 EM iterations). 10 symbols for EM initialization, EM algorithm versus Perfect CSI at 100 Mbps using 4-PPM modulation. Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
26 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
1
I NTRODUCTION AND M OTIVATION
2
E NERGY D ETECTION R ECEIVER
3
P ROBABILISTIC UWB E QUALIZER
4
C HANNEL PARAMETERS E STIMATION
5
A RCHITECTURE F EASIBILITY
6
C ONCLUSIONS
Sami MEKKI
Architecture Feasibility
Digital processing algorithms and architectures for UWB low cost communication system
Conclusions
July 3rd, 2009
27 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
M OTIVATIONS AND P ROPOSALS M OTIVATION Study and reduce the hardware implementation complexity (computational complexity) of the developed algorithm. Optimize the size of variables (number of bits) required for the operations.
T HE PROBLEMS The χ2 pdf will occupy a very large memory (448 Mbit). Keep reasonable amount of operators (3.2 GMultip/s).
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
28 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
M OTIVATIONS AND P ROPOSALS M OTIVATION Study and reduce the hardware implementation complexity (computational complexity) of the developed algorithm. Optimize the size of variables (number of bits) required for the operations.
T HE PROBLEMS The χ2 pdf will occupy a very large memory (448 Mbit). Keep reasonable amount of operators (3.2 GMultip/s).
P ROPOSALS Approximate the χ2 pdf by a function easier to implement (reduce the required memories, i.e. ROM). Compute in the logarithm domain (additions become Log-Sum or max*10 operations and multiplications become additions). Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
28 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
χ2 A PPROXIMATION According to the central limit theorem, for large 2L, the distribution of En,m is asymptotically Gaussian [Marcuse-90]. When φ is used to approximate ψ, it is equivalent to match the mean and variance [Jin-Ting Zhang-05]. Mean and variance of χ2 distribution:
mχ2 σ2χ2
= 2Lσ2 + Bn,m 4
(11)
2
= 4Lσ + 4σ Bn,m
p(En,m |Bn,m ) ≈ e p(En,m |Bn,m ) = q
1 2πσ2χ2
(12)
(En,m − m 2 )2 χ exp − 2σ2
(13)
χ2
i.e. En,m ∼ N (mχ2 , σ2 2 ) χ
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
29 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
χ2 A PPROXIMATION E RROR
WITH
Channel Parameters Estimation
Architecture Feasibility
Conclusions
σ2 = 1 ( LOW Eb /N0 )
10−3
4
3
Error 2
1,000
1 750
Bnm
500 250 00 0
100
300
200
400
500
Enm
Error measured by |p(En,m |Bn,m ) −e p(En,m |Bn,m )| ∀ En,m ≥ 0, ∀ Bn,m > 0 and σ2 = 1. Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
30 / 39
Introduction and Motivation
Energy Detection Receiver
A PPROXIMATED χ2
log(En,m )
ROM
En,m
10x
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
IMPLEMENTATION
En,m − mχ2
ROM
φ(x) =
2 √1 e−x 2π
pe(En,m |Bn,m )
Bn,m
q1 σ2 2
χ
log(2Lσ 2 )
ROM
σχ2 2
mχ 2
10x
2Lσ 2
ROM √1 x
1 L
: Stands for trivial Multiplication by a power of 2 : Stands for non-trivial Multiplication
F IGURE : Approximated Energy distribution architecture for the linear equalizer with χ2 approximation. The linear equalizer requires 12.8 GMultip/s and 30.25 Kbit per level of parallelism. Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
31 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
TC Frame Type 864 with rate 1/2 at 100Mbps (Chi-2 vs Fixed Chi2 approximation) 100
Bit Error Rate (information)
10-1
10-2
CM1-float Chi2 CM1-Fixed Gauss Approx CM2-float Chi2 CM2-Fixed Gauss Approx CM3-float Chi2 CM3-Fixed Gauss Approx CM4-float Chi2 CM4-Fixed Gauss Approx
10-3
10-4
8
10
12
14 Eb/N0 (dB)
16
18
20
χ2 float precision versus the Gaussian approximation in fixed point precision for K = 2. A duo-binary turbo code with encoded data of 864 bits at rate 1/2, 4-PPM modulation at 100 Mbps Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
32 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
L OG -S UM OR max*10 O PERATOR The logarithmic domain allows to replace linear multiplications by log additions and the linear additions have to be replaced by the Log-Sum (max*10 ) operation [Viterbi, 1998]: ∆
max*10 (a, b) = log10 (10a + 10b ) ∆
= max(a, b) + log10 (1 + 10−|a−b| )
(14)
The new probabilistic equalizer in the logarithm domain is
log p(En |xn ) =
*
M K−1 X X log p(En,m |Bn,m ) + log π(xn−k )
max10 xn−1 ,...,xn−K+1 m=1
(15)
k =1
where h i loge p(En,m |Bn,m ) ∝ − 12 max*10 log(2Lσ2 ), log 2 + log Bn,m − 2 lnL10 (En,m − mχ2 )2 10−γ and γ = log(2Lσ2 ) + max*10 [log(2Lσ2 ), log 2 + log Bn,m ]
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
33 / 39
Introduction and Motivation
Energy Detection Receiver
A PPROXIMATED χ2
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
IMPLEMENTATION − 2 lnL10
log(En,m )
10x
2
log(2Lσ )
log Bn,m
ROM
En,m
ROM
2Lσ 2
10x
ROM
Bn,m
10x
log pe(En,m |Bn,m )
mχ 2
γ
−γ
ROM
10x
10−γ
−1 log 2
log 2 + log Bn,m
max∗10
−1/2
: Stands for trivial Multiplication by a power of 2 : Stands for non-trivial Multiplication
max∗10
≡
d
sign
ROM
log(1 + 10−|d|)
F IGURE : Energy distribution architecture for logarithmic equalizer. The logarithmic equalizer requires 6.4 GMultip/s and 16.25 Kbit per level of parallelism. Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
34 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
TC Frame Type 864 with rate 1/2 at 100Mbps and K=2 (Chi-2 vs Fixed Log approximated Chi2) 100
Bit Error Rate (information)
10-1
10-2
CM1-float Chi2 CM1-Fixed CM2-float Chi2 CM2-Fixed CM3-float Chi2 CM3-Fixed CM4-float Chi2 CM4-Fixed
10-3
10-4
8
10
12
14 Eb/N0 (dB)
16
18
20
χ2 float precision versus the logarithmic approximated χ2 in fixed point precision for K = 2. A duo-binary turbo code with encoded data of 864 bits at rate 1/2, 4-PPM modulation at 100 Mbps Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
35 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
A RCHITECTURE F EASIBILITY S UMMARY
TABLE : Complexity Requirement with 4-PPM and K = 2
Functions
Linear
Number of Multiplications Linear Logarithmic χ2 approximated χ2 approximated
(M + K − 2)M K 3.2 GMultip/s 0
p(En |xn ) p(En,m |Bn,m ) Total Equalizer Multiplications Total required memory per level of parallelism
Sami MEKKI
χ2
3.2 GMultip/s
(M + K − 2)M K 3.2 GMultip/s 3M K +1 9.6 GMultip/s 12.8 GMultip/s
2M K +1 6.4 GMultip/s 6.4 GMultip/s
448 Mbit
30.25 Kbit
16.25 Kbit
Digital processing algorithms and architectures for UWB low cost communication system
0
July 3rd, 2009
36 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
1
I NTRODUCTION AND M OTIVATION
2
E NERGY D ETECTION R ECEIVER
3
P ROBABILISTIC UWB E QUALIZER
4
C HANNEL PARAMETERS E STIMATION
5
A RCHITECTURE F EASIBILITY
6
C ONCLUSIONS
Sami MEKKI
Architecture Feasibility
Digital processing algorithms and architectures for UWB low cost communication system
Conclusions
July 3rd, 2009
37 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
C ONCLUSION We proposed the probabilistic UWB equalizer with energy detection that improves the receiver performance at high data rate.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
38 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
C ONCLUSION We proposed the probabilistic UWB equalizer with energy detection that improves the receiver performance at high data rate. We implemented the EM algorithm for channel parameters estimation necessary to equalization.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
38 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
C ONCLUSION We proposed the probabilistic UWB equalizer with energy detection that improves the receiver performance at high data rate. We implemented the EM algorithm for channel parameters estimation necessary to equalization. We defined a special training sequence capable to get a good estimate of channel parameters with very low complexity channel estimation with regards to the EM method.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
38 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
C ONCLUSION We proposed the probabilistic UWB equalizer with energy detection that improves the receiver performance at high data rate. We implemented the EM algorithm for channel parameters estimation necessary to equalization. We defined a special training sequence capable to get a good estimate of channel parameters with very low complexity channel estimation with regards to the EM method. Equalization with the proposed Gaussian approximation achieves the equalization performance.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
38 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
C ONCLUSION We proposed the probabilistic UWB equalizer with energy detection that improves the receiver performance at high data rate. We implemented the EM algorithm for channel parameters estimation necessary to equalization. We defined a special training sequence capable to get a good estimate of channel parameters with very low complexity channel estimation with regards to the EM method. Equalization with the proposed Gaussian approximation achieves the equalization performance. We reduced the memory size and the number of multipliers by performing the calculation in the logarithm domain.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
38 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
C ONCLUSION We proposed the probabilistic UWB equalizer with energy detection that improves the receiver performance at high data rate. We implemented the EM algorithm for channel parameters estimation necessary to equalization. We defined a special training sequence capable to get a good estimate of channel parameters with very low complexity channel estimation with regards to the EM method. Equalization with the proposed Gaussian approximation achieves the equalization performance. We reduced the memory size and the number of multipliers by performing the calculation in the logarithm domain.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
38 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
C ONCLUSION We proposed the probabilistic UWB equalizer with energy detection that improves the receiver performance at high data rate. We implemented the EM algorithm for channel parameters estimation necessary to equalization. We defined a special training sequence capable to get a good estimate of channel parameters with very low complexity channel estimation with regards to the EM method. Equalization with the proposed Gaussian approximation achieves the equalization performance. We reduced the memory size and the number of multipliers by performing the calculation in the logarithm domain. This study shows the feasibility of IR-UWB receiver based on energy detection for high data rate transmission. Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
38 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
O PEN I SSUES
Reduce the energy coefficient in the equalizer to optimize the digital processing cost.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
39 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
O PEN I SSUES
Reduce the energy coefficient in the equalizer to optimize the digital processing cost. Study the feasibility of a multi-user receiver based on energy detection.
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
39 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
O PEN I SSUES
Reduce the energy coefficient in the equalizer to optimize the digital processing cost. Study the feasibility of a multi-user receiver based on energy detection. Extend the application of this work in any field dealing with energy detection (wireless optical communication).
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
39 / 39
Introduction and Motivation
Energy Detection Receiver
Probabilistic UWB Equalizer
Channel Parameters Estimation
Architecture Feasibility
Conclusions
Thank you for your attention! Questions?
Sami MEKKI
Digital processing algorithms and architectures for UWB low cost communication system
July 3rd, 2009
39 / 39