Optical Wireless Communication using Digital Pulse

Download Report

Transcript Optical Wireless Communication using Digital Pulse

Performance of Diffuse Indoor Optical
Wireless Links Employing Neural and
Adaptive Linear Equalizers
Z. Ghassemlooy & S Rajbhandari
Optical Communications Research Group, School of Computing,
Engineering & Information Sciences, University of Northumbria,
Newcastle upon Tyne, UK
ICICS 2007 Singapore
Outline





Optical wireless – introduction
Mutipath induces ISI
ANN based equalizer
Wavelet-ANN receiver
Final comments
Optical Wireless Communication –
What Does It Offer?











Abundance bandwidth
No multipath fading
High data rates
Protocol transparent
Secure data transmission
License free
Free from electromagnetic interference
Compatible with optical fibre (last mile bottle neck?)
Low cost of deployment
Easy to deploy
Etc.
Power Spectra of Ambient Light Sources
Normalised power/unit wavelength
1.2
Pave)amb-light >> Pave)signal (Typically 30 dB with no optical filtering)
Sun
1
Incandescent
0.8
1st window IR
0.6
2nd window IR
Fluorescent
0.4
x 10
0.2
0
1.5
1.4
1.3
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Wavelength (m)
Classification of Indoor OW Links
Directed
TX
Non-directed
Hybrid
TX
TX
Line-of-sight
RX
RX
RX
Non-line-of-sight
(Diffuse)
TX
RX
TX
RX
TX
RX
Indoor OWC - Challenges
Challenges
Causes
(Possible ) Solutions
Power limitation
Eye and skin safety.
Power efficient modulation
techniques/holographic diffuser/
Transreceiver at 1500 nm band
Noise
Intense ambient light
(artificial/ natural)
Optical and electrical band pass
filters, Error control codes
Intersymbol
interference (ISI)
Multipath propagation
(non-LOS links)
Equalization, Multi-beam
transmitter
No/limited mobility
Beam confined to small
area.
Wide angle optical transmitter ,
MIMO transceiver.
Shadowing
blocking
Limited data rate
LOS links
Strict link set-up
LOS links
Diffuse links/ cellular system/ wide
angle optical transmitter
Bandwidth-efficient modulation
techniques/Multiple small area
photo-detector
Diffuse links/ wide angle transmitter
Large area photodetectors
Modulation Techniques
Normalized Power and Bandwidth
Requirement
Normalized bandwidth requirement
20
18
 PPM the most power efficient
while requires the largest
bandwidth
 DH-PIM2 is the most bandwidth
efficient
16
14
12
10
8
6
4
2
OOK
Normalized Power Requirement (dB)
0
2
3
4
5
6
7
8
 DH-PIM and DPIM shows almost
identical bandwidth requirement
and power requirement
Bit resolution, M
0
-2
-4
-6
-8
 There is always a trade-off
between power and bandwidth
-10
-12
-14
-16
2
3
4
5
6
Bit Resolution, M
7
8
Power Spectral Density
6
Notice the DC component:when filtered will result in base
line wander effect
5
4
32-DPIM
P
S
D 3
16-DPIM
8-DPIM
OOK
2
8-PPM
1
0
0
1
2
3
4
Normalised frequency (f/Rb)
5
6
Optical Wireless - Channel Model
 Basic system models – F. R. Gfeller et al 1979, J. M. Kahn et al 1995,
 Measurement studies - H. Hashemi et al 1994, J. M. Kahn et al 1995,
- Diffuse + shadowing
 Statistical models - J.B. Carruthers et al 1997
 Ray tracing techniques (to obtain simulated channel
responses) - J.R. Barry, J.R., et al. 1995, F.J. Lopez-Hernandez, et al,
2000
 Segmentation of reflecting surfaces + ray tracing
techniques to calculate the intensity and temporal
distributions - S. H. Khoo et al 2001
 Fast multi-receiver channel estimation - J.B. Carruthers et al 2002
Channel Model - Ceiling Bounce Model
 Developed by Carruthers and Kahn.
 Impulse response is:
h(t , a) 
6a 6
t  a 7
u (t )
where u(t) is the unit step
function and a is related to
the RMS delay spread D
a 13
D
12 11
LOS
Diffuse
Diffuse shadowed
LOS shadowed
OWC - LOS Links
Rx
 Least path loss
 No multipath propagation
 High data rates
Problems
Tx
 Noise is limiting factor
 Possibility of
blocking/shadowing
 Tracking necessary
 No/limited mobility
OWC - Diffuse Links



Rx
Tx
1.2
Received signal for non-LOS Links
1
Different paths ─>Different path
lengths ─> different delay ─>ISI.
ISI ─> Delay Spread Drms ─>
Room design and size
Impulse response of channel


6
6 0.1D
rms
h (t ) 
7 u (t )
t  0.1Drms 
Amplitude
0.8
Problems:
 High path loss
 Limited data rate due to ISI
 Power penalty due to ISI
0.6
0.4
0.2
0
-0.2
-0.4
0
2
4
6
Normalized Time
8
10
How to Combat Noise and Dispersion?
 Noise Filtering: Optical or Electrical
 Match Filtering:
 Maximises signal-to-noise ratio,
 Modulation: Z. Ghassemlooy et al
 Coding: Block codes, Convolutional and Turbo codes.
 Spread Spectrum
 Tracking Transmitters: D. Wisely et al
 Imaging Receivers: J.M. Kahn et al
 Integrated Optical Wireless Transceivers: D.C. O’Brien
 Equalisation
 Diversity: S. H. Khoo et al 2001
 Wavelet and AI based equalisers: Z. Ghassemlooy et al
Techniques to Mitigate the ISI
 Optimal solution - Maximum likelihood sequence
detection.
- Issues: complexity and delay
 Sub-optimal solution - Linear or decision
feedback equalizer based on the finite impulse
response (FIR) digital filter
- The impulse response of filter c(f) = 1/h(f), where h(f) is
the frequency response of channel
FIR Filter Equalizer
(Classical Signal Processing Tool)
Assumptions
 The statistics of noise is known (normally assume to be
Gaussian)
 The channel is stationary or quasi-stationary
 The channel characteristics are known (at least partially)
 Signals are linear
Problems: Non-linearity, time-varying and non-Gaussianity of
real signals and channel
Solution: Artificial neural network (ANN) based signal
processing which takes into account non-linearity, timevarying and non-Gaussianity of signal and channel
ANN
Output


One or more hidden layer(s)
Output is function of sum and product
of many functions

Useful tool because of learning
and adaptability capabilities
Extensively used as a classifier
Application in many areas like
engineering, medicine, financial,
physics and so on
Neurons
Hidden
layer



Input 
Training is necessary to adjust the
free parameters ( weight) before can
be used as classifier
Supervised and unsupervised
learning (training)
ff ((ZZi)) 11/ 1  exp( Z i )
i
ANN
Weights
Inputs
x1
x1w1
Z
w1
xnwn
xn
Activation
function
Bias bi
∑
f(.)
wn
Output
n
z   wi xi
Activation Function f(.)
• Sigmoid function • Linear function -
y  f (z )
i 1
f ( Z )  1 / 1  exp( Z i )
i
if Zi  0.5 ,
f (Z i )  0 if Z  0.5
• Any function that is differentiable
f (Zi )  1
i
f (Z )  Zi
i
if
 0.5  Z i  0.5
ANN
 Both the multilayer perceptrons (MLP) and the radial basic
function (RBF) have been used for equalization
 RBF requires a larger number of hidden nodes at lower
values of SNR
 The cascaded MLP and RBF outperform both the MLP and
RBF in terms of the BER performance
Learning rules for MLP
• The error-correction: {wij} are renewed after each iteration
- the most simplest
• The Boltzmann
• Hebbian …………
Whichever training rule is used, the basic principle is to
modify {wij} so that the error function is decreased after
each iteration.
ANN Supervised Learning (Training)
Target: to minimize
the error en between
target vector set tn and
neural network output
on for all input vector
set in.
in
Neural
network
on
Comparator
Error signal
en
Algorithms:
 Compare tn and on to determine en (= tn-on)
 Adjust {wn} and bi to reduce the error en
 Continue the process until en is small
tn
OWC System Block Diagram
Input
data
X(t)
Output
data
Tx
h(t)
∑
Rx
Equalizer
Threshold
detector
n(t)
ANN
Equalizer
Adaptive
Linear
Equalizer
For a non-stationary environment
OWC Link
n(t)
M
0100
M
0010
PPM
Encoder
PPM
Decoder
Xj
Optical
Transmitter
Decision
Device
Z (t )  X (t )  h(t )  n(t )
X(t)
Z(t)
h(t)
Yj
Neural
Network
Optical
∑
Receiver
Zj
Zj-1
Zj
Matched
Filter
.
Zj-n
Ts = M/LRb
.
 A feedforward back propagation ANN
 ANN is trained using a training sequence at the operating SNR
 Trained AAN is used for equalization
ANN Training Process
 The channel is time-varying
 To estimate channel parameters, a training sequence is
transmitted at regular interval for tracking changes in the
channel
 The information on channel is stored in the form of weights
that are updated on receiving the training sequence
 The signal flows from input to the output (feedforward) while
the error signal propagates backward, hence the name
feedforward backpropagation NN
 The learning duration and the number of iteration required to
adjust the NN parameters depends on the complexity of
learning task
 Here the aim is not to optimize the learning task but to
send a learning sequence of certain length to allow the
NN to estimate new channel parameters
Simulation Flow Chart
No
Start
Train neural
network
Generate OOK
RZ data stream
Generate OOK
RZ data stream
Generate
multipath h(t)
Generate
multipath h(t)
Convolve data
stream & h(t)
Convolve data
stream & h(t)
Add training
AWGN
Add simulation
AWGN
Window the
data stream
Window the
data stream
Classify using
neural network
Yes
3 sets
3 data sets are required each time
the network is trained.
Threshold
network output
If the network is to be trained with
another noise figure, start again.
Yes
Calculate BER
Train?
No
Save
Res ult
Plot results
Inform user
sim' end
Yes
Loop?
No
End
Train
Yes
Detect
BER?
10 blocks of
data are
processed
before loop
exits.
No
Is the BER target met?
Decrease
AWGN Value
Simulation Parameters
Parameters
Values
Number of layers
2
Number of neurons in each
layer
Training algorithm
Minimum error
36,1
tan-sigmoid,
log-sigmoid
scaled conjugate gradient
algorithm
1-30
Minimum gradient
1-30
Activation function
Simulation Parameters –
Parameters
Values
Data rate, Rb (Mbps)
OOK
PPM
DPIM
150
150
150
3
3
1/ Rb
M/( Rb .2M)
2M/(2M+1) Rb
2000 bits
300 symbols
600 symbols
Bit resolution, M
Slot duration, Ts
Training sequence
RMS delay spread,
Drms(ns)
10
5
2
10
5
Normalized time delay
(Drms/Ts)
1.5
0.75
0.3
4
2
8
4
2
Delayed samples
Contd.
22 11
2
10
5
2
0.75 2.3 1.13 0.45
6
13
7
3
Results and Discussion
Error performance for LOS links (150 Mbps)
0
10
 PPM requires the least
SNR to achieve a
desirable slot error rate
(SER)
-1
10
-2
SER
10
-3
10
-4
10
-5
10
-6
10
0
2
4
6
8
10
12
SNR( dB)
SNR  ( RP)
2
/ 2 Rb N 0
14
 OOK shows the highest
power requirement to
achieve a desirable
SER
Results and Discussion
Unequalized (Rb = 150Mbps, Drms = 5ns)
 Unequalized OOK requires
~27dB more SNR compared to
LOS link at SER of 10-5
0
10
Unequalized DPIM
-1
10
Unequalized PPM
-2
SER
10
 For high values of normalized
delay spread increasing the
optical power will not improve
error performance
-3
10
-4
10
-5
10
LOS
-6
10
0
5
10
15
20
25
SNR( dB)
30
35
40
 PPM suffers the most severely
in a diffuse link because of the
short pulse duration
Results and Discussion
OOK performance (Rb = 150Mbps, Drms = 5ns)
0
10
 ANN equalizer and linear
equalizer shows identical
performance
-1
10
-2
SER
10
 Power penalty is ~6.6 dB
compared to LOS links at
SER of 10-5
-3
10
-4
10
-5
10
-6
10 0
5
10
15
20
25
SNR( dB)
30
35
 SNR gain is ~ 20 dB
compared to unequalized
40
performance at SER of
10-5
Results and Discussion
ANN Equalizer (Rb = 150Mbps, Drms = 5ns)
 Performance of equalized DPIM
and PPM is better than OOK
even in highly dispersive channel
0
10
-1
10
-2
 DPIM show the best SER
performance.
SER
10
-3
10
-4
10
ANN Equalized
-5
10
LOS
-6
10
0
5
10
SNR( dB)
15
20
 Power penalty is ~14.3dB, 9.2dB,
6.7dB for equalized PPM, DPIM
and OOK compared to
corresponding LOS performance
for a SER of 10-5 .
Results and Discussion
ANN Equalizer (Rb = 150Mbps, Drms = 1, 2, &10 ns)
10
10
SER
10
10
10
10
10
0
 Equalized PPM shows the
best performance in less
dispersive channel (Drms<2)
-1
-2
 Equalized DPIM shows the
best SER performance in
highly dispersive channel
(Drms >2)
-3
-4
-5
10ns
1ns
2ns
-6
0
5
10
15
SNR(dB)
20
25
Wavelet-AI Receiver
XPavg
Input
bits
Output
bits
Transmitter
filter g(t)
Slicer
X
Artificial
Intelligence
Diffuse IR
channel h(t)
Wavelet
Analysis
R
noise
n(t)
X
+
ADC
Antialias
(LPF)
 Signal decimated into 3 bit sliding windows.
 Each window is transformed into wavelet
coefficients by the CWT process.
 The coefficients are passed to the neural
network for classification.
Signal Sample ‘The Window’
 For OOK signal decimated into
3 bit windows.
1
2
3
3 bit sliding window
 Each window is processed into
wavelet coefficients by the
continuous wavelet transform
(CWT).
8
Simulation Results Multipath Propagation 3
Equalised traditional receiver
architecture & Wlt-AI
reference (OOK RZ)
Equalised traditional receiver
architecture & Wlt-AI
reference (PPM)
Normalised to: 2.5Mb/s for BER 10-6 OOK RZ
Conclusions
 Artificial neural network as an equalizer shows
similar error performance to the linear equalizer
 Equalized PPM shows the best performance in less
dispersive channel while DPIM shows the best error
performance in highly dispersive channel
 Power penalty for equalized OOK is ~11.5 dB in
highly dispersive channel (Drms = 10 ns) at high data
rate of 150Mbps making it feasible for practical
implementation.
Issues and Future Works



Higher sampling rate (at least 8 samples per bit)
Hardware complexity
The need for parallel processing, at the moment

Adaptive error control decoding using neural
network.
Combine equalization and decoding as a single
classification problem
Wavelet network for equalization and decoding



Development of high performance pointing, acquisition,
and tracking.
Thank you!