MIC-CPE2010, Jordan

Optimizing the Performance of Digital Pulse Interval Modulation with Guard Slots for

Diffuse Indoor Optical Wireless Links

Z. Ghassemlooy and S RajbhandariOptical Communication Research Group,

School of CEIS, Northumbria University, Newcastle upon Tyne, UKhttp://soe.unn.ac.uk/ocr/

MIC-CPE2010, Jordan

Outline

Indoor optical wireless communication Modulation techniques DPIM Techniques to reduce ISI Decoding Scheme for DPIM(nGS) Scheme

Threshold decoding Hybrid decoding scheme Maximizing the likelihood (ML) of a pulse

Results and discussions Conclusions

MIC-CPE2010, Jordan

Optical Wireless System: Overview

1 M. Kavehrad, Scientific American Magazine, July 2007, pp. 82-87.

Typical optical wireless system components

Optical wireless connectivity 1

Uses light beams (visible and infrared) propagating through the atmosphere or space to carry information.

Optical transmitter- Light emitting diodes- Laser diodes

Optical receiver- p-i-n photodiodes- Avalanche photodiodes

Links- Line-of-sight (LOS)- Non-LOS- Hybrid

MIC-CPE2010, Jordan

Digital Modulation Schemes

On-off keying (OOK)

Pulse position modulation (PPM)

Digital pulse interval modulation (DPIM)

Dual-header pulse interval modulation (DH-PIM)

Subcarrier modulation

MIC-CPE2010, Jordan

Digital Pulse Modulation Schemes

DPIM

MIC-CPE2010, Jordan

The DPIM Scheme An anisochronous modulation technique A symbol is composed of a pulse of one slot duration followed

by a series of empty slots:

where dj-1 is j empty slot(s), j = 0, ..., D and D is the decimal value of ai.

DPIM signal is defined as:

p(t) - rectangular pulse shape, Ts - slot duration bi - set of random variables representing a pulse/no pulse

in the nth Ts

𝑺ij=1 {𝒅 j−1 ­}

s (t )= ∑i=−∞

∞

bi p ( t −i T s )

MIC-CPE2010, Jordan

Why DPIM ? An excellent compromise

between the bandwidth and the power efficiencies.

Higher bandwidth efficiency than PPM.

Higher power efficiency than OOK.

Easy to implement compared to more complex modulation scheme like DH-PIM.

2 3 4 5 6 7 80

5

10

15

20

25

30

Bit resolution

Nor

mal

ized

ban

dwid

th re

quir

emen

t

PPM

DH-PIM1

DPIM

DH-PIM2

OOK

2 3 4 5 6 7 8-16

-14

-12

-10

-8

-6

-4

-2

0

Bit Resolution, M

Nor

mal

ized

Pow

er R

equi

rem

ent (

dB)

DH-PIM2

PPM

DH-PIM1

DPIM

MIC-CPE2010, Jordan

Indoor Optical Wireless Links

The key issues:

- Eye safety- shift from 900 nm to 1550 nm - eye retina is less

sensitive to optical radiation- power efficient modulation techniques

- Mobility and blocking- is a problem in diffuse configurations (i.e. Non-

LOS), thus resulting in: - reduced data rates- increased path loss- multipath induced inter-symbol-interference (ISI)

MIC-CPE2010, Jordan

Indoor OWC - Diffuse Links Pulse spreading due to the

different path delays leading to intersymbol interference (ISI)

ISI is the limiting factor in achieving higher data rates

Diffuse links are characterised by RMS delay spread

The impulse response in the Ceiling bounce model is:

LOS

Diffuse

Diffuse shadowed

LOS shadowed

)(1.06

)( 7

6

1.0tu

t

Dth

rms

rms

D

where u(t) is the unit step function

Fig. Impulse response of indoor optical wireless channel

MIC-CPE2010, Jordan

Techniques to Reduce ISI Maximum likelihood sequence detection

The optimum solution to reduce ISI Difficult to implement due to high complexity and large delay Practical implementation is not feasible for DPIM due to variable

symbol length

Equalization Trade-off between complexity and performance Preferred due to lower complexity compared to MLSD Channel estimation is necessary

Guard slots (GSs) Simple to implement without additional complexity Effective in moderately dispersive channel Ineffective in highly dispersive channel

MIC-CPE2010, Jordan

DPIM with Guard Slots to Reduces ISI The postcursor slot immediately

following a pulse is most severely effected due to ISI.

Adding GSs immediately following a pulse can be effective in reducing the ISI.

Clear overlapping in the constellation of DPIM(0GS). Hence difficult to assign a fixed threshold level.

The constellation of 0s and 1s are clearly separated for DPIM(1GS). However, distance is clearly reduced.

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

Fig. Scatter plots of received signals at DT = 0.3 for DPIM(0GS). red= 0, blue =1

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

Fig. Scatter plots of received signals at DT = 0.3 for DPIM(1GS). red= 0, blue =1

MIC-CPE2010, Jordan

Decoding Scheme for DPIM(nGS) Scheme

DPIMencoder

ai bi ib̂Transmitter filterp(t)

n(t)

v(t)

R

z(t)Decoder

1/Ts-DPIM

y(t) yiMatched filter r(t)

x(t) Multipathchannelh(t)

s(t) φ(t)

avgDPIMPL

Decoding schemes:

Threshold decoding

Hybrid decoding scheme

Maximizing the likelihood (ML) of a pulse

Fig. The block diagram of the DPIM system.

MIC-CPE2010, Jordan

Threshold Decoding

A threshold level set at half the peak amplitude is non-optimum in diffuse channel.

ISI reduces the minimum Euclidean distance dmin.

Threshold level needs to be adjusted accordingly.

The optimum threshold level is given by:

where ci is the channel taps

Error probability can be approximated as :

-8 -6 -4 -2 0 2 4 610

-6

10-5

10-4

10-3

10-2

10-1

100

SNR (dB)SE

R

8-DPIM (D

T = 0.01, simulation)

8-DPIM (DT = 0.01, theory)

8-DPIM (DT = 0.1, simulation)

8-DPIM (DT = 0.1, theory)

16-DPIM (DT = 0.01, simulation)

16-DPIM (DT = 0.01, theory)

16-DPIM (DT = 0.1, simulat ion)

16-DPIM (DT = 0.1, theory)

The predicted and simulated SER against the SNR for the 8 and 16- DPIM schemes at Dt­= 0.01 and 0.1.

𝛼 th=(c− 1+c1 )+dmin

2=c−1+c0+c1

2

P seDPIM ≤Q(√ c0− (c−1+c1 ) E2N 0

)

MIC-CPE2010, Jordan

Hybrid Decoding Scheme Soft decoding is difficult to implement

due to non-uniform symbol length. Valid DPIM(1GS) symbol always has a 010

sequence except for the all zero sequence. Unique slot sequence in DPIM(1GS) can

be exploit for hybrid decoding. The decoding algorithm can be summarised

as:

Valid DPIM(1GS) sequence

00000100100100

P ( b̂i=1|bi=1)=P ( y i≥𝛼∨b i=1 ) P ( y i> y i+1 ) P ( y i> y i −1 )

if yi­>&­yi> (yi-1,­yi+1)else

The probability of correctly decoding a pulse is given by:

MIC-CPE2010, Jordan

Maximizing the Likelihood (ML) of a Pulse

In DPIM scheme with n GSs, a pulse should always be followed by n empty slots.

Taking two slots into consideration (00, 01, 10) are the only valid DPIM(1GS) sequence.

The approach taken here is to maximize a-Posterior probability of a pulse.

i.e. If the posterior probability of sequence (10) is greater than posterior probabilities of (00) and (01) sequence, decode the bit sequence as (10) else decode present bit as 0.

if

;else .

MIC-CPE2010, Jordan

Results and Discussions A fixed threshold level of 0.5

demonstrates the worst performance.

The ML detection scheme offers the best performance.

All other decoding approaches show improved performance compared to the DPIM (0GS).

The optimum threshold decoding offer significantly improved performance compared to a fixed threshold decoding.

10-2

10-1

100

-2

0

2

4

6

DT

NO

PR

NGS

1 GS ( = 0.5)

1GS (opt)

1GS (Hybrid)

1GS (ML)

Fig. The NOPRs against the normalized delay spreads for 8-DPIM (0 &1GS) for different decoding algorithms and a SER of 10-6.

MIC-CPE2010, Jordan

Results and Discussions

Hybrid decoding offers improved performance compared to the optimum threshold level.

The advantage of the ML detection scheme can be observed at higher values of DT .

A difference of ~ 3.4 dB and ~ 2.8 dB can be observed between the ML detection and the hybrid decoding at DT = 0.4 for 8 and 16-DPIM(1GS), respectively.

10-2

10-1

100

-4

-2

0

2

4

6

DT

NO

PR

NGS

1 GS ( = 0.5)

1GS (opt)

1GS (Hybrid)

1GS (ML)

Fig. The NOPRs against the normalized delay spreads for 16-DPIM (0 &1GS) for different decoding algorithms and a SER of 10-6.

MIC-CPE2010, Jordan

Conclusion A number of decoding approaches has been proposed and

studied for DPIM(1GS) The decoding algorithm exploits the unique slot sequence

of DPIM(1GS) The fixed threshold based decoding schemes is the non-

optimum for diffuse links. A hybrid decoding scheme surpasses the performance of

the optimum thresholding. The ML decoding of a pulse offered the best

performance. The system complexity using the ML detection scheme is

not significantly higher than that of a threshold detector, ML detection is practically recommendable.

MIC-CPE2010, Jordan

Questions?

Thank you!