Digital Pulse Interval Modulation for Optical Wireless
Communications
Z. Ghassemlooy, A. R. Hayes
Electronics Research Group,
School of Engineering,
Sheffield Hallam University,
Pond St., Sheffield, S1 1WB,
U.K
E-mail:
Z.F.Ghassemlooy@shu.ac.uk
N. L. Seed
Electronic Systems Research
Group, Dept. of Electronic
and Electrical Eng.,
University of Sheffield,
Mappin St., Sheffield, S1
3JD, U.K
E. D. Kaluarachchi
Bell Laboratories,
67 Whippany
Road, Whippany,
NJ 07981, USA.
Abstract
This paper presents a study of digital pulse interval modulation (DPIM) for optical wireless
communications using intensity modulation with direct detection. Expressions are presented for the
code characteristics, error probability and power spectral density. Practical results are given for an
experimental system. We have shown that DPIM has higher transmission capacity compared with
digital pulse position modulation (DPPM) and is less complex to implement.
1. Introduction
In recent years, optical wireless communications has emerged as a suitable alternative to radiofrequency transmission for short-range indoor data communications. A Diffuse system relies on
reflections from the ceiling and other reflectors within a room [1], and is susceptible to ambient light
noise, high signal attenuation, and intersymbol interference caused by multipath propagation. These
factors drive the requirement for emitting high optical power levels. However, optical wireless
transceivers are subject to eye safety regulations, which limit the average optical power level that can
be emitted. Furthermore, power consumption must be kept to a minimum in battery-powered portable
devices, which is another constraining factor. Thus, a power efficient modulation scheme is desirable in
order to maximise the peak to average power level. Usually, diffuse transceivers require the use of
large area photodetectors. However, the high capacitance associated with large area photodetectors
limits the receiver bandwidth and hence, the spectral efficiency of the modulation scheme is also an
important consideration.
DPPM has been used widely in optical wireless communication systems, primarily due to its power
efficiency, and has been adopted by the IEEE 802.11 working group for the infrared physical layer
standard [2]. However, the use of DPPM increases system complexity compared with on-off keying
(OOK), since both slot and symbol synchronisation is required in the receiver. As a potential
alternative to DPPM, DPIM is a technique which displays a higher transmission capacity by virtue of
its anisochronous nature, and requires no symbol synchronisation, since each frame is initiated with a
short duration pulse. The following sections outline code properties, spectral characteristics, error
probability analysis and system implementation of DPIM.
2. DPIM Code Properties
2.1 Frame Structure
Digital pulse interval modulation is an anisochronous pulse time modulation technique in which data is
encoded as a number of discrete time intervals, or slots, between adjacent pulses. The symbol length is
variable and is determined by the information content of the symbol. In order to avoid symbols in
which the time between adjacent pulses is zero, an additional guard slot may be added to each symbol
immediately following the pulse. Thus, to encode M bits of data, the minimum and maximum symbol
lengths are 2Ts and (L+1)Ts respectively, where Ts is the slot duration and L = 2M.
In DPIM, an M bit symbol is represented by a pulse of constant power Ps in one slot, followed by k
slots of zero power, where 1 ≤ k ≤ L. This may be expressed as:
P
S DPIM (t ) =  s
0
nTc ≤ t < ( n + 1)Tc
(1)
( n + 1)Tc ≤ t < ( n + k + 1)Tc
The mapping between source data and transmitted symbols for 4DPPM and DPIM is outlined in Table
1.
Table 1: Comparison of symbol mapping for 4DPPM and DPIM
Source Data
4DPPM
00
01
10
11
1000
0100
0010
0001
DPIM
(without guard slot)
1
10
100
1000
DPIM
(with guard slot)
10
100
1000
10000
For a DPIM system encoding M bits of data per symbol, if the slot duration is chosen such that the
maximum symbol duration is equal to the time taken to transmit M bits of data using OOK, T M,OOK, the
slot duration is given by:
Ts =
TM ,OOK
L +1
(2)
Assuming that the symbol length is random and uniformly distributed between 2 and L+1 slots, the
average bit rate, Rb is given by:
Rb =
M
n Ts
where n =
(3)
L+3
is the mean symbol length in slots.
2
The increase in capacity of DPIM compared with isochronous schemes such as DPPM is dependant on
the statistical properties of the source data. If all symbols are equiprobable, the capacity of DPIM tends
towards twice that of PPM as M increases. For optical wireless communications, this increased
capacity can be utilised in a number of ways. The same average data rate can be supported with
approximately half the slot frequency of DPPM, thereby improving the bandwidth efficiency of the
modulation scheme, though the power efficiency would be reduced due to the increased duty cycle.
Alternatively, a higher number of bits per symbol could be supported without an increase in slot
frequency, thereby improving the power efficiency [3]. Finally, the increased capacity could be utilised
by introducing some redundancy into the code, thereby giving error detection and possibly correction
capabilities to the code and subsequently allowing a lower average power to be used in order to achieve
the same bit error rate as DPPM.
2.2 Spectral Properties
A theoretically infinite length DPIM pulse train can be represented by [4]:
x(t ) =
k −1



−
+

a
g
t
T
2
k
S m 
∑
∑
k 
s
k = −∞
m = −∞



∞
(4)
where g(t) represents the pulse shape of duration γTs (0 < γ ≤ 1) and amplitude ak, Ts is the slot duration
and S (Sm ∈ S) is the stochastic random data sequence representing data coded into DPIM. The power
spectral density (PSD) model of the DPIM coding scheme is given as [4],
2
k −1

 



j
2
fT
2
k
S
−
π
+
s
m 
Ln

∑

 
1


m=0
(
)
Sp ( f ) =
G
f
e


∑
Ln


k =0

Ts  2Ln + ∑ Sm  


m=0


(5)
Where G(f) is the DPIM pulse shape transform, L is the length of the truncated data frame sequence
and Ts is the slot duration.
Since, as with DPPM, DPIM signals with non-return-to-zero rectangular pulse shapes do not contain a
frequency component at the slot clock fundamental frequency, a non-linearity must be introduced in
order to generate the desired frequency component which can then be locked on to by a phase lock loop
(PLL) [5].
3. System Implementation
The optical wireless system is shown below in Fig. 1:
M
Bits
Data
Source
DPIM
Modulator
Optical
Source
(a)
Clock
Recovery
Amplifier
Matched
Filter
Threshold
Detector
DPIM
Demodulator
M
Bits
Data
Sink
(b)
Fig. 1. System block diagram: (a) transmitter, and (b) receiver.
The optical transmitter consists of an array of 6 LEDs, emitting a total radiant power of 200 mW at a
wavelength of 875 nm. All the LEDs are pointing vertically upwards towards the ceiling. The receiver
consists of a 1 cm2 silicon pin photodiode, followed by a 14 kΩ transimpedance preamplifier. A
matched filter is employed to maximise the carrier-to-noise ratio prior to the decision making process.
The threshold detector samples the incoming signal every slot, and compares it with the threshold level.
A logic ‘1’ is assigned to a particular slot if the sampled signal exceeds the threshold level, otherwise a
logic ‘0’ is assumed. The clock recovery circuit uses a PLL to extract the slot clock from the incoming
DPIM pulse train, which is used to generate the sampling points for the threshold detector, and count
the number of slots between adjacent pulses in the demodulator circuit.
The system was set up in Laboratory room with dimensions 6.5 m x 7 m and a ceiling height of 3 m.
The transmitter and receiver were positioned 1m apart on a bench 0.9m high in the centre of the room.
Fig. 2 shows the eye diagrams of the received signal, for slot frequencies ( fs ) of 1.5 MHz and 6 MHz.
For both cases the transmitted data consisted of 4000 random symbols, each encoding 4 bits of data.
The measured irradiance at the receiver was ~ -37 dBm/cm2. The 6 MHz system is currently being
investigated.
(a)
(b)
Fig. 2. Eye diagram: (a) fs = 1.5 MHz, and (b) fs = 6 MHz.
4. Error Probability Analysis
There are three mechanisms which give rise to errors in the received DPIM pulse train, these being
erasure, false-alarm and wrong-slot errors. An erasure error occurs when noise attenuates the pulse
such that, at the sampling instant, the amplitude falls below the threshold level. A false-alarm error is
the opposite of an erasure error, and arises when noise exceeds the threshold level at the sampling
instant. A wrong-slot error occurs when, due to dispersion, a pulse is detected in the preceding or
succeeding time slot. The average symbol error probability of DPIM is given by [4]:
2

T 


− s  



1
CNR 
CNR
CNR  
 ti  
M


Pse = 2 erfc (1 − α )
 + erfc α − e

 + 2 − 1 erfc  α


4
2 
2 
2 





 

(
)
(6)
where α is the threshold level, CNR is the carrier-to-noise ratio at the receiver, and it is the received
full-width half-maximum pulse width of the received Gaussian shaped pulses. The first, second and
third terms represent erasure, false alarm and wrong slot errors respectively.
Whilst DPPM is subject to the same three error sources as DPIM, the loss of data as a result of such
errors is not the same. In DPPM, should any of the above errors occur, only the symbol in which the
error occurs is affected. In DPIM, an erasure error would effectively combine two adjacent symbols
into one longer symbol, resulting in loss of data from both symbols. A wrong slot error would also
affect two adjacent symbols, whilst a false alarm error would effectively split a symbol into two shorter
length symbols.
5. System Implementation Issues
As discussed in section 2.1, a major characteristic of DPIM is that it has a variable symbol length and
hence, the time required to transmit a packet containing a fixed number of bits is not constant. Thus,
when used in a communication network environment, where packets of data are transmitted, DPIM
encoded symbols may result in two possible scenarios:
(i) when the majority of symbols in a packet are longer than the average length, the time required to
transmit that packet is relatively long. In order to accommodate this situation, the transmitter buffer
would need to be large, which is inefficient from a hardware point of view, and (ii) conversely, packets
containing mainly short length symbols will be transmitted relatively quickly, thus resulting in
transmitter buffer underflow and receiver buffer overflow, or a temporary violation of eye safety
requirements [3].
In order to avoid the problems outlined above, it is necessary to employ some form of coding scheme
to limit the variation in packet size whilst still maintaining the increase in capacity over isochronous
modulation techniques. A dual mapping technique could be employed, whereby source bits may be
mapped to symbols either normally or in reverse fashion, where the complement of the source data is
mapped. Each packet would be formed of symbols either normally or reverse mapped, whichever
yields the shortest packet length. Empty slots could then be added to the end of a packet until it reaches
the mean packet duration [3].
6. Conclusions
This paper has outlined the basic principles and characteristics of DPIM. The two main advantages of
DPIM over DPPM, making it an attractive modulation technique for optical communications, are the
increase in capacity and the fact that symbol synchronisation is not required, resulting in a simplified
receiver structure. As mentioned in section 2.1, the increased capacity can be used to improve either the
bandwidth efficiency or average power efficiency of the system. Expressions have been presented to
characterise DPIM and its error probability performance based on a receiver employing a matched
filter. Practical results are given in the form of eye diagrams for a prototype optical wireless system,
thus highlighting the potential of this new scheme for optical transmission systems. Finally, problems
which may arise when using DPIM in network environments, due to its non-uniform symbol structure,
are discussed.
Acknowledgements
A. R. Hayes is financially assisted by a Sheffield Hallam University - University of Sheffield joint
studentship.
References
[1]
J.R. Barry, “Wireless Infrared Communications,” Kluwer Academic Publishers, 1994.
[2]
A.J.C. Moreira, R.T. Valadas and A.M. de Oliveira Duarte, “Performance Evaluation of the
IEEE 802.11 Infrared Physical Layer,” Proc. First Int. Symp. on Communication Systems and
Digital Signal Processing,” Sheffield, U.K., pp. 10-15, 6-8 April, 1998.
[3]
D. Shiu and J.M. Kahn, “Differential Pulse Position Modulation for Power-Efficient Optical
Communication,” available at: http://alder.eecs.berkeley.edu/~jmk/res.areas/irwc.html
[4]
E.D. Kaluarachchi, “Digital Pulse Interval Modulation for Optical Communication Systems,”
PhD thesis, Sheffield Hallam University, U.K., 1997.
[5]
F. M. Davidson and X. Sun, “Slot Clock Recovery in Optical PPM Communication Systems
with Avalanche Photodiode Photodetectors,” IEEE Transactions on Communications, Vol. 37,
No. 11, pp. 1164-1171, November 1989.