Optical Receivers and

4 Transmission Performance
Fundamentals

Signals and noise are important parameters to determine the quality of any engineering system,
especially in the telecommunications field. The modulated signals under different modulation for-
mats will be treated in subsequent chapters. This chapter presents an overview of the noise process
and mechanism in terms of equivalent noise power and contribution to the optoelectronic process at
the optical receiver. Further, the integration of the noise and the statistical process for determination
of the receiver sensitivity under transmission of different modulation formats are described together
with modeling techniques in the MATLAB and Simulink platform. Statistical analyses are also
given for the evaluation of the transmission performance, especially when there are no uniform
probability distributions.

4.1 INTRODUCTION
Optical detection and the noise interference in such processes are critical to the performance of
optical communications systems. A fundamental understanding of the noise and sensitivity of the
receiver, the minimum optical power available at the photodetector, requires evaluation of the trans-
mission performance of different modulation formats. Optical receivers have evolved from binary
digital direct detection to coherent mixing of signals and local lightwave oscillator and detection
with optical preamplification. This chapter presents an overview of the noise process and mecha-
nism in terms of the equivalent noise power and contribution.
A schematic diagram of a single-channel DPSK system is illustrated in Figure4.1.
As the bit or symbol rate of the optical transmission systems is increased, the demand for mod-
eling is intense, especially for a modeling platform that can simulate the real physical photonic
subsystems. Inorder to enhance the effective transmission capacity with minimum renovation of
the photonic and electronic subsystems, there are two possible solutions: (1) increasing the base
bit rate or (2) employing a multilevel modulation technique such as the M-ary amplitude and/or
phase shift keying [1]. Thelatter solution would be preferred because for the same as the bit rate,
the multilevel offers a much higher transmission capacity (in bits/sec.) but with the same baud rate
without increasing the complexity of the transmitters and receivers as well as the dispersion toler-
ance of the transmission fibers which can be single-mode fiber types ITU G652 or G655. However
due to the nonlinear threshold of the fiber the highest level is limited and so an increase in the level
of the multilevel modulation system will lower the opening of the eye diagram of sublevels. Thus
this may reduce bit error rate.
One possible technique that could offer insights into possible solutions that facilitate the simula-
tion of ultra-high-capacity and ultra-high-bit-rate transmission systems is to develop a comprehensive
modeling platform. Further, the modeling platform should take advantage of any user-friendly soft-
ware platform that is popular, easy to use, and easy to extend. This platform would offer the research
community of optical communication engineering a basis for extension and enhance the linkages
between research groups.

135
136 Advanced Digital Optical Communications

Optical Optical dispersion-

transmission fiber compensating fiber
Optical Optical
transmitter receiver
D
M E
U M
X U
Optical amplifier X
. .
. .
. .
N spans

FIGURE4.1 Schematic diagram of a single-wavelength channel optical transmission system.

Thus, one of the principal objectives of this chapter is to present the development of MATLAB
and Simulink 7.0 [2] for computer experiments in optical transmission systems under advanced
modulation formats, especially the amplitude and/or phase shift keying modulation of binary or
multilevel. Indeed, duobinary in association with an appropriate low-pass filter can offer multiple
rates much higher than the basic rate. The advantage of duobinary is that the detection is much
simpler than that of DPSK, DQPSK, or M-ary PSK, in which the receiver is a direct detection type
receiver. In the photonic domain, the combination of phase and amplitude is taken into account for
coding modulation schemes that involve the tri- or higher-order levels. These features can be imple-
mented with ease on the Simulink platform.
We base our system bit rate on a 40 Gbps per channel. The modulation formats of binary RZ,
NRZ, and CSRZ-ASK-DPSK are demonstrated with the modeling of the transmission over stan-
dard single-mode optical fiber (SSMF), as well as the photonic decoding of this optical signaling.
Thetransmission performance of some optical systems is also given in terms of the bit error rate
(BER), the -factor (usually known as Q factor), and the receiver sensitivity.
In this chapter, we present, to the best of our knowledge, the first MATLAB and Simulink
platform simulation test bed for modeling of the transmission of the amplitude and differential
phase modulation incorporating RZ, NRZ, and carrier-suppressed RZ formats for advanced high-
capacity and long-haul optical fiber transmission systems. A novel, modified fiber propagation algo-
rithm has been developed and optimized to minimize the simulation processing time and enhance
its accuracy. The performance of optical transmission systems can be automatically and accurately
evaluated with various methods. Analysis and methodologies for future development of the simula-
tion test bed are also presented.
In recent years, the capacity and reach of optical transmission systems have dramatically
increased owing to the accelerating growth of data usage demand (Internet, peer-to-peer
network,).Therefore, the necessity of upgrading of current dense wavelength division multiplex-
ing (DWDM) 10 Gbps systems to DWDM 40Gbps or even higher bit rates becomes crucial to tele-
communications service providers. Accordingly, several modulation formats have been proposed
and investigated as the alternatives to the current onoff keying (OOK) intensity modulation, which
is severely degraded at high bit rates owing to dispersion and nonlinear effects of the transmis-
sion fibers. Apart from the requirements of robustness to the transmission impairments, the cost-
effectiveness of system upgrades is also significant. Among the candidates, differential binary and/
or quadrature phase shift keying (DPSK/DQPSK) have recently attracted much attention owing to
the following advantages: (1) a 3 dB improvement on receiver sensitivity (if the balanced receiving
technique is used) [1,2]; (2) high tolerance to fiber nonlinearities, especially to intrachannel nonlin-
ear effects, cross-phase modulation (XPM), and four-wave mixing (FWM) [2,3]; (3) superior spec-
tral efficiency (DQPSK), and hence high tolerance to optical filtering [3]; and finally, (4)advantages
in all-optical networks incorporating optical add-drop multiplexers or optical cross-connects [4].
Recently, several experimental demonstrations of DPSK/DQPSK long-haul transmission DWDM
Optical Receivers and Transmission Performance 137

systems for 10 and 40 Gbps have been reported [5,6]. Therefore, a simulation test bed is necessary for
detailed design, investigation, and verification of the benefits and shortcomings of these advanced
modulation formats on fiber-optic transmission systems. In this chapter, we present, to the best of
our knowledge, the first MATLAB-and Simulink-based simulation package for photonic DWDM
systems. The simulator is still in the first phase of the development, with the focus on single-channel
systems, and it is being continuously improved and updated.
This chapter is organized as follows: (i) in Section4.2, receiver noise sources are outlined; noise
analyses and performance of basic binary digital optical receivers are then given in Section4.3;
(ii)Section4.3 presents the architecture and operational principles of the simulator; (iii) Section4.4
presents simulation results of the up-to-date simulator, and the corroboration of the developed
models in comparison with experimental results are presented in Section 4.5 for both ASK and
DPSK modulation formats. Finally, concluding remarks are presented.

4.2 DIGITAL OPTICAL RECEIVERS

This section gives fundamental concepts of receiver structure, sensitivity, and noise. It is mainly
focused on direct detection, which would be integrated into more complex receiver structures for
different modulation formats.

4.2.1 PHOTONIC AND ELECTRONIC NOISE

Before proceeding to the system calculations to determine the performance of optical receivers, it is
necessary to consider the noise generated in the photodetector and the preamplifier front end. One
can, in fact, consider investigating the system calculation and returning back to the noise calcula-
tion provided that they are taking either the total equivalent noise spectral density at the input of the
detector or its noise figure. This section describes all noise mechanisms related to the photodetec-
tion process, including the electronic noise associated with the receiver.

4.2.1.1 Electronic Noise of Receiver

At the receiver, noise sources that are superimposed onto the electrical signals detected after pho-
todiodes include electrical shot noise, dark current noise, and the equivalent noise current as seen
from the input of the electronic preamplifier. The electronic noise sources, represented as the square
of current, can be found by using the noise spectral density as a function of frequency and then
integrating over the bandwidth of the system. They can be given as

2
iNshot = 2qPin Be (4.1)

2
iND 2qI D Be (4.2)

2
iEA ( I eq )2 Be (4.3)

4.2.1.2 Shot Noise

Shot noise and thermal noise are the two most significant noises in optical detection systems. Shot
noise is generated by either a quantum process or electronic biasing. The noise is specified in noise
spectral density, that is, the square of the noise current per unit frequency. Thus, the noise spectral
density is to be integrated over the total amplifier bandwidth to obtain the equivalent noise currents.
Electrical shot noise is generated by the random generation of streams of electrons (current).
In optical detection, shot noise is generated by (1) biasing currents in electronic devices and
(2)photocurrents generated by the photodiode.
138 Advanced Digital Optical Communications

4.2.1.2.1 Biasing Current Shot Noise

Any biasing current I has a spectral current density SI given by

d (iI2 )
sI 2qI in A2 /Hz (4.4)
df

where q is the electronic charge. The current iI represents the noise current generated due to the
biasing current I.

4.2.1.2.2 Quantum Shot Noise

The average current < is2 > generated by the photodetector by an optical signal with an average
optical power Pin is given by

d is2
sQ 2q is2 (4.5)
df

This is the signal-dependent noise, and it is a unique feature of optical communications. When
the avalanche photo-detector (APD) is used, the noise spectral density is given by

d is2
sQ 2q is2 Gn2 (4.6)
df

It is noted here that the dark currents are generated in the photodetector is a natural phenom-
enon in semiconductor material operating at room temperature. This dark current noise must be
included in the total equivalent noise current as referred to the input. These currents are generated
even in the absence of an optical signal. These dark currents can be eliminated by cooling the
photodetector to at least below the temperature of liquid nitrogen (77K).

4.2.1.3 Thermal Noise

At a certain temperature, the conductivity of a conductor varies randomly. The random movement
of electrons generates a fluctuating current even in the absence of an applied voltage. The thermal
noise of a resistor R is given by

sR =
d iR2( ) = 2k T B
(4.7)
df R

where k B is the Boltzmanns constant, T is the absolute temperature (in K), R is the resistance in
ohms, and iR denotes the noise current due to R.

4.2.1.4 ASE Noise of Optical Amplier

The following formulation accounts for all noise terms that can be treated as Gaussian noise due to
the optical amplifier.

NASE = mnsp hv(Gop 1) Bo (4.8)

where Gop = amplifier gain; nsp = spontaneous emission factor; m = number of polarization
modes (1 or 2); Bo = mean noise in bandwidth; optical signal-to-noise ratio (OSNR) at the output
ofEDFA.
Optical Receivers and Transmission Performance 139

4.2.1.5 Optical Amplier Noise Figure

The amplifier noise figure (NF) is defined at the output of the optical amplifier as the ratio of the
output OSNR at the output to that OSNR at the input of the EDFA.

OSNR in
FN = 2nsp for Gop 1 (4.9)
OSNR out

Modern optical amplifiers are optimized, and nsp reaches unity; so the NF can reach 3 dB.

4.2.1.6 Electronic Beating Noise

When an optical preamplifier is used in association with an optoelectronic receiver front end, then
there would be noise generated due to: the signal-dependent shot noise, beat noise from the beat-
ing of the electronic currents generated from the signals, random ASE noise, and the beating noise
between the ASE electronic currents. Usually, the signal ASE noise dominates this noise process.
The beating noise generated from the beating between the signal current and the ASE noise current,
dominates the detection process in an optical preamplifying optoelectronic receiver and is given by

2 Be
isig-ASE = 2(qGop )2 (2nsp ) Pin (4.10)
h

where Be is the bandwidth of the electronic amplifier system.

Under direct detection, the ASE noise appearing at the input of an electronic preamplifier would
follow square detection; that is, the noise vector would be taken with its absolute value, then squared,
and multiplied by the responsibility of the detector to obtain the spectral density noise current.
On the contrary, under coherent detection, the ASE noise is superimposed on the electric field of the
local oscillator and the signals. The local oscillator amplitude is much higher than that of the signal,
and the beating noise between the local oscillator and the ASE noise dominates the noise source pre-
sented at the input of the electronic preamplifier. This is the significant difference between the optically
amplified and the nonoptically amplified receiver for direct and coherent optical receivers, respectively.

4.2.1.7 Accumulated ASE Noise in Cascaded Optical Ampliers

Long-haul optical communications would be structured with dispersion managed and loss equalized
through tens or hundreds of spans. Error-free transmission over several thousands of kilometers with-
out repeaters has been demonstrated. It is critical to account for the accumulated noise sources that
ultimately limit the transmission distance. The amplified spontaneous emission (ASE) noise accumu-
lated is the principal noise source that has been built up over several optical amplifiers in cascade that
would be originally from the photons generated randomly under spontaneous emission. The spontane-
ous noise factor is given as nsp. Practical EDFAs have now reached a mature stage with nsp reaching
unity. In this section, an effective noise factor at the end of the transmission spans or at the input of the
optical receiver is derived.
Let the linear loss factor of the ith transmission fiber be L,I and Gi be the linear gain factor of
the optical amplifier in the span; then a recursive relation between the noise of the first stage and ith
stage can be written as
nsp,1 = nsp,1
(4.11)
nsp,i 1Gi 1
nsp,i = + nsp,i for 2 i N
L,i 1

where nsp,1 is the noise as seen from the input of the amplifier, and nsp,1 is the equivalent noise factor
at the output of the amplifier. Then, the accumulated ASE noise over the transmission spans is given
140 Advanced Digital Optical Communications

by the sum of all the noise sources of the amplifiers. At the output of the final Nth stage, without
taking into account the other noise sources, the equivalent ASE accumulated noise is given as

=
nsp,N ni =1
sp,i

(4.12)
e GN
nsp,N
nsp,i = + nsp,i
L

under the assumption that the gain of optical amplifiers has equalized all the losses of transmission
and dispersion-compensating fibers (DCFs).
When an optical preamplified receiver and two optical filters are placed before and after the
optical preamplifier, the effective ASE noise is given by

nsp,NGN 2 (4.13)
nspe = + nsp
L 1

where 1 and 2 are the time constants of the two optical filters.
The NF of the equivalent accumulated noise is then given by

NF e 10Log10 nspe (4.14)

This is also the amount of degradation of the OSNR due to the accumulated ASE noise at the
front of the photodetector.

4.3 PERFORMANCE EVALUATION OF BINARYAMPLITUDE

MODULATION FORMAT
As described earlier, the noise of the optical receivers consists of the thermal noise and quantum
shot noise due to the bias currents and the photocurrent generated by the photodetector with and
without the optical signals. Thus, this quantum shot noise is strongly signal dependent. These noises
degrade the sensitivity of the receiver [1] and thus a penalty can be estimated. Another source of
interference that would also result in signal penalty is the intersymbol interference (ISI). The goal
of this section is to obtain an analytical expression for the receiver sensitivity of the direct detection
optical receiver ON/OFF keying (OOK) system.

4.3.1 RECEIVED SIGNALS

Assuming that the received signal power is

p(t ) = a h (t jT )
j
j p b (4.15)

The average output voltage at the output of the electronic preamplifier is thus given by

Tb / 2
1
vo = Gn j
aj
Tb hp (t jTb ) dt RI A (4.16)
Tb / 2
Optical Receivers and Transmission Performance 141

where Tb is the bit period, and hp(tjTb) is the impulse response of the system evaluated at each time
interval. RI is the input resistance of the overall amplifier of the system including both the front end
and the linear channel amplifier. It is assumed that the overall amplifier has a flat gain response
Aover the bandwidth of the system.

bo 0
aj t=0 = (4.17)
b1

with b 0 being the energy when a transmitted 0 is received, and b1 is the energy when a transmitted 1
is received. The sum over a number of periods is necessary to take into account the contribution of
adjacent optical pulses. We now have to distinguish between two cases when a 0 or a 1 is transmit-
ted and received.

4.3.1.1 Case 1: OFF or a Transmitted 0 Is Received

Using Equations 4.12 and 4.13, we have

bo
vo o
= voo = GRI A 0 (4.18)
Tb

2
with the total equivalent noise voltage at the output, vNTo is

2
vNTo 2
vNA 2
iNeq RI2 A2 (4.19)

where iNeq is the total equivalent current at the input of the electronic preamplifier. Appendix 2
(Section 4.12) and Annex 4 gives a method for estimating this noise current for any preamplifier
whose equivalent Y-parameters are known.

4.3.1.2 Case 2: ON Transmitted 1 Received

In this case, the average signal voltage at the output is received as

b1
vo 1
= v01 = Gn RI A (4.20)
Tb

with the total noise equivalent mean voltage at the output given by

2 2 2
vNT1 = voSN + vNA (4.21)

where v2osN is the signal-dependent shot noise. v2NA is the amplifier noise at the output and is given by

2 2
vNA iNeq R12 A2 B (4.22)

The signal-dependent noise is, in fact, the quantum shot noise and is given by

B
2
voSN =
o
2q is 1
Gn2 RI2 A2 df (4.23)
142 Advanced Digital Optical Communications

where B is the 3 dB bandwidth of the overall amplifier, <is>1 is the average photocurrent received
when a 1 was transmitted. This current can be estimated as follows

Tb / 2
is 1
= Tb

1

b Tb / 2
hp (t jTb ) dt (4.24)

or

b1
is 1
=
Tb
hp (t ) dt (4.25)

with a normalization that is
Equation 4.21 becomes

hp (t ) dt = 1.

b1
is 1
= (4.26)
Tb

4.3.2 PROBABILITY DISTRIBUTION FUNCTIONS

Though the optical systems under consideration are typical IM/DD systems, they are also suitable
for further calculations for modulation formats in digital optical communications. Further develop-
ment of statistical distributions can be referred to in Section4.4. In these systems, the optical energy
of each pulse period is equivalent to that of at least a few hundred photons. This number is large
enough so that a Gaussian distribution of the probability density is warranted. The probability den-
sity function (pdf) of a 0 transmitted and received by the optical receiver is thus given by

1 ( v0 v00 )2
p [ v0 0 ] = 2 1/2 exp 2 (4.27)
(2v NTO ) 2vNT0

and similarly for a 1 transmitted

1 ( v0 v01 ) 2
p [ v0 1 ] = 1/2 exp 2 (4.28)
(2 vNT1 ) 2v NT1

The total probability of error or a BER is defined as

BER = p(1) p(0 /1) + p(0) p(1/0) (4.29)

where p(1) and p(0) are the probabilities of receiving a 1 and a 0, respectively, and p(1/0) andp(0/1)
are the probabilities of deciding 1 when a 0 is transmitted and vice versa. In an OOK bit stream,
1and 0 are likely to occur equally, that is, p(1) = p(0) = 0.5; then Equation 4.25 becomes

1
BER = p ( 0/1) + p (1/0 ) (4.30)
2
Optical Receivers and Transmission Performance 143

Thus, as an equal probability of transmitting a 0 and a 1 is assumed, for a decision voltage level
of d, the total probability of error PE is the sum of the errors of deciding 0 or 1. Integrating the pdf
over the overlapping regions of the two pdfs gives

d
1 1
p [ v0 0 ] dv0 +
PE =
2
d
2 p[ v

0 1 ] dv0 (4.31)

Substituting for the probability distribution using Equations 4.26 and 4.27 leads to

d
x2 x2
1 1
BER =
2 e
d v 00
2 dx +
2 e
v 01 d
2 dx (4.32)

vNT 0 v NT1

The functions in Equation 4.28 have the standard form of the complementary error function Q()
defined as

1 2
Q ( a) =
2
e x /2
dx (4.33)

and then Equation 4.28 becomes

1 d v d
BER = Q + Q 01 (4.34)
2 vNT0 vNT1

4.3.3 RECEIVER SENSITIVITY

Again, using the condition p[v0/0] = p[v0/1], we have

d v d
= 01 o (4.35)
vNT0 vNT1

Now assuming vo0 = 0, hence

BER = Q() (4.36)

The Marcum Q() function is a standard function, and this curve is shown in Figure4.2. Note
that for a BER = 10 9, the value of is about 6, which is the normal standard for communications
at bit rates of 134 Mb/s40 Gbps.
Thus, by eliminating the decision-level variable d from Equation 4.31, we obtain

v01 vNT0
= (4.37)
v NT1
144 Advanced Digital Optical Communications

100

102

104
Probability error

106

108

1010

1012
0 1 2 3 4 5 6 7

FIGURE4.2 BER as a function of the parameter for NRZ-ASK.

or alternatively

v01 = (vNT0 + vNT1 ) (4.38)

substituting v01, vNT1, vNT0, we have

1/2
b1 b1 2
Gn R1 A = 2q Gn BR12 A2 + v 2NA + vNA (4.39)
Tb Tb

However, the amplifier noise voltage vNA is given by vNA = iNeqRI AB1/2. Thus, by substituting
this noise voltage, eliminating RI A, and solving for the energy required for the 1 transmitted and
received at the photodetector b1, we obtain

2iNeqTb
b1 = q 2G + (4.40)
qG

where we have used the approximation <Gn> = G and <Gn2> = G 2+x with x being the factor depen-
dent on the ionization ratio in an APD. For a pin photodetector, G = 1. The optical receiver sensitiv-
ity in dBm, denoted as RS in dBm, can thus be obtained as

pav
RS 10Log10 in dBm (4.41)
p0

where Pav = b1/Tb and Po = 1.0 mW. This is the optical receiver sensitivity defined as the mini-
mum optical power required for the receiver to operate reliably with a BER below a specific value.
InEquation 4.36, there are two terms clearly specifying the dependence of the signal dependence
Optical Receivers and Transmission Performance 145

and the amplifier noise contribution at a certain data rate and for a certain modulation format.
Recall that the term b1 represents the optical energy of the 1 required for the optical receiver to
detect with a certain BER. The term q/R in the right-hand side of Equation 4.36 is equivalent to the
optical power required to generate one electron, or the number of photons required to generate one
electron. A typical measure for the optical receiver is the number of photons required for it to oper-
ate with a specified BER.
Note that the common term of the energy b1 is q/R, which is equivalent to about one photon
energy; we can rewrite Equation 4.36 by using q/R = h/

h 2 2iNeq Tb
b1 = G+ (4.42)
qG

Thus, we can observe from equation 4.42 when we have an ideal electronic amplifier, that is, a
noiseless amplifier iNeq = 0, then it requires only a number of photon energy of 2 G for error-free
detection. The second term in the bracket is thus the number of photons required to overcome the
amplifier noise. Note that the determination of the BER using the eye diagram when the pdf is not
Gaussian is inaccurate. However, a much more accurate statistical analysis is developed and pre-
sented later in Section4.9.

4.3.4 OSNR AND NOISE IMPACT

4.3.4.1 Optical Signal-to-Noise Ratio
For an optically preamplified receiver, the ASE dominates the noise processes, and an OSNR is
commonly used that is determined by measuring the noise and optical signal levels in both polariza-
tion directions under an optical filtering bandwidth Bo of 0.1 nm [2].

optical signal power

OSNR = (4.43)
optical noise power

For an average output power P0 of the EDFA, the OSNR can be determined by

P0
OSNR = (4.44)
2nsp hfT (G 1) B0

where nsp is the number of spontaneous emissions of the EDFA, h is Planks constant, and G is the
linear power gain coefficient of the EDFA. Then, if the signal ASE beat noise dominates the noises
of the detection process, an NF of the EDFA can be used to determine the ratio of the OSNR at the
input and output as this is much more practical. Thus, the NF can be approximated by Fn ~ 2nsp.
Thus, the OSNR can be rewritten as

P0
OSNR = (4.45)
Fn hfT (G 1) B0

4.3.4.2 Determination of the Impact of Noise

One way to measure the impact of noise on the received signals of optically amplified systems is
by attenuating the signal power at the input of the optical preamplifier and then obtaining the BER.
A BER versus received power plot would be obtained. Assuming that only one span is used and
146 Advanced Digital Optical Communications

the optical gain is much greater than unity, the OSNR can be written for an operating wavelength
of 1550 nm as

OSNR = 58(dBm) Fn (dB) + Pin (dBm) (4.46)

If an eye diagram is used, then the eye opening penalty (EOP) can be determined for the same
BER. The EOP can be written as

EO
EOP = !10Log10 .....dB (4.47)
EO n

4.4 QUANTUM LIMIT OF OPTICAL RECEIVERSUNDERDIFFERENT

MODULATION FORMATS
Instead of using the equivalent noise current density or the noise-equivalent power (NEP), optical
receiver front ends are sometimes also characterized in terms of their receiver sensitivity. While
the receiver sensitivity is undoubtedly of great interest in optical receiver design, it comprises not
only the degrading effects of noise but also encompasses the essential properties of the received
signal, such as the extinction ratio, signal distortions, and ISI, generated either within the transmit-
ter or within the receiver itself. Thus, knowledge of the receiver sensitivity alone does not allow
trustworthy predictions on how the receiver will perform for other formats.
Although electronic noise usually dominates shot noise, it can be squeezed to zero. The signal-
dependent shot noise, however, is fundamentally present. The limit, when only fundamental noise
sources determine receiver sensitivity, is called the quantum limit in optical communications.
Theexistence of quantum limits makes optical receiver design an exciting task, because there is
always a fundamental measure against which practically implemented receivers can be compared.
Note, however, that each class of receivers in combination with each class of modulation formats
has its own quantum limit.
From Equation 4.36, we can observe that when the amplifier is noiseless, the receiver would
require an energy equivalent to that of 2G photon energy for detection. This is the quantum limit
of the receiver. For example, for a BER of 1e9 for a pin detector under ASK modulation, we would
need 36 photon energy per bit for detection when both the 1 and 0 are 5050 randomly received.
Noise plays a major part in the receiving end of any communication system. In optical com-
munications using coherent or noncoherent detection techniques, noise is contributed by (1) the
electronic noise of the electronic amplifications following the optoelectronic processes in the
photodetector, (2) The quantum shot noise due to the electronic current generated by the optical
signals, (3) the quantum shot noise due to the high power of the local oscillator (additional and
dominant source of coherent detection), and (4) the beating of the local oscillator and the optical
signals.
These noise sources vary from one optical receiver to another depending on their structures and
whether they consist of a photodetector and electronic amplifiers with and without optical preampli-
fication under coherent or noncoherent detection. This section thus presents the fundamental issues
of the noise processes and their impact on the sensitivity of receiving systems. In particular, we
examine the quantum limits of the optical receivers, that is, when the electronic noise is considered
to be nullified.
Schematic diagrams of coherent and noncoherent direct detection receivers are shown in
Figure4.3a and b without using an optical amplifier, while Figure 4.3c and d show their counter-
parts with optical amplifiers. Figure 4.3f and g show the balanced and fiber versions of the detec-
tion and receiving systems. The difference between these configurations is the noise generated
Optical Receivers and Transmission Performance 147

after thephotodetection and at the input of the electronic amplifier. Note that coherent systems are
identified with the mixing of the optical signals and local oscillator whose polarization directions
are aligned with each other.

4.4.1 DIRECT DETECTION

Optical detection for optical fiber communications is in the form of direct modulation and direct
detection. Direct detection is the simplest form of detection, which requires only a photodetector
followed by an electronic amplifier and the decision circuitry, clock recovery, and data recovery.

Electronic
preamplifier
S1,0(t)

(a)

S1,0(t) IF filter (heterodyne)

Electronic
Low-pass filter
preamplifier
(homodyne)
PLO(t)

(b)

Electronic
preamplifier
S1,0(t)
OA

(c)

S1,0(t) IF filter (heterodyne)

Electronic
Low-pass filter
preamplifier
ps(t) (homodyne)
PLO(t)
OA

(d)

FIGURE4.3 Schematic diagram of (a) direct detection, (b) coherent detection, (c) direct detection with OA,
and (d) coherent detection with OA. OA=optical amplifier, FC = fiber coupler, PD = photodetector, DSP =
digital signal processor, ADC = analog-to-digital converter. (Continued)
148 Advanced Digital Optical Communications

Modulated lightwave
channels
Decision

Local
oscillator Elect. receiver
(e)

Lightwave channels

Sampler
DSP
ADC

Local
Electronic
oscillator
preamp.

(f )

Decision
Lightwave
channels
PD DSP
FC Electronic
preamp.
Fiber one-bit delay

(g)

FIGURE4.3 (Continued) Schematic diagram of (e) fiber version of coherent receiver, (f) coherent receiver
using balance detection with two photodetectors connected back to back and a local oscillator (laser),
(g) self-homodyne reception balanced receiver with one-bit delay photonic phase comparator. OA=optical
amplifier, FC = fiber coupler, PD = photodetector, DSP = digital signal processor, ADC = analog-to-digital
converter.

For the ASK system, the signals for the 1 and 0 can be expressed in terms of the amount of photon
energy over the entire bit period, contained within a bit period T as

np
for 1
s1,0 (t ) = T 0 t T (4.48)
0 for 0

where np is the number of photons, and the energy of the lightwave is normalized with a single
photon energy at the operating wavelength. Thus, the total optical energy is

Es = nph.T (4.49)

Thus, the electronic current generated after the photodetector, with R being the responsivity of
the photodetection, is
q
is = np h.T = np h.T (4.50)
h
Optical Receivers and Transmission Performance 149

Thus, we could say that np is the number of photons per bit required for the detection of a1 if
there is no noise contributed by the electronic amplifier or detection.
With the probability of 1 and 0 being equal (50%), the probability of error of the detection is

1 np
Pe = e (4.51)
2

Thus, for a BER of 1e9 the argument np = (32)2 or np = 18 with an allowance of a factor of
thesingle-sided estimation so that np = 20 for the full detection error. This is the super quantum
limit. We also assume a unity responsibility of the photodetection.

4.4.2 COHERENT DETECTION

In the case of coherent homodyne detection with a local oscillator whose optical power PLO is very
much larger than the signal average power, the quantum shot noise current dominates the noise pro-
cess. The detected electronic current is the beating current between the local oscillator lightwave
and that of the signal, and thus we have

2 q 1 1 (4.52)
iN(LO) = 2qPLO = 2q 2 PLO
h T h T
and the SNR is given by

is2 4 2 ps (t ) PLO /T
SNR = 2
= = 2np (4.53)
iN(LO) 2qPLO / T

Pe erfc(2np ) (4.54)

Thus, for 10 9 BER, we have 2np = 3(2)1/2 .

4.4.3 COHERENT DETECTION WITH MATCHED FILTER

Now, it is assumed that a matched filter is inserted after the coherent receiver of Figure4.3b, and dif-
ferent modulation formats are used for transmission over long-haul optically amplified fiber systems.
In general, and under the assumption that the noise process in the optical detection is Gaussian,
the BER is given by
d
BER = erfc (4.55)
N 0

where d is the signal power separation between the average level of the 1 and 0 for binary sys-
tems and the equidistance between the constellation points of the modulation scheme as shown in
Figure4.4. Let E1 and E0 be the field amplitudes of the signals 1 and 0; then, the Euclidean distance
d is given by
d 2 = E12 + E02 2E1E0
(4.56)
T
2
with =
T s (t ) s (t ) dt
0
1 0

is the correlation coefficient between the two logic levels or, alternatively, the Euclidean angle
between the two vector signals as represented on the scattering plane of the constellation.
150 Advanced Digital Optical Communications

d
0

FIGURE4.4 Signal constellation and energy level and the geometrical distance between 1 and 0.

d
0 1 and E1

FIGURE 4.5 Signal constellation, energy level, and the geometrical distance between 1 and 0 of ASK
system.

4.4.3.1 Coherent ASK Systems

In heterodyne detection with an IF frequency region, the two-bit signals (see also the constellation
in Figure4.5) are given by

2np
cos IF t for 1
s1,0 (t ) = T 0t T (4.57)
0 for 0

where the amplitude of the lightwave-modulated signal is expressed in terms of the energy of the
photons over the time interval, and thus the square root of this amount is the amplitude of the field
Optical Receivers and Transmission Performance 151

of the lightwave. Naturally, the characteristic impedance of medium is set at unity. Thedistance is
then d = np.
Thus, the BER is given by

np
BER = erfc (4.58)
2N0

By setting N0 = 1, then for a BER of 10 9 the required number of photons is np = 4 9 2 = 72 for

the heterodyne receiver, while a 3 dB improvement for the homodyne receiver requires 36 photon
energy under the assumption that no electronic noise is contributed by the electronic amplifier.
These are the quantum limits of ASK heterodyne and homodyne detection when the power of the
local oscillator is much larger than that of the signal.

4.4.3.2 Coherent Phase and Frequency Shift Keying Systems

2np
cos 1t for 1
T
s1,0 (t ) = 0t T (4.59)
2np
cos 0 t for 0
T

The FSK modulation scheme with two distinct frequencies f1 and f 2 can be represented with a
constant envelope and variation in the carrier frequency, or continuous phase between the two states
as shown in Figure4.6.
The modulation index can be defined as

1 0
m= T (4.60)
2

0 1 and E1

FIGURE4.6 Signal constellation and energy level evolution of the signal envelope and the continuous phase
between the 1 and 0 of FSK system.
152 Advanced Digital Optical Communications

If the two frequencies are large enough and the second and higher harmonics are outside the
detection region, the signal correlation coefficient is given by

T T
2 2
=
T
0
s1 ( t ) s0 ( t ) dt =
T cos ( t ) cos ( t ) dt
1 0
(4.61)
sin(2 m)

2 m

Thus, the BER is given as

1 n sin(2 m)
BER = erfc p 1 (4.62)
2 2 2 m

Thus, for BER of 1e9, we have

np sin(2m)
BER = 10 9 1 =3 2
2 2m
(4.63)
36
then, np =
sin(2m)
1
2m

For MSK, the modulation index is 0.25, leading to the required number of photon energy per
bit of 6070, much higher than that of 0.8 at which only 30 photons energy is required per bit. Itis,
however, shown that the MSK can be optimum for the transmission over a dispersive medium owing
to the optimum bandwidth of the modulation scheme and hence minimum dispersive effects on
thephase of the carrier.
The FSK can be implemented with continuous phase frequency shift keying (CPFSK); that is,the
phase of the carrier is continuously chirped. Assume that the phase is linearly chirped such that
thephase variation for the 1 and 0 are given by

2np
cos 1 (t ) for 1
T
s1,0 (t ) =
2np
cos 0 (t ) for 0
T
where

m
IF t + t for 1 (4.64)
T T
1,0 (t ) = 0t
m 2m
IF t t for 0
T

IF t + for 1
2 T
1,0 (t ) = tT
2m
IF t for 0
2
Optical Receivers and Transmission Performance 153

The correlation coefficient is then given by

1 1 1
BER = 10 9 = erfc np 1 np 1
=3 2
2 4 m 4 m
(4.65)
18
then, np =
1
1
4m

The variation in the photon energy with the modulation index for this linear CPFSK is shown in
Figure4.7.
The FSK can be modified with the control of the relative phase of the two carriers between the
two bits; thus, the data bits can be written as

2np
cos(1t + ) for 1
T
s1,0 (t ) = 0t T (4.66)
2np
cos 0t for 0
T

where the phase angle t can be chosen as

m for p m 2 p + 1 p = 0,1,2....
1,0 (t ) = (4.67)
m for 2 p + 1 m 2( p + 1)

50

45

40
No. of photon energy

35

30

25

20
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Modulation index (m)

FIGURE4.7 Fundamental limits of the coherent phase and frequency shift keying detection for (a) mid-gray
o CPFSK, (b) light gray + FSK (1 and 2) with optimized correlation coefficient with phase control, and
(c) black * linear DMPSK.
154 Advanced Digital Optical Communications

with the correlation coefficient

T T
2 2
=
T
0
s1 (t )s0 (t )dt =
T cos( t + ) cos( t)dt
0
1 0

(4.68)
sin(m)

m

Thus, this gives a BER for a phase-controlled FSK as

1 n sin ( m )
BER = erfc p 1 +
2 2 m
(4.69)
36
np = for a....BER = 10 9
sin( m)
1+
m

The required number of photon is plotted against the modulation index and shown in Figure4.7b
black*.

4.5 BINARY COHERENT OPTICAL RECEIVER

Another way of amplifying the signal while boosting the accompanying noise above the electronic
noise floor is known as coherent detection [3]. A coherent receiver, as depicted in Figure 4.8,
combines the signal with a local oscillator (LO) laser by means of an optical coupler. Upon detec-
tion, the two fields beat against each other during the electronic generation process within the
photodetector, and the average electrical signal results in a mixing product between these time-
dependent signals and the local CW. Electronic signals are generated at an intermediate frequency
(IF) between the signal and the LO because the IF signal is usually mixed down to the baseband
after the photodetection, using standard microwave techniques. The splitting ratio of the optical
coupler has to be chosen as high as possible so as not to waste too much signal power and as low
as possible to allow sufficient LO power to reach the detector to achieve shot-noise-limited per-
formance (see the following explanation). The heterodyne efficiency accounts for the degree of
spatial overlap as well as for the polarization match between the LO field and the signal field. If both
LO and signal are provided co-polarized in single-mode optical fibers, then approaches unity.

Delay interferometer Electrical

(MZDI) amplifier

Electrical Data
filter recovery
Tb

one-bit delay

FIGURE4.8 MZIM optical balanced receiver for optical DPSK detection.