Calculation problems and short questions

from old exams.

Solutions and answers

CALCULATION PROBLEMS AND SHORT QUESTIONS ............................................................ 1

2. OPTICAL FIBRES ............................................................................................................................ 2
Problems ......................................................................................................................................... 2
Short questions ................................................................................................................................ 5
SOLUTIONS ........................................................................................................................................... 8
3. OPTICAL SOURCES AND TRANSMITTERS ............................................................................ 30
Problems ....................................................................................................................................... 30
Short questions .............................................................................................................................. 33
SOLUTIONS ......................................................................................................................................... 34
4. OPTICAL DETECTORS AND RECEIVERS............................................................................... 50
Problems ....................................................................................................................................... 50
Short questions .............................................................................................................................. 53
SOLUTIONS ......................................................................................................................................... 54
5. LIGHTWAVE SYSTEMS ............................................................................................................... 66
Problems ....................................................................................................................................... 66
Short questions .............................................................................................................................. 67
SOLUTIONS ......................................................................................................................................... 68
6. OPTICAL AMPLIFIERS ................................................................................................................ 73
Problems ....................................................................................................................................... 73
Short questions .............................................................................................................................. 76
SOLUTIONS ......................................................................................................................................... 77
7. DISPERSION COMPENSATION .................................................................................................. 93
Problems ....................................................................................................................................... 93
SOLUTIONS ......................................................................................................................................... 94
8. MULTICHANNEL SYSTEMS ....................................................................................................... 99
Problems ....................................................................................................................................... 99
Short questions .............................................................................................................................. 99
SOLUTIONS ....................................................................................................................................... 101
9. SOLITONS...................................................................................................................................... 103
Problems ..................................................................................................................................... 103
Short questions ............................................................................................................................ 103
SOLUTIONS ....................................................................................................................................... 104
10. COHERENT LIGHTWAVE SYSTEMS ................................................................................... 106
Problems ..................................................................................................................................... 106
Short questions ............................................................................................................................ 106
SOLUTIONS ....................................................................................................................................... 108

Almost all of the problems and solutions are translations, from Swedish into English,
of old exams. The problems is ordered in chapters, which follow the course book.

1
2. Optical fibres
Problems

Wave propagation
2.1 A single mode fibre is to be dimensioned. The normalised frequency is 80 % of
the maximum allowed for single mode condition. The relative refractive index
difference is 0.005 and the refractive index of the core is 1.5. The wavelength is
1.5 μm. Determine the radius of the core.

Calculate numerically also the ratio between the intensity at the centre of the core
and the intensity at the surface of the cladding.
(Exam. 960830)

2.2 A standard single mode fibre is to be characterised. The refractive index of the
core is 1.44 and the relative refractive index difference is 10-3. The fibre is to be
used at 1.7 μm and has the normalised frequency of 2.

How large is the diameter of the core? How much of the power is transported in
the core? What is the material dispersion parameter? What is the minimal
attenuation?
(Exam. 970411)

2.3 Calculate numerically the field intensity of the fundamental mode in a fibre,
however normalised to the field intensity of the midpoint of the core. The radius
of the core is a = 5 μm and its refractive index is 1.5. The relative refractive
index difference is 0.005<<1. Normalised frequency is 2. Calculate the
normalised values for 0a, 0.5a, a, 1.5a.
(Exam. 990409)

2.4 Some features of the fundamental mode shall be determined of a waveguide

from the wave equation and boundary conditions. The wave guide is an infinite
slab in the x-axis and z-axis direction and has thickness d in the y-axis direction.
The coordinate system is symmetrically oriented with respect to the slab. The
slab has the refractive index n1 and is surrounded on both sides of a refractive
index of n2<n1. The mode propagates in the z-axis direction. The angular
frequency is ω. It can be assumed that the electrical field only has one field
component apart from zero and that the field is directed in the x-axis direction.
Further, the field component is an even function of y and is independent of x.
Derive the dispersion relation, i.e. the relation between the propagation constant
β and the free-space wave number k0. How does the E-field vary in the y-axis
and the z-axis directions?
(Exam. 010423)

2
Dispersion in Single-Mode Fibers
2.5 A step-index fibre with a pure silica core has a core refractive index of 1.5000, a
cladding refractive index of 1.4967, and is used to transmit light at 1.55 μm. The
core radius is 5 μm.
a) How many modes can propagate in the fibre?
b) Which are the two main causes of dispersion in this fibre?
c) Determine the dispersion due to these two mechanisms.
For waveguide dispersion, make the (realistic) assumption that the material
dispersion can be neglected (dn/dω=0).

2.6 A standard single-mode optical fibre has minimum dispersion at λ≈1.3 μm and
minimum attenuation at λ≈1.5 μm. To be able to make use of the low
attenuation we would like to design a dispersion shifted optical fibre with zero
dispersion at 1.5 μm. To facilitate coupling of light into the fibre, the core radius
should be as large as possible. By minimising the core-cladding index
difference, the core radius will be sufficiently large. Determine the core radius
and the core-cladding index difference. The fibre material is SiO2 and the
refractive index is approximately 1.5. Neglect the material dispersion when
calculating the contribution of waveguide dispersion.

Dispersion-Induced Limitations
2.7 An ideal light source generates the wavelength 1.55 μm when it is unmodulated.
The light source is then intensity modulated and the optical power varies as

P0 1   cos t  , ε<<1 and Ω/2π is in the MHz-region. The signal is transmitted on
a standard fibre, and higher order dispersion can be neglected. The power
attenuation coefficient is α. How does the detected power vary as a function of the
length L of the fibre? The detector is ideal.
(Exam. 970829)

2.8 A Gaussian pulse with the RMS-width, σ0, of the intensity and with the chirp
parameter C is transmitted through a standard fibre. It is found that the pulse can
be compressed in the time-domain. Derive an expression for the position of the
fibre where the compression is maximal and how large it is. What is the time
derivative of the instantaneous frequency at this position? What is the RMS-
width σω of the optical spectrum (i.e. the spectrum which is measured with an
optical spectrum analyser) at this position of the fibre compared to the optical
spectrum of the input pulse? Begin with equation (2.4.16) or previous equations.
(Exam. 970306)

2.9 A lossless fibre with negligible higher order dispersion than β2 is excited by a
Gaussian pulse. The power of the pulse has the RMS-width σ0. The spectral
width of the optical pulse, i.e. the RMS-width which is measured with an optical
spectrum analyser (OSA), is σω. The complex amplitude or the complex

3
 envelope or the slowly varying amplitude, has the phase equal to 0 at the
position of the excitation. At what distance along the fibre does the complex
amplitude show the largest second order derivative of the phase with respect to
the time? Calculate this time derivative, and the RMS-width of the power, and
the spectral RMS- width at the corresponding position.(Exam. 980305)

2.10 A chirped Gaussian pulse is created in the following way. A laser with
constant output power is coupled to an external intensity modulator, which in turn
is coupled to a single mode fibre with the dispersion parameter β2. At the output
of the modulator the light power is Gaussian shaped and the pulse has no chirp
(C0=0). At the output of the fibre a Gaussian shaped power with RMS-width σ1,
chirp parameter C1 and peak power P1 is desired. How long should the fibre be
and what are the RMS-width σ0 and the peak power P0 at the modulator output?
Assume the fibre attenuation α is zero.
(Exam. 000901)

2.11 A standard single mode fibre is excited with a pulse with carrier frequency
194 THz. The power of the pulse is Gaussian shaped and has an RMS-width of
0.2 ns. The optical field is without chirp at the exciting position. Where along
the fibre has the time derivative of the instantaneous frequency a maximal
absolute value and what is the maximal absolute value? Find also the RMS-
width of the pulse power for this position.
(Exam. 010309)

2.12 A single mode fibre has the length L and the power attenuation coefficient α
per unit length (not dB/unit length). The propagation constant β can be written
in a Taylor series β(ω0+Δω)= β0+ β1Δω + β2(Δω)2/2. The fibre is excited with a
light pulse which is chirp free and its power varies as Gaussian function with
peak power P and RMS-width σ0. The angular frequency of the carrier is
ω0.What is the peak power, the RMS-width of the power, the time delay and the
time derivative of the instantaneous frequency of the pulse at the output?
(Exam. 010831)

2.13 Estimate the limiting bit rate for a 60 km single-mode fibre link at 1.3 and
1.55 μm wavelength assuming transform-limited 50 ps (FWHM) Gaussian input
pulses. Assume β2=0 and -20 ps2/km and β3=0.1 ps3/km and 0 at 1.3 and 1.55
μm wavelengths, respectively.

2.14 A laser diode is current modulated so that Gaussian shaped pulses with RMS-
width of T/8 are obtained, where T is the bit slot. The laser is assumed to have a
Gaussian shaped spectrum when it is unmodulated, the linewidth of the
spectrum is 5 MHz at the full width half max. The linewidth enhancement factor
of the laser is +4 or -4. The pulses are launched into a standard single mode
fibre at 1.55 μm. How long can the fibre be in the two cases if the dispersion
criterion σ/T<1/4 is used? The bit rate is 2.4 Gbit/s.
(Exam. 950601)

4
 2.15 Consider a single mode fibre system from a dispersion point of view. The
system works at minimum dispersion parameter D, and the transmitting laser is
direct modulated and has a chirp parameter = 6. The unmodulated laser has the
linewidth 50 MHz at full width at half height of the spectral density. The data
rate is 1/T = 10 Gbit/s. The launched pulse is Gaussian with the width of the
optical intensity 35 ps at full width at half height. Determine the maximal length
of the fibre with the condition that the dispersion condition is fulfilled, i.e. the
received RMS-width of the optical signal <T/4.
(Exam. 950831)

2.16 The dispersion limit, i.e. the relation between the bit rate and the length of the
link, for a system with a typical direct modulated light emitting diode and a
typical direct modulated single mode laser diode should be compared
numerically. In both cases the chirp parameter is assumed to be zero. The carrier
wavelength is chosen so that the dispersion parameter is zero for the standard
single mode waveguide used. The transmitted pulse width is assumed to be half
of the permitted width at the receiver. Typical values can be found in the
literature.
(Exam. 000505)

2.17 A light source with negligible linewidth generates Gaussian pulses with the
chirp parameter, C>0. The pulses are transmitted on a single mode fibre, which
is characterized by the group velocity parameter β2<0 and the length L. The
presence of a pulse in a symbol slot represents a ONE and the absence of a pulse
represents a ZERO. What is the maximal bit rate if the common dispersion limit
criterion is accomplished?
(Exam. 020308)

2.18 Consider a single mode fibre system from a dispersion point of view. The
system works at 1.55 μm and uses a standard fibre. The linewidth of the laser is
100 MHz at full width half maximum of the spectrum, which is considered also
to have a Gaussian shape. The laser is direct modulated and has a chirp
parameter, C  4 . The bitrate is 1/T=2.5 Gb/s. Assume the transmitted pulse is
a Gaussian pulse with the width equal to 140 ps at full width half maximum of
the optical power at the transmitter. Calculate the maximum length of the fibre,
providing normal dispersion limit criterion is fulfilled.
(Exam. 020823)

Short questions

2.19 What is waveguide dispersion and what is its cause?

(Exam. 970306)

5
2.20 An advantage with an optical fibre is its immunity against electromagnetic
interference even without metallic shield. Explain why outer light disturbances,
for example lightning, which penetrate into the fibre at midpoint of the fibre
link, are harmless.
(Exam. 000307)

2.21 What is multimode dispersion?

(Exam. 000901)

2.22 What is and how does polarization-mode dispersion behave?

(Exam. 010309)

2.23 Sketch the field distribution in core and cladding for the fundamental mode in
a planar waveguide when the normalised frequency
(1) goes to infinity
(2) goes to zero
(3) a typical intermediate frequency
(Exam. 020308)

2.24 The illustration below shows two curves which show some relations between
log(length)- and log(bit rate)-diagram for a system with a standard fibre with
1.55 μm.
log(length)

log(bit rate)

How do the curves in principle change if the system is upgraded with a dispersion
compensating fibre roll incoupled at the receiver. Note that the length variable is
only referring to the length of the standard fibre.
(Exam. 980305)

2.25 In a diagram which shows length vs bit rate, the dispersion limit curve and the
attenuation limit curve cross in one point. How should the point of intersection
be explained with respect to length, bit rate and bit error rate and penalty?
(Exam. 020308)

2.26 What is the main mechanism for attenuation in a standard fibre at 1.55 μm?
Describe this mechanism physically.

6
 (Exam. 010831)

2.27 The dispersion limit in a link with a standard single mode fibre has different
slopes in a log(length)-log(bitrate)-diagram for different light sources. What is
the slope when the light source is a LED and a singlemode laser with small
spectral width and without chirp. The wavelength is 1.55 μm.
(Exam. 020408)

2.28 The student with name Osquar makes a fibre optic experiment. He increases
the transmitter power in steps of 1 dB from an ideal unmodulated laser with
narrow spectral width. He observes that received power, after transmission
through the fibre, also increases in steps of 1 dB initially, but thereafter less.
What has happened?
(Exam. 020408)

2.29 A normal single mode laser is placed in a module and is positioned so that the
light is optimally coupled into a single mode fibre. After some time the light
power is zero in the fibre and it is suspected that the laser became defect.
However, an investigation shows that the laser has started to lase in another
mode, which has two variations in the lateral direction instead of one, which is
normal. Is this a possible explanation to the zero power? Motivate you answer.
(Exam. 990309)

7
Solutions

2.1 The condition for single-mode is the normalised frequency V<2.405

Equation (2.2.35):
2
V an1 2  0.8  2.4


V 0.8  2.4 1.5

a   3.06 μm
2  n1 2 2 1.5 2  0.005

The field in the core is given by a Bessel function, equation (2.2.41). The intensity
is the squared absolute value. The ratio, Q, is

2
 J ( p  0) 
Q   0 
 J 0 ( pa) 

Use

p 2  n1 2 k 0 2   2 equation (2.2.25)

q 2   2  n2 2 k 0 2 equation (2.2.26)

V 2 / a2  p2  q2 equation (2.2.35)

 2 / k 0 2  n2 2 (qa) 2
b  equation (2.2.36)
n1 2  n 2 2 V2

 ( pa) 2  V 2 (1  b)

b is achieved from Figure 2.5 or equation (2.2.40)

b  1.1428  0.9960 / V 2  {1.5  V  2.5}  0.39

pa  V (1  b)  0.8  2.4  0.61  1.5

Table gives J 0 (0)  1 , J 0 (1.5)  0.51

 Q  3.7

8
 Answer: a  3.06 μm and Q  3.7

2.2 Equation (2.2.35) gives

2 V 2 1.7
V an1 2  a    8.4 μm  diameter  16.8 μm
 2  n1 2 2 1.44 2 10 3

Equation (2.2.46) gives the power inside the core

 2 
  1  exp  
 w / a 2 
 

w / a  is achieved from equation (2.2.45)

w / a   0.65  1.619  2 3 / 2  2.879  2 6  1.27

Pcore  2 
 1  exp    71 %
Ptotal  1.61 

The material dispersion is achieved from Figure 2.10.

DM=28 ps/km nm

The attenuation is read off from Figure 2.15. The minimal fibre attenuation is
from Rayleigh scattering and Infrared absorption 0.1+0.3=0.4 dB/km

Answer: Diameter of the core: 16.8 μm, power transported inside the core: 71 %,
material dispersion: DM=28 ps/km nm, minimal attenuation: 0.4 dB/km

2.3 While the relative refractive index difference Δ<<1, the E and H fields are
linear, and E ┴ H, and E/H is a constant. Thus it is sufficient to calculate,
according to equation (2.2.41),

E x (0)  E0 J 0 (0) / J 0 ( pa)

E x (  ) J 0 ( p )
 , a
E x (0) J 0 (0)

E x (  ) K 0 ( ) J 0 ( pa)
 , a
E x (0) K 0 (a) J 0 (0)

9
 Use p 2  n1 2 k 0 2   2 equation (2.2.25)
q    n2 k 0
2 2 2 2
equation (2.2.26)

V 2 / a2  p2  q2 equation (2.2.35)

 2 / k 0 2  n2 2 (qa) 2
b  equation (2.2.36)
n1 2  n 2 2 V2

( pa) 2  V 2 (1  b)

Figure 2.5 or equation (2.2.40)

b  1.1428  0.9960 / V 2  0.4158

qa  V 2 b  1.2897

pa  V 2 (1  b)  1.5292

From a mathematical table, e.g. Beta.

ρ pρ qρ J0(pρ) K0(qρ) Ex(ρ)/Ex(0)

0a 0 1 1
(0.25a) 0.3823 0.9637 0.9637
0.5a 0.7646 0.8591 0.8591
(0.75a) 1.1469 0.6972 0.6972
a 1.5292 1.2897 0.4955 0.2821 0.4955
(1.25a) 1.6121 0.1851 0.3251
1.5a 1.9346 0.1234 0.2168

10
 2.4

y
n2

n1
d
n2 x
z

According to the assignment E  E x e x where E x ( x, y, z ) can be written as

F ( x)( y) exp(  jz ) exp( jt ) , where F (x) = constant due to the planar geometry, i.e.

 0.
x
( y) is even.

The wave equation (2.2.16) is valid for all field components and gives for Ex
 2 E x  n1 2 k 0 2 E x  0 , y  d 2
 2 E x  n2 2 k 0 2 E x  0 , y  d 2

where

 2
 (n1 2 k 0 2   2 )  0 , y  d 2
y 2

 2
 (n2 2 k0 2   2 )  0 , y  d 2
y 2

The solutions to the above stated equations are trigonometrical or exponential. In

order to have a bounded mode the fields must  0 when y   , why we put
exponential decreasing functions for y  d 2 .
Introduce q 2   2  n 2 2 k 0 2 in accordance with equation (2.2.26). We then achieve

yd 2
F  A1 exp( q( y  d / 2)) ,
F  A1 exp(q( y  d / 2)) , y   d 2 because F is even

For the field in the slab an even trigonometric function is chosen

F  A2 cos( py) , y  d 2

where p 2  n12 k 0 2   2  0 according to equation (2.2.25).

In order to connect the solutions at the boundaries, we use the fact that the tangential
components of the E- and H-fields should be continuous. For example, Ex should be
continuous at y  d 2

11
 A2 cos( p d 2)  A1

We can express the H-field from the E-field

rot E   j 0 H

ex ey ez
 E x
0  j  (0, jE x , )   j 0 ( H x , H y , H z )
y y
Ex 0 0

F
While Hz is a tangential component also should be continuous at y  d 2
y
  pA2 sin( p d 2)  qA1  qA2 cos( p d 2)

q  p tan( p d 2) , which is the dispersion relation and gives the relation between β and
k0.

Answer: q  p tan( p d 2) , where p 2  n12 k 0 2   2  0 , q 2   2  n 2 2 k 0 2

E x  A2 cos( p d 2) exp( q( y  d 2)) exp(  jz ) , yd 2

E x  A2 cos( py) exp(  jz ) , y  d 2

E x  A2 cos( p d 2) exp(  q ( y  d 2)) exp(  jz ) , y  d 2

2.5 n1  1.5000
n 2  1.4967
0  1.55 μm
2a  10 μm

2
a) V  a n1 2  n 2 2  2.0  2.405 
0
 1 mode (single-mode), which has two orthogonal polarizations

b) 1. Material dispersion
2. Waveguide dispersion

1 dn g
c) 1. D M 
c d

  ZD   ps 
equation (2.3.12): D M  1221    nm  km 
    

 ps 
pure silica:  ZD  1.276 μm  D M  21.6 
 nm  km 

12
 n  Vd 2 Vb  dn
2. DW   2  , while 2 g  0
c dV 2 d

Vd 2 Vb 
From Figure 2.9  0.22 V  2.0
dV 2

n n
 1 2  0.0022
n
1

 ps 
 D  1.6  
W  nm  km 

 ps 
Dtot  D M  DW  20  
 nm  km 

2.6
n2 cladding

a n1 a: core radius

n n
 1 2
n
1

According to Figure 2.10 and 2.11:

D M  19 ps/ nm  km  

at 1.55 μm
D W  19 ps/ nm  km 

n d 2 Vb 
DW   V
c
 2
dV

minimised maximised

dn 2 g
where  0 , n 2 g  n 2  n  1.5
d

13
 According to Figure 2.9:

 d 2 Vb  
V   0.96 for V  1.32
 dV 2  max


D  c   19  10  6  3  10 8  1.55  10  6
 6.1  10  3
W
  
min 0.96 n 1.5  0.96

2  a 2  a
V n1 2  n 2 2  n1 2
 

V 1 1.32 1.55 10 6

a    2 μm
2  n 2 2 1.5  2  6.1 10 3

2.7 Transmitted power P0 1   cost 

Transmitted field normalized as fieldstrength  power

2

( j0t )   ( j0t )
 P0 1   e ( jt )  e ( jt )  e

P0 1   cost e
 4 

Received field intensity


 j (0t  0l )   j (0 )t   (0 )l j ( )t   ( )l    l
P0  e  e e 0 0  e 2
 4  

2
Taylor expanded:  (  )   0  1   2
2

j ( t   l )  l   jt   l    l   jt   l    l 
2 2
0 0 2 1 2 2 1 2 2
 P0 e e 1  4 e  e 
 4 
 
 2 
    l
j (0t   l )  l  
 
j
0 2 1  cos t   l e 2 2 
 P0 e  2 
1
 
 

  
Answer: Received power  P0 e l 1   cos t  1l cos  2


  
2
2

l  

  

14
 i.e. the signal power is attenuated as e l but has also a dispersive effect
  2 
cos  2 l ;  l corresponds to the group delay.
 2  1
 

2.8 Start with equation (2.4.16)

A0 T0  (1  iC 0 )t 0 2 
exp  
A( z, t ) 
T0 2  i 2 z (1  iC 0 )  0 2 
 2 T 2  i z (1  iC )
0  


Expand the exponent

(1  iC )t 2 
(1  iC )t 2 T0 2   2 zC  i 2 z    T 2
 
  2 zC   2 zC t 2  it 2 CT0 2   2 zC 2   2 z 

   0

2 T0  i 2 z (1  iC )
2



2 T0 2   2 zC 
2
  2 z 2 
 

2 T0 2   2 zC 
2
  2 z 2 


The RMS-width of the intensity of the pulse   T / 2 .

Identify

T2 
T 0
2
  2 zC 2
  2 z  2
T0 2

2 2 2 2
 C 2 z    2 z     
2

T2
 1      1  C 2 z     2 z 
 02 T0 2  T0   T0 
2 2  2   2 2
2 
  0   0 

Differentiate with respect to z and set to zero

d   2   
  21  C 2 z  C 2  2  2 z  2  0
!

dz   0 2 


 T0 2  T0 2 T0 2 T0 2


 1 C 2 2
z  C
T0 2

C 2  2 0 
2
 z opt    , ok according to equation (2.4.18)
1  C 2   2 2 


We understand that C 2  0 in order to achieve compression

2 2 2
 min 2  C  C 2  T0    2
 T0 2 
    2  C 2
2 2 2
     1   C   1 
 1  22            
 02  T0  1 C   2  T 2   1 C   2  1  C 2  1  C 2  1  C 2 
2 2 2
  0   2 

The instantaneous frequency derivative is proportional to the second derivative of

the phase with respect to time.

15
 ~

2 CT0 2   2 zC 2   2 z     C 2  T0 
~ CT0 2   2 z 1  C 2  {z  zopt }  CT0 2   2 (1  C 2 )     0
2

 
2 T0 2   2 zC   2 z 2 

2


2 
 1  C   2 

i.e. at maximal compression (z=zopt) the pulse has no chirp.

~ 2
The spectrum analyser measures A( z,  ) , i.e. the squared absolute value of the
Fourier transform of the complex envelope. The complex envelope:

~ ~ i 
A( z ,  )  A(0,  ) exp  2 2 z  equation (2.4.15)
2 

i.e. the squared absolute value does not change and neither does the RMS-width,
σω. At zopt the pulse is transform limited, i.e.  ( z  z opt )    ( z  z opt ) is minimal.

2.9 The complex amplitude at the position of excitation is

 1 t0 
P  A(0, t )  A0 exp  
 2 2 2 
 0 

From page 49, T0 2  2 0 2 , inserted in equation (2.4.16), {i=-j}

A0 T0 
exp 
t2 
 A0 T0
exp 

 t 2 T0 2  j 2 z  
A( z, t ) 
T0 2  j 2 z  0 
 2 T 2  j z
2  
 T0 2  j 2 z  
 2 T 4   z  2
0 2 
The time dependent phase is

t22z
 (t ) 

2 T0 4    2 z  2 
d 2 2z

dt 2
T0    2 z  2
4


4 2

d d 2  2 T0   2 z    2 z 2  2 z 2 !
0
dz dt 2 T0 4   2 z 2
2
 

T0 2
 z0  , which is called the dispersion length
2

16
 T0 2
2
d 2 2 1
2
 2
 sgn(  2 )
dt  T 2  2T0 2
  2 0 
max
T0 4
 2 
 

 t 2 T0 2 
The intensity A( z, t ) 2 is proportional to exp  

 T 0
4
  2 z  2
 


T0 4    2 z  2
The variance is  T0 2 .
2T0 2

The RMS-width is T0.

~ 2
It remains to calculate the spectral width, which is given by A( z,  ) , where
~ ~ ~
A( z,  ) is the Fourier transform of A( z,  ) . While A ( z ,  )  HA (0,  ) , where H is
the transfer function of the fibre and H 2  1 , the spectral width does not change
along the fibre. That is obvious while the power is preserved and the
“frequencies” are not attenuated differently. However, the phase relations changes
between them (i.e. the “frequencies”) and it is showed by the pulse broadening
and chirp.

Answer: RMS-width of the intensity: T0  2 0

2 0 2
Optimal position z 0 
2

d 2 1
Maximal derivative of second order  sgn(  2 )
dt 2
max
4 0 2

Spectral RMS-width: σω

2.10 Equation (2.4.16) illustrates how the slowly varying amplitude varies in a
single mode fibre. In (2.4.16) C is the chirp parameter at the input of the fibre
and is zero in our case, equation (1). Thus, at the output of the fibre
A0 T0  t2 
exp  
A( z , t ) 
T 0
2
 i 2 z 
1/ 2
 0 
 2 T 2  i z
2  

(1)

In order to determine the RMS-width and the chirp parameter equation (1) is
reconfigured and in accordance with equation (2.4.10)

17
  1  iC1   t 2 
A( z , t )  A1 exp  
 2T1 2 
 
where C1 and 2T1 are the chirp parameter and the RMS-width of the intensity
respectively, at the output of the fibre.

  i z  
 t 2 1  22  
A0 T0
exp 

 t 2 T 0 4  i 2 z    A0 T0 
exp 
 T0  

T 0
2
 i 2 z 1/ 2
 0 
 2 T 4   z  2
2  T 0
2
 i 2 z 
1/ 2   T 4   z  2  
 2 0 2

  2
 
  T 0

Identification gives

2z T0 4    2 z  2 A0 T0
C1  2
, T1 2  2
, A1 
T0 T0 T0 2  i  2 z

or  1 2   0 2 1  C1 2  , A1 
A0
1  iC1

Thus,
1
0 
1  C  1
2

2 1 2 C1
zL
1  C1 2  2

2
The peak power at the output is A( z,0) ~ P1

2
And at the input of the fibre A0 ~ P0

P0 T0 2 P0
P1  
T0   2 z
4 2 2
1  C1 2

1 2 1 2 C1
Answer: P0  P1 1  C1 2 ,  0  , L
1  C1 2 1  C1 2  2

18
2.11 The frequency 194 THZ  λ=1.55 μm. The dispersion is then dominated by
β2 and β3 can be neglected. Typical values of D are in the range 15-18 ps/(km-
2
nm) near 1.55 μm. If 16.5 ps/(km-nm) is chosen, then  2   D {equation
2c
(2.3.5)}
2
  2  21 ps /km

The power is Gaussian shaped which results in that the slowly varying amplitude
of the field also is Gaussian with the RMS-width T0  2  0.2 10 9  0.28 ns.

Equation (2.4.16) with C=0 gives

A0 T0  t2  A0 T0  t2  T0 2  j 2 z 
A( z, t )  exp     
T
0
2
 j 2 z 
1/ 2
 0
 2 T 2  j z
2  

 T0  j 2 z
2 1/ 2

exp
 2

 T 4   z  2
 0 2



The chirp or the instantaneous frequency  (t ) = phase derivative with respect to

d  t2 (  2 z )  t z
time=   4 2 
 2 T0  (  2 z )  T0  (  2 z ) 2
dt 4 2

d  z d 2z
 (t )  4 2 , which is maximal when 0
dt T0  (  2 z ) 2 dz T0 4  (  2 z ) 2


T 0
4

 (  2 z ) 2  2   2 z2  2 z  2
0
T 0
4
 ( 2 z) 2 2

 (  2 z ) 2  T0 4

T0 2 0.08 (ns) 2
z opt    3800 km
2 21 ( ps) 2 km 1

T0 2
2
d 2 1 -2 Grad / s GHz
 (t )  4  sgn(  2 )  6.2 ns = -6.2  1 
dt max T0  T0 4
2T0 4 ns ns

 The absolute value ≈1 GHz/ns

The RMS-width is derived from the squared absolute value of A(z,t)

 
 
2  t 2
2T0 2   t2 1   T0 2   t2 1 
A( z , t ) ~ exp    exp     z    exp  
 2 T0
4
 (  2 z ) 
2
 2 T0  (  2 z )
4 2
  2  2
 2 T0 
 2T0 2 
 

19
 10 24
 RMS-width = T0   2 z opt  21   3800 10 3  0.28 ns
10 3

Answer: zopt=3800 km, maximum absolute value= 1 GHz/ns, T0=0.28 ns

2.12 The pulse will be attenuated, broadened, chirped and delayed as it propagates
through the fibre. The propagation can be expressed as

exp(  j ( )  )L (1)
2
The power attenuation is given by the factor exp( L) .
The delay is given by the group velocity time τ=L/vg={2.3.1}=Lβ1
It remains to calculate the broadening and the chirping.
Use the formulas for chirped Gaussian pulses (without attenuation and delay).
Equation (2.4.16) gives
A0 T0  t2 
A( L, t )  exp  , (2)
(T0 2  j 2 L)1 / 2  2(T 2  j L) 
 0 2 
because C=0.

A0 is identified with P0 and T0 2  2 0 2 , because A0 , T0 are field quantities.

The power P( L, t )  A( L, t ) 2 and the peak power output

A0 2 T0 2 P0
P1  P ( L,0)   , which also should be multiplied with
T0  (  2 L )
4 2
  L 
2

1  2 2 
 2 
 0 
exp( L) .

In order to get the broadening and chirp, equation (2) gives:

   2 2 
exp 
t2   exp  t (T0  j 2 L) 
 2(T 2  j L) 
 0 2 
 2 T 4  (  L) 2 
 0 2   
We can identify from the real part

2
  L
T1 2  T0 2   2  ,
 T0 
2   
2
1   2 L 
2
2  2L   L
i.e.  1   0
2 2
   0      0 2 1   2  
      
2  2 0   2 0    2 0
2
 

The time derivative of the instantaneous frequency is derived from the Im-part of
the exponential function

20
 t22L 1  1 t 2 L
 , f 

2 T0  (  2 L )
4 2
 2 2 T0 4  (  2 L) 2

1  L 1 2L 1
f   4 2  
2 T0  (  2 L) 2 2 4 0 4   L 
2

1  2 2 
 2 
 0 
  
2
P0 2  L ,
Answer: P1  exp( L) ,  1   0
2
1  2  τ = Lβ1,
  2 2  
  0 
2
  L  
1  2 2 
 2 
 0 
1 2L 1
f   
2 4 0 4   L 
2

1  2 2 
 2 
 0 

2.13 Consider equation (2.4.23)

Because the pulses are transform-limited  chirp=0

V  1

Let C=0 and V=0 in equation (2.4.23)

2 2
   L  1  L 
 1  2 2    3 3 
0  2  2  4 0 
 0 

where L is the transform-limitation distance,

 is the  RMS after propagation and  0 is the  RMS before propagation

intensity intensity

σ0 σ

time time
Before fibre After fibre
propagation propagation

Case 1.

21
   1.3 μm
 2  0 ps 2 / km
 3  0.1 ps 3 / km
L  60 km

T FWHM 50 10 12

0    21.23 ps
2  2 ln 2 2 1.665

1
 2
 2
 
0.1  60  
  21.23  1  
1
 21.23 ps
2  4  21.233  
    
 1 

TB 1
   Bmax  11.8 Gbit/s
4 4B

Case 2.

  1.55 μm
 2  20 ps 2 / km
 3  0 ps 3 / km
L  60 km

1
 2
 2
1  20  60  
  21.23  1     35.34 ps
 2 2  21.232  
   
 

1
 Bmax   7.1 Gbit/s
4

2.14 Consider equation (2.4.23), which is equal to

2 2 2
 2  C 2 L 
0 2
 1
  2 
   L
 



 1  V 2  2 2   1  C 2  V 2
2

2 1  L 
 3 
 4 3 
 2 0   0 
2  0 

Let us firstly calculate the parameters:

T 1 1000
 ,T s ps
4 2.4 10 9 2.4

 0  from the text 

T
 52 ps
8

22
 From equation (2.3.5):
2
2   D
2c

From Figure 2.11 or page 41, D=17ps/nm km

 2 
1.55 10  6 2

17 10 12

10  12 2  ps 2 
 20  
2  3 10 8 10 3 10 9 10 3  km 

From page 50
2  5 10 6 10 12 1000
V  2   0  2    1.39 10 3
8 ln 2 8  2.4

 V 2 can be neglected.

C    c  4 page 113

3
 3  0.1 ps /km from page 51

2 2  0
Calculate  2 L2  4 0 2  20  52
3
   1
2 0 3L 3 0.1

i.e. the last term in equation (2.4.23) can be neglected.

2 2
2  C 2 L    2 L 
 2  2 2  1     
0   2   2 2 
 2 0   0 

 
  L 2
C 2 L 4  20  L  20  2
 1  C2 1  2 2   1  17   L  1  0.0296  L  2.3  10  4  L2
 02  2  2
  2

 0  52 2 52

L2  128  L  13100  0

L  64  4100  13100  64  131

For  c  4  64  131  67 km

For  c  4  64  131  195 km

Answer: 67 km resp. 195 km

2.15 Consider equation (2.4.23) and note that β2=0 because the system works at
minimum dispersion. Thus,

23
 2

  
2  3 L 
2
 
   1  1  C 2  V 2
  3 
 0   4 2 0 

  
2  2 16 0 6
 L2     1
  0


  
 1 C V 2
2
 
2
 32

Find the different parameters:

  T 4  25 ps

 0 is calculated in the following way. Launched Gaussian pulse has the shape


exp  t 2 2 0 2   t2
 ln( 2)
2 0 2

  full pulse width at half height   2 0 2 ln 2


0   0.42  {from text :  35ps}  15 ps
8 ln 2

C=6

From page 50
 
V  2   0     0.42  50 10 6  2  1.35 10 8  2 1.35 10 8 15 10 12  1

i.e. V can be neglected.

3
 3  0.1 ps /km from page 51

 25  2  2 16 15 6 ps 6
L2     1 2
 
km 2  L  7000 km
 15   37 0.12 ps 6  

2.16 Use the pulse broadening formula in equation (2.4.23) with C=0, β2=0,
1
  2 0 ,   , λ=1.3μm
4B

  2

    3 L  
2
  
  4  1  1  V 2
2
   3  
 0 
   4 2 0  

24
    L
6  1  Vm 2  3 3
 4
 


 0 

2
 2 
 3  {at λ=1.3 μm β2 =0}  S  

equation (2.3.13)
 2c 

Table 2.1 gives S=0.09 ps/km nm2

Consider firstly the light diode case and assume for the present that Vω2>>1,
where V     0

3L  33L
6  4  2 0 2 
4 0 3 0

2c
On page 51,    15 nm     
2

2
 2c 
2
 2  L L  1 
6   2    2 S     2S   0      SL8B
2
 
   2c   0 0  8B 

6 6
BL    1.5 10 13 m b/s = 1.5 1010 km b/s = 15 km Gb/s
8  S 2

8  15 10 
9 2
 0.09 10 3

We check that Vω2>>1 is fulfilled.

V     0 , and the worst case occurs when  0 is minimal, i.e. when B is

maximal. We know that the modulation bandwidth of a typical light diode is ≈
100 MHz (see page 91) and B is of the same magnitude, for example 200 Mb/s.

1 2  3 10 8
V  2   15 10 9  1
4  8  200 10 6 1.3 10 9  2

Consider secondly the single mode laser diode case and assume for the present
that Vω2<<1.

3L  3 L(8B) 3
6 
4 0 3 4

B3L 
6

6  (2c) 2

6  2  3 10 8  
2
 26 10 37 m(b/s)3 =
 
128  3 128  S   2 2 128  0.09 10 3 1.3 10 6  
4

= 2.6 10 35 km(b/s)3 = 2.6 10 8 km(Gb/s)3

25
 Let us check if Vω2<<1. The linewidth is about 10 MHz, according to page 117,
1
which results in    2  5 10 6  3 10 7 . Vω=1 when  6 10 7 , i.e. when
0
B  0.75 10 7 b/s

 The formula B 3 L  2.6 10 8 km (Gb/s) is thus not valid for small B. It is then
limited by the linewidth. However, that case is not interesting in practice.

Answer: Light diode: BL=15 km Gb/s, Single mode laser: B3L=2.6·108 km(Gb/s)3

2.17 Consider the “superformula” (2.4.23) with β3=0 and Vω=0. The dispersion
T 1
limit criterion,    . We can minimise σ2 with respect to  0
4 4B
2 2
 C 2 L    2 L 
 2    0     
 2 0   2 0 

d 
2
 2  0 
C 2 L  C 2 L 
 1
 L  L
 2 2  2 2
  2

  2    2 L  1  C
2

d 0 
2 0    2 0  2 0   0

 2 0 
2
  4 0 3 

d 2  2 L 1 C 2
 0   02 
d 0 2

d 2 2
 0  minimum
d 0 2

 2 2
 2 L 1  C 2  2C 2 L
 
  2 2 L

 
  2
 1     
2  2  2 L 1  C   2  2 L 1  C 
2 2
 
 

 2 L 1 C 2  2
2C 2 L

1  C L 1  C 2  C 
1 
    2
2  1 C 2  L 1 C 2 1 C 2   
 2 

because C  0 and  2  0

 2 L 1  C 2  C 
1
Thus, we have 
4B  

Compare with equation (2.4.30):

1
B
4  2 L 1  C 2  C 
 

26
 T
2.18 Consider the “superformula” (2.4.23) and the dispersion limit rule,   .
4
Further we have to find various values and possibly make some approximations.
 The wavelength λ=1.55 μm and standard fibre results in that the β3-term can
be neglected. β2=-20 ps2/km, according to page 53.
1

2
The linewidth 100MHz  indicates that V can be neglected.
140 ps

Let us calculate σ ω and σ 0.

t2

2 0 2
Transmitted power has the formula e . FWHM   is then 2 0 2 ln(2)


0   0.43  0.43  140  60 ps
8 ln(2)

Similarly,    0.43  (2  100)  270 Mrad/s

V  2   0  2  270  10 6  60  10 12  0.03  1  Vω can be neglected.

Thus,

2 2
 
2
 C 2 L    2
   1     2L    T 

    4 
 0   2 0 2   2 2
 0

  0 

 
2 2
 T   T 
   C  C 2    1 C 2 1 C 2
2L  4   1  
C C2  0  4 0 
    
2 0 2 1 C 2 1  C  2 2 
1 C 2
 1 C 2

 
2
 T 
 C    1  C 2 1

2 0 2  4 0 
L
2 1 C 2

Choose the minus sign because for C=0, L has to be >0 but β2<0

Numerically,

2
 
 1  16   1
1
4 
L

2  60  10 12 2
 12
 4  2.5  10  60  10 
9

 24 1  16
 20  10

 59 km

27
2.19 The waveguide dispersion, DW, is one part of the chromatic dispersion, and is
caused by a field redistribution between core and cladding when frequency
changes and hence the group velocity changes with frequency.

2.20 Light which penetrates into the optical fibre from outside will gradually leak
out because the condition for total reflection is not fulfilled.

lightning

fibre

2.21 In a multimode fibre the light propagates in several modes (hundreds) which
have different group velocity, and hence a pulse will be smeared out when it
propagates trugh the fibre. The mode dispersion is much larger than the
intramode dispersion and is dependent on the bandwidth of the pulse. The
smallest mode dispersion is achieved with a parabolic attenuation profile.

2.22 The two orthonormal modes do not have the same group velocity in a physical
single mode fibre. Random deformations in the fibre make the modes couple to
each other randomly which result in that the dispersion increases as length .

2.23
E-field E-field E-field

V  V 0 V

2.24

28
 log(length)

log(bit rate)

2.25 Let the point of intersection be denoted by BxLx. If you want to operate at
BxLx, with unchanged power, the eye, i.e. the Q-value, is decreased with about 1
dB and the BER is increased. Alternatively, the output power can be increased
by about 1 dB and the Q-value and BER are unchanged.

2.26 Rayleigh scattering. Small “in-frozen” density fluctuations give rise to

variations in index. When the light reaches these the light is scattered (similar to
light in fog). Attenuation ~1/λ4.

1
2.27 A LED has a large spectral width  L ~  the slope is -1 decade/decade.
B
1
A single mode laser  L ~  -2 decade/decade. See equations (2.4.26) and
B2
(2.4.30).

2.28 The limit for Brillouin scattering is reached, i.e. the light is increasingly
scattered backwards as the power increases.

2.29

The laser mode has got an odd field distribution compared to the fibre mode,
which results in that the laser mode cannot excite the fibre mode (they are
orthogonal).

29
3. Optical sources and transmitters

Problems

3.1 A laser is biased at 40 mA and has the threshold current 20 mA. The reflectance
of the mirrors is 0.3. The length of the cavity is 500 μm and the internal loss
coefficient is 10 cm-1. The reflectance of the left hand side mirror is increased to 1
by depositing a coating. How many percent does the output power increase from
the right hand side mirror compared to the first case? The bias current is still 40
mA. The spontaneous emission is neglected and the stimulated emission is
proportional to the number of the charge carriers, i.e. the transparency value is 0.
(Exam. 950601)

3.2 Two FP-lasers differs only by having different lengths. One of them is 300 μm
and the other one is 600 μm. The bias current is well above threshold current. For
a small change of the bias current of 1 mA, a change of the output power
throughout one of the mirrors is measured of 0.24 mW and 0.19 mW. What is the
internal efficiency and internal loss coefficient for the optical mode? Assume the
reflectivity is 0.32. The wavelength is 1.55 μm.
(Exam. 970306)

3.3 A Fabry-Perot laser diode shall be used as a pump laser for an optical fibre
amplifier. The given bias current I shall optimally be used, so that the optical,
outcoupled power through one of the mirrors is maximum. When manufacturing
the laser diode it is possible to choose the mirror reflectance within the interval 0
to 1. Both mirrors should be equal. What is the maximal outcoupled power to the
fibre amplifier? Use the linear gain model in equation (3.5.3), where GN and N0
are considered known. The charge carrier lifetime is τc. Internal losses/(unit
length) is αint. Rsp is neglected. The group velocity is vg. The confinement factor,
Г, and the internal efficiency, η, are 1.
(Exam. 010309)

3.4 A laser diode (LD), constructed as a Fabry-Perot etalon, is to be used as a pump

laser for an optical fibre amplifier. The given bias current, I, is to be converted in
the most beneficial way, with one of the mirrors as the outcoupling mirror, into
optical pump power to the fibre amplifier. When the laser diode is manufactured
the reflectivity of the two different mirrors are chosen in the interval 0 to 1, the
two mirrors can be chosen to have different values of the reflectivity. What is the
maximum outcoupled power to the fibre amplifier? Use a linear laser-gain model
where GN and N0 are assumed known. The carrier lifetime is  c . Internal
losses/duration is αint. Rsp can be neglected. The group velocity is vg. Internal
efficiency factor, η, is 1.
(Exam. 020823)

30
3.5 Calculate the so called turn on time for a semiconductor laser. If a laser is biased
under threshold and the current is increased by a current step so that the total
current exceeds the threshold current, then after certain time, the turn on time, the
laser starts to lase. This is due to that it takes time to fill up with charge carriers so
that the threshold is reached, whereafter the laser can be considered to
immediately start to lase. Assume that the spontaneous recombination into the
laser mode is negligible. Thus, calculate the turn on time for Ibias=5 mA,
Ithreshold=10 mA and Istep=10 mA. Further, the life time of the charge carriers is 1 ns
and 1 ps for the photons.
(Exam. 990830)

3.6 Consider an FP-laser with nonlinear gain parameter according to equation

(3.5.15). Assume that the bias current increases very little to a new stationary
value. This causes a change of the stationary number of photons and charge
carriers. Calculate these changes as a function of the current change. Assume that
the spontaneous emission into the lasing mode is negligible.

(Also the frequency is changed because of the linewidth factor. This is significant
for frequency modulation at low frequencies. However this matter is not included
in the assignment)
(Exam. 970411)

3.7 Derive an expression for the transfer function of the instantaneous frequency and
output power from one of the mirrors of a FP-laser. The bias values for the
number of photons, the number of charge carriers, current, phase and output
power through one of the mirrors are denoted Pb, Nb, Ib, Φb, Ppower,b.
G  G N ( N  N 0 ) and τp and τc are constants. Rsp and εNL are neglected.
(Exam. 980305)

3.8 When an ordinary semiconductor laser is current modulated it will chirp. Small
signal analysis can be used to achieve an understanding how the chirp is
dependent on the shape of the modulation. The assignment is to derive the transfer
function between the instantaneous angular frequency and the current as a
function of the modulation angular frequency.

Assume that the nonlinear-gain parameter is zero. Further, the laser is biased, with
the current Ib, essentially above threshold, It, which results in that the number of
charge carriers for transparency and the spontaneous emission into the lasing
mode can be neglected.

The answer may not contain the variables the number of photons or the number of
charge carriers in order to achieve maximal credits for this assignment.
(Exam. 960309)

31
3.9 The first part of the assignment is to use small signal analysis to find the transfer
function between the number of charge carriers and the current for a
semiconductor laser for low frequencies where resonance phenomena are
negligible. The bias values for the number of photons, the number of the charge
carriers, current and phase are Pb, Nb, Ib and Φb, respectively. The gain is
G  G N ( N  N 0 ) . The photon lifetime τp, and the carrier lifetime τc are constant.
The second part of the assignment is to find (1) The transfer function between the
number of photons and the current. (2) The transfer function between the
frequency chirp and the current for low modulation frequencies. Rsp and εNL are
neglected.
(Exam. 990309)

3.10 The modulation bandwidth for direct modulation of the intensity in a

semiconductor laser is approximately the same as the resonance frequency. The
resonance frequency is here defined as the frequency when the absolute value of a
transfer function is maximal. Derive an expression for the resonance frequency
with given parameters beginning with the rate equations. Assume the spontaneous
emission in the lasing mode is zero. Similarly the so-called “nonlinear-gain
parameter” equal to zero. Determine a numerical value of the resonance frequency
when the lifetime of the charge carriers is 1 ns, the photon lifetime is 1 ps and the
laser is biased at the double threshold current. G  G N ( N  N 0 ) where GN=6*103 s-1
and N0=108.
(Exam. 000901)

3.11 Small signal analysis for frequencies essentially lower than the resonant
frequency gives a good insight of how, for example, the clamping of the charge
carriers works. The first part of the assignment is to use small signal analysis to
find the resonance frequency of the transfer function between the number of
photons and the current for a semiconductor laser expressed in bias current Ib,
photon lifetime τp, carrier lifetime τc and differential gain GN. The second part of
the assignment is to find, for low frequencies where the resonance phenomena are
negligible, the following: (1) The transfer function between the number of photons
and the current. (2) The transfer function between the number of charge carriers
and the current. G  G N ( N  N 0 ) , where N 0 for simplification is set to 0. Further, τp
and τc are constants and τp << τc. Ib is twice the threshold current. Rsp can be
neglected.
(Exam. 020308)

32
3.12 For a laser the curves below have been measured.

normalised modulation response

at 50 mA bias current
output power
[mW]
10 5

10 20 30 40 50 current 8 frequency
[mA] [GHz]

Use these in order to determine the life time of the charge carriers τc and of the
photons τp. Use a simplified gain model, G  G N N , neglect Rsp and εNL.
(Exam. 970829)

3.13 The number of photons in a semiconductor laser varies according to

P(t )  P0 exp t 2 2 2 . How does the time derivative of the phase as a function of
time varies? Start with the rate equations for P, N, and dφ/dt. Neglect the
spontaneous emission in the lasing mode and use a linear gain model. Note that a
small signal analysis is not applicable because P0 is not considered small.
(Exam. 000307)

Short questions

3.14 Why is not the more mature Si-technology used for semiconductor
manufacturing?
(Exam. 960410)

3.15 Why is the modulation bandwidth of a LED much smaller than of a LD?
(Exam. 980828)

3.16 What are the similarities and differences in the construction between a DFB-
laser and a DBR-laser?
(Exam. 950831)

33
Solutions

3.1 The coating decreases the threshold and increases the differential quantum
efficiency

From equation (3.5.6):

q 1
I th   , N0=0
 c GN p

The photon lifetime is given by equation (3.5.4)

1  1 
 p 1  v g ( m   int ) with  m  ln  
2 L  RL RR 

The coating (RL=1) in case 2 will decrease the mirror loss coefficient αm

I th 2  p1  m 2   int 1  1  1  1 
  with  m1  ln   and  m 2  ln  
I th1  p 2  m1   int 2 L  RL RR  2 L  RR 

Equations (3.5.7) and (3.5.8) gives output power from theright mirror
p p
Pout1   g  m1 I  I th1  and Pout 2   g  m 2  I  I th 2 
1
2 q q

The factor 1/2 stems from the fact that only half of the power is emitted from right
mirror in case 1 (before the coating) whereas all the power is emitted in case 2

 m 2   int
I  I th1
Pout 2     int  m1   int
 2 m 2  m1 
Pout1  m1  m 2   int I  I th1 

1  1  1  1 
 m1  ln  2   ln    24 cm and  m 2  12 cm
-1 -1

2 L  0.3  0.05  0.3 

 m 2   int 12  10
 int  10 cm -1 and   0.65 we obtain
 m1   int 24  10

Pout 2 2 1 40  20  0.65
   2
Pout1 2 0.65 40  20

Answer: 2

34
3.2 Consider equation (3.5.10)
 int  mir
d 
 mir   int

1 1 1  int
   
d  int  int  mir

Equation (3.3.4) gives

1 1
 mir  ln  
L R
 1
 m
 int 
 int
1 1 1
    L  m  kL where 
d  int  int ln1 / R   1  int
k  
  int ln 1 / R 

Equation (3.5.9) gives

1 hc I 1 1
   0.8   1.67 for 300 μm
d 2q P 2 0.24

1 hc I 1 1
   0.8   2.10 for 600 μm
d 2q P 2 0.19

1/ηd

2 x
x

L
300 μm 600 μm

Extrapolation to L=0 gives 1 /  int  1.25 , i.e.  int  0.8

2.10  1.67 1  int

k  
300 μm  int ln1 / R 

0.43  0.8  ln1 / 0.32 

 int   1226 /m  12.3 /cm
300 10 6

35
3.3 Equation (3.5.9) gives

h  mir
Pe  ( I  I th )
2q  mir   int

q  1 

Equation (3.5.6): I th  
N0 

c  GN p 

Equation (3.5.4):  p 1  v g ( mir   int )

Equation (3.5.3): G  G N ( N  N 0 )

h  mir  qN 0 vg q 
 Pe  I 
   mir   int  (1)
2q  mir   int  c GN c 

From equation (1) we see that when αmir = 0 (R=1) minimal Ith is achieved, but
also minimal differential efficiency. If αmir is very large (R  0 ) the differential
efficiency is maximised, but then Ith is large. Thus, an optimal αmir exists.

Let
h qN 0 vg q
a, I b, c
2q c G N c

 mir
Pe  ab  ac mir
 mir   int

dPe  int
 ab  ac  0 
d mir  mir   int 2

b
  int   mir   int
c

  
2 
  b    b 
 int   int  c c  int
b  
  
  ac b     c 
int
Pe  a
c   
c
b
 int  1 
 c int int 
  int 
b
 int 
c
    int 
 
c    
 

36
 2
 
2  2
ac  b  
 1
 int  
 c 

  int   ab1  b  int
 int  c   
  b
 int   
 
 c 
2
 
 
h  qN 0   int vg q 
Answer: Pe   I  1 
  
2q   c  qN 0 G N  c

I
 c 
 

b
Comment: In order to achieve αmir>0 then it must hold  mir   int   int  0 , i.e.
c
b b  qN 0  v g q int  qN 0  c
 int   int ,   int   I  
   I  
 qv G N   int
c c   c  G N c  c  g

i.e. I must be at least as large so that the gain compensates for αint. Then αmir=0
and Ith is minimal.

3.4 Consider equation (3.5.9), which can be modified to be valid for different mirrors
h 1
Pe  I  I th  ,
q  1   2   int

where  1 and  2 are the mirror loss due to facet 1 and 2, respectively. The total
mirror loss is given by:
1 1
 mir   1   2  ln( ) , according to equation (3.3.4)
2L R1 R2

Equation (3.5.6):

q  1 

I th  N0 
 c  GN p 


Equation (3.5.4) 

  p 1   g  int   1   2 

h 1  gq 
 Pe  I 

qN 0
 1   2   int 
q  1   2   int  c GN c 

Observe that for a given α1, Pe is larger when  2  0 , that is R2  1 , that is no

light is coupled out from the second mirror.

h 1   gq 
 Pe  I 

qN 0
 1   int 
q  1   int  c GN c 

37
 h qN 0  gq
Let  a, I   b and c
q c GN c

1
Pe  a b  ac 1
 1   int
dPe  int
 ab  ac
d 1  1   int 2

b b
 int  1   int  1   int   int
c c

b
 int   int  
2
  int c 
Pe  a
c  b  c b  int   ab1 
 c   b 
b
 int    
c
 
 
h  qN 0   int gq 
 Pe   I  1  
 
 c 
q  qN 0 GN c 
I
  
 c 

3.5 Consider the rate equations (3.5.1)-(3.5.2). Before lasing, i.e. below threshold, the
number of photons, P, is zero, i.e.
dN I N
  , N<Nth
dt q c

The differential equation is analogous with the charging of a capacitor, via a

resistor

R
dQ Q
R u
u C dt C

Nth Nstep
Lasing starts
Nbias

0 T t

38
 dN I N I 
Before the current step ,  0  bias  bias  N  N bias  bias c
dt q c q

If there was no threshold, N  N bias  N step , at t  

I th c
But N clamps at the threshold N  N th 
q

The charging is governed by the time constant  c (compare with RC)

 N th  N bias  N step (1  exp( T /  c ))

N th  N bias  N step  I step 

  exp( T /  c )  T   c ln 
N step  I bias  I step  I th 
 

Note that if Ibias = Ith the turn on time is zero

10
In our case T  1 ln  0.7 ns
5  10  10

Answer: T=0.7 ns

3.6 Consider the rate equations (3.5.1) and (3.5.2). Complement with equation
(3.5.15), which gives

dP p  1 
 GP   G   P (1)
dt  P  P 
dN I N
   GP (2)
dt q c

G  G N ( N  N 0 )1   NL P  (3)

We solve the transfer functions and use small signal analysis and consider steady
d
state, i.e. 0.
dt

P  Pb  P
N  N b  N
I  I b  I

Equation (1) and (2) then result in

Pb  P
0  G N ( N b  N  N 0 )1   NL Pb  P Pb  P   (4)
p

39
 I b  I N b  N
0   G N ( N b  N  N 0 )1   NL ( Pb  P Pb  P  (5)
q c

Linearise and separate

Equation (4) results in

P
0  G N ( N b  N 0 )P  2 NL Pb P   G N N 1   NL Pb Pb  (6)
p

(5)+(4):

I N P
0   (7)
q c  p

1
Simplify equation (6) by using G  0 above threshold:
p
0  GN ( Nb  N0 ) NL Pb P  GN N 1   NL Pb Pb (8)

Solve for ΔP:

G N 1   NL Pb Pb 1   NL Pb 
 P  N  N (9)
G N ( N b  N 0 ) NL Pb ( N b  N 0 ) NL

Insert equation (9) in equation (7) and solve for ΔN:

I 1
 N  
q 1 1 1   NL Pb 

c  p ( N b  N 0 ) NL

which inserted in equation (9) gives:

I 1
 P  
q ( N  N 0 ) NL 1
b

1   NL Pb  c  p

As a confirmation we see that when  NL  0 , ΔN is not changed, i.e. the charge

carriers are clamped.

3.7 Consider equations (3.5.1)-(3.5.3) and (3.5.16), which gives

dP p
 GP  (1)
dt P
dN I N
   GP (2)
dt q c

d 1  1 

 c G  (3)
dt 2  p 


40
 G  GN (N  N 0 ) (4)

Instantaneous frequency,

1 d
f 
2 dt

and it holds Ppower~P

equations (1) and (3) gives

dP
d 1
  c dt
dt 2 P

d
Replace  with  b   , with j etc. Linearise and separate
dt

1 jP 1 jP 1 jPpower

 j  c  {linearise}   c  c  2f
2 Pb  P 2 Pb 2 Ppower ,b

f ( )  j
Answer:  c
Ppower ( ) 4 Ppower ,b

3.8 Consider the rate equations,

dP p
 GP  equation (3.5.1)
dt P
dN I N
   GP equation (3.5.2)
dt q c

d 1  1 

 c G  equation (3.5.16)
dt 2  p 


d
We solve the transfer functions and use small signal analysis. Let  j .
dt
P  Pb  P
N  N b  N
I  I b  I
   b  

where P, N and I are time dependent. and Pb , N b and I b are independent of
time. Separate the rate equations in time independent and time dependent parts
and neglect the 2nd order terms

1
jP  G N ( N t  N 0 )P  G N NPb  P (1)
p

41
  1 
0  G N ( N t  N 0 )   Pb (2)
  p 

I N
jN    G N ( N t  N 0 )P  G N Pb N (3)
q c

Ib Nb
0   G N ( N t  N 0 ) Pb (4)
q c

1
   c G N N (5)
2

1
0  GN (N t  N 0 )  (6)
p

We want  / I , and so we eliminate ΔP and express ΔN in ΔI.

Equation (6) or (2) in (1) results in

jP  G N NPb (7)

Equation (7) in (3) results in

 1 G P  I
 j   G N Pb  G N ( N t  N 0 ) N b   N  (8)
  c j   q

Equation (8) in (5) gives

 1 1 j
  cGN (9)
I 2 q  1 
 j  2
   G N Pb  j  G N N t  N 0 G N Pb
 c 

Express Pb in current variables

p
G N 2 N t  N 0 Pb  {equation (3.5.7) or (4) and (2)}  G N 2 N t  N 0  I b  I th  
q

 p  Ib  I th 2 p c  Ib 
 G N 2 N t  N 0 I th   1  {N t  N 0 }  G N 2  
 I  1  {equation (3.5.6)} 
q  I th 
 q2  th 

2
 q
2 1   p c  Ib  1  Ib 
 GN     
  c GN p  q2  I  1     I  1
   th  p c  th 

and

42
  1  1 p p  Ib 
   G N q I b  I th     G N q I th  I  1  {N t  N 0 } 
1
  G N Pb    

 c  c c  th 

1 p q 1  Ib  Ib
  GN  
c q  c GN p  I  1   I
 th  c th

 1 1 j
Answer:   cGN
I 2 q 1 Ib  Ib 
 j  2  j 
1
 
 I  1
 c I th  p c  th 

3.9 Consider the rate equations,

dP p
 GP  equation (3.5.1)
dt P
dN I N
   GP equation (3.5.2)
dt q c

d 1  1 

 c G  equation (3.5.16)
dt 2  p 


We solve the transfer functions and use small signal analysis with the time
d
dependence e jt   j . Let therefore
dt

P  Pb  P
N  N b  N
I  I b  I
G  Gb  G  G N ( N b  N  N 0 )  Gb  G N N
1 d
   b   ; is the frequency chirp = Δν
2 dt

where P, N and I have the time dependence e jt and Pb , N b and I b are
independent of time. Separate the rate equations in time independent and time
dependent parts and neglect the 2nd order terms

 1 
0   Gb  Pb (1) jP  G N Pb N (2)
  p 

I N I N
0  b  b  G b Pb (3) jN    Gb P  G N Pb N (4)
q c q c

1
   c G N N (5)
4

If we find N / I , the others will be easy to evaluate.

43
  1 G N Pb  I
(4) and (2) result in  j   G N Pb  Gb  N 
 c j  q

N 1/ q
  { (1) }
I 1  G P
j    G N Pb   Gb N b
 c  j
j / q j / q j p
  { small}   , i.e. N is clamped
 1  G P G N Pb /  p G N Pb q
     G N Pb  j  N b
2

 c  p
when   0

P P N G N Pb j p p
    , i.e. for each electron which is fed in by the
I N I j G N Pb q q
current, one photon is created.

  N 1 j p j c p
   cGN 
I N I 4 G N Pb q 4Pb q

Thus, ΔP is in phase with ΔI and ΔN and Δν are in phase with the derivative of ΔI

N j p P  p  j c p
Answer:  ;  ; 
I G N Pb q I q I 4Pb q

3.10 Equations (3.5.1)-(3.5.3) with Rsp=0

dP P
  GP 
dt p

dN I N
   GP
dt q c

G  GN (N  N 0 )

We want to find the transfer function resonance peak between the output power
and current. While the output power is proportional to the number of photons it is
sufficient to find ΔP(ω)/ΔI(ω), where ΔP and ΔI are small sinusoidal variations.
Replace for example P with Pb+ΔP, where Pb is time independent bias value and
ΔP is time dependent exp(jωt). Separate the equations in time independent (a,b)
and time dependent (c,d) parts and neglect the second order terms.

Pb 1
a) 0  Gb Pb   Gb 
p p

44
 Ib Nb p  N q Nbq
b) 0    Gb Pb  Pb   Ib  b  , where I th 
q c   c  c
q 

P
c) jP  G N NPb  G b P   G N NPb
p

I N
d) jN    G N N b P  G N NPb
q c

Eliminate ΔN I c), and d)

P G N Pb / q
 
I  
   2  G N Pb   j  1  G N Pb 
  p   
  c 

P
The resonance frequency is here then defined as
I max

We differentiate the absolute value squared of the denominator with respect to ω

and set it to zero.

 
2
G P
2   2  N b  2   2  1  G N Pb   0
 p   
   c 

2
G N Pb 1 1 
 res 2     G N Pb 
p 2   c 


Pb can be expressed by Ib and Ith according to b) above

2
G N I  1 1 p I 
 res 2
 N b  b  1    G N N b  b  1 

 c  I th  2   c  c 
 I th 

 1 
and Gb  G N N th  N 0   G N N b  N 0  
1
 GN Nb    GN N 0 
p  p 
 

2
 1  1  Ib  1  1  p  1  I b  
 res 2
  GN N 0     
 I  1  2        G N N 0  I  1 
 p c  th   
   c c  p  th

 
 res 2  1012  6 1011 10 9 2  1  
1 9
2
  2

10  10 3 1012  6 1011 2  1  16 10 20

   4 1010 rad/s

Answer: fres = ωres/2π = 6.4 GHz

45
3.11 Consider equations (3.5.1)-(3.5.3) with the given conditions,
dP p
 GP 
dt P
dN I N
   GP
dt q c

G  GN N

We solve the transfer functions and use small signal analysis with the time
dependence e jt . Let therefore

P  Pb  P
N  N b  N
I  I b  I

where P, N and I have the time dependence e jt and Pb , N b and I b are
independent of time. Separate the rate equations in time independent and time
dependent parts and neglect the 2nd order terms

Pb P
0  G N N b Pb  (a) jP  G N N b P  G N NPb   G N NPb (c)
p p
Ib Nb I N
0   G N N b Pb (b) jN    G N N b P  G N NPb (d)
q c q c

The goal is to eliminate N b , Pb and N

1
(a) results in G N N b 
p
Pb Ib 1 1 I th
(b) results in   , where is identified with
p q G N  p c G N  p c q
(c) results in jP  G N NPb
 1  I P
(d) results in  j   G N Pb N  
 c  q p
 1  I P
where  j   G N Pb  jP / G N Pb 
 c  q p
P G N Pb / q
 
I G N Pb  1 
 2   j   G N Pb 
p 
 c 
GN
 p ( I b  I th )
q2

GN  1 GN p 
 2  ( I b  I th )  j   ( I b  I th ) 
q  c q 

46
 1  Ib 
  1

q c  I th 
 (1)
1  I  j I b
 2   b  1 
 p c  I th   c I th

2
P
The resonance frequency is achieved when is maximal, i.e. when the
I
denominator is minimal. Equation (1) gives

 2 2
1  Ib     I b  
d    2  
I  1 
   I  

  p c  th    c th 
   
2 2
d I   Ib 2  Ib  !
  2  
1
  2   2  
I  1

I
 0

d d   p c  th   c2  th 


2
1  Ib  1 1  Ib 
 res 2    1   
 p c  I th  2  c 2 I
 th



1  Ib  1 Ib
Answer:  res 2    1  for τp << τc and 2

 p c  I th   p c I th

Thus, ωres coincides approximately with the imaginary part of the pole of the
transfer function.

P  p
For low frequencies 
I q

P  p
Further, jP  G N NPb gives with 
I q

G N NPb
j p N 1 p j j jG N  p c
   j   
I q I G N Pb q G N I b  I th  I  I 
G N I th  b  1 G N q b  1
 I th   I th 

N j p c
 , i.e. N  0 when   0 , i.e. N is clamped
I q

3.12 Consider equation (3.5.24), ГR<<ΩR, the resonance angular frequency is then
GN I 
r2  I th  b  1
q  I th 

Equation (3.5.6), with N0 negligible

47
 1  Ib 
 r2    1

 p c  I th 

According to the curves

 r  2  8 10 9 rad/s and I b / I th  50 / 15

1  50 
  p c    1
2  8 10  9 2  15 

The right curve is determined by equation (3.5.19). From the curve we can read
r
out H ( r )  5
2 R

1
2R  {equation (3.5.20) with P  0 }   G N Pb  {equation (3.5.7) and (3.5.6)} 
c

1 p I  1 Ib
  GN I th  b  1 
c q  I th   c I th

 r c
 H ( r )  5
I b I th

5 50
c   0.33 ns
2  8 10 
9 2 15

  p  2.8 ps

3.13 The stated conditions gives

dP P
 GP  equation (3.5.1)
dt p

dN I N
   GP equation (3.5.2)
dt q c

d 1  1 
    c  G   equation (3.5.16)
dt 2  p 


G  GN (N  N 0 ) equation (3.5.3)

R sp and  NL = 0

48
   2
P  ln P    G 
dP d 1  
dt dt    c
 p 

Thus,    c
1 d
ln P 
2 dt

When P(t )  P0 exp t 2 2 2  

1  t 
Answer:    c   
2  2

Note: If  is replaced by the width of the field: T0  2 and compare with

equation (2.4.12)

 ct Ct
   2
 , i.e.  c  C !
T0 T0 2

3.14 Si (and Ge) has an indirect bandgap and the internal quantum efficiency is
very poor, <<1. For direct bandgap semiconductors, such as GaAs the internal
quantum efficiency ≈0.5 and approaches 1 when stimulated emission dominates.

3.15 For a LED the modulation bandwidth is determined by the lifetime of the
charge carriers, τc. For a LD the corresponding life time of the charge carriers can
be much shorter, due to the stimulated recombination.

3.16 Both lasers have Bragg gitters. In the DBR case they are passive and works as
mirrors. In the DFB case the gitter is active and distributed through the whole
active cavity.

49
4. Optical detectors and receivers

Problems

4.1 Consider the PIN-diode detector structure in Figure 4.5, i.e. a detector where the
incoming light reaches the i-region after passing the p-region. It is assumed that
the absorption only takes place in the i-region. The rise time of a detector can be
described by a time constant τ, which is the rms-value of a total impulse response.

 2   RC 2   tr 2

 RC is the rms-vale of the impulse response for the RC-circuit which consists of
the capacitance of the PIN- diode and a resistance R=50 Ω in the outer circuit, and
 RC =RC. The capacity C of the diode is calculated in the same way as for a planar
capacitor. The i-region has the length W and the area is A = 100 μm2. The relative
dielectric constant is εr = 9.  tr is the rms-value of the impulse response for the
outer current, which is created by the generated charge carriers of an optical
impulse, if the RC-filter is omitted. In this assignment this impulse response is
approximated with a rectangular pulse with the width T, where T is the time for
charge carriers with the velocity vs=107 cm/s to completely pass the I-region.
a) Show that  tr 2  T 2 / 12
b) Assume that w is chosen so that τ is minimised. What is then the efficiency if
the depletion region has the absorption coefficient 104 cm-1.
(Exam. 980423)

4.2 Find the impulse response for a pin-diode which is caused by the drifting charge
carriers in the intrinsic region, i-region. RC-effects and other effects are neglected.
Study the pin-diode detector in Fig. 4.5. At time t=0 the detector is illuminated
with a very short light pulse, containing N photons, each with the energy hν. This
corresponds to a δ-pulse with the area Nhν. It is assumed that at t=0, ηN electron-
hole pairs are instantaneously produced in the i-region, which is W long, and that
these are uniformly distributed in the i-region. The efficiency is η<1. The drifting
velocity for both holes and electrons is v. What is the impulse response for the
current in the outer circuit? Hint: It holds that a charge carrier gives rise to a
current as long as it is drifting, and that current is qv/W. (Exam. 010423)

4.3 Consider the PIN-diode detector structure in Figure 4.5(a). The P-region and the N-
region are negligible and the absorption within these regions can be neglected.
The I-region has the absorption coefficient α and the length w, which is a few μm.
At the time instant t=0, the detector is illuminated from the P-region with a very
short light impulse, which contains N photons. Each photon which is absorbed
creates at t=0 immediately an electron-hole pair (i.e. the time the photons travels
through the I-region is neglected). The drifting velocity for both holes and
electrons is v. What is the efficiency, η? What is the current in the outer circuit at

50
 t=0? How does the current vary with time in the outer circuit? Hint: A charge
carrier gives rise to a current as long as it is drifting and that current is qv/w.
Ignore RC-effects.
(Exam. 020408)

4.4 We have an loss limited digital fibre optical link with the following data:

Data rate = 2 Gbit/s

Code: NRZ, equal probability for 1 and 0
Output power from the laser transmitter when a 1 is transmitted = 10 mW
Extinction ratio = 0.1
Wavelength = 1.3 μm
Coupling loss between laser and fibre = 3 dB
Fibre attenuation = 1 dB/km
Fibre length = 34 km
Coupling loss between fibre and detector = 0 dB
p-i-n photodetector quantum efficieny = 80%
Receiver noise equivalent bandwidth = 1 GHz
Temperature = 290 K
Equivalent load resistance = 1 kΩ
Amplifier noise factor = 3 dB

a) Determine the bit error rate (BER) with optimum threshold.

b) We now replace the p-i-n photodetector by an APD with FA(M)=M0.7 and

apply a reverse bias voltage so that M = 300. We also adjust the threshold so
that the error probability is equal for a received 1 and 0. Assume that the
quantum efficiency, the coupling losses, the receiver bandwidth, and the
equivalent load resistance remain the same. What is now the BER?

4.5 Determine the minimum detected power for a ONE in order to achieve BER=10-9
by choosing optimal APD amplification. Also determine the optimal
amplification.

The bit rate B is 2.5 Gb/s. The ONE bit is rectangular and has the length 1/B. The
ZERO bit does not have any power. Noise bandwidth is 0.7B. The electronic
amplifier has the noise factor 3 dB and the load resistance 1 kΩ. The responsivity
is 1 A/W (exclusive the APD amplification) and the multiplication factor is M0.5
where M is the APD amplification. The dark current can be neglected.
(Exam. 000901)

4.6 A pin-diode detector and an APD-detector are to be compared. Calculate the

minimum detected power for ONES that are needed in respective case in order to
achieve BER=10-9. ZEROS have zero power. The data rate is 2.5 Gb/s and the
signal format is NRZ. The noise equivalent bandwidth is assumed to be 2 GHz.
The load resistance of the diode is 1000 ohm. The noise factor of the amplifier is
not given, on the other hand the spectral efficiency of the equivalent noise current

51
 at the input of the amplifier is given to be 9 pA2/Hz. Both detectors have a
quantum efficiency of 70%. The APD has an amplification of 100 and the
ionisation coefficient for the holes is zero. The dark current can be neglected. The
wavelength is 1.55 μm.
(Exam. 000505)

4.7
a) Calculate some characteristics for a receiver with an APD. Determine the
minimum detected power for a ONE signal in order to achieve a BER=10-9 by
using an optimal APD-amplification. Also determine the ratio between the
threshold level and the mean level for the sampled ONE signal. Calculate also
the thermal noise percentage of the total noise in a ONE signal.

 The ONE signal has a rectangular shape and the duration of the pulse
is T=1 μs, where T is bit-slot
 No power in the ZERO signal
 The bandwidth for the noise calculations Δf=1/2T
 The amplifier has the noise factor 2
 The excess noise factor, FA(M)=M
 The responsivity exclusive internal amplification  =0.5 A/W
 The load resistance RL=1 kΩ
 Use Gaussian statistics
 The dark current can be neglected

b) What are the corresponding three quantities for a receiver where the APD has
been replaced with a PIN-diode and otherwise the same conditions.

(Exam. 950601)

4.8 A direct detection receiver has an APD (dark current is zero) and is followed by
an electronic amplifier. The later provides with additive noise, which has, related
to the receiver input, the single sided spectral density S [A2/Hz]. The noise
bandwidth is Δf. Derive an expression for the optimal multiplication factor M for
a BER=10-9. The ionization coefficient ratio kA is in the interval 0 to 1. The
extinction ratio between 0 and 1 bits is zero. Start with the equation (4.5.16) or
earlier equations. Assume S>>q2Δf.

Assume further that the APD has the worst possible noise features. What is
optimal Mopt and minimal input power for a 1 bit?

Assume instead that the APD has the best noise features possible. What is the
value of M, expressed in Mopt, in order to achieve a less input power than in the
previous case above?
(Exam. 970411)

52
4.9 A direct detection receiver has an APD with the best possible excess noise factor
FA and multiplication factor M and zero dark current. The detector is followed by
an electronic amplifier which provides an additive noise current, which has,
related to the receiver input, the single sided spectral density S [A2/Hz]. The noise
bandwidth is Δf. Derive an expression for the minimal detected optical power for
a 1 bit for BER=10-9 when the APD-gain is >>1. The extinction ratio between 0
and 1 bits is   0 .
(Exam. 970829)

Short questions

4.10 By decreasing the intrinsic region of a PIN-diode the bandwidth may increase.
At the expense of what?
(Exam. 950601)

4.11 The receiver unit between the detector and the sampler acts as a filter. What
are the purposes of the filter?
(Exam. 980305)

4.12 How should an APD be designed so that the excess noise factor is as small as
possible, and how small is it for large M?
(Exam. 990830)

53
Solutions

4.1
optical power

~exp(-αx)

0 W
x

According to the assumptions the efficiency is

optical power into the I  region  optical power out from the I  region
  1  e W
optical power into the I - region

Let us find  RC and  tr .

The impulse response for a RC-circuit is ~ exp( t / RC ) , t>0, and its rms-value is
RA r  0
 RC  RC 
W

The impulse response for the outer current of the charge carriers is approximated
W
with a rectangle with the width T  .
vs

Normalise the area of the impulse response to 1 and place t=0 symmetrically, the
second moment is then

vs/W
W / 2v s 3 v s  W 2

}
W / 2vs
 tr 
2
{ t
2 
0
2v
t s dt  2
W 3

W  v s
 1

 12  T / 12

2

W/vs

k1 W
Thus,  RC  , where k1  RA r  0 and  tr  , where k 2  v s 12
W k2

2 2
k  W 
  2   1    
 W   k2 

d 2
2
2k 2W !
  13  2  0
dW W k2

 W 4  k1 2 k 2 2  W  k1 k 2

54
 10 9
W  50 100 10 12  9  10 5  12  3.7 10  7 m  3.7 10 5 cm
36

 W  3.7 10 5 10 4  0.37

   1  exp(0.37)  0.31

4.2 At the time t=0, N photons, with the total energy Nhν, excite ηN electrons and
holes with the following density distribution:

holes/unit length electrons/unit length

N W N W t=0

0 W W

At the time W/v·1/2 the distribution is

holes/unit length electrons/unit length

N W N W t=W/v·1/2

0 W W

We see that the number of electrons and holes which are drifting are both
 
 
 t 
decreasing as N 1  ; t<W/v
 W 
   
  v 
τ=W/v
While holes and electrons are of opposite charge and goes in opposite directions
the outer current can be added

 
 
qv  t  q  t
i  2  N 1    2  N 1   ; t<τ
W  W    
   
  v 

The impulse response is normalised to a unit impulse by dividing with Nhν

55
 q  t 
i2 1   , t<τ,
h   
q
where is recognized as the DC-responsivity
h

q  t 
Answer: i  2 1  
h   

dn( x)
4.3 At t=0 photons are absorbed according to  n( x)  n( x)  n(0)e x
dx

absorbed photons and el-hole

pairs created at t=0
~exp(-αx)

0 W

 n(0)e
x
dx n(0)
1  e 
W

 0

 
n(0)

 1  e W 
 n(0)e
x
dx
0


n ( 0)
Observe that N


The holes drift to the left and the electrons drift to the right. At t=0 there are ηN
holes and ηN electrons, which are drifting with v and give rise to the current
q q
ηN . Total current at t=0 is thus 2 ηN .
W W

W W
At 0  t    the distributions are according to the figure. At   all the
 
charge carriers are drifted out.

56
 holes electrons

t/τ·W t/τ·W

0 W 0 W

Number of drifting holes, 0<t<τ

 N e t / W  e W 
 1 
W


x
N e dx  N  e x 
t/ W    t / W

Number of drifting electrons, 0<t<τ

1t /  W
N e
x

dx N 1  e 1t /  W 
0

Answer: I tot t   N
q

e t / W
 e W  1  e 1t /  W 
Following figures illustrate some cases
I
I I 
 
p-current e-current
e current h- and e-current
h-current

τ t τ t τ t
intermediate α    0

4.4
a) Attenuation between laser-detector: 3  34 1  37 dB  a factor of 5000

rex  0.1

10 10 3
Received power in “1”:  2 μW
5 10 3

1
Received power in “0”:  2 μW  0.2 μW
10

Signal currents (before amplifier)

57
 q
I 1  P1  P1  1.68 μA
hc

I 0  0.17 μA

1 2  2qP1 f 
4k B TFn f
RL
 5.4 10 16  32 
. 10 14
 A
2
 
dominating

02 
1
10
 5.4 10 16  32 
. 10 14
 A
2
 
dominating

 1   0   T  and 0.18 μA, i.e. 3.2  10 14

I 1  I 0 
lowest BER when I D 
2

BER 
1


exp Q 2 / 2   Q  I 1  I 0 
 4.2  1.4 10 5

2
Q
 2 T 

b) M = 300
FA(M) = M0.7
I 1  P1 M  500 μA
1
I0   500 μA  50 μA
10

1 2  2qP1 M 2.7 f   T 2  2
. 9
10
6   3.2 10
14
A2  
do min ating

 0 2  26 
. 10
10  3.2 10
-14
A2  
do min ating

I1  I 0
Q  6.72
1  0

 BER  9.3 10 10

4.5 Use Gaussian statistics

1  Q2 
BER = 10-9  exp    Q=6
 2 
2Q 2  

I1  I 0 I1
Q 
1  0 1  0

58
 4k B TFn f
I1  M    P ,  02  ,  1 2   0 2   s 2   0 2  2q  M 2.5    P  f
RL

Square:
( I 1  Q 0 ) 2  (Q 1 ) 2

I 1 2  2IQ 0  (Q 0 ) 2  (Q 0 ) 2  Q 2 s 2

( M    P) 2  2  M    P  Q   0  Q 2  2q  M 2.5    P  f

Q 2  2q  M 1.5  f  2Q 0
P  aM 1 / 2  bM 1
M 

dP 1 !
 aM 1 / 2  bM  2  0
dM 2

 2Q 0 
2  2/3
   2 0
2/3
 2b  
 M opt     2    Qqf

 a  Q 2qf
   
  

Q 2 2qf 2Q 0
Pmin  M opt1 / 2  2 M opt 1
 

Q 2 2qf 36  2 1.6 10 19  0.7  2.5 10 9

  2.02 10 8
 1

4 1.38 23  290  2  0.7  2.5 10 9

226
2Q 0 1000
2   5.68 10 6
 1

2/3
 5.68 10 6 
M opt     43
 2.02 10 8 
 

Pmin  2.02 10 8  6.56  5.68 10 6 / 43  0.26 μW

Answer: Mopt=43, Pmin=0.26 μW

4.6 Use the Gaussian statistics

I1  I 0 I1 -9
Q   6 for BER=10
1  0 1  0

I 1    P1  M

q

h

59
M  1 for the case with the pin-diode and M  100 for the case with the APD.

4k B Tf
 02   S i f
RL

 1 2   0 2  2qP1 M 2 FA (M )f

FA ( M )  1 for the case with the pin-diode and equation (4.4.18) gives

FA ( M )  k A M  (1  k A )(2  1 / M ) , but k A  0  FA ( M )  2  1 / M  {M  1}  2

Let us calculate P1min when Q=Q0=6

I1
Q0   Q 0  1  I 1  Q0  0
1  0

 
Q0 2  0 2  2qP1 M 2 FA ( M )f  P1 M 2  (Q0 0 ) 2  2P1 MQ0 0

Divide by P1 M :

Q0 2 2qMFA ( M )f  P1 M  2Q0 0

P1 
1
 M

Q0 2 MF A ( M )f  2Q0 0  
Q 2 0 
  2qF A ( M )Qf  , compare with equation (4.5.17)
 M 

 q 0.7 1.6 10 19 1.55 10 6

   0.88 A/W
hc 6.626 10 34  3 10 8

4k B Tf
 02   S i f 
RL

4 1.38 10 23  290  2 10 9

 9 10  24  2 10 9  3.20 10 14  1.8 10 14  5 10 14 A
2

1000

The pin-diode case

6  
P1  2 1.6 10 19 1  6  2 10 9 
0.88 
2  5 10 14
1
 
  6.8  1.2 10 9  4.5 10 7  {thermal

 
noise dominates}  6.8  4.5 10 7  3.110 6 W

60
 The APD case


P1  6.82 1.6 10 19  2  6  2 10 9 
4.5 10 7 
 9
 9 8
  6.8  7.7 10  4.5 10  10 W
 100 

Answer: pin-diode 3.110 6 W, APD 10 8 W

4.7

σ0
σ1

I0 Ith I1
a)

I 1  P1 M

I0  0

 1 2   s 2   T 2  2qP1 M 3 B   T 2

4k B TFn B
 02   T 2 
RL

The threshold Ith is determined approximately by letting the error areas be equal


 2  I th
  I  I 1 2   Q2 
1   I dI 
  2 2 
exp
1

   2 2 dI  BER  10 
exp -9
 1
exp   
2 0
2
I th  0  2 0
2
 0  2Q 2  2 

I th I 1  I th
Q6 
0 1

 0 I1 I1
 I th  and  Q  6  compare with equations (4.5.9) and (4.5.11)
 0 1  0 1

Hence,
I1  Q 0 2  Q 1 2

I 1 2  2I 1Q 0  Q 0 2  Q 0 2  Q s 2

P1 M 2  2P1 M Q 0  Q 2 2qP1 M 3 B

61
 dP1  02 Q
 0 , gives M opt  and Pmin  4 qBQ 0
2
Differentiation and
dM qBQ  0 
Numerically,

4 1.38 10 23  290  2  0.5 10 6

 02  3
 1.6 10 17
10

qBQ 0  1.6 10 19  0.5 10 6  6  4 10 9  1.9 10 21

1.6 10 17

M opt   91
1.9 10  21

Q 6
PAPD  4 qBQ 0  4 1.9 10  21  2.1 nW
 0.5

2qB  0.5  2 1.6 10 19  0.5 10 6  0.8 10 13

Noise quotient:
1 1 1
q APD     0.14  14 %
PAPD M 3
2qB  1
2.1 10 9
 91  0.8 10
3 13 1  6.3
1
0 2
1.6 10 17

I th 1
  0.27
I APD 1
1
q APD

b)
Here M=1
PPIN  2
Q
 0

qBQ 0    0 2  2 6
0.5  4 10

9

1.9 10  21  1.6 10 17  9.6 10 8  96 nW 

Noise quotient:
1 1
q PIN    1  100 %
2qPPIN B 96 10 9
 0.8 10 13
1 1
 02 1.6 10 17

I th 1
  0. 5
I PIN 1
1
q PIN

I th
Answer: a) PAPD  2.1 nW ,  0.27 , q APD  14 %
I APD
I th
b) PPIN  96 nW , q PIN  100 % ,  0.5
I PIN

62
 Note in the APD-case both signal and shot noise are amplified and the shot noise
will reach the same order as the thermal noise.

I1  I 0 I1
4.8 BER=10-9  Q=6=  {I 0  0} 
1   0 1   0

I 1  P1 M

 0 2   T 2  Sf

 1 2  2qP1 M 2 FA (M )f   T 2

P1 M
Q
2qP1 M 2 F A ( M )f   T 2   T

 Q 2qP1 M 2 F A ( M )f   T 2  P1 M  Q T

Square:
 
 Q 2 2qP1 M 2 FA ( M )f   T 2  P1 M 2  Q 2 T 2  2P1 MQ T

Divide with P1 M 2

2Q T
 Q 2 2qFA ( M )f  P1 
M

1 2 2Q T 
 P1   Q 2qF A ( M )f  
  M 

Compare with equation (4.5.17). Differentiate

dP1 Q 2 2qf dF A ( M ) 2Q T Q 2 2qf  1  2Q T !

     {equation (4.4.18) }   k A  1  k A   0
dM  dM M 2   M2 M

M opt 2 Q 2qfk A  2 T  k A  1Q 2qf

1 2 T 1 2 Sf 1 S 1 S
 M opt    k A 1    k A 1    k A 1  
kA Q 2qf kA Q 2qf kA Qq f kA Qq f

because S>>q2Δf and kA<1 and Q=Order(1)

The largest noise is achieved for kA=1.

63
 S
 M opt  and FA  M opt
Qq f

1  2 2Q Sf 

 Pmin  Q 2qM opt f 
 M opt 
 

The smallest noise is achieved for kA=0, and M opt   . However, only a limited
M is required in order to achieve less than Pmin in the previous case.

 1 
FA   2  2, due to M>>1 is assumed.
 M 

1  2 2Q Sf 

P Q 2q 2f 
 M 
 

Let Pmin  P 

2Q Sf 2Q Sf

M

 Q 2 2q M opt  2 f   M opt

1

Qqf

M opt  2 
1

M Sf M opt

 
1

1
M M opt 2

M opt  2 
1

2

M opt M opt
1  1
 M opt


 

M opt 1 M opt
M 
2   2
1  1 
 M opt 
 

I1  I 0
4.9 BER=10-9  Q=6=
1  0

I 1  P1 M

I 0  P1 M

 1 2   s 2   T 2  2qP1 M 2 FA (M )f  Sf

 0 2   s 2   T 2

64
 where FA (M ) is the best possible, i.e. FA (M )  2 see page 160.

I 1  I 0  Q 0  Q 1

Take the square

I1  I 0 2  2I 1  I 0 Q 0  Q 2 I 1  I 0 2qMFA f

1   I1  Q 2 2qMFA f  2Q 0

Square again:
1   2 I12  21   I 1Q 2 2qMFA f  Q 4 2qMFA f 2  4Q 2 2qMFA fI 1   T 2 

1   2 I12  21   I 1Q 2 2qMFA f  Q 4 2qMFA f 2  4Q 2 T 2

P1 M  I 1 
1   Q 2 2qMFA f

1   Q 2
2qMF A f 
2

4Q 2 T 2  Q 4 2qMF A f 2

1   2 1   4 1   2

1   Q 2 2qMFA f  4Q 4 2qMF A f 2  4Q 2 T 2 1   2


1   2

1   Q 2 2qFA f  4Q 4 2qFA f 2  4Q 2 T 2 1   2 M 2

Answer: P1 
1   2 

4.10 The depletion width determines drift-time and efficiency. Decreased drift-time
 Decreased efficiency. Also the capacitance increases, which increases the RC-
constant which counteract the bandwidth increase.

4.11 Between the detector and the sampler there is a lowpass filter in order to
maximise the SNR at the sampler.

4.12 The avalanche should for example start with only injected holes and the holes
only should create new electron-hole pairs. Then FA = 2 when M is large. See
Figure 4.16.

65
5. Lightwave systems

Problems

5.1 What is the maximum transmission distance in the fibre-optical communication

system specified below? Is the system loss or dispersion limited? What is the real
system margin?

Transmitter: 1.55 μm DFB laser with 0 dBm output from the fibre pigtail.
The spectral width is ΔfFWHM≈10·B, where B is the bit rate.
Rise time = 0.1/B
Receiver: Sensitivity = -30 dBm for the required BER.
Rise time = 100 ps.
Fibre: Attenuation = 0.3 dB/km.
Splice loss = 0.1 dB.
Dispersion = 18 ps/(nm km) at 1. 55 μm.
Available fibre length = 10 km (full lengths have to be used)
System: One fibre connector at the transmitter and one at the receiver
with connector loss = 1 dB each.
Minimum system margin = 10 dB.
Modulation: 1 Gbit/s RZ.

5.2 Calculate the mean number of photons, which are needed in order to achieve a bit
error rate of 10-9 for two different cases.

In the first case the detector is an APD with a very large multiplication factor and
best possible excess noise factor and the efficiency=1

In the second case there is an optical preamplifier with very large gain and the
best possible noise factor. The optical bandwidth Δνop <<P/(hν). The detector is a
PIN-diode with the efficiency=1.

ONEs have the power P through the whole symbol slot T and ZEROs have the
power 0.5P during the whole symbol slot. Both receivers have the noise
bandwidth 1/(2T). Gaussian statistics is assumed. Dark currents are neglected.
(Exam. 960309)

5.3 Derive an expression for the loss limit and an expression for the dispersion limit,
i.e. the length L as a function of the bit rate B. The fibre is a standard single mode
fibre with the loss 0.3 dB/km and the wavelength 1.55 μm. The receiver is a PIN-
detector with the quantum efficiency 1. The noise bandwidth is 0.7B. The load
resistance is 50 Ω and the noise factor of the amplifier is 3 dB. The shot noise is
neglected. ONE bit has the mean power 1 mW at the transmitter and ZERO bit has
the mean power 0 mW. The bit error rate is 10-9.

66
 The unmodulated light source has negligible linewidth. The transmitted pulse
shape is Gaussian and has no chirp and the RMS-value is σ=1/(8B)
(Exam. 980423)

5.4 Consider an intensity modulated direct detection (IM-DD) system. The link is
L=100 km standard single mode fibre and the wavelength is 1500 nm. The
attenuation is 0.3 dB/km. The launched peak power is 10 mW for one bits and 0
mW for zero bits and the bit rate is B. The detector is a pin-diode with the
efficiency 100%. The load resistance is 50 ohm. The noise from the electronic
amplifier is neglected. The noise bandwidth is 0.7B. How large can B be if
BER<10-9? Is the dispersion criterion fulfilled for this bit rate? Assume that the
linewidth for the light source is zero and that the chirp of the launched pulse is
zero, then holds approximately B2L<(8πβ2)-1
(Exam. 980828)

Short questions

5.5 You want to improve the bit error rate of a fibre optic link. There is an optical
amplifier available. Should it be positioned at the transmitter end or at the receiver
end in order to achieve the best result? Motivate your answer. Saturation effects in
the amplifier and nonlinearities in the fibre can be omitted.
(Exam. 970411)

5.6 Consider Figure 5.7. Describe what happens with the impulse when the dispersion
parameter goes to 0.25.
(Exam. 990409)

5.7 Sketch the eye-diagram at the sampler for a receiver where the light pulse
corresponding to 1 is rectangular and T long and the impulse response of the
receiver filter is also T long. The power of the light pulses corresponding to 0 is
zero. The bitrate is 1/T.
(Exam. 020823)

5.8 What are the disadvantages with the LED if you wish links with high bitrate-
distance-product in a standard single-mode fibre.
(Exam. 020823)

67
Solutions

5.1
C = 1 dB S = 0.1 dB
Tx x x x Rx

splice
α=0.3 dB/km

Power budget:

Total loss  2  N  3  N - 1  0.1  1.9  3.1  N dB 

N : loss per fibre 

Allowed loss  transmitte r power - receiver sensitivit y - margin  
dBm dBm dB 

0 - 30  10 

 N  5.8  N max  5  Lmax  50 km

Loss margin 50 km = 0+30-(19+3.1·5)=12 dB

Rise time budget, according to 5.2:

0.35
RZ format, Tr   350 ps
B

Tr  Ttr 2  T f 2  Trec 2

0.1
where Ttr   100 ps , Trec  100 ps
B

2 2 
T L  D L FWHM  D L f FWHM  D L 10 B 
c c 
  L max  221 km

T f max  Tr 2  Ttr 2  Trec 2  320 / 10 ps 


Answer: The system is loss limited: 50 km

I1  I 0
5.2 BER=10-9  Q=6 
1  0

a) I 1  PM

I 0  0.5I 1

68
 4k B TFn f
 1 2   s 2   T 2  2qPM 2 F A ( M )f   {M  1  thermal noise,  T , is neglected } 
RL

 2qPM 2 FA (M )f  {the best possible F ( M )  2}  {f  1 2T } 

 2qPM 2 T

 0 2  0.5 1 2

PT
0.5MP 0.5 h   PT  N  number of photons in a 1 bit  
6  {  1}   
1  0.5  2qM 2
P T 1  0.5
2
PT  h 
h

0 .5 N

1  0 .5 2

2
 1  0.5 
N  36  2     839
 0.5 
 

1.5
N N  630
2

b) I 1  PG

I 0  0.5I 1

 1 2 consists of noise terms according to equations (6.5.12)-(6.5.14) and thermal

noise. While G>>1 and equations (6.5.12)-(6.5.14) contain G or G2, they will
dominate over thermal noise.

Further,  s 2   sig  sp 2

 sig  sp 2 2qG 2 Fn Ps f h 2 Ps
   1
 sp  sp 2
qGFn  2
 opt f Fn  opt h

i.e.  sp  sp 2   sig  sp 2 and  sig  sp 2 remains.

q2 2 1
 1 2    1  2q 2 G 2 Fn Pf h  Fn is at most 2  2
1
G P  2qG 2 P
h T T

69
 which is differing from the noise term in a) only by changing M to G, i.e. the same
answer.

Answer: a) and b) number of photons 630

5.3 The loss limit

BER<10-9  Q>6, according to Figure 4.19.

I1  I 0 I1 I1
Q   {assume that thermal noise is dominating} 
1  0 1  0 2 1

q q 0
I 1  Pmin ,  
hv hc

 1 2  4k B TGBFn , G=1/50 Ω, Fn=3 dB=2 times

1 hc
Pmin  Q2 1  Q 2 4k B TGB 0.7 Fn
 q 0

Ptransm
Fibre attenuation: , and the fibre length thus becomes
Pmin
1 P 
L 10  log 10  transm  [km]
 dB  Pmin 

 10 -3 1.6 10 19 1.55 10 6 

10  log 10  
1
L
0.3   23 
 6.63 10  3 10  6  2  4 1.38 10  290  0.02  0.7  2 10 B Gb / s
-34 8 9


 
 33 log 10 1.55 / B Gb / s  332.14  0.5 log 10 ( BGb / s )   72  16.5 log( BGb / s )

shot noise
(We can check if the shot noise is negligible, form the ratio
thermal noise

1
2q Q 2 10.7 B
2qPmin 0.7 B  16.8 Bq 16.8 1.6 10 19  BGb / s 10 9
     1,
 12 1 6.69 10 7 BGb / s 6.69 10 7 BGb / s
for B<62.5 Tb/s)

The dispersion limit

T 1
We choose the most common criterion  tot  
4 4B
Use the “superformula” (2.4.23) with β3≈0, Vω2≈0, C=0

70
   2
 tot 2  L   L  64  B 2 
2
 8B 
 1   2 2       1   2 
 02   2    4 B  2 
  0    

3 3
 2 L  32  B 2  3  L   { 2  20 ps 2 /km}   2700 /BGb/s 2
32  2 B 2
32  20 10 - 24
 BGb / s 10
2 18

5.4 BER<10-9  Q>6

I1  I 0 I1 I1
Q   {assume that thermal noise is dominating} 
1  0 1  0 2 1

1  1.5
I 1  P1  ,   , Loss=30 dB=100
Loss 1.24 1.24

4 k B T 0.7 B
 12 
RL

1
P1
Q6 Loss
4 k B T 0 .7 B
2
RL

2
 
 
 P1 1 
B  
 Loss  2Q 4 k B T 0. 7 
 RL 
 

2
 1.5 10  2  50
   4.6 Gb/s
 1.24 10 3  2  6  4 1.38 10  23  290  0.7
 

We check that the shot noise is negligible

1.5
 s 2  2qI 0.7 B  2 1.6 10 19  10 3 10  2  4.6 10 9  0.7  1.3 10 14 A
2
1.24

4k B T 0.7 B 4 1.38 10 23  290  0.7  4.6 10 9

T 2    1 10 12 A2 >> 1.3 10 14 A2 ok.
RL 50

Answer: 4.6 Gb/s

Only the attenuation criterion has been considered above. Let us now investigate
if the dispersion criterion is fulfilled. The given expression is derived from
1
T  , T   2 L   2 2BL
4B

71
 1 1
B2L   {page 53}   2 10 24 , which is compared to
8  2 8 20 10  24
/ 10 3

4.6 10 
9 2
10 5  2.1 10 24

Thus, not fulfilled. Because the expression is a rough estimation, a more accurate
analysis should be done while the limit is very near. However it is not included in
the problem.

5.5 At the transmitter. In both cases the received signal power is the same. In both
cases the noise power leaving the optical amplifier is also the same. But if the
amplifier is placed after the transmitter this noise is attenuated before it reaches
the detector.

5.6 The penalty is infinite at BLDσλ=0.25, because then the width of the sent pulse is
zero, and there is no power in the pulse. Note that it is assumed that the peak
power is constant and that the spectral width is determined by the unmodulated
linewidth of the light source and not by the pulse.

5.7

Before the filter

After the filter

T
Eye-diagram

5.8 Disadvantages with LEDs for links

 Large linewidth results in short dispersion limit.
 The low power, which can be coupled in, results in short loss limit.
 The LED has low modulation bandwidth

72
6. Optical amplifiers

Problems
See “Read this first” in the solutions

6.1 When the optical output power from an optical amplifier is detected by a
photodetector with unity quantum efficiency, the mean square noise current
generated in the detector is given by

 2   T 2   s  sig 2   s  sp 2   sig  sp 2   sp  sp 2

Show that, if the signal-spontaneous beat noise dominates, the noise figure
defined as SNRin / SNRout can be expressed as

Fn  2n sp

6.2 A linear optical amplifier with the gain G>>1 and the noise factor = 2 is
positioned just after the transmitter or just before the receiver in a IM-DD link. At
the output of the detector, which has the efficiency 1, several noise contributions
appear which values should be compared (the contribution from the thermal noise
is omitted). A transmitted ONE has the power P1 at the transmitter. The losses in
the link, L>1, is much larger than G. The signal power at the detector is larger
than the spontaneous emission power at the detector. Which one of the eight noise
current variances has the least value and where is the linear optical amplifier then
positioned?
(Exam. 990830)

6.3 A receiver consists of an optical preamplifier and a pin-diode detector followed by

an electronic amplifier. The ONE bits of the received signal have the power P1
and the ZERO bits of the received signal have the power P0  0.1 P1 . The bit rate B
is 10 Gb/s and the electrical noise bandwidth is assumed to be B/2. The optical
bandwidth is assumed to be small enough so that the amplified output signal from
the optical preamplifier is much larger than the power from the amplified
spontaneous emission. The optical amplification is 30 dB and nsp is 1. The
efficiency of the pin-detector is 1 and its load resistance is 1000 Ω. The noise
factor of the electronic amplifier is a factor 3. The wavelength is 1.55 μm. How
large has P1 to be in order to achieve a BER = 10-9?

The optical preamplifier and the pin-detector are replaced by an APD with an
amplification of 30 dB and the best possible excess noise factor. What is the value
of P1 in this case?

Do not forget to motivate your approximations.

(Exam. 000307)

73
6.4 How many kilometres longer can a link be with a good standard fibre if the PIN-
detector is preceded by an EDFA with the best possible noise factor and 25 dB
power amplification. The bit rate is 2.5 Gb/s and the noise bandwidth in the
receiver is 2 GHz. The efficiency of the PIN-detector is 1 and the load resistance
is 50 Ω and the noise factor of the electrical amplifier is 3 dB. The optical
bandwidth is 30 nm. The wavelength is 1.55 μm. ZEROS have 0 power. BER<10-
9
is required.

Do not forget to motivate your approximations.

(Exam. 010831)

6.5 A fibre optic system A consists of a transmitter and a receiver and NA

concatenated links, which each consists of a fibre followed by an optical
amplifier. The length of each fibre is 100 km and the fibre has the attenuation 0.2
dB/km and the optical amplifier has an amplification, which just compensates for
the fibre attenuation for 100 km. The launched power for a ONE is P and zero for
a ZERO. The total system is dimensioned so that the BER=10-9. A system B has
NB intermediated coupled links, where the fibre length is 50 km and the amplifier
just compensates the fibre attenuation for 50 km. In all other respects the systems
are equal, for example the same BER. How long is the total system length in
system B, compared to system A? Dispersion and saturation effects are omitted.
(Exam. 010309)

6.6
a) A fibre optic link consists of a light source with NRZ-pulses with 2.5 Gbit/s at
1.55 μm. The ZERO has 0 mW optical power. The receiver uses a pin-
detector, which has the efficiency 0.8 and the load resistance 50 Ω. The noise
bandwidth of the electrical amplifier is 2 GHz and the noise factor is 3 dB.
What is the minimal received power for a ONE to achieve a BER=10-9? Also
calculate the shot noise and the thermal noise, which is expected to be
dominant.

b) In order to upgrade the link an optical preamplifier with the amplification 20

dB, noise factor 2 and the optical bandwidth 1 THz is used. The fibre length is
increased to start with, so that the increased fibre attenuation is 20 dB. Several
new noise terms are added. These should be determined, and also how they are
related to the shot noise and the thermal noise, which both are unchanged. (If
the added noise terms are not negligible the fibre length must be decreased to
keep BER=10-9 but it is not included in the exercise to calculate.)
(Exam. 010423)

6.7 A fibre-optical system at 1.55 μm consists of a transmitter and a receiver.

Between these there is a regularly pattern with N+1 equal long fibre modules and
N equal optical amplifiers. Closest to the transmitter and the receiver are fibre
modules. The transmitter transmits ONES with power 10 mW into the fibre and
ZEROS with power 1 mW. The amplifiers have the best possible noise factor and
have the power amplification 25 dB over the entire optical bandwidth. The

74
 amplification just compensates for the fibre losses. How large is the single sided
optical spectral density for the noise of both polarisations respectively, and what is
the signal power right after the last optical amplifier? The same questions but right
before the receiver? How large (numerical) can N be for BER<10-9 and an
electrical noise bandwidth of 10 GHz? The dominating noise term is assumed to
be the signal-spontaneous emission noise. Saturation and dispersion can be
omitted.
(Exam. 990409)

6.8 In a receiver, consisting of a PIN-detector, followed by an electrical amplifier, the

sensitivity is mainly determined by the thermal noise and the noise of the
electrical amplifier. By connecting an optical pre-amplifier, the sensitivity is often
determined by the noise of the optical amplifier. In the detector, four mixing noise
terms occur in the latter case. The dominating noise term of the four is signal-
spontaneous beat noise if there is a sufficiently narrow optical filter (not valid if
the ZERO has zero power). The problem is to calculate the received power for a
ONE that is needed for the case without an optical amplifier and the received
power, which is needed for the case with an optical amplifier with sufficiently
narrow optical filter. A ZERO has 10% of the power of the ONE. The bit error
rate should be less than 10-9. The power amplification is 25 dB. The wavelength is
1.55 μm. The efficiency of the detector is 1. The load resistance is 50 ohm and the
electrical amplifier adds a noise term with the spectral density 10-22 A2/Hz. The
bitrate is 1 Gb/s. The noise bandwidth is 0.7 GHz.

Do not forget to motivate your approximations.

(Exam. 020823)

6.9 A fiber-optic link consists of a transmitter, which transmits ONES and ZEROS
with the power of 1 mW and 0 mW respectively. The bit rate is 2.5 Gb/s. The
receiver has a pin- detector with the efficiency 1. The noise factor is 3, the noise
bandwidth is 2 GHz and the load resistance is 50 Ω. Between the transmitter and
the receiver there are a number of N sections with each 100 km of fibre, which has
the attenuation 0.2 dB/km, in the end of the last section there is an optical
amplifier, which has a gain that exactly compensates for the fibre losses. The
amplifiers do not show any saturation effects and have the best possible noise
factor. In order to decrease the optical noise there is an optical filter with
bandwidth 100 GHz. How large can N be to achieve a BER < 10-9? Assume
Gaussian statistics. Assume no dispersion problems.
Do not forget to motivate your approximations.
(Exam. 020308)

6.10 A laser diode (LD), constructed as a Fabry-Perot etalon, can be used as an

oscillator, an amplifier and an attenuator, depending on how large the bias current
is. The assignment is to find the bias current interval where the laser diode works
as an amplifier.

75
 Start with equation (6.2.1) and chose the optimal frequency in order to receive the
maximal amplification.
Differential gain σg is 10-16 cm2 and transparency carrier density NT is 1018 cm-3.
The laser length is 200 μm and its volume is 2000 μm3. Waveguide losses or
internal losses is 0 and the confinement factor is 1. The power reflectivity of the
mirrors is 0.3 and the charge carrier life time τc is 1μs. The number of photons in
the cavity can be neglected.
NOTE: A letter can have different meanings at different places in the book.
chapter.
(Exam. 010831)

Short questions

6.11 Study Figure 6.17b. Explain why the graphs start to decrease after a certain
length.
(Exam. 010423)

6.12 State at least three important advantages for a system, which includes an
EDFA (erbium-doped fiber amplifier) in comparison with an SLA (semiconductor
laser amplifier), in a WDM-system at 1.55 μm.
(Exam. 020823)

76
Solutions

Notations used in chapter 6 solutions

Some frequently occurring noise formulas are collected below. (The corresponding
formulas in Agrawal’s 2nd and 3rd editions have some printing errors and they are also
presented differently in 2nd and 3rd editions.)

It is assumed that the signal entering the optical amplifier can be treated as sinusoid
with frequency ν , power Ps, and linearly polarized. Figure 1 shows the assumed single
sided optical spectrum in signal polarization plane. The optical spectral density in the
orthogonal polarization plane is the same but without the impulse.

Optical spectral
density in the
signal Impulse with area Ps
polarization
plane
Psp=ΔνoptSsp

Ssp=(G-1)nsphν

Ssp
h
2
frequency
ν

Δνopt
Zero point fluctuations
(which can be viewed as the
cause of shot noise). They can
only be detected as beating terms
Δνopt is the bandwidth of the optical filter.

After detection the single sided electrical spectral density from both polarizations is as
in figure 2

77
 Electrical
spectral density
A is signal-shot (beat) noise
E B is ASE-ASE (beat) noise
C is signal-ASE (beat) noise
C D is ASE-shot (beat) noise
E is an impulse with area (GRPs)2
A , i.e. the signal current squared
Δf is the electrical equivalent
D noise bandwidth
B

Δf Δνopt/2 Δνopt frequency

part of (6.5.7) A1:  s  sig  2qPs Gf  2q 2GPs f / h

2
:A2

(6.5.8) B1:  sig  sp  4 2 GPs S sp f  2qG  Fn Ps f / h :B2 (6.5.13)

2 2

 qGFn   opt f
2
(6.5.9) C1:  sp  sp 2  4 2 S sp2  opt f :C2 (6.5.14)

part of (6.5.7) D1:  s  sp 2  4q S sp  opt f  2q 2GFn  opt f :D2

With normal values of G, Fn, and η one can often neglect A in comparison with B and
neglect D in comparison with C.

In 3rd edition A and D are added in (6.5.7) and called shot noise σs2. In (6.5.12) D is
neglected but still called shot noise σs2 . One can also include the contribution to shot
noise from the dark current 2qI d f

In the solutions we denote the equations with A1, A2, etc.

6.1

Pinopt Poutopt =·G·Pinopt

G

SNR in Pinelect /Pnoise,

elect
in
Fn   elect elect
SNR out Pout /Pnoise, out

78
 The electrical powers are proportional to the currents squared, hence we derive the
corresponding currents

2
q   q P opt 
(Input signal) ~ P2
 
opt 2
in

       out 
hν   hν ηin  G 


 
opt opt
Pout 2q 2 Pout
Input current noise (only shot noise) ~ 2q P opt
in Δf  2qG Δf   Δf
G hν G

(Output signal) 2 ~ Pout

opt
 
2  q opt 
   Pout 
 hν 

Output current noise ~ σ sig-sp  equation (6.5.8)  4 2 Pout S sp Δf  S sp  G  1n sp hν 

2 opt

4q 2
 2 2
opt
Pout G  1n sp hf
h ν

2
 q Pout  4q 2
     Pout (G  1)n sp f
h G  h 2(G  1)n sp
 Fn     G  1, because  sig-sp dominates  2n sp
 2q Pout f
2
 q G
2
  Pout   
 h  h G

6.2 Equations (6.5.12)-(6.5.14) give the noise variances directly when the amplifier is
positioned at the detector, which we call case 1.

P1 f
 s  sig1 2  2q 2G A2
L h

P1 f
 sig  sp1 2  2qG 2 Fn B2
L h

 spsp12  qGFn 2  opt f C2

 s  sp1 2  2q 2GFn  opt f D2

In case 2, the linear optical amplifier is positioned at the transmitter.

 s sig 2 2 =A2, because signal is the same at the detector in both cases.

79
  sig  sp 2 2 =B2/L, the signal has not been attenuated at the detector, but the
spontaneous emission has.

2
 sp sp 2 2 =C2/L , this is a mixing product from the amplified spontaneous emission
with itself. The amplified spontaneous emission is created in the linear optical
amplifier and is attenuated L when it reaches the detector.

 s sp 2 2 =D2/L, this is the spontaneous emission shot noise. The amplified

spontaneous emission is attenuated L.

In order to calculate which noise term has the lowest value, we only need to
consider the terms in case 2, while they are less than or equal to the corresponding
terms in case 1.

P1 f
2 q 2 G
Compare
A2
 L h 
L

L
1
P f 1 GFn 2G
2qG 2 Fn 1
B2 / L
L h L

i.e. B2 / L is less.

1
(qGFn ) 2  opt f
C / L2
Compare 2  L2  Fn  1  1
D2 / L 1 2L L
2q 2GFn  opt f
L

i.e. C2 / L2 is less

Compare
B /L P1 P1G PG PG / L
. 2 2    G  1  1  1  { according to
C2 / L  opt h  opt h (G  1) S sp S sp / L
the assignment }  1

i.e C 2 / L2 is less, which means  sp sp 2 2 has the lowest value and the amplifier is
placed at the transmitter.

I1  I 0
6.3 Q 
1  0

I 1  I 0  0.9    P1  G

 1 and  0 depend on several terms. Let us approximate.  0 has the least signal
dependent noise and is determined firstly.

80
 0 2   T 2   s  sig 2   sp  sp 2   sig  sp 2   s  sp 2

According to equations (6.5.12)-(6.5.14) together with G>>1 and

(G  1)
Fn  2n sp  2 , result in following
G

 s  sig 2   sig  sp 2 and  s  sp 2   sp  sp 2

 sp  sp 2 qGFn 2  opt f  opt hFn  opt h 2n sp (G  1) S p  opt

Let      1
 sig  sp 2 2qG 2 Fn P0 f / h 2 P0 2 P0 G P0,out

according to the assumptions.

Thus, it remains  0 2   T 2   sig  sp 2

Assume for the time being that  T 2 is negligible.

0.9    P1  G
Then Q={BER=10-9}=6=
1  
0.1 2(qG) 2 Fn P1 f / h

0.9 2 P1 36 1.316 2  2  Fn  h  f
Q 2  36    P1 
1.316 2 2  Fn hf 0.9 2

Now, compare  sig  sp 2 with  T 2 when P=P0

 sig  sp 2 2qG 2 Fn f / h  0.1  36 1.316 2  2  Fn  h  f qG 2 Fn 2 f  0.1 36 1.316 2

  
T 2 4k B T0 fFn / R L  0.9 2 k B T0 Fn / R L  0.9 2

(1.6 10 19 1000 ) 2  5 10 9  0.1  36 1.316 2  4

  320  1
290 1.38 10  23  3 / 1000  0.9 2

i.e.  T 2 can be neglected as well 

36 1.316 2  2  6.63 10 34  3 10 8  5 10 9  2

 P1   197 nW
0.9 2 1.55 10 6

For an APD, if  T 2 is neglected

0.9P1 M 0.9 2 P1
Q  Q2   {F A ( M )  2 according to
1.316 2qP1 M 2 F A ( M )f 1.316 2 F A ( M )hf 2

equation 4.4.18 }  the same result as above

Answer: P1=197 nW for both pin and APD receivers

81
6.4 Consider firstly the case without optical amplifier. BER=10-9 results in Q=6,
according to Fig.4.19.

I1  I 0 I1
Q   {assume, until further notice, that the thermal noise>> shot
1  0 1  0
I1 I
noise}   1
2 0 2 T

I 1 2 T Q
Pmin   ,
 
 11.55
where   
1.24 1.24

4k B T0 fF 4 1.38 10 23  290  2 10 9  2

T2    1.3 10 12 A
Rl 50

2 1.3 10 12  6 1.24

Pmin 1   1.1 10 5 W
1.55

We have to check the shot noise

 s  sig 2  2qPmin f 

1.55
 2 1.6 10 19   3 10 5  2 10 9  2 10 14  1.3 10 12 A
1.24
i.e. the assumption is proved to be valid.

Consider secondly the amplifier case. According to equation (6.5.6) and A,B,C,D
(from the “Read this first”) the noise variance is

 2   T 2   s  sig 2   sig  sp 2   sp  sp 2   s  sp 2

Equations (6.5.7)-(6.5.9) together with G=316>>1 result in that  s  sig 2   sig  sp 2

and  sp  sp 2   s  sp 2 .
I1
Q 6, if also σT is negligible then
 0 1
 1 2   sig  sp 2   sp  sp 2 and  0 2   sp sp 2

Equation (6.5.16) then gives

  opt   
3 10 8 9 3 10 8  30 10 9 
Pmin 2  2hFn f Q 2  Q   2  6.626 10 34   2  2  10 36  6


f 
 1.55 10 6 
   2
1.55 10 6  2 10 9 

 3.110 7 W

82
 We have to check if  T 2   sp  sp 2

 sp  sp 2  (qGFn ) 2  opt f 


 1.6 10 19  316  2 2 3 10 8  30 10 9
 2 10 9  7.6 10 11
1.55 10 6 2

12
 1.3 10 A

Thus we can neglect thermal noise. The extra length is then

1.1 10 5
10 log10
3.1 10 7  77 km
0.2

Where we have assumed that a good fibre has an attenuation of 0.2 dB/km.

6.5

1. 2. NA.

Tx GA GA GA Rx

SspA SspA SspA

LA=100 km

1. 2. NB.

Tx GB GB GB Rx

SspB SspB SspB

LB=50 km

N B LB
We shall calculate , where BER=10-9 in both systems.
N A LA

I 1A  I 0 A I 1B  I 0 B
BER=10-9  Q=6= 
 1A   0 A  1B   0 B

Tx is transmitting P. After each link in A or B the power is again P, i.e. I1A=I1B and
I0A=I0B. This gives,  1A   0 A   1B   0 B

Consider equation (6.5.6), together with A, B, C, D:

 2   T 2   s  sig 2   sp  sp 2   sig  sp 2   s  sp 2 . In order to achieve the same noise,
according to the equations, SsptotA=SsptotB

83
 S sptotA  {equation(6.1.15)}  N A ( G A  1 )n sp h  S sptotB  N B ( G B  1 )n sp h

N B L B L B G A  1 50 10 1000.2 / 10  1 1 99
     5.5
N A L A L A G B  1 100 10 500.2 / 10  1 2 9

Answer: 5.5

6.6
a) First part without optical amplifier.
I  I0 I1
BER=10-9  Q=6= 
1  0 1  0

4k B T0 Fn f
 02  T2
RL

 1 2   s  sig 2   T 2 {from equation (6.1.14)}  2qIf   0 2

I 1 Q 2 T
Assume firstly that  T 2   s  sig 2  P1  
 
 0.8 1.55
   1.0
1.24 1.24

4 1.38 10 23  290  2  2 10 9

T 0   1.13 10 6 A
50

6  2 1.13 10 6
P1   13.6 μW
1.0

Check that σs-sig2<< σT2

13.6 10 6
 s  sig 2  2 1.6 10 19   2 10 9  8.7 10 15 A
2
1.0
 T 2  1.28 10 12 >>  s sig  Hypothesis accepted
2

b) The second part with the amplifier included.

 T 2 is the same  1.28 10 12 A

2

Study for example equations (6.5.12)-(6.5.14) for the additional noise terms.
Equation (6.5.12) is identical with the shot noise in a) above.

 q
 s sig 2  equation A 2   2q 2GPs f / h    and GPs  P1 due to extra fibre
 h
attenuation  G  2qP1f  8.7 10 15 A2

84
 
 sig sp 2  equation B 2   2(qG) 2 Fn Ps f / h  2qGFn P1 f  compare with  s -sig 2 
 0.8 100  2  8.7 10 15  1.39 10 12 A , which is not negligible compared to  T 2 .
2

 sp  sp 2  equation C 2 
 (qGFn ) 2 v opt f  (1.6 10 19  0.8 100  2) 2 10 12  2 10 9  1.31 10 12 A2, which is
not negligible compared to  T 2 .

 s  sp 2  {D 2 }  2q 2GFn v opt f  2  (1.6 10 19 ) 2  0.8 100  2  2 10 12  2 10 9  1.64 10 14 A 2

We can conclude that  T 2 ,  sig  sp 2 and  sp sp 2 are of the same magnitude, but
much larger than  s sig 2 and  s sp 2 . The fibre length must be decreased in
order to achieve BER=10-9.

Answer: P1  13.6 μW,  T 2  1.28 10 12 A2,  s  sig 2  8.7 10 15 A2,

 sig  sp 2  1.39 10 12 A ,  sp  sp 2  1.31 10 12 A and

2 2

 s  sp 2  1.64 10 14 A

2

6.7
1. 2. N.

Tx G G G Rx
L L
L
P1=10 mW Ssp Ssp Ssp
P0=1 mW

Each amplifier emits at its output the single sided spectral density.

S sp  (G  1)n sp h

for each polarisation component.

Because the fibre attenuation L=G,

S spN  N  (G  1)n sp h

is emitted after output of the last amplifier.

At the input of the receiver the spectral density is

S sp .rec.  N  ( G  1 )n sp h / L

85
 The signal powers are P1 and P0, and P1/L and P0/L, respectively.

Calculations of N.

BER=10-9  Q=6

I1  I 0
Q
1  0

I 1  10 I 0  P1 / L

Consider equation (6.5.8),

1 2   sig  sp 2  4 2  G / L  P1 / L  N ( G  1 )n sp h / L  f  10 0 2 , while the noise
contributions from all other noise terms can be neglected.

GPs in equation (6.5.8) is the output signal power at the amplifier, but we are
interested in the signal power at the receiver  GPs/L, but Ps=P1,0/L  GP1/L2. In
the same manner the noise is to be evaluated at the receiver.

 S sp .rec.  N  ( G  1 )n sp h / L

0.9P1 / L
Q 
1.316 4 G / L  P1 / L  N (G  1)n sp h / L  f
2

0.9 P1

1.316 4 N (G  1)n sp hf

10 2
2
1  0.9 
N    81
36  1.316  1.24
4  315 1  1.6 10 19 10 10
1.55

 P1
Answer: At the amplifier Ssp= N·(G-1)nsphν, Ps= 
 P0
 P1 / L
At the receiver Ssp= N·(G-1)nsphν/L, Ps= 
 P0 / L
N≈81

6.8 Observe that there is both additive and signal dependent noise. The additive noise
is

4k B T0 f
t2   S I f
RL

The signal dependent noise in the first case is

86
2qP1,0 f

In the second case, consider equations (6.5.12-14). Because of the conditions

it remains

 sig  sp 2  2  qG 2 Fn P1,0 f / h  2qG 2 Fn P1,0 f

where Fn  21    2
1
 G

We calculate for the second case. The first case is then obtained by letting
G  1 and Fn  1 .

BER  10 9  Q  6

I1  I 0
Q
1   0

where I 1  P1G , I 0  P0 G , P0  0,1P1

 1 2  2qG 2 Fn P1 f   t 2
 0 2  2qG 2 Fn P0 f   t 2

I1  I 0   Q 0   1

Square:

I1  I 0 2  2  I1  I 0  Q 0  Q 2 0 2  Q 2 1 2

I1  I 0 2  2  I1  I 0   Q 0  Q 2 I1  I 0   k

where k  2qGFn f

I1  I 0   Q 2 k  2Q 0 (1)

Square again:

I1  I 0 2  2  I1  I 0   Q 2 k  Q 4 k 2  4Q 2 kI 0   t 2 

Let I 0  pI1

1  p 2 I12  2  1  p   I1Q 2 k  Q 4 k 2  4Q 2 kpI1   t 2 

87
 1  p Q 2 k  4 pQ 4 k 2  4Q 2 t 2 1  p 2
I1 
1  p 2

We now could calculate I 1 directly, but we chose to make some comparisons.

Let us begin with the first case and investigate which noise dominates.

“Thermal noise”:
 4  1.38  10 23  290  0.7  10 9 
4Q 2 1  p 2  t 2  4  36  0.81    10 22  0.7  10 9  
 50 
 
 117  2.24 10 13  0.7 10 13  343 pA 2

“Shot noise”:

4 pQ 4 k 2  4  0.1  6 4  2 1.6 10 19 10 9  0.7 
2
 0.026 fA 2  343 pA 2

 Shot noise can be neglected

2  Q  t 1.55 2  6  2.94 10 13

a) I 1   P1  I 1    9.0 W
1  p  1.24 0.9

For the second case,

k 2 is increased by the factor

G 2 Fn 2  4 10 7  4 pQ 4 k 2  1.04 nA 2  343 pA 2 

1  p  Q 2 2
k
1.55 1.316

I 1 1.24 0.81

36  2 1.6 10 19  316  2  0.7 10 9
b) I 1   P1    33 nW
1  p 2 G 316

Answer: 9.0 μW without optical amplifier and 33 nW with optical amplifier

6.9
P1, P0 1/G 1/G P1, P0

Tx G G G Rx

I1  I 0
We use Q  6  , which gives BER 10-9
1   0
Because the amplifiers compensate for the losses, the received power is P1 and
0, respectively, and I 1  P1 , where  is the responsivity. I 0  0 .

88
The total noise variance is given by equation (6.5.6). In order to approximate,
study equations A, B, C, and D. Observe that we must modify Fn to N  Fn ,
because the ASE noise is accumulated N stages.

 s  sig 2 2q 2GPs f / h 1
  {  1}  ; Fn  2
 sig  sp 2
2qG  NFn Ps f / h
2 NFn G

1
 1 if N  1 .Thus, we can neglect  s sig .
2
N  200

 s  sp 2 2q 2NFn  opt f 2
Compare  
 sp  sp 2
qGNFn  2
 opt f GNFn

We neglect for the same reason as above  s sp 2

Compare  sig  sp 2 and  sp sp 2 for a ONE, i.e. Ps  P1 / G

 sp  sp 2 qGNFn 2  opt f GNFn  opt h  q 1.24 

   h  
 sig  sp 2
2qG 2 NFn P1 / G  f / h 2 P1  1.55 

100  N  2 10 11 1.6 10 19 1.24

  N 10 3
2 10 3 1.55

If N<< 1000,  sp sp 2 can be neglected.

Compare  T 2 and  sp sp 2

T 2 4k B T0 Ff / R L 4 1.38 10 23  3 / 50 300

  
 sp  sp 2
qGNFn  2
 opt f 1.6 10 19
2
 2 N 10 2 11 N2

If N 2  300  T 2 can be neglected.

Let us assume that  T 2 can be neglected but  sp sp 2 cannot. Hence:

P1 c
Q 
 sig  sp   sp  sp   sp  sp
2 2 2
aN  bN 2  bN 2

Q aN  bN 2  c  Q bN 2

 
Q 2 aN  bN 2  c 2  2cQ bN 2  Q 2 bN 2

Q 2 aN  c 2  2cQ bN 2

Q 4 a 2 N 2  c 4  2c 2 Q 2 aN  4c 2 Q 2 bN 2

89
  
N 2 Q 4 a 2  4c 2 Q 2 b  N 2c 2 Q 2 a  c 4  0

N
c 2Q 2 a  c Q a 
2 2 2

 c 4 Q 4 a 2  4c 2 Q 2 b c Q 2 2
a  c 3 2Q b
Q 4 a 2  4c 2 Q 2 b Q 4 a 2  4c 2 Q 2 b

The minus sign is chosen to achieve N>0 for a=0

 c 3 2Q b 
c 2 Q 2 a 1  2 2 
 c Q a 
 c2 1
N  2
 4c 2 Q 2 b  Q a  2c b 
Q 4 a 2 1   1  
 4 2   Q 2 a 
 Q a  

c2 q 2 h 2 P1 2 P1 10 3
   
Q2a Q 2  2  qG 2 Fn P1 / G  f / h hQ 2 2GFn f 1.6 10 19
1.24
 36  2 100  2  2 10 9
1.55
= 273

c2 b P1 2qGFn  opt f   opt f  opt 10 11

    =1.17
QA Q 2qG  Fn P1 / G  f / h Qq / h  f
2
f  Q 2
2 10 9  36

1
N = 273 = 126
1  1.17

Thus, the hypothesis to neglect  T 2 was correct.

Answer: N=126

c
NOTE: If instead the use an expression as ,  T 2 can
aN  bN  d  bN  d
2 2

be included with the same solving method.

6.10 Start with equation (6.2.1). Optimal frequency occurs when    m , where m is
chosen so that G( m ) is maximal and is denoted G.

(1  R ) 2 G
G FP is then ,
(1  GR) 2
G  exp( g   )l  {  0}  exp gl , from equation (6.1.7)
g   g (n  nT ) , from equation (3.3.1)

When the bias current is increased from zero, attenuation is first achieved (g<0),
then amplification (g>0) and finally oscillation (lasing) (g=gth), when the
threshold current is reached.

90
 The threshold condition for lasing GR=1, which means G FP   , which means
oscillation.

The attenuation means that G  1

G  2RG  R 2 G  1  2GR  (GR) 2

1 R 2 1
G2  2
G 0
R R2

G
1 R 2

1  R 
2 2

1

1 R 2

1  2R 2  R 4  4R 2


1 R 2  1 R 2 
2 4 2 2 4 2
2R 4R R 2R 4R 2R
 1
 2 , above threshold
R

 1 , is chosen



It remains to find the current interval where 1  G  1 R

d
The rate equation (3.5.2) when  0 and P  0 below threshold gives:
dt
q q
I N V n
c c

where N is the carrier number and n is the carrier density. n is determined by

g   g (n  nT ) , where G  expgl 

The limit G=1  g=0  n=nT=1024 m-1

1 ln 1  
R  ln 3.33  6.01 10 3 m-1
The limit G  g
R l 200 10 6
g 6.01 10 3
 n  nT   10 24  1.6 10 24
g 10  20

Respective currents are then

1.6 10 19

I  2 10 15 10 24  0.32 mA
10 6

1.6 10 19

I  2 10 15 1.6 10 24  0.51 mA
10 6

Answer: 0.32  I  0.51 mA

6.11 Assume that the pump power is injected at the same port as the signal. The
pump power is first used to excite the Er3+-ions (inversion) The inversion used for

91
 amplification of the signal. After a certain distance the pump power is not
sufficient to excite the Er3+-ions (no inversion).The signal will now be absorbed
instead of being amplified, the amplification decreases.

6.12 Advantages for EDFAs compared to SLAs in a WDM-system

 Polarisation independent
 Small channel cross talk
 No rippel
 Smaller noise factor
 Larger amplification

92
7. Dispersion compensation

Problems

7.1 The pre-chirp technique is used for dispersion compensation in a 10-Gbit/s

lightwave system operating at 1,55 μm and transmitting the 1 bits as chirped
Gaussian pulses of 40 ps power-width (FWHM). Pulse broadening by up to 50 %
can be tolerated. What is the optimum value of the chirp parameter C, and how far
can the signal be transmitted for this optimum value?

7.2 The magnitude of the slowly varying amplitude of a pulse is Gaussian and the
RMS-width is T0. The pulse is launched into a standard single mode fibre with the
GVD-parameter β2<0. The launched pulse has a linear frequency chirp at the input
of the fibre which is described by the chirp parameter C>0. The pre-chirp
technique can be used for dispersion compensation. At what distance along the
fibre does the time derivative of the chirp obtain its maximal negative value and
what is the value? Begin with equation (2.4.16) or previous equations.
(Exam. 980828)

7.3 A standard single mode fibre has length L and power attenuation coefficient α per
unit length and GVD-parameter β2. The fibre is excited with a light pulse which is
chirp free, i.e. the chirp parameter C=0, and the power is Gaussian distributed
with the peak power P and the RMS-width σ. What is the peak power, the RMS-
width and the chirp parameter of the pulse at the output?

The fibre is perfectly joined with a second fibre, i.e. the pulse propagates further
into the next fibre without losses and changes due to the joint. The second fibre
has the same values as the foregoing fibre, however the GVD-parameter has
opposite value. What is the peak power, the RMS-width and the chirp parameter
when the pulse has reached the output of the second fibre?
(Exam. 000307)

93
Solutions

7.1
Total chirp is zero
High Low
frequencies frequencies

Slow Fast
frequencies frequencies

The chirp at input has opposite sign of β2 because they can counteract each other
and make the allowed distance longer.

T1  1.5T0

TFWHM  40 ps  T0  width at intensity 1/e 

40
ps  24 ps
1.665

Equation (2.4.17), {β3 = 0}:

 
  L / L 
 
2 2
T1  C 2 L    2 L  D
T1
 1        1  C 2   2
T0   T0 
2 2
T0   2
 T0
T
 LD  0 
  2 

T1
 1.5  1.5  1  C    2
2 2

T0

 
1.25  1  C 2  2  2C

C  2.25C 2  1.25

1 C 2

d
maximise ξ:  0 is calculated numerically
dC

C opt  0.667   max  1.5

94
 Lmax
  1.5  Lmax  1.5  L D
LD

T0 = 24 ps

Consider standard single-mode fibre:

 2  20 ps 2 /km  D  17 ps/(nm km)

 L D  28.9 km  L max  43.3 km

7.2 Equation (2.4.16) gives {i=-j}

A0 T0  (1  jC )t 2 
exp  
A( z, t ) 
T0 2  j 2 z (1  jC )  0 2 
 2 T  j z (1  jC )
2
 


 (1  jC )t 2 
A(0, t )  A0 exp  
 2T0 2 
 

where T0 is the rms-value of the amplitude, T0  2 0

While we want the second order time derivative of the phase, we first part the
argument of the exponential into real and imaginary parts, where the imaginary
part is the phase.


(1  jC )t 2 T0   2 zC  j 2 z
2
   T 2
  
  2 zC   2 zC  j C T0   2 zC   2 z
2
t

  2T   zC    z  
0 2

2 T0   2 zC   2 z 
2

2 2
0
2
2
2
2
2

The second order derivative of the phase,  (z ) , is

( z ) 
C T 0
2

  2 zC   2 z  (1)

 T 2   zC
 0 2 
2
  2 z 2 


C
From (1) we see that (0)  . Further we realise that at a certain value of zzero
T0 2
( z zero )  0 and thereafter negative because β2<0. When z goes to infinity
( z )  0 , see illustration below

95
 (z )

zopt

min

Thus, we want to find min and zopt

Take the derivative of (z ) with respect to z and set the numerator to zero


 T 2   zC
 0 2 
2


   
  2 z 2   2   2 C 2  2 T0 2   2 zC  2 C  2  2 2 z C T0 2   2 zC   2 z  0    
After simplification

  22 1 C 2  z
2 2

 2T0 2  2 C 1  C 2 z  T0 4 1  C 2  0   
Let x   2 1  C 2 z  0

 x 2  2T0 2 Cx  T 4 1  C 2  0  
x  T0 2 C  T 4 C 2  T 4 1  C 2  T0 2 C  1

From the illustration zopt>zzero and hence xopt < xzero

xzero  CT0  xopt  T0 (C  1)
2 2

T0 2 (C  1)
Thus, z opt  

 2 1 C 2 
 T 2 C  1C  T0 2 C  1
C  T0 2  0 
min 

 1  C 2  1 C 2    
 2 T0 2 C  1C   T0 2 C  1 
2 2
T    
 0
 1  C 2
   1 C 2 
    


  
C T0 2 1  C 2  T0 2 C  1C  T0 2 C  1  1  C  
2

T 1  C  T
0
2 2
0
2
C  1C 
2

 T0 C  1 2
 2


 C 2 1  2
C 2
1 
 2

T0 1  C   C  1
2 2
 
2T0 2

96
 Answer: ( z ) 
C T0 2   2 zC   2 z  
2
T 0
2
  2 zC    2 z  2

T0 2 (C  1)
z opt  

 2 1 C 2 

min  
C 2
1 
2
2T0

7.3 Start with equation (2.4.16) with T0 2  2 2 , A0  P , C  0 and complement with

 
the attenuation factor exp  z
 2 
 1 
A( z, t )  {i   j} 
A0 T0
exp  t 2 2
1  exp   z 
 2 T  j z   2 
T0 2  j 2 z  0 2 

A0 2 T0 2
The peak value: A(0, z ) 2  exp( z )
T0 4   2 z 2

1 1  jC z
Write on the form , compare with equation (2.4.10)
T 0  j 2 z
2
Tz 2

2z
1 j
1  jC z T0  j 2 z
2
T0 2
 
Tz 2 T0 4  (  2 z ) 2  z
2
T0 2   2 
 T0 

2
 z 2z
2 z  T z  T0
2 2 2
  2  , and C z 
 T0  T0 2

We may guess that the pulse evolution is the opposite in the second fibre. The
transfer function of the slowly varying amplitude with suppressed group velocity
is

    
0  z  L , H , z   exp  j  2  2 z  exp  z 
~ 1
 2   2 

L  z  2L ,
         
H , z   exp  j  2  2 L  exp  L   exp  j  2  2 ( z  L)  exp  ( z  L)  
~ 1 1
 2   2   2   2 

97
  1    
 exp  j  2  2 (2 L  z )  exp  z  (1)
 2   2 

We notice that the pulse width and chirp, which are determined by the first factor
in equation (1) are equal at z and 2L-z. Especially at 0 and 2L:
C2L = 0, T2L = T0, but P2L=Pexp(-α2L)

Answer: At L: peak power:

P  2  exp( L) ,
2

 2    L
4
2
2

2
 2z 
RMS-width:  2    ,
 2 
2L
chirp parameter:
 2  2

At 2L: peak power: P  exp( 2L) ,

RMS-width: σ,
chirp parameter: 0

98
8. Multichannel systems

Problems

8.1 The low-loss region of a silica fibre extends from 1.5 to 1.6 μm. How many
channels can be transmitted by using optical FDM when the channel spacing is
10 GHz? If each channel is operated at 2 Gb/s with a power budget of 30 dB
allocated to fibre loss, calculate the effective bit rate - distance product BL of
the multichannel system by assuming a loss of 0.2 dB/km. Assume loss
limitation.

Short questions

8.2 Consider a multichannel system. Does the crosstalk increase or decrease due
to four wave mixing if the dispersion parameter is decreased in the fibre?
Motivate your answer.
(Exam. 970411)

8.3 In frequency division multiplexed systems there are different kinds of

crosstalk, e.g. stimulated Raman scattering. Give an explanation to why that
crosstalk is proportional to the frequency spacing (which is assumed to be
equidistant) between the channels.
(Exam. 970306)

8.4 Figures 8.10, 8.14 and 8.16 show different optical passive components. Which
one or which ones of these can in principle be used as a wavelength
demultiplexer?
(Exam. 980305)

8.5 Figure 8.10 shows a multiplexer. What is the time delay in the MZ3 if the
channel spacing is 100 GHz?
(Exam. 980423)

8.6 One can choose between an OTDM-system of 100 Gb/s or a 10-channel

WDM-system, where each channel has a bit rate of 10 Gb/s. The total power is
the 3.16 larger in the 100Gb/s system . Both systems work at about 1.55 μm.
Which system can have the longest standard fibre between transmitter and
receiver? Intensity modulation and direct PIN-diode detection is assumed.
Thermal noise dominates. Motivate your answer.
(Exam. 990830)

99
8.7 What is a broadcast and select system?
(Exam. 000505)

8.8 What is the expression for the filter function in Figure 8.8(b)?
(Exam. 000901)

8.9 How does the component in Figure 8.27(b) work? The function of delay, XPM
and clock signal should also be explained.
(Exam. 020308)

8.10 How does the component in Figure 8.14 work?

(Exam. 970829)

100
Solutions

8.1 Δλ=100 nm (1.6 μm-1.5 μm)

3  108
    1550 nm 
c
   100  10 9  12.5 THz
2 1.55  10 
6 2

number of channels:

 12.5 10 12
N channels    1250
 channel 10 10 9

30 dB
Transmission distance:  150 km (loss limited transmission)
0.2 dB/km

BL - product  N channel  B  distance  1250  2  10 9  150  375 Tbitkm/s

8.2 The condition for four wave mixing is good phase matching. Decreased
dispersion  increased phase matching  increased FWM  increased
crosstalk

8.3 Long-wavelength channels are amplified by the short-wavelength channels as

long as the wavelength difference is within the bandwidth of the Raman gain.
The gain is roughly proportional to the wavelength difference, see figure 2.18.

8.4 Figure 8.10, based on MZ filters can be used as a demultiplexer, Figure 8.14
cannot be used as a demultiplexer since star couplers are not wavelength
selective. Figure 8.16 uses a wavelength router, which can be used as a
demultiplexer.

8.5 The time delay,  m  2 m  ch 

1
in equation (8.2.3), m=1   1  5 10 12 s

8.6 Both systems need same minimum power at the receiver, which gives the
same length if the systems are loss limited. However, dispersion criterion
gives longer length for the WDM, up 100 times more if the lasers are good.
So if the systems are dispersion limited, then the WDM system is better

8.7 Broadcast-and-select means that the transmitter transmits signals to all

receivers of a cluster, even if the signal is meant just for one receiver. The
receiver itself has to select its channel, which for example means that it is
tuned for a specific receiver frequency.

101
8.8
T
1.

2. 4. 6.

3. 5. 7.

a is the amplitude entering port 1

1. a
2. a / 2 3.  ja / 2
4. a / 2 exp(  jT ) 5.  ja / 2
a a a a
6. exp(  jT )  7.  j  j exp(  jT )
2 2 2 2

6. ja exp(  jT / 2) sin(T / 2)  S 61  sin 2 (T / 2)

2

7.  ja exp(  jT / 2) cos(T / 2)  S 71  cos 2 (T / 2)

2

see also (8.2.3)

8.9 Suppose that a clock signal is generated at the same time as a signal
propagating in the clockwise direction passes the “clock signal generator”.
When the clock signal and the clockwise propagating signal reach the XPM,
the clock signal induces a π-phase shift to the clockwise propagating signal
due to cross-phase modulation (XPM).

The counterclockwise signal reaches the XPM later because of the delay and
there is no clock signal at that time, so no phase shift is induced. When the
counterclockwise and clockwise signals have passed the whole loop, the signal
goes to the right due to the phase shift. The signal without the phase shift goes
to the left.

8.10 On each output there is 1/8 power from each input.

102
9. Solitons
Problems

9.1 A soliton communication system is operating at 1. 55 μm by using dispersion-

shifted fibres with D = 2 ps/(nm km). The effective core area of the fibre is 50
μm2. Calculate the peak power and the pulse energy required for launching
fundamental solitons into the fibre. Assume hyperbolic secant pulses with a
FWHM of 30 ps.

9.2 A soliton transmission system is intended to operate at 30 Gbit/s. For this purpose,
10 ps wide (FWHM), transform-limited sech2-shaped power-pulses are used.
What is the required soliton peak power and maximum amplifier spacing in the
two cases below if we neglect fibre loss and assume that the soliton period should
be no smaller than twice amplifier spacing?

a) Standard fibre with D=16 ps/(nm km)

b) Dispersion-shifted fibre with D=1 ps/(nm km)

The operating wavelength is 1550 nm, the effective area of the fibre is 50 μm2,
and the nonlinear index of refraction is 3·10-16 cm2/W.

Short questions

9.3 Assume you have access to the following components: EDFA, standard fibre,
dispersion shifted fibre (2 ps/nm km at 1.55 μm), semiconductor laser at 1.3 μm
and a semiconductor laser at 1.55 μm. A system for soliton communication with
long reach is to be built. Which components should be chosen and why?
(Exam. 970306)

9.4 Consider a soliton system. Why is it better to have closer spacing between the
amplifiers with lower amplification, than the contrary? No economical
circumstances have to be considered.
(Exam. 990830)

9.5 What happens with a soliton pulse when it propagates in a fibre with normal
attenuation?
(Exam. 010831)

9.6 Why should a DSF (Dispersion shifted fibre) rather than a standard single mode
fiber (SMF) be used in a practical soliton system?
(Exam. 020308)

9.7 Solitons interact. What are the consequences for a soliton link?
(Exam. 020408)

103
Solutions

9.1
a)

N  1  P0 LD 
  2  2n2  Aeff  2
  P0     
2
 T
2

 Aeff 
 2n T
2
T 0 2 0
LD  0 
 2 

TFWHM
T0   17 ps
1.763

2
2  D  2.55 ps 2 /km
2  c

 P0  3.4 mW

b)

  
2 
dt  P0 T0  sech 2 x dx  P0 T0  2  0.116 pJ
t
E p   P(t )dt  P0  sech  
   0
T  

9.2 Equations (9.1.7), (2.3.5) and (9.2.4) gives:

0.322 2 c
z0  T FWHM 2
2 D

Equations (9.2.3), (9.1.7) and the nonlinear parameter introduced in section 2.6
gives:

Aeff
Psol 
4n 2 z 0

a) D = 16 ps/(nm km)
z0 = 2.5 km  LA  1.25 km (L<0.5z0)
Psol = 260 mW

b) D = 1 ps/(nm km)
z0 = 40 km  LA  20 km
Psol = 16 mW

104
9.3 Choose 1.55 µm in order to use EDFA to compensate for losses. Further choose
dispersion shifted fibre to maximise amplifier spacing. Then it is necessary to
choose laser diode at 1.55 µm.

9.4 A soliton should have N=1 (first-order soliton). When the soliton is amplified
N  1 , (N ≤1.5 is ok) and by creating dispersive waves the soliton again can
achieve N  1 . In order to avoid significant levels of dispersive waves the
amplification spacing << dispersion length, LA<<LD.

9.5 The soliton characteristics remain, i.e. the pulse is still a soliton. However,
because the power is attenuated the peak power is decreased and the pulse is
  PT 2
broadened in order to maintain N 2  1  . See Figure 9.9.
2

9.6 See page 420. For DSF the amplifier spacing is then reasonably large and fewer
amplifiers are needed in comparison with SMF. DSF gives a cheaper system.

9.7 The total transmission distance, LT, must be made much shorter than the
oscillation period, LT<< ξLD, in order to avoid soliton interaction. See page 414.

105
10. Coherent lightwave systems

Problems

10.1 A fibre optic link consists of a transmitter with wavelength 1.55 μm, which
transmits ONES and ZEROS with the bitrate 2.4 Gb/s in ASK-format. The
detector is shown in Figure 10.1. The optical field falling on the photodetector
is mixed with the local oscillator signal and the signal in the following way.
When a ONE is received, the amplitude of the optical field at the detector is
ELO+E1 during the entire bit slot and both the fields have the same phase,
frequenc, and polarisation. It can be assumed that ELO>>E1. When a ZERO is
received, the amplitude of the field at the detector is ELO during the entire bit
slot. The detector is a PIN-detector with the efficiency 1. The noise factor is 3,
the noise bandwidth is 1.2 GHz and the load resistance is 50 Ω. What is the
thermal noise? How many photons does the signal, which has the amplitude
E1, contain for a BER = 10-9? The power of the field with the amplitude
ELO=1mW. Assume Gaussian statistics. All approximations must be
explained.
(Exam. 020408)

Short questions

10.2 The shot noise is increased when the power of the local oscillator is increased
in coherent receivers. Why is it still useful to have a powerful local oscillator?
(Exam. 960830)

10.3 A balanced mixer or receiver in coherent systems is said to save certain

advantages. Why does not the receiver in Figure 10.1 have that?
(Exam. 990309)

10.4 In a coherent receiver the phase changes from 0 to 45 degrees between the
polarisation of the local oscillator, which is linear, and the polarisation of the
signal, which is also linear. How does the signal current from the detector
change?
(Exam. 990409)

10.5 A coherent system uses binary PSK-modulation. The phase between the
polarisation of the local oscillator, which is linear, and the polarisation of the
signal, which is also linear, is 90 degrees. How does the total detected current,
with one and zero bits, vary? Motivate your answer.
(Exam. 990830)

106
10.6 How does a polarisation-diversity in the receiver in a coherent system work?
(Exam. 000307)

10.7 State three important advantages with coherent system in comparison with
direct detection without optical amplifiers.
(Exam. 000505)

10.8 State one advantage and one disadvantage for a heterodyne asynchronous
demodulation compared to homodyne demodulation for a ASK-system.
(Exam. 010309)

10.9 State two advantages for using a balanced receiver in coherent systems.
(Exam. 010831)

10.10 Study formula (10.1.3). Is Ps the power of the received signal and PLO the
power of the local oscillator according to Figure 10.1?
(Exam. 020823)

107
Solutions

10.1
λ=1.55 μm
RL=50 Ω
E1,0
F=3
Δf=2·109 GHz
η=1
ELO

The problem deals with homodyne ASK.

I1  I 0
BER=10-9  Q=6 ; Q  , Gaussian statistics
1   0

q q 
I 0  PLO , where    
hv hv 1.24

2
PLO ~ E LO


I 1   PLO  2 P1 PLO  P1 , E , E 1 LO have the same phase, frequency, and
polarisation.

Assume PLO  P1  I1   PLO  2 P1 PLO  

4k B T0 Ff
 0 2   s 0 2   T 2 ;  s 0 2  2qPLO f ;  T 2 
RL

 12   s12   T 2 ;  s1 2  {Assume PLO  2 P1 PLO }   s 0 2

  12   0 2

2 P1 PLO
Q
2  T 2   s0 2

Q2  4k T Ff 
P1   B 0  2qPLO f 
 PLO
2
 RL 

4 1.38 10 23  290  3 1.2 10 9

T 2   1.15 pA 2
50
1.55
 s0 2  2 1.6 10 19  10 3 1.2 10 9  0.48 pA 2
1.24

108
 Consequently,  T 2 cannot be neglected.

P1 
36
2
 
 1.15  0.48  10 12  37 nW
 1.55  3
  10
 1.24 

Does PLO  2 PLO P1 hold?  10 3  12  10 6  Yes!

Now to the number of photons of the signal.

P1 P1  37 10 9 1.55
N    120
B  hv B 1.24  q 2.4 10 9 1.24 1.6 10 19

Answer: Thermal noise:  T 2  1.15 pA 2

Number of photons: N = 120

NOTE: P1 and PLO are not the powers before the semitransparent mirror, beam
combiner. They are P1/k and PLO/(1-k), respectively, where 0<k<1.

10.2 If LO-power is sufficiently high then the thermal noise is negligible in

comparison with shot noise. The system approaches the quantum limitation,
e.g. 9 signal photons for homodyne PSK, as also the signal current increases.

10.3 The beam combiner also couples the signal and LO powers vertically. These
powers are lost because there is no detector in this direction. Compare with
Figure 10.11. A balanced receiver has two detectors and uses all power.

10.4 See section 10.5.3. Only components of the same polarisation cooperate.
According to equation (10.1.7) i.e. 2 Ps P LO cos( IF t   s   LO ) , this results in
the signal current

2 Ps P LO cos( p ) cos( IF t   s   LO )

Thus, the signal current is decreased with 2

10.5 According to chapter 10.5.3, the interference is zero if the polarisations are
orthogonal. The detected current will be proportional to Ps + PLO. For PSK
modulation Ps is equal for a one bit and a zero bit. Thus the total current is not
changed.

109
10.6 Incoming signal power Ps with unknown polarisation angle φ is divided into
two orthonormal polarisations, e.g. e x , e y . The LO has controlled polarisation
and is similarly divided. Then the mixing products ~ PLO / 2  Ps cos( ) and
~ PLO / 2  Ps sin( )
are achieved in the two detectors. The photocurrents are
squared and added which results in φ-independence. (See Figure 10.12.)

10.7 Four important advantages with coherent system are

1) About 20 dB more sensitive
2) More dense channel packaging possible
3) Dispersion compensation in the IF domain possible
4) Allows phase modulation

10.8 The sensitivity is worse: 40 compared to 18 photons, on average, for a BER

= 10-9, which is a disadvantage. An advantage is that phase control is not
needed.

10.9 The intensity noise at the LO is eliminated.

The LO-and signal power are effectively utilized.

Ps
10.10 No! Received optical power is  Ps ' and the power of the local oscillator
k
PLO
is  PLO ' , because of the beam combiner where 0  k  1 . This explains
1 k
the paradox that P can be larger than PLO+PS in 10.1.3.

110