Rose-Hulman Institute of Technology

Rose-Hulman Scholar
Graduate Theses - Physics and Optical Engineering Graduate Theses

5-2018

Gain Modeling of Erbium-Doped Fiber Amplifiers

Pumped at 980nm
Deepak Charles Baskar

Follow this and additional works at: https://scholar.rose-hulman.edu/optics_grad_theses

Part of the Optics Commons

Recommended Citation
Baskar, Deepak Charles, "Gain Modeling of Erbium-Doped Fiber Amplifiers Pumped at 980nm" (2018). Graduate Theses - Physics and
Optical Engineering. 23.
https://scholar.rose-hulman.edu/optics_grad_theses/23

This Thesis is brought to you for free and open access by the Graduate Theses at Rose-Hulman Scholar. It has been accepted for inclusion in Graduate
Theses - Physics and Optical Engineering by an authorized administrator of Rose-Hulman Scholar. For more information, please contact weir1@rose-
hulman.edu.
Gain Modeling of Erbium-Doped Fiber Amplifiers Pumped at 980nm

A Thesis

Submitted to the Faculty

of

Rose-Hulman Institute of Technology

by

Deepak Charles Baskar

In Partial Fulfillment of the Requirements for the Degree

of

Master of Science in Optical Engineering

May 2018

© Deepak Charles Baskar

i
 ABSTRACT

Baskar, Deepak Charles

M.S.O.E.

Rose-Hulman Institute of Technology

May 2018

Gain Modeling of Erbium-Doped Fiber Amplifiers Pumped at 980 nm

Thesis Advisor: Dr. Sergio Granieri

Erbium-Doped Fiber Amplifiers (EDFA) are one of the most widely used optical amplifiers

in the field of optical communications and fiber lasers. Theoretical models based on the rate

equations, therefore, were developed to predict the performance of such amplifiers. The goal of

this thesis is to provide a numerical model for EDFAs and verify its validity through experimental

measurements. Two computer programs based on two different numerical methods (the Finite

Difference method and the 4th Order Runge-Kutta Method) to solve differential equations were

written. The different fiber parameters to build the model including absorption and emission cross-

sections and scattering losses were experimentally determined. Two different optical amplifiers

were built using different lengths of doped Erbium fiber. Experimental output signal optical power

and gain of the two amplifiers were measured for different values of input signal power and pump

power. These results were predicted by the numerical model with a considerable degree of

accuracy.

Keywords: Numerical models, Erbium-Doped Fiber Amplifier, finite difference method

To my parents, my uncle, and Sai
 ACKNOWLEDGEMENTS

I would like to express my sincere gratitude to my advisor, Dr. Sergio Granieri for his continued

support, insight, and patience throughout my research. His guidance was instrumental to the

nuances of my work, and I could not imagine having another advisor and mentor.

In addition, I would like to extend my heartfelt thanks to the rest of my thesis committee: Dr.

Robert Bunch, Dr. Richard Liptak, and Dr. Joo for their encouragement and critical insight. I would

also like to thank Dr. Joo for his support during my exchange student program in Seoul, Korea.

My sincere thanks go to Karen DeGrange, Dr. Craig Downing, and my counselors for supporting

me and helping me through the tough times.

I would also like to thank my friends who have helped me stay on track and focused: Srirram

Sridhar, Rama Krishnan, Léa Vincent, Vishnu PG, Gautham Chinnaswami, Prasanna KN, Preshya

Stanley, Prithiviraj Shanmugam, Tahereh Naderishahab, Amanda Hinson, Korey Jividen, Ani,

Hobbs, Ana, and Tayler Evans.

I am forever indebted to my parents, Charles Baskar and Arputhamani Briskillah, and my uncle

Charles Joenathan who have stayed by my side and supported me through thick and thin.

Last but not least, I would like to thank Sai Krupa for making my life what it is.

iii
Contents

1.1 A Brief History of Rare-Earth-Doped Fibers Amplifiers................................................ 3

1.2 Motivation for this thesis .............................................................................................. 4

1.3 Organization of the Thesis ............................................................................................ 5

2.1 Derivation of the Equations .......................................................................................... 6

2.1.1 Derivation of the Population Equation .................................................................................. 8

2.1.2 Derivation of the Coupled Differential Power Equations ........................................................ 9

2.1.3 Amplified Spontaneous Emission ........................................................................................ 12

2.2 Assumptions ............................................................................................................... 13

2.2.1 Bessel functions.................................................................................................................. 13

2.3 Methods to ascertain Cross-Sections ........................................................................... 14

2.3.1 Fuchtbauer-Landenberg Analysis ........................................................................................ 15

2.3.2 Gain-Loss and Saturation Power Measurements ................................................................. 19

2.3.3 McCumber Relation ............................................................................................................ 23

3.1 Modeling .................................................................................................................... 25

3.1.1 Finite Difference Method.................................................................................................... 26

iv
 3.1.2 4th Order Runge-Kutta Method: .......................................................................................... 27

3.2 MATLAB Coding ...................................................................................................... 28

3.2.1 Finite Difference Method.................................................................................................... 28

3.2.2 4th Order Runge-Kutta Method ........................................................................................... 35

3.3 Adjustments to model with MATLAB ........................................................................ 39

3.4 Simulations................................................................................................................. 40

3.4.1 Test #1 ............................................................................................................................... 40

3.4.2 Test #2 ............................................................................................................................... 43

3.5 Data Required for Simulation...................................................................................... 45

4.1 Overview .................................................................................................................... 47

4.2 Fiber Information........................................................................................................ 47

4.3 Characterization of the components ............................................................................ 48

4.3.1 Characterization of the WDM ............................................................................................. 49

4.3.2 Mode Field Diameter Mismatch Loss .................................................................................. 51

4.3.3 OSA vs. Power Meter Calibration ........................................................................................ 52

4.4 Absorption Spectrum and Cross-Section: .................................................................... 53

4.5 Emission Spectrum and Cross-Section: ....................................................................... 58

5.1 General Procedure ...................................................................................................... 61

5.1.1 Amplified Spontaneous Emission ........................................................................................ 63

5.2 EDFA based on Sample #1 ......................................................................................... 65

5.2.1 Absorption Results for Sample #1 ....................................................................................... 66

5.2.2 Test #1 with Sample #1 ....................................................................................................... 68

v
 5.2.3 Test #2 with Sample #1 ....................................................................................................... 69

5.2.4 Test #3 with Sample #1 ....................................................................................................... 71

5.3 EDFA based on Sample #2 ......................................................................................... 72

5.3.1 Absorption Results for Sample #2 ....................................................................................... 74

5.3.2 Test #1 with Sample #2 ....................................................................................................... 75

5.3.3 Test #2 with Sample #2 ....................................................................................................... 77

5.4 Adjustment of the MATLAB Model ........................................................................... 78

vi
 List of Figures

Figure 2-1: Er3+ 3-Level System with 980nm pump .....................................................................6

Figure 2-2: Er3+ 3-Level System with both 980nm pumps and ‘two’ levels..................................7

Figure 2-3 Recreation of Gain/Loss data plotted against P/Pth [10] ........................................... 20

Figure 3-1 Flowchart describing execution process for Finite Difference Method based
MATLAB code ......................................................................................................................... 29

Figure 3-2 Code snippet that shows the computation of MFD for both pump and signal lasers in
the un-doped and the doped fibers ............................................................................................. 30

Figure 3-3 Code snippet showing the computation of optical power that enters the WDM ......... 31

Figure 3-4 Flowchart describing process flow for the Main function in Finite Difference model 32

Figure 3-5 Code snippet showcasing the computation of the rates in the Main function of the
Finite Difference program ......................................................................................................... 33

Figure 3-6 Code snippet of the Amplified Spontaneous Emission computation in the Finite
Difference model ...................................................................................................................... 33

Figure 3-7 Code snippet of the computation of γr and γe for the Pump and Signal in the Finite
Difference model. ..................................................................................................................... 34

Figure 3-8 Code snippet of the computation of the Signal and Pump Optical Power for the next
‘step’ in the Finite Difference Model ......................................................................................... 34

Figure 3-9 Flowchart describing execution process for Runge-Kutta based MATLAB code ...... 36

Figure 3-10 Flowchart showing the execution of Gamma Function ........................................... 37

Figure 3-11 Code snippet showcasing the decision making statements of the Gamma Function in
the RK4 method ........................................................................................................................ 38

Figure 3-12 Code snippet showcasing the function handles that call the two Gamma functions . 38

Figure 3-13 Code snippet of the Main function for the 4th Order Runge-Kutta based solution .... 39

Figure 3-14 The Variation of Small Signal Gain with respect to length from [11] ...................... 41
vii
Figure 3-15 The variation of Large Signal Gain with respect to length from [11]....................... 41

Figure 3-16 Gain vs. Length for Small Signal Input................................................................... 42

Figure 3-17 Gain vs. Length for Large Signal Input................................................................... 42

Figure 3-18 Comparison between results obtained from the simulations and the results from
Mohammad [12] for an input pump power of 8.148mW ............................................................ 44

Figure 3-19 Comparison between results obtained from the simulations and the results from
Mohammad [12] for an input pump power of 5.397mW ............................................................ 45

Figure 4-1 Final Schematic for WDM ....................................................................................... 49

Figure 4-2 Experimental Setup to measure the Input to the Erbium-Doped Fiber ....................... 53

Figure 4-3 Output of the Inphenix LED Source ......................................................................... 54

Figure 4-4 Output of the Broadband Source .............................................................................. 54

Figure 4-5 Experimental Setup for measuring cross-sections ..................................................... 55

Figure 4-6 Absorption Spectra for the Er110-4/125 Fiber using the LED Source ....................... 56

Figure 4-7 Absorption Spectra for the Er110-4/125 Fiber using the Broadband Source.............. 57

Figure 4-8 Comparison between experimental and vendor Absorption Cross-Section of Er110-
4/125 ......................................................................................................................................... 58

Figure 4-9 – Normalized Amplified Spontaneous Emission with a Pump Current of 35mA ....... 59

Figure 4-10 Comparison between experimental and vendor Emission Cross-Section of Er110-
4/125 ......................................................................................................................................... 60

Figure 5-1 Schematic for the EDFA experiments ...................................................................... 62

Figure 5-2 Experimental Amplified Spontaneous Emission measured at a pump optical power of
102.51mW ................................................................................................................................ 64

Figure 5-3 Simulated Amplified Spontaneous Emission at a pump optical power of 102.51mW 64

Figure 5-4 Output Signal Powers vs. Input Signal Powers for 13.3cm EDF for different Pump
Input Powers ............................................................................................................................. 66

viii
Figure 5-5 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for No Pump Power Input (13.3 cm) ............................................................................... 67

Figure 5-6 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for an input Pump Power of 25.7 mW (13.3 cm) ............................................................. 69

Figure 5-7 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for an input Pump Power of 83.1 mW (13.3 cm) ............................................................. 70

Figure 5-8 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for an input Pump Power of 102.5 mW (13.3 cm) ........................................................... 72

Figure 5-9 Output Signal Powers vs. Input Signal Powers for 27.7 cm EDF for different Pump
Input Powers ............................................................................................................................. 73

Figure 5-10 Experimental Output Power vs. Simulated Output Power as a function of Input
Signal Power for no input Pump Power (27.7 cm) ..................................................................... 75

Figure 5-11 Experimental Output Power vs. Simulated Output Power as a function of Input
Signal Power for an input Pump Power of 83.1 mW (27.7 cm) .................................................. 76

Figure 5-12 Experimental Output Power vs. Simulated Output Power as a function of Input
Signal Power for an input Pump Power of 102.5 mW (27.7 cm) ................................................ 78

Figure 5-13 Default Input Signal Optical Mode ......................................................................... 79

Figure 5-14 Modified Input Signal Optical Mode ...................................................................... 80

Figure 5-15 Experimental Output Power vs. Adjusted Simulated Output Power as a function of
Input Signal Power for no input Pump Power (13.3 cm) ............................................................ 81

Figure 5-16 Experimental Output Power vs. Adjusted Simulated Output Power as a function of
Input Signal Power for an input Pump Power of 25.7 mW (13.3 cm) ......................................... 81

Figure 5-17 Experimental Output Power vs. Adjusted Simulated Output Power as a function of
Input Signal Power for an input Pump Power of 83.1 mW (13.3 cm) ......................................... 82

Figure 5-18 Experimental Output Power vs. Adjusted Simulated Output Power as a function of
Input Signal Power for an input Pump Power of 102.5 mW (13.3 cm) ....................................... 82

Figure 5-19 Experimental Output Power vs. Adjusted Simulated Output Power as a function of
Input Signal Power for an input Pump Power of 83.1 mW (27.7 cm) ......................................... 83

Figure 5-20 Experimental Output Power vs. Adjusted Simulated Output Power as a function of
Input Signal Power for an input Pump Power of 102.5 mW (27.7 cm) ....................................... 84
ix
List of Tables

Table 2-1 Results of the Fuchtbauer-Landenberg Analysis for different fiber types in [10] ........ 18

Table 2-2 Cross-Section Ratio Comparison [σE/σA] ................................................................... 21

Table 2-3 Pump Absorption Cross-Sections as obtained by Barnes et al. [10] ......................... 22

Table 2-4 Emission and Absorption Cross-Sections, as derived using the Saturation Method [10]
................................................................................................................................................. 23

Table 4-1 Port Efficiencies when P2 and P3 are input ports for Signal Wavelength ................... 50

Table 4-2 Port Efficiency when P2 is the input port for Pump Wavelength ................................ 50

Table 4-3 Final WDM test displaying the average output percentage at the output port for both
inputs wavelengths .................................................................................................................... 51

Table 4-4 Splice Losses when SMF 28e is spliced with Er110-4/125......................................... 52

Table 4-5 OSA and Power Meter Calibration ............................................................................ 52

Table 5-1 Output Signal Powers vs. Input Signal Powers for 13.3cm EDF for different Pump
Input Powers ............................................................................................................................. 65

Table 5-2 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for No Pump Power Input (13.3 cm) ............................................................................... 67

Table 5-3 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for an input Pump Power of 25.7 mW (13.3 cm) ............................................................. 68

Table 5-4 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for an input Pump Power of 83.1 mW (13.3 cm) ............................................................. 70

Table 5-5 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for an input Pump Power of 102.5 mW (13.3 cm) ........................................................... 71

1
Table 5-6 Output Signal Powers vs. Input Signal Powers for 27.7 cm EDF for different Pump
Input Powers ............................................................................................................................. 73

Table 5-7 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for no input Pump Power (27.7 cm) ................................................................................ 74

Table 5-8 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for an input Pump Power of 83.1 mW (27.7 cm) ............................................................. 76

Table 5-9 Experimental Output Power vs. Simulated Output Power as a function of Input Signal
Power for an input Pump Power of 102.5 mW (27.7 cm) ........................................................... 77

2
 INTRODUCTION

1.1 A Brief History of Rare-Earth-Doped Fibers Amplifiers

Doped fiber amplifiers are a type of optical amplifier that use rare earth metals like Erbium

and Thulium to provide the medium for the stimulated emission that amplifies the input optical

signal. The first doped fiber amplifier was a Neodymium-doped fiber operating at 1.06 um devised

by E Snitzer, in 1964 [1].

This work lay dormant after a demonstration of its abilities until the advent of silica glass

fibers that could be used for telecommunications. [2] Almost a decade later, rare-earth-doped lasers

were inspected as a possible device for transmission purposes. [3] In 1983, single mode rare-earth-

doped fibers were exhibited by Broer and Simpson at Bell Laboratories. [4] The fiber was doped

with Neodymium at a concentration of 10ppm, and it helped in the study of the relaxation

mechanisms of rare-earth ions in an amorphous medium.

A few years later, the development of fiber amplifiers occurred because of improvements

to fabrication techniques. [5] Erbium-doped fiber amplifiers were fabricated simultaneously by

both AT&T Bell Laboratories and the University of Southampton. [2] The key advancement here

was the identification of the Er3+ ion for its ideal transition wavelengths. This triggered research

into these fibers, and it became the catalyst to develop a new generation of transmission networks.

In the present day, fiber amplifiers are essential to the transmission with massive undersea cables

and networks being a good example of how integrated into our lives it has become.

3
1.2 Motivation for this thesis

The widespread usage of fibers in optical communication required characterization of the

fibers. Therefore, theoretical models were developed based on existing optical theory and these

models most often involve complex differential equations, even if the amplifier was operating

under ideal conditions.

The computational intensity of such models was high with regards to the performance of

existing computers. Pedersen et al. mention [6] that they would have to solve 402 coupled

differential equations to solve just the four equations that describe the power and the noise along

the length of the fiber. This is excluding the 40 steps that they had to consider for the optical mode

of the fiber. This meant that the rate equations had to be computed for 40 separate points.

There is a huge level of computation required for this, and therefore it can be extrapolated

that some level of approximation had to be done to make sure that the results were obtained in

time. This thesis attempts to recreate the models that were proposed during the rise in popularity

of fiber amplifiers. It follows the model that Pedersen et al. [6] describe closely. This model is in

itself a reduction of a model that Desurvire and Giles present in [7].

Considering the advancements in computation and programming since the early 90’s, the

researcher hoped that any inaccuracies that were inherent to older means of computation could be

overcome. The goal of this thesis is to model an Erbium-Doped Fiber Amplifier accurately which

would then be used to model a fiber Q-switched laser with a saturable absorber.

4
1.3 Organization of the Thesis

The thesis is presented in the following manner:

Chapter 2 deals with the theory needed to understand the working of laser amplifiers. The

atomic rate equations and the equations that describe the power and the population density in the

fiber are derived based on the information and theory mentioned by Pedersen et al. in [6] and Giles

and Desurvire in [7].

Chapter 3 deals with the process of building the model using MATLAB, the mathematical

theory behind the solution of the equations described in Chapter 2 and the difficulties that were

faced while programming. The chapter describes two of the techniques used to solve the coupled

differential equations that were derived in Chapter 2, the adjustments to the code, and then shows

the testing of the model with data from papers that performed simulations.

Chapter 4 involves experiments conducted to characterize both the fiber and the total

system that is built. It includes a verification of the data procured from the vendor.

Chapter 5 deals with experiments conducted with the fiber when it works as an amplifier.

This chapter describes how the final experiments were conducted, the issues faced, and a showcase

of the simulated results for different lengths of the fiber.

The conclusion and future work sections deal with possible adjustments that can be made

to the code and possible additions to the code with regards to non-linear optical effects that were

excluded from this model and possible tweaks that could be done to make the model more accurate.

5
 THEORY OF ERBIUM-DOPED FIBER AMPLIFIERS

2.1 Derivation of the Equations

A three-level laser system is the standard model for Erbium-Doped Fiber Amplifiers

(EDFAs). There are two possible three-level configurations. The first configuration, as shown in

Figure 2-1 has the first level as the ground state, the third level as the short-lived excited state or

the pump state, and the second level as the metastable state, which is characterized by the lifetime

τ. Another three-level system that exists has the metastable state as the third level instead of the

second [8]. Since the former system corresponds to the case of the Er3+ ion, it is used for the model.

Figure 2-1: Er3+ 3-Level System with 980nm pump

The lifetime τ typically is a few milliseconds, while the decay rate from the short-lived

excited state is extremely low, most often about 10-8 seconds. Therefore, the three-level system

can be approximated by a two-level system, as shown in Figure 2-2.

6
 Figure 2-2: Er3+ 3-Level System with both 980nm pumps and ‘two’ levels

To derive a list of equations that would characterize the two-level system depicted in Figure

2-2, the following terms used in the equations must be defined. The subscript ‘p’ defines any rate

for the pump laser, while the subscript ‘s’ is for any signal laser and to differentiate the two further,

the letter ‘R’ is used in any instance of rates involving the pump, while ‘W’ is used for rates linked

to the signal. Since there are two types of transitions that can happen, they are also defined by the

appropriate subscripts; ‘a’ for absorption, and ‘e’ for emission.

Therefore, we can write the transitions as follows:

Rpa: Pump absorption rate from level 1 to level 2

Rpe: Pump emission rate from level 2 to level 1

Wsa: Stimulated absorption rate from level 1 to level 2

Wse: Stimulated emission rate from level 2 to level 1

and we define Ae as the spontaneous emission rate (which occurs from level 2 to level 1).

The population density at level 1 is defined by 𝑁1 while the density at level 2 is defined as 𝑁2 . The

total density of the Erbium ions is given by 𝜌𝐸𝑟 .

7
2.1.1 Derivation of the Population Equation

The equations for the population density of both the energy levels can be derived from the

rates. The reduced two-state model is considered for this derivation. The atomic rate equations

with respect to the two population levels can be written as follows:

𝑑𝑁2 𝑁2
= 𝑁1 𝑅𝑝𝑎 + 𝑁1 𝑊𝑠𝑎 − 𝑁2 𝑅𝑝𝑒 − 𝑁2 𝑊𝑠𝑒 − 2-1
𝑑𝑡 𝜏

𝑁1 + 𝑁2 = 𝜌𝐸𝑟 2-2

Equation 2-1 is evaluated in the steady-state regime where the rate is zero. Therefore, 2-1

can be equated to zero which lets the equation be rearranged as follows:

𝑁2
𝑁1 𝑅𝑝𝑎 + 𝑁1 𝑊𝑠𝑎 = 𝑁2 𝑅𝑝𝑒 + 𝑁2 𝑊𝑠𝑒 + 2-3
𝜏

Factoring out 𝑁1 and 𝑁2 from 2-3 and rearranging the resultant equation, the following

expression for 𝑁2 can be obtained:

(𝑅𝑝𝑎 + 𝑊𝑠𝑎 )
𝑁2 = 𝑁1
1 2-4
(𝑅𝑝𝑒 + 𝑊𝑠𝑒 + )
𝜏

Now, using the above expression in 2-2 will result in an expression for 𝑁1 in terms of the

rates after simplification as seen in 2-5:

8
 (𝑅𝑝𝑎 + 𝑊𝑠𝑎 )
𝑁1 + 𝑁1 = 𝜌𝐸𝑟
1
(𝑅𝑝𝑒 + 𝑊𝑠𝑒 + )
𝜏

1
(𝑅𝑝𝑒 + 𝑊𝑠𝑒 + )
𝑁1 = 𝜌𝐸𝑟 𝜏
1 2-5
(𝑅𝑝𝑎 + 𝑊𝑠𝑎 + 𝑅𝑝𝑒 + 𝑊𝑠𝑒 + )
𝜏

An expression for 𝑁2 can be obtained by using 2-2 and 2-5to obtain the expression below:

(𝑅𝑝𝑎 + 𝑊𝑠𝑎 )
𝑁2 = 𝜌𝐸𝑟
1 2-6
(𝑅𝑝𝑎 + 𝑊𝑠𝑎 + 𝑅𝑝𝑒 + 𝑊𝑠𝑒 + )
𝜏

Equations 2-5 and 2-6 together define the population density at steady state. They are used

in the coupled differential equations that define the pump and the signal powers. This thesis uses

the approach that Pedersen et al. use in [6]. The next section derives these equations along with

the expressions for Amplified Spontaneous Emission.

2.1.2 Derivation of the Coupled Differential Power Equations

This section will closely follow the equations described in [6]. The various pump and signal

absorption and emission rates must first be defined. The pump absorption rate is defined as follows:

𝑃𝑝 (𝑧) 01
𝑅𝑝𝑎 (𝑟, 𝑧) = 𝜎𝑝𝑎 𝐼 (𝑟)
ℎ𝜈𝑝 𝑝 2-7

9
where σpa is the absorption cross-section at the pump wavelength, 𝑃𝑝 (𝑧)is the pump power at ‘𝑧’,

ℎ is Planck’s constant, 𝜈𝑝 is the pump frequency and 𝐼𝑝01 is the normalized pump LP01 mode that

satisfied the following equation.

∞
2𝜋 ∫ 𝐼𝑝01 (𝑟)𝑟𝑑𝑟 = 1
2-8
0

Similarly, the emission rate for the pump can be defined:

𝑃𝑝 (𝑧) 01
𝑅𝑝𝑒 (𝑟, 𝑧) = 𝜎𝑝𝑒 𝐼 (𝑟)
ℎ𝜈𝑝 𝑝 2-9

In the case of this thesis, the pump laser operates at 980 nm. Therefore, the emission cross-

section for the pump is considered to be zero, and so Rpe is zero. The same absorption and emission

cross-sections can be defined for the signal wavelength as well. These rates, however, will include

the effect of Amplified Spontaneous Emission (ASE):

𝑃𝑠 (𝑧) ∞ 𝜎 (𝜈 )
𝑎
𝑊𝑠𝑎 (𝑟, 𝑧) = [𝜎𝑎 (𝜈𝑠 ) +∫ 𝑆𝐴𝑆𝐸 (𝜈, 𝑧)𝑑𝜈] 𝐼𝑠01 (𝑟)
ℎ𝜈𝑠 0 ℎ𝜈 2-10

∞
𝑃𝑠 (𝑧) 𝜎𝑒 (𝜈 )
𝑊𝑠𝑒 (𝑟, 𝑧) = [𝜎𝑒 (𝜈𝑠 ) +∫ 𝑆𝐴𝑆𝐸 (𝜈, 𝑧)𝑑𝜈] 𝐼𝑠01 (𝑟)
ℎ𝜈𝑠 0 ℎ𝜈 2-11

where σe(ν) and σa(ν) are the emission and absorption cross-section while νs is the signal laser

frequency. SASE(ν,z) is the amplified spontaneous emission spectral density at position ‘z’. The

10
spontaneous emission is amplified in both the forward and backward directions. Therefore,

SASE(ν,z) has to be determined from the backward, and the forward traveling amplified

spontaneous emission spectrum:

− (
𝑑𝑆𝐴𝑆𝐸 𝜈, 𝑧) − (
= −2ℎ𝜈𝛾𝑒 (𝜈, 𝑧) − [𝛾𝑒 (𝜈, 𝑧) − 𝛾𝑎 (𝜈, 𝑧)]𝑆𝐴𝑆𝐸 𝜈, 𝑧) 2-12
𝑑𝑧

+ (
𝑑𝑆𝐴𝑆𝐸 𝜈, 𝑧) + (
= +2ℎ𝜈𝛾𝑎 (𝜈, 𝑧) + [𝛾𝑒 (𝜈, 𝑧) − 𝛾𝑎 (𝜈, 𝑧)]𝑆𝐴𝑆𝐸 𝜈, 𝑧) 2-13
𝑑𝑧

The absorption and emission factors are determined from the overlap integral between the signal

mode and the population density of the ground and the excited states respectively:

𝑎𝑑
𝛾𝑒 (𝜈, 𝑧) = 𝜎𝑒 (𝜈 )2𝜋 ∫ 𝑁2 (𝑟, 𝑧)𝐼𝑠01 (𝑟)𝑟 𝑑𝑟 2-14
0

𝑎𝑑
𝛾𝑎 (𝜈, 𝑧) = 𝜎𝑎 (𝜈 )2𝜋 ∫ 𝑁1 (𝑟, 𝑧)𝐼𝑠01 (𝑟)𝑟 𝑑𝑟 2-15
0

where ad is the Erbium doping radius. Therefore, the signal power amplified in the forward

direction is given by

𝑑𝑃𝑠 (𝑧)
= [𝛾𝑒 (𝜈𝑠 , 𝑧) − 𝛾𝑎 (𝜈𝑠 , 𝑧)]𝑃𝑠 (𝑧) 2-16
𝑑𝑧

11
 And similarly, the pump power in the forward direction is given by:

𝑑𝑃𝑝 (𝑧)
= [𝛾𝑒 (𝜈𝑝 , 𝑧) − 𝛾𝑎 (𝜈𝑝 , 𝑧)]𝑃𝑝 (𝑧) 2-17
𝑑𝑧

2.1.3 Amplified Spontaneous Emission

Spontaneous Emission is a phenomenon present in all optical amplifiers. There is the

possibility that any ion in the excited state could spontaneously decay to the lower energy state by

emitting a photon that is not coherent with the photons emitted via a stimulated process.

It can be understood that the phenomenon exists in fiber amplifiers as well. The original

spontaneous photon can cause emission of photons that are coherent with itself and not the main

signal. This can occur within the entire signal bandwidth and therefore can cause a drastic

reduction in gain. It is referred to as Amplified Spontaneous Emission or ASE with regards to

EDFAs since the spontaneous emission evolves along the entire length of the fiber, both in the

forward and the backward directions. This, therefore, means that the ASE can also be defined by

a set of coupled equations, 2-12 and 2-13.

12
2.2 Assumptions

The equations used by Pedersen et al. is based on certain assumptions which have been

adopted. They are:

 It assumes that pumping is done at 980nm which would mean that the emission

cross-section for the pump is zero, as mentioned earlier when deriving the coupled

differential equations.

 There is negligible Excited State Absorption while modeling.

 Other non-linear optical effects, like two-photon absorption, non-linear dispersion

are assumed to be negligible.

2.2.1 Bessel functions

The mode for the fiber is simply defined as a normalized LP01 mode in [6]. This is

insufficient to model the fiber accurately, and therefore we use the equations that Giles and

Desurvire use in [7] to model the fiber mode:

𝑢𝑟 2
1 𝑢 𝐽0 ( 𝑎 )
𝐼𝑘 (𝑟) = [ ] 𝑟<𝑎
𝜋 𝑎𝑉 𝐽1(𝑢) 2-18

𝑣𝑟 2
1 𝑢 0 𝑎)
𝐾 (
( )
𝐼𝑘 𝑟 = [ ] 𝑟≥𝑎
𝜋 𝑎𝑉 𝐾1 (𝑣 ) 2-19

13
 J0,1 and K0,1 are the Bessel and modified Bessel functions, V is the Fiber Number, and the

variables u, and v are defined based on the value of V, and r is the radius of the fiber. The V number

is given by the equation where λk is the chosen wavelength:

1
2𝜋𝑎
𝑉= (𝑛𝑐𝑜𝑟𝑒 2 − 𝑛𝑐𝑜𝑟𝑒 2 )2 .
𝜆𝑘

If V lies between 1 and 3, u and v can be approximated as [8]

𝑣 = 1.1428𝑉 − 0.9960
1
𝑢 = ( 𝑉 2 − 𝑣 2 )2 .

The calculation requires knowledge of the index of refraction of both the core and the

cladding of the fiber and the numerical aperture is used in the code in the place of the refractive
∞
index expression. This mode is normalized and therefore 2𝜋 ∫0 𝐼𝑘 01 (𝑟)𝑟𝑑𝑟 = 1 still holds.

2.3 Methods to ascertain Cross-Sections

There are a handful of theoretical and experimental techniques that help in the

determination of the absorption and emission cross-sections, of which this thesis will detail three.

 Fuchtbauer Landenberg Analysis:

o This theoretical analysis is based on the Einstein Coefficients. It is agreed that the

analysis over-estimates the cross-sections by up to 50%.

14
  Gain-Loss measurements and Saturation Power measurements:

o Gain-Loss measurements help in determining the ratio between the emission and

absorption cross-sections.

o The Saturation Power measurements independently measure the absorption cross-

section and so, with the aid of the previous measurements, the emission cross

section can be determined.

 McCumber Theory:

o This theory is added only for the sake of completion as it is by far the most
accurate.

2.3.1 Fuchtbauer-Landenberg Analysis

The Fuchtbauer-Landenberg analysis uses equations derived from the Einstein relations for

the A and B coefficients for a two-level system. [10] This thesis will conduct the experiment using

this analysis.

The analysis assumes that

 The ratio between the cross-sections is equal to the ratio between the effective

linewidths.

 The population of the ions on the Stark Levels is nearly equal. The field correlation

factors for g1 and g2 manifolds must be identical.

 There is no inhomogeneous broadening.

15
 The equations for cross-sections can be written as follows [10]:

 The emission cross-section:

𝜆2
𝜎21 = 𝐴 𝑔(𝜈 )
8𝜋𝜇 2 21 2-20

 The absorption cross-section:

𝑔2 𝜆2
𝜎12 = 𝐴 𝑔′(𝜈 )
𝑔1 8𝜋𝜇 2 21 2-21

Here, λ is the wavelength of peak emission or absorption, g(ν) and g’(ν) are the respective

line shape functions, μ is the refractive index, g1 and g2 are the level degeneracies while A21 is the
1
Einstein coefficient, the spontaneous decay rate. If there is no non-radiative decay, 𝐴21 =
𝜏𝑓𝑙

where τfl is the fluorescence lifetime of the fiber. The lineshape function is given by:

𝐼𝑝𝑘
𝑔(𝜈) =
∫ 𝐼𝑑𝜈

The effective linewidth is defined because the lineshape function is complex:

1 ∞
Δλ𝑒𝑓𝑓 = ∫ 𝐼 (𝜆)𝑑𝜆
𝐼𝑝𝑘 0 2-22

16
 The lineshape now can be given by:

𝜆2 1
𝑔(𝜈 ) =
𝑐 Δ𝜆𝑒𝑓𝑓 2-23

Now, the two cross-section equations can be re-written as

𝜆4 1
𝜎21 =
8𝜋𝜇 2𝑐 𝜏𝑓𝑙 Δλ𝐸 2-24

𝑔2 𝜆4 1
𝜎12 =
𝑔1 8𝜋𝜇 2𝑐 𝜏𝑓𝑙 Δλ𝐴 2-25

At this point, the ratio between the two cross-section equations is computed to give an

expression that can be used in the analysis. The level degeneracies are given by 2J+1 where J is

the angular quantum number. For the higher level, i.e., 4I13/2, the degeneracy is 7, while it is 8 for

the lower level. Since the medium is silica, the bandwidths already have the effect of degeneracy

built into them. Therefore, the ratio is given by:

𝜎12 Δλ𝐸
=
𝜎21 Δλ𝐴 2-26

Barnes et al. measure fluorescence lifetimes with three sets of fibers in [10]. The fibers

have a higher concentration of Erbium ions, thus avoiding issues with quenching. They use short

lengths of the fiber and pumped it with 800nm lasers and obtained fluorescence data and the

17
spectral response. Absorption data for both the signal and pump bandwidths are obtained by using

the Cutback technique. The linewidths may be calculated from Eqn. 2-22 which is used to calculate

the cross-sections. These results that Barnes et al. obtained in [10] are shown below in Table 2.1

Table 2-1 Results of the Fuchtbauer-Landenberg Analysis for different fiber types in [10]

FIBER TYPE ABSORPTION CROSS-SECTION (x1025 m2) EMISSION CROSS-SECTION (x1025 m2)
+0.8 +0.6
GeO2-SiO2 7.9 6.2
-0.5 -0.3
+0.3 +0.3
Al2O3-SiO2 5.6 4.7
-0.2 -0.2
+0.3 +0.3
GeO2-Al2O3-SiO2 5.8 4.9
-0.2 -0.2

The Fuchtbauer-Landenberg analysis has some issues with accuracy on account of the

aforementioned assumptions.

 It can only be applied when the population of the different Stark levels is close in

value. But in reality, this condition does not hold because of the effect of erbium

doping in silica glass. The equations hold as long as the ΔE ≪ kBT or even when

ΔE ≅ kBT, where ΔE is the total Stark Splitting Energy. But in silica glass doped

with Er3+ ions, ΔE > 2kBT and so the condition is not satisfied. This causes the Stark

levels to have an unequal population which causes the probability of transition to

be different and therefore the analysis does not result in accurate results.

 The local field correlation factors for J and J’ must be identical but it is not. [8]

 Inhomogeneous broadening exists and weights the probability of transition.

18
2.3.2 Gain-Loss and Saturation Power Measurements

In the Gain-Loss measurement analysis, the assumption taken is as follows:

 Concentration does not vary radially as well and does not vary along the entire

length of the fiber

The terms used in this analysis is defined below:

 W13 is the pump rate

 τ is the radiative lifetime

 nT is the concentration of Er3+

 The pump power to achieve bleached condition (no gain/no loss) is Pth

𝜎
𝑊13 𝜎𝐸 − 𝐴
𝐺 = 𝑒𝑥𝑝 [𝑛 𝑇 𝑙 𝜏]
1 2-27
𝑊13 +
𝜏

𝜎
The pump rate for the bleached condition is given by 𝜏𝜎𝐴 and so (2-25) can be rewritten as
𝐸

𝑃 − 𝑃𝑡ℎ
𝐺 = 𝑒𝑥𝑝 [𝑛𝑇 𝑙𝜎𝐸 𝜎 ]
𝑃 + 𝑃𝑡ℎ 𝐸
𝜎𝐴

19
 Setting R=P/Pth and (nT*l*σE) as Gmax, the equation can be written as shown below, if

expressed in terms of decibels.

𝑅−1
𝐺 = [𝐺𝑚𝑎𝑥 𝜎 ]
𝑅+ 𝐸 2-28
𝜎𝐴

In Equation 2-28, the value of σe/σa is set to 1, and then G is predicted after the value of

Gmax is obtained from the experimental data. Lmax is the loss when the fiber is not pumped, and the

ratio between Gmax and Lmax provides the value for the ratio of the cross-sections for the next

iteration. As the value of R increases, the accuracy increases and this can be seen from the graph

that details the gain-loss data.

Gain/Loss data for GeO2-SiO2

10.0
Gmax

5.0
Gain/Loss (dB)

0.0

-5.0

-10.0
0 5 10 15 20 25
P/Pth (no unit)

Figure 2-3 Recreation of Gain/Loss data plotted against P/Pth [10]

20
 The ratio converges to give a better value for the ratio of the cross-sections because it does

not assume as many criteria as the Fuchtbauer-Landenberg analysis. This can be clearly seen from

the comparison of the ratios of the cross-sections [10].

Table 2-2 Cross-Section Ratio Comparison [σE/σA]

FIBER TYPE Fuchtbauer Landenberg Method Spectroscopic Measurement

GeO2-SiO2 0.7848 1.28±0.13

Al2O3-SiO2 0.8392 1.20±0.11

GeO2-Al2O3-SiO2 0.8448 1.17±0.10

In Saturation Power measurements, the population of an infinitesimal section of the fiber

can be written as:

𝑊13 𝑛 𝑇
𝑛2 =
1
𝑊13 + 𝜏

nT
W13 = 1/τ is the rate that is needed to get an inversion of . The stimulated rate between
2

two levels ‘i’ and ‘j’ is given by:

𝜎𝑖𝑗 𝑃𝑖𝑗
𝑊𝑖𝑗 =
ℎ𝜐𝑖𝑗 𝐴

where 𝜎𝑖𝑗 is the cross section of the transition, 𝑃𝑖𝑗 is the optical power, ℎ𝜐𝑖𝑗 is the photon energy,

and A is the area of the segment.

21
 The saturation power, the power required to reach half inversion, is given by:

ℎ𝜈13 𝐴
𝑃𝑠 =
𝜎13 𝜏

In [10], the fluorescent power is measured as a function of the input pump power, in a

similar vein to the gain/loss measurement exercise, only the graph is plotted with the x-axis being

the input pump power. The saturation power can be used to find the absorption cross-section at the

pump wavelength, which can then be used to determine the absorption cross-section at the signal

wavelength by using the information about the spectral attenuation at different bands. The

absorption cross-section at the pump wavelength is shown below:

Table 2-3 Pump Absorption Cross-Sections as obtained by Barnes et al. [10]

FIBER TYPE CROSS-SECTION (x1025 m2)

+0.03
GeO2-SiO2 2.52
-0.03
+0.3
Al2O3-SiO2 1.9
-0.3
+0.3
GeO2-Al2O3-SiO2 1.7
-0.3

The absorption cross-section at 980 nm can then be used to extrapolate the absorption

cross-section at the 1500 nm bandwidth, which then can be used to calculate the emission cross-

section. This is tabulated in Table 2-4:

22
 Table 2-4 Emission and Absorption Cross-Sections, as derived using the Saturation Method [10]

FIBER TYPE ABSORPTION CROSS-SECTION (x1025 m2) EMISSION CROSS-SECTION (x1025 m2)
+0.3 +0.2
GeO2-SiO2 6.7 7.9
-0.3 -0.2
+0.6 +0.6
Al2O3-SiO2 4.4 5.1
-0.6 -0.6
+1.0 +0.8
GeO2-Al2O3-SiO2 4.4 4.7
-1.0 -0.8

2.3.3 McCumber Relation

The McCumber Relation is a relation between the emission and absorption cross-section

that is used in McCumber’s theory of phonon-terminated optical masers [8]. This relation is given

as:

ℎ(𝜈 − 𝜀 )
𝜎𝑎 = 𝜎𝑒 (𝜐)exp { }
𝑘𝐵 𝑇

where hϵ represents the thermodynamic free energy required to move an Erbium ion from the

lower energy level to a higher one while the lattice temperature is constant. The expression for the

free energy involves knowing the energy differences between the Stark sublevels with respect to

the lowest energy level in the corresponding manifold:

𝛿𝐹(𝑁1 , 𝑁2 , 𝑇) 𝛿𝐹(𝑁1 , 𝑁2 , 𝑇)
ℎ𝜀 = [ ] −[ ]
𝛿𝑁2 𝑇
𝛿𝑁1 𝑇

23
 Since the Stark energies are not easy to calculate, two approaches can be taken with the

McCumber Relation:

1. An average separation is assumed between the Stark sublevels

2. The phenomenological values of λpeak, ηpeak is used to calculate the value of free
energy

Since the first assumption still requires computation of energies, the second assumption is

studied. In this situation, the peak wavelength and the ratio between the cross-sections are used to

calculate the free energy as given by the following equation:

ℎ𝑐 ℎ𝑐 𝑘𝐵 𝑇
ℎ𝜀 = = {1 + 𝜆𝑝𝑒𝑎𝑘 log(𝜂𝑝𝑒𝑎𝑘 )}
𝜆 𝜆𝑝𝑒𝑎𝑘 ℎ𝑐

Therefore, the MC relation can be written as follows:

𝜎𝑒 (𝜐) ℎ(𝜈 − 𝜈𝑝𝑒𝑎𝑘 )

𝜎𝑎 = exp { }
𝜂𝑝𝑒𝑎𝑘 𝑘𝐵 𝑇

where νpeak is the frequency at λpeak.

It is shown in [8] that the McCumber relation produces results that are accurate to a very

high degree.

24
 COMPUTATIONAL MODELING

3.1 Modeling

There are a multitude of ways to solve coupled differential equations. The coupled

differential equations this thesis attempts to solve are too complex to solve with the inbuilt

functions that MATLAB possesses, and so, the model is hard-coded. As mentioned earlier this will

later be incorporated into the program that models the laser cavity.

Two basic approaches exist for solving differential equations. They can be solved either

analytically or numerically. When the equations in Pedersen et al. use in [6] are attempted to be

solved analytically, it leads to a complicated set of equations, even under zero-input conditions.

Therefore a numerical approach is chosen with three distinguishable attempts in hard-coding the

solutions to the equations.

The first one is an iterative method that attempted to ‘step’ through the fiber to compute

the power at the next ‘step’ by using the parameters at the current ‘step.’ This attempt is a

rudimentary form of the finite difference method, and therefore it did not work. So, the two

established numerical methods are chosen instead.

 4th Order Runge-Kutta Method

 Finite Difference Method

25
 The finite difference method is a derivation from Taylor’s theorem while the 4th Order

Runge-Kutta method is one of the Runge-Kutta family of numerical methods. Both methods have

their advantages and disadvantages which will be discussed in the following subsections.

3.1.1 Finite Difference Method

The finite difference method, a derivation from Taylor’s theorem, is a technique to solve

differential or partial differential equations. The assumption made is that the function in question

can be expanded as a Taylor’s series. If the function can indeed be expanded as a series, the formula

for the finite difference method can be derived as follows. The function is first written in the form

of a Taylor’s series.

𝑓 ′(𝑥0 ) 𝑓 ′′(𝑥0 ) 2 𝑓 𝑛 (𝑥0 ) 𝑛

𝑓 (𝑥0 + ℎ) = 𝑓 (𝑥0 ) + ℎ+ ℎ + ⋯+ ℎ
1! 2! 𝑛!

If the step size is small enough, the value of the higher order differential terms can be

assumed to be very small and thus, everything except the first two terms can be discarded.

𝑓 (𝑥0 + ℎ) = 𝑓 (𝑥0 ) + 𝑓 ′ (𝑥0 )ℎ 3-1

Solving for f’(x0) the expression or the finite difference method is obtained.

𝑓 (𝑥0 + ℎ) 𝑓 (𝑥0 )
𝑓 ′ (𝑥0 ) = − 3-2
ℎ ℎ

26
 The finite difference method inherently involves a truncation error because the higher order

terms are neglected. The other error that can arise in the result is directly proportional to the step

size ‘h.’ The accuracy of this method is highly dependent on the value of ‘h’ and the number of

differential terms that can be used in the equation and therefore, a smaller step size would increase

the accuracy. At this junction, it is worth mentioning that the step size chosen in the final version

of the program is 1 millimeter.

3.1.2 4th Order Runge-Kutta Method:

The 4th order Runge-Kutta method is the most widely known of the Runge-Kutta methods.

It’s an iterative numerical method that attempts to predict the next value based on a weighted

average of four ‘jumps’ that it calculates. The equations for the 4th order Runge Kutta method are:

𝑥𝑛+1 = 𝑥𝑛 + ℎ 3-3

1
𝑦𝑛+1 = 𝑦𝑛 + (𝑘1 + 2𝑘2 + 2𝑘3 + 𝑘4 ) 3-4
6

where

𝑘1 = ℎ𝑓 (𝑥𝑛 , 𝑦𝑛 )

ℎ 𝑘1
𝑘2 = ℎ𝑓 (𝑥𝑛 + , 𝑦𝑛 + )
2 2

ℎ 𝑘2
𝑘3 = ℎ𝑓 (𝑥𝑛 + , 𝑦𝑛 + )
2 2

𝑘1 = ℎ𝑓 (𝑥𝑛 + ℎ, 𝑦𝑛 + 𝑘3)
27
 The advantage that the Runge-Kutta method has over the Finite Difference method is the

quicker convergence that it offers at the cost of its complexity.

3.2 MATLAB Coding

3.2.1 Finite Difference Method

The Finite Difference method is chosen because of its simplicity. The algorithm of the

method lends to a very linear execution meaning that the values of the variables can be extracted

at any point during the execution which makes debugging incredibly easy.

The algorithm is hard coded because all the in-built ODE functions (such as ode23 or

ode45) that MATLAB uses are based on the Runge-Kutta equations. This code is developed in

concurrence with the code that used the Runge-Kutta method.

Both 32-bit and 64-bit versions of MATLAB are used and initial tests using the 32-bit

version of MATLAB failed. This is caused by the length of the fiber chosen in the analysis that

Du et al. conduct in [11]. Since the step size is 1mm and the length of the fiber is 60m, the size of

each of the variables caused a memory overflow.

28
Figure 3-1 Flowchart describing execution process for Finite Difference Method based MATLAB code

29
 The memory overflow is solved by using a 64-bit version of MATLAB. Initial testing is

done using a flat mode for the fiber which is updated to a Bessel mode as mentioned in Section

2.2.1 with the equations used from [7]. While both the developed programs are added in the

appendix, it is prudent to discuss some of the sections of the code here.

%% Mode Mismatch Calc

wavelength_1 = 974E-9;
wavelength_2 = 1550E-9;

%SMF-28E
smf_r = (8.20E-6)/2;
smf_NA = 0.14;
smf_V_974 = (2*pi*smf_r*smf_NA)/wavelength_1;
smf_V_1550 = (2*pi*smf_r*smf_NA)/wavelength_2;

%MFD
smf_w_974 =
smf_r*(0.65+(1.619/smf_V_974^1.5)+(2.879/smf_V_974^6));

smf_w_1550 =
smf_r*(0.65+(1.619/smf_V_1550^1.5)+(2.879/smf_V_1550^6));

smf_MFD_974 = 2*smf_w_974;
smf_MFD_1550 = 2*smf_w_1550;

%Er110-4/125
Er_r = (3.50E-06)/2;
Er_NA = 0.2;
Er_V_974 = (2*pi*Er_r*Er_NA)/wavelength_1;
Er_V_1550 = (2*pi*Er_r*Er_NA)/wavelength_2;

%MFD
Er_w_974 = Er_r*(0.65+(1.619*Er_V_974^-1.5)+(2.879*Er_V_974^-6));
Er_w_1550 = Er_r*(0.65+(1.619/Er_V_1550^1.5)+(2.879/Er_V_1550^6));
Er_MFD_974 = 2*Er_w_974;
Er_MFD_1550 = 2*Er_w_1550;

Figure 3-2 Code snippet that shows the computation of MFD for both pump and signal lasers in the un-doped and
the doped fibers

The code snippet in Figure 3-1 is a section from the “Mode Mismatch Calculation” block

in the flowchart. In this snippet, the mode field diameters (MFD) of both the pump and the signal

lasers are calculated, both in the SMF-28 and the Er110 fibers. This helps in computing the Mode
30
Mismatch Loss which will be discussed in detail in Chapter 4. The loss arises because the light

from the first fiber is not coupled into the second fiber owing to the mismatched MFDs.

%% Fiber Loss Terms

L_mismatch_974 = -10*log10((4/((smf_MFD_974/Er_MFD_974)+...
(Er_MFD_974/smf_MFD_974))^2));
L_mismatch_1550 = -10*log10((4/((smf_MFD_1550/Er_MFD_1550)+...
(Er_MFD_1550/smf_MFD_1550))^2));
T_WDM_974 = 1; %ratio
T_WDM_1550 = 0.7; %ratio
splice_loss = 0.85; %ratio

% Pp1 and Ps1 are input pump and signal optical powers measured in Watts
% that is sent into the WDM

Pp(1) =(10^(-0.1*L_mismatch_974))*(splice_loss*(T_WDM_974*Pp1));
Ps(1) =(10^(-0.1*L_mismatch_1550))*(splice_loss*(T_WDM_1550*Ps1));

Figure 3-3 Code snippet showing the computation of optical power that enters the WDM

The Main function of the Finite Difference method is straightforward. It uses the equations

described in Chapter 2 to calculate the values of all the physical constants like the rates and the

population to calculate the incremental change in optical power for the specified step size. Then

the power at the next step is calculated using the incremental change for both the pump and the

signal powers. This process is repeated until the end of the fiber is reached.

Amplified Spontaneous Emission (ASE) is computed as a function of the chosen

wavelengths at each and every step as seen in Equations 2-12 and 2-13. The sum of the solution to

2-12 and 2-13 is referred to as SASE. This term is used in the computation of the Signal emission

and absorption rates. A flowchart describing the process flow is shown in Figure 3-4.

31
 Figure 3-4 Flowchart describing process flow for the Main function in Finite Difference model

Since the code for the main function is long, it will be split into parts. The first part detailing

the computation of the rates and the population terms is given below. As seen in the flowchart, the

noise is integral to calculating the Signal emission and absorption rates.

32
%RATE EQUATIONS:

Rpa(i,j) = ((sigpa*Pp(i))/(Ep))*Ip01(1,j);
Rpe(i,j) = ((sigpe*Pp(i))/(Ep))*Ip01(1,j);

Wsa(i,j) = (((sigsa*Ps(i))/(Es))+...
sum(siga(1,:).*Sase(i,:).*dv(1,:)))*Is01(1,j);
Wse(i,j) = (((sigse*Ps(i))/(Es))+...
sum(sige(1,:).*Sase(i,:).*dv(1,:)))*Is01(1,j);
%POPULATION EQUATIONS:

N2(i,j) = rhoer*((Rpa(i,j)+Wsa(i,j))/...
(Rpa(i,j)+Rpe(i,j)+Wsa(i,j)+Wse(i,j)+(1/tau)));
N1(i,j) = rhoer*((Rpe(i,j)+Wse(i,j)+(1/tau))/...
(Rpa(i,j)+Rpe(i,j)+Wsa(i,j)+Wse(i,j)+(1/tau)));

Figure 3-5 Code snippet showcasing the computation of the rates in the Main function of the Finite Difference
program

The forward and backward ASE term is set to zero at the start of the fiber. The total SASE

term slowly is amplified along the length of the fiber because of forward ASE. All the rates are

then used to compute the population of the ground and excited state. The code that details the

computation of the SASE term is given below.

%ASE EQUATIONS:

for v=1:1:length_lambda
Ge(v) = sige(v)*2*pi*sum...
(N2(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);
Ga(v) = siga(v)*2*pi*sum...
(N1(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);

dSase_n(i,v) = -2*h*freq(v)*...
Ge(v)-(Ge(v)-Ga(v))*Sase_n(i,v);
dSase_p(i,v) = +2*h*freq(v)*...
Ge(v)+(Ge(v)-Ga(v))*Sase_p(i,v);

Sase_n(i+1,v) = Sase_n(i,v)+dSase_n(i,v)*dzz;
Sase_p(i+1,v) = Sase_p(i,v)+dSase_p(i,v)*dzz;

Sase(i+1,v) = (Sase_p(i+1,v)+Sase_n(i+1,v))*S_Flag;
end

Figure 3-6 Code snippet of the Amplified Spontaneous Emission computation in the Finite Difference model

33
 As can be seen, the ASE is computed for a range of signal wavelengths that is chosen

before in the initialization and this is computed for a particular point in the fiber. This is then used

in the next iteration, i.e., used for the next step in the fiber. The next step is the computation of the

γe and γa terms from Eqns. 2-14 and 2-15 for both the pump and the signal lasers as shown below.

% computation of gamma for emission and absorption

Ges(i) = sigse*2*pi*sum(N2(i,r_len:(r_len...
+r_temp)).*Is01(1,r_len:(r_len+...
r_temp)).*r(1:(r_temp+1))*r_step);
Gas(i) = sigsa*2*pi*sum(N1(i,r_len:(r_len...
+r_temp)).*Is01(1,r_len:(r_len+...
r_temp)).*r(1:(r_temp+1))*r_step);

Gep(i) = sigpe*2*pi*sum(N2(i,r_len:(r_len...
+r_temp)).*Ip01(1,r_len:(r_len+...
r_temp)).*r(1:(r_temp+1))*r_step);
Gap(i) = sigpa*2*pi*sum(N1(i,r_len:(r_len...
+r_temp)).*Ip01(1,r_len:(r_len+...
r_temp)).*r(1:(r_temp+1))*r_step);

Figure 3-7 Code snippet of the computation of γr and γe for the Pump and Signal in the Finite Difference model.

The final snippet is shown below and it describes the computation of the power at the ‘next

step’ as based on the Finite Difference method of solving differential equations for both the Pump

and the Signal laser. This entire process is repeated for the entire length of the fiber.

%computation of differential power

%i.e. change of power in the step

dPp(i) = ((Gep(i)-Gap(i))*Pp(i));
dPs(i) = ((Ges(i)-Gas(i))*Ps(i));

%computation of power in the next step

Pp(i+1) = (Pp(i)+dPp(i)*dzz);
Ps(i+1) = (Ps(i)+dPs(i)*dzz);

Figure 3-8 Code snippet of the computation of the Signal and Pump Optical Power for the next ‘step’ in the Finite
Difference Model

34
 The disadvantage of a signal pass is that the backward ASE cannot be calculated. Therefore

to calculate that a three pass execution is done, the first in forward direction to calculate the forward

ASE, the second in the backward direction to calculate the backward ASE, and a final forward

direction with the computed value of backward ASE. This approach is used in [6] by Pedersen et

al. From simulations later done for the cases chosen for this thesis, the impact of SASE is minimal.

Therefore, single pass execution is preferred in most situations.

3.2.2 4th Order Runge-Kutta Method

Implementing the 4th order Runge-Kutta method involves coding functions for some of the

computations and this leads to a much cleaner program as opposed to the finite difference method.

Two functions are written, one each for the pump and the signal emission absorption factors as

mentioned in Section 2.1.2.

The RK4 algorithm executes the equations in a specific order and in the code, these

equations use the function handles to call the functions which would then compute the values for

equations 2-14 and 2-15 which define the γ term in the equations described by Pedersen et al. [6].

The flowchart for the RK4 program is shown in Figure 3-3.

35
Figure 3-9 Flowchart describing execution process for Runge-Kutta based MATLAB code

36
 The code for the Runge-Kutta based MATLAB code and the code for the Finite Difference

method based MATLAB code share many similar sections. This is because the data initialization

and the code for the approximation of the losses will not change depending on the method used.

Therefore, the code snippets prudent to the Runge-Kutta code is only mentioned in this

section while the others have been explained in the section detailing the Finite Difference model.

As can be seen in Figure 3-9, the execution process for the Runge-Kutta method is reliant on the

two functions that are defined for the Pump and Signal γ terms.

Figure 3-10 Flowchart showing the execution of Gamma Function

37
 As seen in the flowchart above, the function process is simple; an IF statement to check

whether the requirement is to compute the γ for the absorption or the emission term and to execute

the required section. The function for the Pump Gamma is shown in the code snippet below.

if(P=='E')
Gp = sigpe*2*pi*...
sum(N2(101:121).*Ip01(1,101:121).*r(1:21)*r_step);
end

if(P=='A')
Gp = sigpa*2*pi*...
sum(N1(101:121).*Ip01(1,101:121).*r(1:21)*r_step);
end

Figure 3-11 Code snippet showcasing the decision making statements of the Gamma Function in the RK4 method

The variables used in this code snippet are either sent to the function from the Main

function or is computed during the execution of the function. The two functions are called with

the help of the function handles shown below. The code in the main function uses these handles to

compute the values required for the Runge-Kutta formula

%Function Handles

dPp = @(dzz,Pp,Ps)(Gp('E',Pp,Ps,r,Is01,Ip01)-...
Gp('A',Pp,Ps,r,Is01,Ip01))*Pp;

dPs = @(dzz,Pp,Ps)(Gs('E',Pp,Ps,r,Is01,Ip01)-...
Gs('A',Pp,Ps,r,Is01,Ip01))*Ps;

Figure 3-12 Code snippet showcasing the function handles that call the two Gamma functions

The code used in the Main function is given below. In this, it can be seen that the Runge-

Kutta factors, k1, k2, k3, and k4 are computed both for the Signal and the Pump laser. And as

mentioned earlier, these factors use the function handles to call the Gamma functions. Once the

factors are computed, the change in the Pump or the Signal term, ‘dP’ is calculated using the

38
Runge-Kutta formula as mentioned in Section 3.1.1. As can be seen in the code snippet, the code

is executed for each small section of the fiber until the end of the fiber as mentioned in the

flowchart and once this is done, the data is presented in the form of graphs of Pump and Signal

power vs. length and Signal Gain vs. length.

for i=1:1:L

k1Pp = dPp(dzz, Pp(i), Ps(i));

k1Ps = dPs(dzz, Pp(i), Ps(i));

k2Pp = dPp(dzz, Pp(i)+dzz/2*k1Pp, Ps(i)+dzz/2*k1Ps);

k2Ps = dPs(dzz, Pp(i)+dzz/2*k1Pp, Ps(i)+dzz/2*k1Ps);

k3Pp = dPp(dzz, Pp(i)+dzz/2*k2Pp, Ps(i)+dzz/2*k2Ps);

k3Ps = dPs(dzz, Pp(i)+dzz/2*k2Pp, Ps(i)+dzz/2*k2Ps);

k4Pp = dPp(dzz, Pp(i)+dzz*k3Pp, Ps(i)+dzz*k3Ps);

k4Ps = dPs(dzz, Pp(i)+dzz*k3Pp, Ps(i)+dzz*k3Ps);

dPp_dz(i) = dzz/6 * (k1Pp + 2*k2Pp + 2*k3Pp + k4Pp);

dPs_dz(i) = dzz/6 * (k1Ps + 2*k2Ps + 2*k3Ps + k4Ps);

Pp(i+1) = Pp(i) + dPp_dz(i);

Ps(i+1) = Ps(i) + dPs_dz(i);

end

Figure 3-13 Code snippet of the Main function for the 4th Order Runge-Kutta based solution

3.3 Adjustments to model with MATLAB

Multiple iterations through the coupled equations for SASE are not implemented in the some

of the testing since the impact of SASE for the situation is extremely minimal. For simulations with

Du et al. [11] and Mohammed [12], the numerical aperture is not known. Therefore, a value of

N.A. is chosen so that the overlap matched the mentioned value as close as possible.

39
3.4 Simulations

For the purpose of validating the accuracy of the code, two papers are chosen, one by

Mohammed [12] and the other by Du and Chen [11]. Both papers model EDFAs pumped at 980nm

but [12] uses a formula to calculate gain and then compare it with experimental results while [11]

simulates the model using the two equations previously derived. The first test of the code is done

by comparing the simulations with the results from Du and Chen [11] and the second test is done

by comparing the simulations to the results from Mohammed [12].

3.4.1 Test #1

Du and Chen in [11] use the following values for simulating their calculations.

 λp = 980nm

 λs = 1550nm

 Γp = Γs = 0.6

 No = 2.0*10^24 m-3

 a = 2.0μm

 Small Signal Launching Power = 1μW

 Large Signal Launching Power = 1mW

 σpa = 3.8*10^-25 m2

 σsa = 3.1*10^-25 m2

 σsa = 2.7*10^-25 m2

 τ = 11.4ms

40
 In [11], Du and Chen mention that they do not consider the effects of ASE, in addition to

disregarding the effects of ESA. The fiber parameters, σpa, σsa, σse, and τ are used from [13]. They

go on to perform simulations under different conditions and provide graphs that detail their results

about gain, pump threshold, and optimum length, while also explaining said results. The Fiber

Length (m) vs. Gain (dB) graphs taken from [11] are shown below. They detail multiple simulation

results for small and large signal pumping with changes in the value of the Er3+ concentration. The

labels 1, 2, 3, and 4 indicate various Er3+ doping concentrations,

Figure 3-14 The Variation of Small Signal Gain with Figure 3-15 The variation of Large Signal Gain with
respect to length from [11] respect to length from [11]

41
Figure 3-16 Gain vs. Length for Small Signal Input

Figure 3-17 Gain vs. Length for Large Signal Input

42
3.4.2 Test #2

In this paper [12], Mohammed uses three formulae for computing gain to choose the best

option for simulations. He splits the formula into three sections to compute which one gives the

lowest gain value for a chosen large signal input. Out of the three sections, he concludes that the

third part of the equation is the least and hence the most effective. This, however, does not help as

the equation does not directly involve the final signal powers.

𝑠 𝑝 𝜂 −𝜂
𝜆𝑝 𝑃𝑝 (𝜌𝜎𝐿) ( 1+𝜂 𝛼𝑠 𝐿)
𝐺 ≤ 𝑚𝑖𝑛 (1 + ,𝑒 ,𝑒 𝑝 )
𝜆𝑠 𝑃𝑠 3-5

This is the data that Mohammad uses in [12].

 λp = 980nm

 λs = 1550nm

 No = 4.86*10^24 m-3

 Small Signal Launching Power = 1μW to 12μW

 σpa = 5.8*10^-25 m2

 σsa = 2.9*10^-25 m2

 σse = 3.47*10^-25 m2

 τ = 10ms

The issue faced is that the paper does not provide the data for some of the terms that is

required for the simulation. The missing terms are the value of the overlap integral and the fiber

radius. These two terms have a huge impact on the equations and so, the values for the two terms
43
are replaced from the terms used in Du and Chen [11]. The graphs for the gains are calculated for

each power input and as we can see, the simulated graph matches the graph in the paper very

closely. The graph for an input power of 8.148mW is shown.

Validating Simulated Results with Results from

Mohammad (8.148mW)
40

35
Gain (dB)

30

25

20

15
0 2 4 6 8 10 12
Input Signal Power (µW)

Simulated Results Simulated Results from Mohammad

Experimental Results from Mohammad

Figure 3-18 Comparison between results obtained from the simulations and the results from Mohammad [12] for an
input pump power of 8.148mW

Shown below is another set of simulations that are conducted with an input pump optical

power of 5.397mW with the same range of input signal optical powers. It can be seen the

simulations slightly over-estimate the value of the gain, but the simulations are a better fit to the

experimental curve Mohammad’s simulations. This lack of accuracy can be attributed to the fact

that some relevant data is missing in this paper.

44
 Validating Simulated Results with Results from
Mohammad (5.397mW)
40

35
Gain (dB)

30

25

20

15
0 2 4 6 8 10 12
Input Signal Power (µW)

Simulated Results Simulated Results from Mohammad

Experimental Results from Mohammad

Figure 3-19 Comparison between results obtained from the simulations and the results from Mohammad [12] for an
input pump power of 5.397mW

3.5 Data Required for Simulation

The code has now been validated with two separate theoretical papers. Therefore, the

model can be tested experimentally. In order to model said experiments, the following data is

required.

 Erbium Doping Concentration

 Cross-Sections of the fiber for the Pump and the Signal Inputs

 Fluorescence Lifetime

 Numerical Aperture

 Radius of the Fiber

45
 Optical fibers are sold by their concentrations, i.e., the fiber that is used in this thesis is an

Er110-4 fiber from nLight. Therefore, there is no need to perform experiments to ascertain the

doping concentration. Almost all the other data can also be obtained from the company’s

datasheets, but one must also be able to ascertain this data with the help of experiments.

Experiments can help determine the cross-section areas and the fluorescence lifetime of

the fibers.

The following chapter will deal with the experiments that are conducted on the fiber to

determine the previously mentioned values.

46
 CHARACTERIZATION OF THE SYSTEM

4.1 Overview

In this chapter, experiments are conducted with the Erbium fiber to measure the various

optical characteristics. This ranges from the absorption and the emission spectrums to the various

losses in the system, thereby characterizing the system as accurately as possible.

4.2 Fiber Information

The Erbium-doped fiber available is an nLight fiber (Er110-4/125). The specification sheet

for the fiber is available in Appendix B. The specification sheet along with the technical data

present on the website list all the information that is prudent to the simulation except the absorption

and emission cross-section data which is obtained from the vendor. This data is extensively used

in my simulations.

 Radius = 3.75±0.5 μm

 Numerical Aperture = 0.2

 τ = 11.4 ms


ρ = 8.4 * 10^25 m-3

47
4.3 Characterization of the components

The system in its simplest form involves two lasers being multiplexed into a single beam

that is then fed into the EDF. The output from the EDF is measured and studied. Therefore, these

components and the other additional components of the system must be characterized so the

simulations can be as accurate as possible.

All experiments conducted used the following equipment.

 ILX Lightwave Laser Diode Controller 3900 (S/N: 39002458)

 ILX Lightwave Optical Power Meter FPM 8210 (S/N: 82101308)

 Wavelength Division Multiplexer (S/N: 0A2081)

 JDS Uniphase HA9 Attenuator (S/N: KE09319)

 3 Window Coupler (S/N: L050049083)

 JDS Uniphase BBS Broadband Source (S/N: FD110946)

 Inphenix LED Light Source (S/N: 868512485)

 Agilent Optical Spectrum Analyzer 86143B (S/N: DE44103038)

The following Laser Diodes are used in all experiments.

 974 nm - Bookham LC95A74-20 (S/N: B243446.001)

 1550 nm - JDS Uniphase CQF933/408-19340 (S/N: 461784)

 1534 nm - LMI A1905 3CN00410CDAA (S/N: 990419413)

48
4.3.1 Characterization of the WDM

The Wavelength Division Multiplexer (WDM) is a passive device that joins two or more

wavelengths into a single fiber output. The WDM that is to be used in the experiments has four

ports, two clear ports, one black port, and one red port. The final schematic for the WDM after

testing is shown in Figure 4-1.

Figure 4-1 Final Schematic for WDM

This WDM is chosen because it possesses a transmission window wide enough to meet the

requirement of the experiments this thesis conducts. Unfortunately, there is no specification sheet

to be found for this WDM, nor a make or model number since it is appropriated from an existing

system. The WDM must, therefore, be characterized. WDMs are bidirectional, and this is used to

test it by sending known optical power values of the pump and the signal laser to the each of the

output ports to measure the output on the other side. This lets us know what ports could be used

as inputs to the system that this thesis wants to use.

Initially, the WDM is tested by sending in the signal laser via the P2 and P3 ports and then

the output at ports P0 and P1 are measured. The input optical power is varied as a function of the

diode drive current. Table 4-1 shows two measurements using the signal laser, one with P2 as the

input port while the second one uses P3 as the input port.

49
 Table 4-1 Port Efficiencies when P2 and P3 are input ports for Signal Wavelength

1550 nm (100 mA/13 mW Signal Input Optical Power)

P2 (red) (Input) P3 (clear) P0 (clear) P1 (black)
W dBm W dBm % W dBm % W dBm %
1.33E- 0.853 31.1 5.83
11.23 -30.69 0.01 -15.1 0.23 7.66 43.8
02 E-06 E-06 E-03
P3 (clear)
P2 (red) P0 (clear) P1 (black)
(Input)
W dBm % W dBm W dBm % W dBm %
1.13E- 13.3 5.01 16.1
-29.5 0.01% 11.23 6.99 37.7% -17.9 0.12
06 E-03 E-03 E-06

From Table 4-1, it can be seen that the loss is lower when P2 is used as the input port for

the signal wavelength. The same measurements are now repeated with the pump laser. P2 is used

as the input for this measurement since it proved to be the most efficient path. Table 4-2 shows the

measurements made for an input pump optical power of 120.82 mW which is the optical power

output from the laser when the pump laser is driven at 240 mA.

Table 4-2 Port Efficiency when P2 is the input port for Pump Wavelength

P2 (red) P0 (clear) P1 (black)

(W) (dBm) (W) (dBm) % W dBm %
1.21E-01 20.82 1.03E-01 20.13 85% 2.02E-04 -6.95 0.17%

It is seen that the output at P0 is the highest and therefore, it can be assumed that sending

in the pump laser via the P0 port and the signal laser via the P1 port would be the most efficient

setup. A final test of the WDM is conducted to verify the previous measurements. The procedure

is similar; known values of optical power, both in the pump and the signal wavelength range are

sent into the respective ports, and the output is measured at port P2 using a spectrum analyzer. The

ratio of output power to input power for both the wavelengths is calculated and is seen to be

approximately constant (Figure 4-3), thereby validating the previous measurements.

50
 Table 4-3 Final WDM test displaying the average output percentage at the output port for both inputs wavelengths

974nm
Current P0 (clear) P2 (red)
(mA) (W) (dBm) (W) (dBm) %
240 1.21E-01 20.8 7.90E-02 18.975 65.37%
AVERAGE 63.70%

1534nm
Current P1 (black) P2 (red)
(mA) (W) (dBm) (W) (dBm) %
100 1.34E-02 11.266 1.03E-02 10.109 77%
AVERAGE 77.82%

4.3.2 Mode Field Diameter Mismatch Loss

When two fibers with different core radii are spliced together, a loss is introduced into the

system that cannot be prevented since all the light from one fiber is not coupled into the next one.

This loss must be factored into the model with all the other losses that are inherent to a fiber optic

system like reflectance losses and scattering losses.

The loss that is introduced into the system can be calculated using the formula given below

in equation 4-1 [14]. It depends on the Mode Field Diameter (MFD) which in turn is different for

both the signal and pump wavelengths. Therefore it is prudent to code the computation of the MFD

in the model.

4
𝑆𝑝𝑙𝑖𝑐𝑒 𝐿𝑜𝑠𝑠 = −10 ∗ log𝑒 ( 2 )
𝑀𝐹𝐷 𝑀𝐹𝐷 2 4-1
(𝑀𝐹𝐷1 ) + (𝑀𝐹𝐷2 )
2 1

51
 The code that computes the splice loss is mentioned in Section 3.2.1. If a situation arises

needing the replacement of either the Er110 fiber or the SMF fiber, the appropriate values in the

code snippet must be altered. The mode mismatch loss is calculated for the un-doped and the doped

Erbium fiber and shown is below in Table 4-4.

Table 4-4 Splice Losses when SMF 28e is spliced with Er110-4/125

Loss in the Pump Wavelength (dB) 1.39

Loss in the Signal Wavelength (dB) 0.37

4.3.3 OSA vs. Power Meter Calibration

The Optical Spectrum Analyzer (OSA) and the Power Meter (PWM) have different

wavelength resolutions and also different methods of measuring the power. The signal laser is sent

into the erbium fiber and output at the end of the fiber is measured using the OSA and the PWM.

This lets the measuring devices measure just the signal laser because if the pump laser had been

used, the ASE would have interfered with the measurements. As can be seen, there is a loss of

15% incurred when using the OSA to measure the output of the system.

Table 4-5 OSA and Power Meter Calibration

Laser Diode
Input Power Power Meter
Controller OSA (W) Ratio
to EDF (W) (W)
Current (mA)
40 1.45E-03 4.12E-04 3.49E-04 0.84
50 2.39E-03 8.69E-04 7.45E-04 0.85
60 3.28E-03 1.39E-03 1.20E-03 0.86
70 4.20E-03 1.98E-03 1.70E-03 0.85
80 5.10E-03 2.56E-03 2.22E-03 0.86
90 6.00E-03 3.20E-03 2.76E-03 0.86
100 6.95E-03 3.83E-03 3.31E-03 0.86
AVG 0.85

52
4.4 Absorption Spectrum and Cross-Section:

Measurement of the absorption or emission spectra requires the knowledge of the power

being sent into Erbium-doped fiber. The required data is collected by sending known values of the

pump and the signal lasers into the WDM. The output of the WDM is measured and exported to

an Excel file. The schematic for this setup is shown below in Figure 4-4.

Figure 4-2 Experimental Setup to measure the Input to the Erbium-Doped Fiber

Two input sources are used; one being the LED Source and the other is the Broadband

Source. The output spectra for both these sources are measured using the setup shown in Figure 4-

2. The LED source is driven at 7 different values of diode current and the output spectrum of the

LED source before the WDM is shown in Figure 4-5. The same procedure is repeated for the

Broadband source to give Figure 4-6. The broadband source is connected as the input to port P1

and the output is measured at port P2 and saved to an Excel file with the help of the spectrum

analyzer. Both input sources have been sampled every 0.1nm.

53
 Output Spectrum of the LED Source
0

450mA
-10

400mA
Power (dBm)

-20
350mA

-30 300mA
250mA

-40 200mA

150mA
-50
1400 1425 1450 1475 1500 1525 1550 1575
Wavelength (nm)

Figure 4-3 Output of the Inphenix LED Source

Output Power of the Broadband Source (dBm) vs.

Wavelength (nm)
0

-5
Power (dBm)

-10

-15

-20

-25
1520 1530 1540 1550 1560 1570 1580
Wavelength (nm)

Figure 4-4 Output of the Broadband Source

54
 Now that the input to the Erbium Fiber is known, the absorption experiment can be

conducted. One end of the Erbium fiber is spliced to the output port of the WDM, P2, while the

other end of the EDF is spliced to a pigtail connector which is then connected to the spectrum

analyzer. The general procedure for measuring the absorption spectrum would be to find the

amount of power absorbed over the range of wavelengths. This will then be used in conjunction

with the Fuchtbauer Landenberg Analysis to compute the cross-sections. The schematic for the

experiment is shown below in Figure 4-5.

Figure 4-5 Experimental Setup for measuring cross-sections

The procedure is repeated with both the input sources. The LED source is first connected

and driven at the chosen values of currents, and the output from the doped fiber is measured and

saved as an Excel file. The same procedure is then repeated with the Broadband source. Once both

the measurements are made, the values of the output power are then used to compute the absorption

coefficient.

55
 𝑃
log 𝑒 ( 𝐼𝑁 )
𝑃𝑂𝑈𝑇
𝛼= 4-2
𝐿

PIN is the input optical power and POUT is the output optical power and L is the length of

the erbium fiber. This absorption coefficient α is then used in the equations derived from the

Fuchtbauer-Landenberg analysis to compute the cross-section. This is done by measuring the

FWHM of the absorption spectra. The FWHM is then used in equation 2-22 which then lets us

compute the value of the absorption cross-section. Both the absorption spectra for the LED source

and the BBS is shown below. The absorption spectra of the LED is given in Figure 4-8. As can be

seen, the maximum absorption seems to occur at 1530nm which is expected.

Absorption Spectrum of the EDF with the LED Source

0
-5
-10
-15 450mA
400mA
Power (dBm)

-20
350mA
-25
300mA
-30
250mA
-35
200mA
-40
-45 150mA
-50
1400 1420 1440 1460 1480 1500 1520 1540 1560
Wavelength (nm)

Figure 4-6 Absorption Spectra for the Er110-4/125 Fiber using the LED Source

56
 Absorption Coefficient vs. Wavelength for the
Broadband Source
0.006

0.005

0.004
Power (dBm)

0.003

0.002

0.001

0
1525 1530 1535 1540 1545 1550 1555 1560 1565 1570 1575
Wavelength (nm)

Figure 4-7 Absorption Spectra for the Er110-4/125 Fiber using the Broadband Source

Using the Fuchtbauer-Landenberg relations, the absorption cross-sections can be computed

and they are presented below. As can be seen, the results deviate at the point where the two sets of

data from the different sources are concatenated. If a better degree of accuracy is required, a wider

broadband source must be used in place of the dual sources used in the experiments.

57
 Absorption Cross-Section of Er110-4/125
8E-25

7E-25
Vendor Absorption
6E-25 Cross-Section
Cross-Section (m2)

5E-25

4E-25 Experimental
Absorption Cross-
3E-25 Section

2E-25

1E-25

0
1350 1400 1450 1500 1550 1600 1650
Wavelength (nm)

Figure 4-8 Comparison between experimental and vendor Absorption Cross-Section of Er110-4/125

4.5 Emission Spectrum and Cross-Section:

The experiment for the emission cross-section is slightly different on account of having to

use a shorter piece of the fiber to measure the emission spectra. In this case, an extremely short

piece of the fiber is spliced instead of the previous 35.5cm. The experimental setup is virtually

same as that of the one used for the absorption spectra measurement.

The experiment for the emission spectra involves measuring the output for the signal

wavelength ranges for low input pump optical powers. And to avoid the influence of Amplified

Spontaneous Emission, an extremely short piece of fiber is cleaved and spliced to the output port

P2 of the WDM. The other end is spliced to the pigtail which is connected to the OSA. The pump laser

is turned on and fed into the system. For the emission spectra, the pump is driven at a low current

58
and the output from the fiber in the signal wavelength range is measured. There is no input from

the signal laser.

The experiments are conducted at several different values of the pump current. The cross-

sections computed from this curve are also slightly skewed, showing an increased cross-section

value as compared to the vendor’s cross-section. This can be attributed to various losses and such

in the system, and therefore the vendor cross-section data is used in both cases.

Normalized Amplified Spontaneous Emission

1.2

0.8
ASE (no unit)

0.6

0.4

0.2

0
0 200 400 600 800 1000 1200
Wavelength (nm)

Figure 4-9 – Normalized Amplified Spontaneous Emission with a Pump Current of 35mA

59
 Emission Cross-Section of Er110-4/125
7E-25

6E-25

5E-25
Cross-Section (m2)

Experimental
Emission Cross-
4E-25 Section

3E-25
Vendor Emission
Cross-Section
2E-25

1E-25

0
1480 1500 1520 1540 1560 1580 1600 1620 1640
Wavelength (nm)

Figure 4-10 Comparison between experimental and vendor Emission Cross-Section of Er110-4/125

It can be seen from Figure 4-10 that the cross-section data measured is comparable to the

data obtained from the vendor. Therefore, the cross-section data acquired from the vendor will be

used in the model.

60
 VERIFICATION OF THE MODEL

The data acquired from the vendor is validated in the previous chapter by conducting

absorption and emission experiments. The system is also characterized and the various losses such

as splice loss and mode mismatch loss are factored into the MATLAB model. This is, however,

discounting other losses inherent to an experiment such as unclean connectors.

The experimental setup is checked to make sure that no unaccountable but preventable

losses are present which meant making sure that the pigtail connectors are clean and properly

inserted and there are no sharp bends in the fiber to prevent bending losses. Once this is verified,

the system is set up as shown in Figure 5-1. The signal laser is connected to the attenuator which

is then connected to a 90/10 Optical Coupler which helps make sure that the predefined amount of

optical power is input into the WDM.

Two sets of experiments are conducted, one with an erbium fiber of length 13.3 cm (Sample

#1) while the other experiment used an erbium fiber that is 27.7 cm long (Sample #2). These two

lengths are compatible with the fiber length in the fiber laser that is to be modeled later.

5.1 General Procedure

In order to test the model for its accuracy, the experiment has to be repeated under different

conditions. Therefore, an experimental setup is designed which lets the erbium fiber be tested for

multiple lengths. As mentioned in the earlier two EDFAs are constructed with the help of the two

samples.

61
 The pump laser output is spliced to port P0 of the WDM while the signal laser output is

connected to an attenuator and the attenuator is in turn, connected to the 90/10 Coupler. The output

from the 90% port is connected as the input to the port P1 of the WDM. The 10% port is connected

to an optical power meter. The port P2 of the WDM is spliced to the chosen Erbium fiber while

the other end of the Erbium fiber is spliced to a pigtail which can be connected to the spectrum

analyzer to measure the output from the EDF. The schematic for this is shown in Figure 5-1.

Figure 5-1 Schematic for the EDFA experiments

Three input pump optical powers are chosen, each corresponding to the current at which

the pump laser is driven. These currents are 100 mA, 250 mA, and 300 mA which correspond to

input pump optical powers of 25.654 mW, 83.081 mW, and 102.51 mW respectively.

Now that the input pump optical powers are chosen, the input signal optical powers can be

decided. Since it is preferred that both the small signal gain and the large signal gain be tested, a

range of optical power inputs is chosen: 1 μW, 10 μW, 25 μW, 50 μW, 100 μW, 250 μW, 500 μW,

and 1 mW.

62
 Regardless of the length of the fiber chosen, the general procedure is as follows. The input

pump optical power is set a particular value with the help of the laser diode controller. Once this

is done, the input signal optical power into the WDM is adjusted using the attenuator. The chosen

values of optical power are sent in and the output signal optical power is measured using the

spectrum analyzer. This is repeated for other values of the input pump optical powers and then the

erbium fiber is replaced with another fiber of differing length.

Once this is done, simulations are conducted to obtain the output signal optical power at

the same length as that of the fiber. With these values, the experimental and the simulated output

signal optical powers are calculated. These results are then compared so that conclusions can be

drawn about the accuracy of the model.

5.1.1 Amplified Spontaneous Emission

During each of the experiments, the noise value aka the Amplified Spontaneous Emission

at the end of the fiber is measured by using the spectrum analyzer. This is done to compare the

ASE computation that is done by the MATLAB program. An example ASE measurement is shown

below in Figure 5-2. The ASE data is then extracted from the data output from the code and it is

seen that the model predicts an extremely low value of ASE, on the order of 10 -13W while the

experimental data show that the maximum ASE at 1530 nm is almost 2.5µW. The ASE code is

then decided not to be included in the execution of the code as the impact of ASE is extremely

minimal and therefore, the execution time of the simulations was drastically reduced.

63
 Amplified Spontaneous Emission at 102 mW
2.5

1.5
ASE (uW)

0.5

0
1480 1500 1520 1540 1560 1580 1600 1620
Wavelength (nm)

Figure 5-2 Experimental Amplified Spontaneous Emission measured at a pump optical power of 102.51mW

Figure 5-3 Simulated Amplified Spontaneous Emission at a pump optical power of 102.51mW
64
5.2 EDFA based on Sample #1

A piece of the erbium-doped fiber is cut from the spool. This is Sample EDF #1, measured

to be 13.3 cm and used in the experimental setup to build the EDFA. The fiber is then cleaved and

spliced into the system with one end of the fiber spliced to the P2 lead of the WDM and the pigtail

connector is similarly spliced to the other end of the fiber.

Four experiments are conducted, one with no pump while the other three are with the input

pump optical powers as mentioned before. The results are presented in Table 5-1 along with a

graph that shows the output signal optical power as a function of the input signal optical power.

Table 5-1 Output Signal Powers vs. Input Signal Powers for 13.3cm EDF for different Pump Input Powers

Input Signal Powers Output Signal Power (dBm)

25.7 mW 83.1 mW 102.5 mW
Input Signal Input Signal No Pump
Pump Input Pump Input Pump Input
Power (mW) Power (dBm) Input Power
Power Power Power
1E-03 -30.00 -39.45 -28.36 -26.47 -26.17
10E-03 -20.00 -29.49 -18.67 -16.74 -16.56
25E-03 -16.02 -25.38 -14.67 -12.82 -12.58
50E-03 -13.01 -22.57 -11.73 -9.78 -9.48
100E-03 -10.00 -18.96 -8.76 -6.88 -6.57
250E-03 -6.02 -14.37 -4.90 -2.95 -2.59
500E-03 -3.01 -10.70 -2.11 -0.05 0.25
1 0.00 -6.79 0.52 2.63 2.99

65
 Input Signal Power vs. Output Signal Power (13cm)
10

0
Output Power (dBm)

-10

-20

-30

-40

-50
-35.0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 0.0
Input Signal Power (dBm)

No Pump Input 25.7 mW 83.1 mW 102.5 mW

Figure 5-4 Output Signal Powers vs. Input Signal Powers for 13.3cm EDF for different Pump Input Powers

5.2.1 Absorption Results for Sample #1

The EDFA is operated without the pump laser turned on, meaning that the experiment is

an absorption experiment. As mentioned in Section 5.1, 7 values of optical power in the signal

wavelength are sent into the fiber and the output is measured using the spectrum analyzer.

Once these values are measured, simulations are conducted with MATLAB to compute the

value of the output signal optical power at the same length as that of the experimental fiber, which

in this case is 13.3 cm by changing the value of the input signal optical power in the code.

Table 5-2 lists the data recorded when the EDF is operated without the pump laser

operating. It is seen that the simulated output optical power is close in value to that of the

66
experimental values. The accuracy of the simulations is further seen in Figure 5-5 where the results

are presented in the form of a chart.

Table 5-2 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for No Pump
Power Input (13.3 cm)

Input Signal Power Input Signal Power Experimental Output Simulated Output
(mW) (dBm) Power (dBm) Power (dBm)
0.00 -30.00 -39.45 -40.58
0.01 -20.00 -29.49 -30.53
0.03 -16.02 -25.38 -26.46
0.05 -13.01 -22.57 -23.31
0.10 -10.00 -18.96 -20.03
0.25 -6.02 -14.37 -15.30
0.50 -3.01 -10.70 -11.22
1.00 0.00 -6.79 -6.67

Input Signal Power vs. Output Signal Power

(No Pump Power Input)
0
Output Signal Power (dBm)

-10

-20

-30

-40

-50
-35 -30 -25 -20 -15 -10 -5 0
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-5 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for No
Pump Power Input (13.3 cm)

67
5.2.2 Test #1 with Sample #1

The pump laser is driven at 100 mA which translates to an input pump optical power of

25.7 mW being sent into the WDM. With the input pump optical power set at that value, the input

signal optical power is varied with the help of the attenuator to reach the same 7 values as

mentioned in Section 5.1 and the output signal optical power is measured using the spectrum

analyzer. Table 5-3 lists this data along with the simulated output signal optical power.

The procedure to obtain the simulated results are similar to the previous experiment.

Instead of using a value of zero for the input pump optical power, the value mentioned earlier,

25.7 mW is used and the value of the input signal optical power is changed as before.

Table 5-3 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for an input
Pump Power of 25.7 mW (13.3 cm)

Input Signal Power Input Signal Power Experimental Output Simulated Output
(mW) (dBm) Power (dBm) Power (dBm)
0.00 -30.00 -28.36 -23.16
0.01 -20.00 -18.67 -13.20
0.03 -16.02 -14.67 -9.28
0.05 -13.01 -11.73 -6.36
0.10 -10.00 -8.76 -3.53
0.25 -6.02 -4.90 -0.03
0.50 -3.01 -2.11 2.34
1.00 0.00 0.52 4.44

It is seen that there is a difference in the simulated output optical powers and the

experimental output powers. This discrepancy, as will be seen to exist in the latter experiments,

will be investigated in Section 5.4

68
 Input Signal Power vs. Output Signal Power
(25.7 mW Pump Power Input)
Output Signal Power (dBm) 10

-10

-20

-30

-40
-35 -30 -25 -20 -15 -10 -5 0
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-6 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for an input
Pump Power of 25.7 mW (13.3 cm)

5.2.3 Test #2 with Sample #1

Test #2 involves driving the pump laser at a higher current, now at 250 mA which translates

into an input pump optical power of 83.1 mW sent into the WDM. The same procedure is repeated

as in the previous experiment with the pump optical power set and the input signal optical power

varied with the help of the attenuator.

The simulations are similarly performed by setting the value of the input pump optical

power to the WDM as 83.1 mW in the code. The results are tabulated in Table 5-4 and presented

in the form of a chart in Figure 5-7.

69
 It is seen that the discrepancy continues to exist with the simulated output signal optical

power and the experimental output signal optical power. It can be noted that the difference between

the two is approximately 4 dB.

Table 5-4 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for an input
Pump Power of 83.1 mW (13.3 cm)

Input Signal Power Input Signal Power Final Output Power Simulated Output
(mW) (dBm) (dBm) Power (dBm)
0.00 -30.00 -26.47 -22.06
0.01 -20.00 -16.74 -12.07
0.03 -16.02 -12.82 -8.12
0.05 -13.01 -9.78 -5.15
0.10 -10.00 -6.88 -2.21
0.25 -6.02 -2.95 1.54
0.50 -3.01 -0.05 4.22
1.00 0.00 2.63 6.68

Input Signal Power vs. Output Signal Power

(83.1 mW Pump Power Input)
10
Output Signal Power (dBm)

-10

-20

-30
-35 -30 -25 -20 -15 -10 -5 0
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-7 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for an input
Pump Power of 83.1 mW (13.3 cm)

70
5.2.4 Test #3 with Sample #1

Test #3 is conducted with a different input pump optical power. The pump laser is driven

at 300 mA which translates into an input pump optical power of 102.5 mW. The experimental

procedure continues to be the same with the input pump optical power being constant while the

input signal optical power to the WDM varied with the help of the attenuator.

The simulations are similarly performed using MATLAB by modifying the value of the

input pump and signal optical powers for each of the trials. These results are again tabulated in

Table 5-5 and presented in a chart in Figure 5-8. It is seen that the discrepancy between the

experimental results and the simulated results is a little higher.

Table 5-5 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for an input
Pump Power of 102.5 mW (13.3 cm)

Input Signal Power Input Signal Power Final Output Power Simulated Output
(mW) (dBm) (dBm) Power (dBm)
0.00 -30.00 -26.17 -21.96
0.01 -20.00 -16.56 -11.97
0.03 -16.02 -12.58 -8.02
0.05 -13.01 -9.48 -5.04
0.10 -10.00 -6.57 -2.09
0.25 -6.02 -2.59 1.69
0.50 -3.01 0.25 4.42
1.00 0.00 2.99 6.95

71
 Input Signal Power vs. Output Signal Power
(102.5 mW Pump Power Input)
Output Signal Power (dBm) 10

-10

-20

-30
-35 -30 -25 -20 -15 -10 -5 0
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-8 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for an input
Pump Power of 102.5 mW (13.3 cm)

5.3 EDFA based on Sample #2

Sample #2 is an EDF that is 27.7 cm long. Chronologically this is the first experiment that

is conducted, and at that time, the procedure is not decided upon. Therefore, the absorption

experiment is conducted without the attenuator. The control variable, in this case, is the signal

laser drive current on the Diode Controller. The Gain experiments that followed used the attenuator

and therefore the data for them is much more organized. Therefore the absorption curve is not

shown in Figure 5-9, with only the experimental results for tests #1 and #2 displayed.

72
Table 5-6 Output Signal Powers vs. Input Signal Powers for 27.7 cm EDF for different Pump Input Powers

Input Signal Powers Output Signal Powers (dBm)

Input Signal Input Signal 83.1 mW Pump Input 102.5 mW Pump Input
Power (W) Power (dBm) Power Power
1.00E-05 -20.0000 -12.89 -11.46
2.50E-05 -16.02 -8.46 -7.66
5.00E-05 -13.01 -5.70 -4.96
1.00E-04 -10.00 -2.76 -1.96
2.50E-04 -6.02 0.84 1.39
5.00E-04 -3.01 3.44 4.04
1.00E-03 0.00 6.14 6.84

Input Signal Power vs. Output Signal Power (27 cm)

10

5
Output Power (dBm)

-5

-10

-15
-25.0 -20.0 -15.0 -10.0 -5.0 0.0
Input Signal Power (dBm)

83.081mW 102.51mW

Figure 5-9 Output Signal Powers vs. Input Signal Powers for 27.7 cm EDF for different Pump Input Powers

73
5.3.1 Absorption Results for Sample #2

This experiment is conducted without the attenuator in place and therefore the input signal

optical power is varied with the signal laser drive current, with each value of input signal optical

power corresponding to a value of a laser drive current. For example, an input signal optical power

of 6.95 mW is achieved by setting the signal laser drive current to be 100 mA.

The simulations are done in a similar fashion as seen in the previous section. It is seen that

the simulated absorption results are not close to the experimental output optical power values like

the case for the absorption results for Sample #1.

This plays an important factor in adjusting the model as will be elaborated in Section 5.4.

The results are tabulated and displayed Table 5-7. The results are also presented in the form of a

chart as seen in Figure 5-10.

Table 5-7 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for no input
Pump Power (27.7 cm)

Input Signal Input Signal Power Experimental Output Simulated Output

Power (mW) (dBm) Power (dBm) Power (dBm)
0.55 -12.12 -12.12 -18.49
1.45 -5.41 -5.41 -9.59
2.39 -2.07 -2.07 -4.45
3.28 0.05 0.05 -1.53
4.20 1.63 1.63 0.49
5.10 2.82 2.82 1.92
6.00 3.81 3.81 3.03
6.95 4.64 4.64 3.98

74
 Input Signal Power vs. Output Signal Power
(No Pump Power Input
Output Signal Power (dBm) 10

-10

-20
-4 -2 0 2 4 6 8 10
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-10 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for no
input Pump Power (27.7 cm)

5.3.2 Test #1 with Sample #2

Sample #2 is used in this experiment. The chosen input pump optical power is 83.1 mW

which is set using the laser drive current which is 250 mA. The procedure is similar to the previous

experiments with the input signal optical powers adjusted with the help of the attenuator.

Once the experimental output signal optical power is measured and saved, the simulations

are conducted with the only change in the procedure being the change of the length of the fiber to

27.7 cm to reflect the length of Sample #2. Both the experimental and the simulated results are

tabulated as shown in Table 5-8, and a chart displaying the two sets of data is seen in Figure 5-11.

75
 Table 5-8 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for
an input Pump Power of 83.1 mW (27.7 cm)

Input Signal Power Input Signal Power Experimental Output Simulated Output
(mW) (dBm) Power (dBm) Power (dBm)
0.01 -20.0000 -12.89 -0.82
0.25 -16.02 -8.46 2.82
0.05 -13.01 -5.70 5.34
0.10 -10.00 -2.76 7.55
0.25 -6.02 0.84 9.92
0.50 -3.01 3.44 11.28
1.00 0.00 6.14 12.33

As seen clearly from Figure 5-11, a discrepancy continues to exist, between the

experimental results and the simulated results. But in this case, it can be seen that the gulf between

the two continues to increase with the greatest difference between the experimental and the

simulated output signal optical powers being about 12 dB.

Input Signal Power vs. Output Signal Power

(83.1 mW Pump Power Input)
20
Output Signal Power (dBm)

10

-10

-20
-25 -20 -15 -10 -5 0
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-11 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for an
input Pump Power of 83.1 mW (27.7 cm)

76
5.3.3 Test #2 with Sample #2

Sample #2 is used in a final experiment where the input pump optical power is set to 102.5

mW by choosing the laser drive current as 300 mA. The input signal optical powers are varied

with the help of the attenuator. The simulations are done similarly. The experimental and the

simulated output signal optical powers are then tabulated and shown in Figure 5-12. It is seen here

that the difference between the results is still similar to the previous result.

Table 5-9 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for an input
Pump Power of 102.5 mW (27.7 cm)

Input Signal Power Input Signal Power Experimental Output Simulated Output
(mW) (dBm) (dBm) (dBm)
0.01 -20.0000 -11.46 -0.57
0.25 -16.02 -7.66 3.12
0.05 -13.01 -4.96 5.70
0.10 -10.00 -1.96 8.00
0.25 -6.02 1.39 10.49
0.50 -3.01 4.04 11.93
1.00 0.00 6.84 13.05

The chart as shown in Figure 5-12, showcases the difference between the experimental and

the simulated output signal optical powers effectively. It can be seen in Figure 5-12 and the

previous figure that in each case, the difference between the experimental and the simulated results

appear to be constant. This is an important inference drawn from the results which will be pivotal

in the hypothesis for the adjustments made in Section 5.4.

77
 Input Signal Power vs. Output Signal Power
(102.5 mW Pump Power Input)
Output Signal Power (dBm) 20

10

-10

-20
-25 -20 -15 -10 -5 0
Input Signal Power (mW)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-12 Experimental Output Power vs. Simulated Output Power as a function of Input Signal Power for an
input Pump Power of 102.5 mW (27.7 cm)

5.4 Adjustment of the MATLAB Model

It is seen that the simulated curve does not match the experimental output graph in most of

the figures shown above. Therefore, it is hypothesized that the value of the overlap integral is not

constant, but it varies with the length as a result of the pump and the signal powers being absorbed

in the fiber. This conclusion is drawn after comparing the absorption experiments for samples #1

and #2.

Therefore, the code is modified to introduce a factor that could change the value of the

signal overlap integral to a chosen value by modifying the mode. With this modified code, the

simulations are repeated for the two sets of experimental data, one for the Sample #1 and the other

one for Sample #2. The value of the mode is reduced by one-half. The results which are shown

78
below, showcase a much more accurate prediction. Figures 5-15, 5-16, 5-17, and 5-18 showcase

the previous results as seen in Section 5.2 but with the simulated results as obtained from the

modified code. As is seen from all these charts, the adjusted code delivers a much more accurate

simulation. The change in the mode, as a result, is shown in Figure 5-14 while the original mode

is shown in Figure 5-13.

Figure 5-13 Default Input Signal Optical Mode

79
 Figure 5-14 Modified Input Signal Optical Mode

Figure 5-15 shows the simulation and experimental results for the absorption experiment

for Sample #1 while Figure 5-16 compares the results for Test #1, Figure 5-17 compares the results

for Test #2, and Figure 5-18 compares the results for Test #3.

80
 Input Signal Power vs. Output Signal Power
(No Pump Power Input) Sample #1
Output Signal Power (dBm) 0

-10

-20

-30

-40

-50
-35 -30 -25 -20 -15 -10 -5 0
Input Signal Power (mW)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-15 Experimental Output Power vs. Adjusted Simulated Output Power as a function of Input Signal Power
for no input Pump Power (13.3 cm)

Input Signal Power vs. Output Signal Power

(25.7 mW Pump Power Input) Sample #1
10
Output Signal Power (dBm)

-10

-20

-30
-35 -30 -25 -20 -15 -10 -5 0
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-16 Experimental Output Power vs. Adjusted Simulated Output Power as a function of Input Signal Power
for an input Pump Power of 25.7 mW (13.3 cm)

81
 Input Signal Power vs. Output Signal Power
(83.1 mW Pump Power Input) Sample #1
Output Signal Power (dBm) 10

-10

-20

-30
-35 -30 -25 -20 -15 -10 -5 0
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-17 Experimental Output Power vs. Adjusted Simulated Output Power as a function of Input Signal Power
for an input Pump Power of 83.1 mW (13.3 cm)

Input Signal Power vs. Output Signal Power

(102.5 mW Pump Power Input) Sample#1
10
Output Signal Power (dBm)

-10

-20

-30
-35 -30 -25 -20 -15 -10 -5 0
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-18 Experimental Output Power vs. Adjusted Simulated Output Power as a function of Input Signal Power
for an input Pump Power of 102.5 mW (13.3 cm)

82
 As seen from these charts, the adjustment improved the accuracy of the simulation. It is

evident that this adjustment is not completely satisfactory as the simulation predicts a lower value

of output signal optical power at higher input pump optical power values.

Figures 5-19 and 5-20 similarly showcase the simulated results obtained from the adjusted

code for the EDFA based on Sample #2. The adjustment for the value of the mode is made on the

absorption curve for Sample #2 which is why the charts are realistic for Sample #2 with the

simulations over-estimating the gain.

Input Signal Power vs. Output Signal Power

(83.1 mW Pump Power Input) Sample #2
10
Output Signal Power (dBm)

-10

-20
-25 -20 -15 -10 -5 0
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-19 Experimental Output Power vs. Adjusted Simulated Output Power as a function of Input Signal Power
for an input Pump Power of 83.1 mW (27.7 cm)

83
 Input Signal Power vs. Output Signal Power
(102.5 mW Pump Power Input) Sample #2
Output Signal Power (dBm) 10

-10

-20
-25 -20 -15 -10 -5 0
Input Signal Power (dBm)

Experimental Output (dBm) Simulated Output (dBm)

Figure 5-20 Experimental Output Power vs. Adjusted Simulated Output Power as a function of Input Signal Power
for an input Pump Power of 102.5 mW (27.7 cm)

The conclusion drawn from these charts is that the value of the mode integral changes as a

function of the optical power present at that particular section of the fiber, i.e., the local optical

power. This means that to accurately define the model, multiple tests must be conducted with

varying lengths of the EDF so that the value of the overlap integral can be estimated.

Since the value of the local optical power does not change drastically, a singular adjustment

made to the code appears to work for both the samples. The lengths of the samples are comparable

to the lengths of the fiber in the Q-switched laser and therefore, it can be said that the goal of the

thesis is achieved in that a MATLAB model that predicts the value of the output signal optical

power for an EDFA operating at 980 nm is coded and verified.

84
 CONCLUSIONS

The motivation for this thesis is to design and validate a MATLAB program that would

model an Erbium-Doped Fiber Amplifier to some acceptable degree of accuracy so that the

program can be incorporated into the model for the Q-switched laser. This thesis presents a series

of equations that model the EDFA in Chapter 2 that includes the coupled differential equations

that define the pump and the signal optical power and the coupled differential equations that define

the Amplified Spontaneous Emission.

From the set of equations that are defined in Chapter 2, two MATLAB models were built,

each one based on a numerical method of solving differential equations. One model was based on

the finite difference method, and the other model was based on the 4th Order Runge-Kutta method.

Both models were validated with theoretical and experimental results from [11] and [12]

respectively. In Chapter 3, in addition to solving the equations using the two mentioned methods,

the data required for simulating a practical system was defined.

In Chapter 4, the system was characterized which brought to light the various nuances

inherent to the system. This helped refine the MATLAB model, making the accuracy of the

simulations increase. The various losses inherent to a fiber amplifier system such as loss due to

mismatched mode field diameters and Rayleigh scattering losses were characterized and

incorporated into the model. In addition, the absorption and emission cross-section data acquired

from the vendor was validated by conducting absorption and emission experiments.

85
 In the final chapter, various experiments were conducted and the comparison between the

experimental and simulated results were shown, which helped this thesis refine the developed

model which is presented in Appendix A. The final model presented in this thesis has been shown

to predict the values of the output signal power with good accuracy as seen in Section 5.4.

The adjustment done to the MATLAB model was done by analyzing the equations. It was

seen from the equations that define the signal power, Eqn. 2-16 that the only factor that could be

modified to account for the difference between the experimental output signal optical power and

the simulated output signal optical power was the mode overlap function. A variable was

introduced which modified the mode overlap integral to let the value of the mode overlap integral

be defined by the user. It was seen that reducing the value of the overlap integral by a factor of

50% made the simulations become more accurate, as shown in Section 5.4.

In regards to future work, there are a few things that can be suggested. The model as of

now is tested for lengths that are compatible with the lengths of the fiber present the Q-switched

fiber laser. It would be interesting to conduct experiments with fibers of longer length to test the

accuracy of this model. It would also be interesting to test various lengths of the fiber and

characterize the mode overlap as a function of the local signal optical power. In addition, extreme

situations can be considered with values of extremely high signal powers and low pump powers.

In conclusion, this thesis presents two different working models that model Erbium-Doped

Fiber Amplifiers, one based on the Finite Difference method of solving differential equations while

the other is based on the 4th Order Runge-Kutta method. Both models are rigorously tested with

theoretical [11]and experimental [12] results from academic papers. An experimental setup is built

86
to verify the data procured from the vendor which allowed for characterization of the setup. Once

this was done, various experiments are conducted to verify the model which leads to the hypothesis

about the mode overlap function which helps solve the issue with the accuracy of the program.

Therefore, the final model that is presented in Appendix A achieves the goal of this thesis

to build a MATLAB model that would model the fiber amplifier in the Q-switched laser.

87
 APPENDIX A

Finite Difference Model:

tic

clc
clear all
close all
disp ('Initialising Variables')

%% Scale Factors

scale_factor_um = 1E6;
scale_factor_mm = 1E3; %step size
dzz = 1E-3*scale_factor_um; %step size for
Power

%% Constants

%Physical Constants
h = 6.636E-34; %Planck constant
(Js)
c = 3E8; %speed of light
(m/s)

%Fiber Constants
L = 2*scale_factor_mm; %length
(mm)
ad = 1.75E-6*scale_factor_um; %fiber radius
(um)
tau = 9E-3; %transition time
(s)
rhoer = 8.4e25*scale_factor_um^-3; %Er doping conc.

%Pump and Signal Constants

lambdap = 974E-9; %pump wavelength
(m)
lambdas = 1550E-9; %signal
wavelength(m)
nup = c/lambdap; %pump frequency
(Hz)
nus = c/lambdas; %signal frequency
(Hz)
Ep = h*nup;
Es = h*nus;

88
%% Signal Absorption Cross-Section

siga=[1.92e-26,1.90e-26,1.88e-26,1.87e-26,1.88e-26,1.88e-26,1.88e-26,...
1.89e-26,1.91e-26,1.93e-26,1.95e-26,1.98e-26,2.01e-26,2.05e-26,...
2.09e-26,2.15e-26,2.20e-26,2.26e-26,2.32e-26,2.39e-26,2.48e-26,...
2.56e-26,2.65e-26,2.75e-26,2.85e-26,2.96e-26,3.08e-26,3.21e-26,...
3.35e-26,3.50e-26,3.67e-26,3.84e-26,4.02e-26,4.21e-26,4.42e-26,...
4.66e-26,4.89e-26,5.16e-26,5.44e-26,5.73e-26,6.03e-26,6.36e-26,...
6.73e-26,7.10e-26,7.48e-26,7.88e-26,8.34e-26,8.82e-26,9.29e-26,...
9.73e-26,1.03e-25,1.08e-25,1.14e-25,1.19e-25,1.25e-25,1.31e-25,...
1.37e-25,1.43e-25,1.49e-25,1.56e-25,1.63e-25,1.69e-25,1.75e-25,...
1.82e-25,1.89e-25,1.97e-25,2.02e-25,2.09e-25,2.16e-25,2.22e-25,...
2.27e-25,2.35e-25,2.43e-25,2.47e-25,2.53e-25,2.60e-25,2.66e-25,...
2.71e-25,2.75e-25,2.81e-25,2.87e-25,2.91e-25,2.95e-25,2.99e-25,...
3.02e-25,3.08e-25,3.11e-25,3.12e-25,3.17e-25,3.18e-25,3.20e-25,...
3.22e-25,3.25e-25,3.27e-25,3.29e-25,3.30e-25,3.32e-25,3.33e-25,...
3.34e-25,3.36e-25,3.37e-25,3.39e-25,3.41e-25,3.43e-25,3.45e-25,...
3.47e-25,3.50e-25,3.54e-25,3.57e-25,3.63e-25,3.66e-25,3.74e-25,...
3.81e-25,3.87e-25,3.96e-25,4.05e-25,4.14e-25,4.27e-25,4.41e-25,...
4.58e-25,4.79e-25,5.16e-25,5.45e-25,5.77e-25,6.08e-25,6.39e-25,...
6.63e-25,6.80e-25,6.90e-25,6.92e-25,6.83e-25,6.64e-25,6.39e-25,...
6.05e-25,5.66e-25,5.25e-25,4.88e-25,4.58e-25,4.37e-25,4.23e-25,...
4.12e-25,4.05e-25,3.97e-25,3.89e-25,3.79e-25,3.69e-25,3.57e-25,...
3.46e-25,3.36e-25,3.26e-25,3.16e-25,3.08e-25,3.00e-25,2.93e-25,...
2.86e-25,2.78e-25,2.72e-25,2.64e-25,2.56e-25,2.48e-25,2.37e-25,...
2.27e-25,2.17e-25,2.06e-25,1.95e-25,1.85e-25,1.74e-25,1.64e-25,...
1.54e-25,1.44e-25,1.35e-25,1.26e-25,1.18e-25,1.10e-25,1.04e-25,...
9.69e-26,9.14e-26,8.58e-26,8.12e-26,7.66e-26,7.29e-26,6.93e-26,...
6.60e-26,6.34e-26,6.03e-26,5.86e-26,5.58e-26,5.41e-26,5.21e-26,...
5.00e-26,4.83e-26,4.66e-26,4.51e-26,4.36e-26,4.21e-26,4.07e-26,...
3.94e-26,3.80e-26,3.68e-26,3.56e-26,3.44e-26,3.33e-26,3.22e-26,...
3.12e-26,3.01e-26,2.91e-26,2.81e-26,2.72e-26,2.63e-26,2.54e-26,...
2.45e-26,2.36e-26,2.29e-26,2.21e-26,2.13e-26,2.05e-26,1.97e-26,...
1.91e-26,1.84e-26,1.78e-26]*scale_factor_um^2;

%% Signal Emission Cross-Section

sige=[9.09e-28,9.22e-28,9.35e-28,9.55e-28,9.80e-28,1.01e-27,1.03e-27,...
1.06e-27,1.10e-27,1.14e-27,1.18e-27,1.23e-27,1.28e-27,1.34e-27,...
1.40e-27,1.47e-27,1.55e-27,1.63e-27,1.71e-27,1.81e-27,1.92e-27,...
2.03e-27,2.15e-27,2.29e-27,2.43e-27,2.59e-27,2.76e-27,2.95e-27,...
3.16e-27,3.38e-27,3.62e-27,3.88e-27,4.16e-27,4.47e-27,4.81e-27,...
5.18e-27,5.57e-27,6.02e-27,6.50e-27,7.02e-27,7.56e-27,8.16e-27,...
8.84e-27,9.55e-27,1.03e-26,1.11e-26,1.20e-26,1.30e-26,1.41e-26,...
1.51e-26,1.63e-26,1.75e-26,1.89e-26,2.03e-26,2.18e-26,2.34e-26,...
2.49e-26,2.66e-26,2.85e-26,3.04e-26,3.25e-26,3.45e-26,3.67e-26,...
3.90e-26,4.15e-26,4.41e-26,4.64e-26,4.91e-26,5.18e-26,5.45e-26,...
5.72e-26,6.06e-26,6.39e-26,6.65e-26,6.97e-26,7.33e-26,7.66e-26,...
7.98e-26,8.29e-26,8.66e-26,9.04e-26,9.38e-26,9.73e-26,1.01e-25,...
1.04e-25,1.09e-25,1.12e-25,1.15e-25,1.19e-25,1.23e-25,1.26e-25,...
1.30e-25,1.34e-25,1.38e-25,1.42e-25,1.45e-25,1.49e-25,1.53e-25,...
1.57e-25,1.62e-25,1.66e-25,1.70e-25,1.75e-25,1.80e-25,1.85e-25,...
1.90e-25,1.96e-25,2.02e-25,2.09e-25,2.17e-25,2.24e-25,2.33e-25,...
2.42e-25,2.52e-25,2.64e-25,2.75e-25,2.88e-25,3.03e-25,3.20e-25,...

89
 3.39e-25,3.62e-25,3.98e-25,4.30e-25,4.65e-25,5.01e-25,5.37e-25,...
5.69e-25,5.96e-25,6.18e-25,6.33e-25,6.38e-25,6.33e-25,6.22e-25,...
6.02e-25,5.75e-25,5.44e-25,5.17e-25,4.95e-25,4.82e-25,4.76e-25,...
4.75e-25,4.76e-25,4.76e-25,4.76e-25,4.74e-25,4.70e-25,4.65e-25,...
4.61e-25,4.56e-25,4.51e-25,4.47e-25,4.44e-25,4.42e-25,4.40e-25,...
4.38e-25,4.36e-25,4.34e-25,4.31e-25,4.26e-25,4.21e-25,4.10e-25,...
4.01e-25,3.91e-25,3.79e-25,3.66e-25,3.54e-25,3.40e-25,3.26e-25,...
3.13e-25,2.99e-25,2.86e-25,2.73e-25,2.61e-25,2.49e-25,2.38e-25,...
2.27e-25,2.18e-25,2.08e-25,2.01e-25,1.94e-25,1.87e-25,1.82e-25,...
1.76e-25,1.72e-25,1.68e-25,1.65e-25,1.61e-25,1.58e-25,1.55e-25,...
1.52e-25,1.50e-25,1.47e-25,1.45e-25,1.42e-25,1.40e-25,1.38e-25,...
1.36e-25,1.33e-25,1.31e-25,1.29e-25,1.26e-25,1.24e-25,1.22e-25,...
1.20e-25,1.17e-25,1.14e-25,1.12e-25,1.09e-25,1.06e-25,1.03e-25,...
9.99e-26,9.66e-26,9.32e-26,8.93e-26,8.54e-26,8.13e-26,7.68e-26,...
7.24e-26,6.75e-26,6.30e-26]*scale_factor_um^2;

%% Fiber Cross-Sections for P&S Wavelengths

sigpa = 2.35e-25*scale_factor_um^2; %cross sec - pump abs

sigpe = 0; %cross sec -
pump em
sigsa = 3.16e-25*scale_factor_um^2; %cross sec - sig abs
sigse = 4.47e-25*scale_factor_um^2; %cross sec - sig em

%% Signal Wavelengths and dv calc

lambda=[1400,1401,1402,1403,1404,1405,1406,1407,1408,1409,1410,1411,...
1412,1413,1414,1415,1416,1417,1418,1419,1420,1421,1422,1423,1424,...
1425,1426,1427,1428,1429,1430,1431,1432,1433,1434,1435,1436,1437,...
1438,1439,1440,1441,1442,1443,1444,1445,1446,1447,1448,1449,1450,...
1451,1452,1453,1454,1455,1456,1457,1458,1459,1460,1461,1462,1463,...
1464,1465,1466,1467,1468,1469,1470,1471,1472,1473,1474,1475,1476,...
1477,1478,1479,1480,1481,1482,1483,1484,1485,1486,1487,1488,1489,...
1490,1491,1492,1493,1494,1495,1496,1497,1498,1499,1500,1501,1502,...
1503,1504,1505,1506,1507,1508,1509,1510,1511,1512,1513,1514,1515,...
1516,1517,1518,1519,1520,1521,1522,1523,1524,1525,1526,1527,1528,...
1529,1530,1531,1532,1533,1534,1535,1536,1537,1538,1539,1540,1541,...
1542,1543,1544,1545,1546,1547,1548,1549,1550,1551,1552,1553,1554,...
1555,1556,1557,1558,1559,1560,1561,1562,1563,1564,1565,1566,1567,...
1568,1569,1570,1571,1572,1573,1574,1575,1576,1577,1578,1579,1580,...
1581,1582,1583,1584,1585,1586,1587,1588,1589,1590,1591,1592,1593,...
1594,1595,1596,1597,1598,1599,1600,1601,1602,1603,1604,1605,1606,...
1607,1608,1609,1610,1611,1612,1613,1614,1615,1616,1617,1618,1619]*1e-9;
freq = c./lambda;
length_lambda = length(lambda);

%calculation of dv for the ASE loops

for i=2:1:length_lambda
dv(i) = (c/lambda(1,i))-(c/lambda(1,i-1));
end

%% Mode Mismatch Calc

wavelength_1 = 974E-9;
wavelength_2 = 1550E-9;
90
%SMF-28E
smf_r = (8.20E-6)/2;
smf_NA = 0.14;
%V Numbers
smf_V_974 = (2*pi*smf_r*smf_NA)/wavelength_1;
smf_V_1550 = (2*pi*smf_r*smf_NA)/wavelength_2;
%MFD
smf_w_974 =
smf_r*(0.65+(1.619/smf_V_974^1.5)+(2.879/smf_V_974^6));
smf_w_1550 =
smf_r*(0.65+(1.619/smf_V_1550^1.5)+(2.879/smf_V_1550^6));
smf_MFD_974 = 2*smf_w_974;
smf_MFD_1550 = 2*smf_w_1550;

%Er110-4/125
Er_r = (3.50E-06)/2;
Er_NA = 0.2;
Er_V_974 = (2*pi*Er_r*Er_NA)/wavelength_1;
Er_V_1550 = (2*pi*Er_r*Er_NA)/wavelength_2;
%MFD
Er_w_974 = Er_r*(0.65+(1.619*Er_V_974^-1.5)+(2.879*Er_V_974^-
6));
Er_w_1550 =
Er_r*(0.65+(1.619/Er_V_1550^1.5)+(2.879/Er_V_1550^6));
Er_MFD_974 = 2*Er_w_974;
Er_MFD_1550 = 2*Er_w_1550;

%% Bessel Function Definition

r_step = 0.1; %radius step

r = 0:r_step:10;
r_temp = round(ad*10,0);
r_len = length(r);

v1 = 1.1428*Er_V_974-0.996;
%Approximation from
u1 = (Er_V_974^2-v1^2)^0.5; %Jeunhomme
v2 = 1.1428*Er_V_1550-0.996; %Single Mode
Fiber
u2 = (Er_V_1550^2-v2^2)^0.5; %1983

%for r<ad
%choose the final value for i based on the value of the radius of the
fiber

for i=1:1:r_temp+1
m(i) = (1/pi)*((v1/(ad*Er_V_974))*...
(besselj(0,(u1*r(i))/ad)/besselj(1,u1)))^2;
n(i) = (1/pi)*((v2/(ad*Er_V_1550))*...
(besselj(0,(u2*r(i))/ad)/besselj(1,u2)))^2;
end

%for r>ad

91
for i=r_temp+2:1:r_len
m(i) = (1/pi)*((u1/(ad*Er_V_974))*...
(besselk(0,(v1*r(i))/ad)/besselk(1,v1)))^2;
n(i) = (1/pi)*((u2/(ad*Er_V_1550))*...
(besselk(0,(v2*r(i))/ad)/besselk(1,v2)))^2;
end

%flipping the values

m_new = fliplr(m);
n_new = fliplr(n);
m_new = m_new(1:end-1);
n_new = n_new(1:end-1);

%% Modes
%concantenation of the two arrays to create the total mode
Ip01 = [m_new,m];
Is01 = [n_new,n];

R = length(Ip01);
LL = 277;

%% Fiber Loss Terms

L_mismatch_974 = -10*log10((4/((smf_MFD_974/Er_MFD_974)+...
(Er_MFD_974/smf_MFD_974))^2));
L_mismatch_1550 = -10*log10((4/((smf_MFD_1550/Er_MFD_1550)+...
(Er_MFD_1550/smf_MFD_1550))^2));
T_WDM_974 = 1; %temp percent
T_WDM_1550 = 0.7; %percent
splice_loss = 0.85; %percent

%% Initialisation
%Power Initialization
%To be entered in mW

%Array Initialization

Rpa =zeros(L+1,R); Rpe =zeros(L+1,R);

Wsa =zeros(L+1,R); Wse =zeros(L+1,R);
N1 =zeros(L+1,R); N2 =zeros(L+1,R);

Pp =zeros(1,L+1); Ps =zeros(1,L+1);
dPp =zeros(1,L+1); dPs =zeros(1,L+1);
Pp1 =0.083081; Ps1 =250e-6;

Pp(1) =(10^(-0.1*L_mismatch_974))*(splice_loss*(T_WDM_974*Pp1));
Ps(1) =(10^(-0.1*L_mismatch_1550))*(splice_loss*(T_WDM_1550*Ps1));

Sase_p =zeros(L+1,length(lambda)); dSase_p =zeros(L+1,length(lambda));

Sase_n =zeros(L+1,length(lambda)); dSase_n =zeros(L+1,length(lambda));
Sase =zeros(L+1,length(lambda));

92
%Flag for ASE %1=ON
%0=OFF
S_Flag =-1;

t_init =toc;

formatSpec ='Variables initialized at %f seconds\n';

fprintf(formatSpec,t_init)
switchintegral = 0;

%
tic

%% First forward execution

for i=1:1:L

for j=1:1:R

%RATE EQUATIONS:

Rpa(i,j)=((sigpa*Pp(i))/(Ep))*Ip01(1,j);
Rpe(i,j)=((sigpe*Pp(i))/(Ep))*Ip01(1,j);

Wsa(i,j)=(((sigsa*Ps(i))/(Es))+...
sum(siga(1,:).*Sase(i,:).*dv(1,:)))*Is01(1,j);
Wse(i,j)=(((sigse*Ps(i))/(Es))+...
sum(sige(1,:).*Sase(i,:).*dv(1,:)))*Is01(1,j);

%POPULATION EQUATIONS:

N2(i,j)=rhoer*((Rpa(i,j)+Wsa(i,j))/...
(Rpa(i,j)+Rpe(i,j)+Wsa(i,j)+Wse(i,j)+(1/tau)));
N1(i,j)=rhoer*((Rpe(i,j)+Wse(i,j)+(1/tau))/...
(Rpa(i,j)+Rpe(i,j)+Wsa(i,j)+Wse(i,j)+(1/tau)));

%ASE EQUATIONS:

for v=1:1:length_lambda
Ge(v)=sige(v)*2*pi*...
sum(N2(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);
Ga(v)=siga(v)*2*pi*...
sum(N1(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);

dSase_n(i,v)=-2*h*freq(v)*Ge(v)-(Ge(v)-Ga(v))*Sase_n(i,v);
dSase_p(i,v)=+2*h*freq(v)*Ge(v)+(Ge(v)-Ga(v))*Sase_p(i,v);

Sase_n(i+1,v)=Sase_n(i,v)+dSase_n(i,v)*dzz;
Sase_p(i+1,v)=Sase_p(i,v)+dSase_p(i,v)*dzz;

Sase(i+1,v)=(Sase_p(i+1,v)+Sase_n(i+1,v))*Sase_Flag;
end
93
 end

%computation of gamma for emission and absorption

Ges(i)=sigse*2*pi*sum(N2(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);
Gas(i)=sigsa*2*pi*sum(N1(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);

Gep(i)=sigpe*2*pi*sum(N2(i,101:121).*Ip01(1,101:121).*r(1:21)*r_step);
Gap(i)=sigpa*2*pi*sum(N1(i,101:121).*Ip01(1,101:121).*r(1:21)*r_step);

%computation of differential power

%i.e., change of power in the step

dPp(i)=((Gep(i)-Gap(i))*Pp(i));
dPs(i)=((Ges(i)-Gas(i))*Ps(i));

%computation of power in the next step

Pp(i+1)=Pp(i)+dPp(i)*dzz;
Ps(i+1)=Ps(i)+dPs(i)*dzz;

end

t_ffe=toc;
formatSpec='First forward execution concluded at %f seconds';
fprintf(formatSpec,t_ffe)

%%
tic

%% First backward execution

%calculates backward ASE

for i=L+1:-1:1
if(i~=1)
for j=1:1:R

%RATE EQUATIONS
Rpa(i-1,j)=((sigpa*Pp(i))/(Ep))*Ip01(1,j);
Rpe(i-1,j)=((sigpe*Pp(i))/(Ep))*Ip01(1,j);

Wsa(i-1,j)=(((sigsa*Ps(i))/(Es))+...
sum(siga(1,:).*Sase(i,:).*dv(1,:)))*Is01(1,j);
Wse(i-1,j)=(((sigse*Ps(i))/(Es))+...
sum(sige(1,:).*Sase(i,:).*dv(1,:)))*Is01(1,j);

%POPULATION EQUATIONS
N2(i-1,j)=(1/10)*rhoer*((Rpa(i-1,j)+Wsa(i-1,j))/...
(Rpa(i-1,j)+Rpe(i-1,j)+Wsa(i-1,j)+Wse(i-1,j)+(1/tau)));
N1(i-1,j)=(1/10)*rhoer*((Rpe(i-1,j)+Wse(i-1,j)+(1/tau))/...
(Rpa(i-1,j)+Rpe(i-1,j)+Wsa(i-1,j)+Wse(i-1,j)+(1/tau)));
94
 %ASE EQUATIONS
for v=1:1:length_lambda
Ge(v)=sige(v)*2*pi*...
sum(N2(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);
Ga(v)=siga(v)*2*pi*...
sum(N1(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);

dSase_n(i,v)=-2*h*freq(v)*Ge(v)-(Ge(v)-Ga(v))*Sase_n(i,v);
dSase_p(i,v)=+2*h*freq(v)*Ge(v)+(Ge(v)-Ga(v))*Sase_p(i,v);

Sase_n(i-1,v)=Sase_n(i,v)+dSase_n(i,v)*(-1)*dzz;
Sase_p(i-1,v)=Sase_p(i,v)+dSase_p(i,v)*(-1)*dzz;

Sase(i-1,v)=(Sase_p(i-1,v)+Sase_n(i-1,v))*Sase_Flag;
end

end

%EMISSION AND ABSORPTION FACTORS

Ges(i-1)=sigse*2*pi*...
sum(N2(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);
Gas(i-1)=sigsa*2*pi*...
sum(N1(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);

Gep(i-1)=sigpe*2*pi*...
sum(N2(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);
Gap(i-1)=sigpa*2*pi*...
sum(N1(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);

%FINITE DIFFERENCE EQUATIONS

dPp(i-1)=(Gep(i-1)-Gap(i-1))*Pp(i);
dPs(i-1)=(Ges(i-1)-Gas(i-1))*Ps(i);

Pp(i-1)=Pp(i)+dPp(i-1)*dzz;
Ps(i-1)=Ps(i)+dPs(i-1)*dzz;
end

end

t_fbe=toc;
formatSpec='First backward execution concluded at %f seconds';
fprintf(formatSpec,t_fbe)

%% Backward Plot

x=(1:1:L+1)/scale_factor_mm;
figure(1)
plot(x,Pp,x,Ps)
legend('Pump','Signal')
title('Inverse Pump and Signal powers');

95
xlabel('Distance (m)');
ylabel('Power (mW)');

%%
tic

%% Resetting Terms

Pp(1)=Pp1; Ps(1)=Ps1;

Sase_p=zeros(L+1,length(lambda)); dSase_p=zeros(L+1,length(lambda));

Sase_n(2:L+1,:)=0;
dSase_n=zeros(L+1,length(lambda));

Sase=zeros(L+1,length(lambda));

%% Final Iteration

for i=1:1:L

if switchintegral == 1
%Ip01 = (1/0.7984)*[m_new,m];
%Is01 = (10/0.4870)*[n_new,n];
end

for j=1:1:R

%RATE EQUATIONS:

Rpa(i,j) = ((sigpa*Pp(i))/(Ep))*Ip01(1,j);
Rpe(i,j) = ((sigpe*Pp(i))/(Ep))*Ip01(1,j);

Wsa(i,j) = (((sigsa*Ps(i))/(Es))+...
sum(siga(1,:).*Sase(i,:).*dv(1,:)))*Is01(1,j);
Wse(i,j) = (((sigse*Ps(i))/(Es))+...
sum(sige(1,:).*Sase(i,:).*dv(1,:)))*Is01(1,j);

%POPULATION EQUATIONS:

N2(i,j) = rhoer*((Rpa(i,j)+Wsa(i,j))/...
(Rpa(i,j)+Rpe(i,j)+Wsa(i,j)+Wse(i,j)+(1/tau)));
N1(i,j) = rhoer*((Rpe(i,j)+Wse(i,j)+(1/tau))/...
(Rpa(i,j)+Rpe(i,j)+Wsa(i,j)+Wse(i,j)+(1/tau)));

%ASE EQUATIONS:

for v=1:1:length_lambda
Ge(v) = sige(v)*2*pi*sum...
(N2(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);

96
 Ga(v) = siga(v)*2*pi*sum...
(N1(i,101:121).*Is01(1,101:121).*r(1:21)*r_step);

dSase_n(i,v) = -2*h*freq(v)*...
Ge(v)-(Ge(v)-Ga(v))*Sase_n(i,v);
dSase_p(i,v) = +2*h*freq(v)*...
Ge(v)+(Ge(v)-Ga(v))*Sase_p(i,v);

Sase_n(i+1,v) = Sase_n(i,v)+dSase_n(i,v)*dzz;
Sase_p(i+1,v) = Sase_p(i,v)+dSase_p(i,v)*dzz;

Sase(i+1,v) = (Sase_p(i+1,v)+Sase_n(i+1,v))*S_Flag;
end

end

% computation of gamma for emission and absorption

Ges(i) =
sigse*2*pi*sum(N2(i,r_len:(r_len+r_temp)).*Is01(1,r_len:(r_len+r_temp)).*r(1:
(r_temp+1))*r_step);
Gas(i) =
sigsa*2*pi*sum(N1(i,r_len:(r_len+r_temp)).*Is01(1,r_len:(r_len+r_temp)).*r(1:
(r_temp+1))*r_step);

Gep(i) =
sigpe*2*pi*sum(N2(i,r_len:(r_len+r_temp)).*Ip01(1,r_len:(r_len+r_temp)).*r(1:
(r_temp+1))*r_step);
Gap(i) =
sigpa*2*pi*sum(N1(i,r_len:(r_len+r_temp)).*Ip01(1,r_len:(r_len+r_temp)).*r(1:
(r_temp+1))*r_step);

%computation of differential power

%i.e., change of power in the step

dPp(i) = ((Gep(i)-Gap(i))*Pp(i));
dPs(i) = ((Ges(i)-Gas(i))*Ps(i));

%computation of power in the next step

Pp(i+1) = (Pp(i)+dPp(i)*dzz);
Ps(i+1) = (Ps(i)+dPs(i)*dzz);

% if i>1
% if Ps(i)<Ps(i-1)
% switchintegral = 1;
% end
% end

end

t_sfe=toc;

97
formatSpec='Second forward execution concluded at %f seconds\n';
fprintf(formatSpec,t_sfe)

tic
x = (1:1:L+1)/scale_factor_mm;

%% Figure 2
figure(2)
plot(x,Pp,x,Ps)
legend('Pump','Signal')
title('Pump and Signal powers');
xlabel('Distance (m)');
ylabel('Power (mW)');

%% Figure 3
figure(3)
for i =1:1:L+1
Gain(i)=10*log10(Ps(i)/Ps(1));
end
x=(1:1:L+1)/scale_factor_mm;
plot(x,Gain)
title('Gain vs Length');
xlabel('Distance (m)');
ylabel('Gain (dB)');

%% Figure 4
figure(4)
% Lambda
lambda=[1400,1401,1402,1403,1404,1405,1406,1407,1408,1409,1410,1411,...
1412,1413,1414,1415,1416,1417,1418,1419,1420,1421,1422,1423,1424,...
1425,1426,1427,1428,1429,1430,1431,1432,1433,1434,1435,1436,1437,...
1438,1439,1440,1441,1442,1443,1444,1445,1446,1447,1448,1449,1450,...
1451,1452,1453,1454,1455,1456,1457,1458,1459,1460,1461,1462,1463,...
1464,1465,1466,1467,1468,1469,1470,1471,1472,1473,1474,1475,1476,...
1477,1478,1479,1480,1481,1482,1483,1484,1485,1486,1487,1488,1489,...
1490,1491,1492,1493,1494,1495,1496,1497,1498,1499,1500,1501,1502,...
1503,1504,1505,1506,1507,1508,1509,1510,1511,1512,1513,1514,1515,...
1516,1517,1518,1519,1520,1521,1522,1523,1524,1525,1526,1527,1528,...
1529,1530,1531,1532,1533,1534,1535,1536,1537,1538,1539,1540,1541,...
1542,1543,1544,1545,1546,1547,1548,1549,1550,1551,1552,1553,1554,...
1555,1556,1557,1558,1559,1560,1561,1562,1563,1564,1565,1566,1567,...
1568,1569,1570,1571,1572,1573,1574,1575,1576,1577,1578,1579,1580,...
1581,1582,1583,1584,1585,1586,1587,1588,1589,1590,1591,1592,1593,...
1594,1595,1596,1597,1598,1599,1600,1601,1602,1603,1604,1605,1606,...
1607,1608,1609,1610,1611,1612,1613,1614,1615,1616,1617,1618,1619];
% End Lambda
plot(lambda,sum(Sase))
title('Sase vs. Wavelengths');
xlabel('Wavelength (nm');
ylabel('Sase (mW)');

%%
% Gain(L)
t_ffe=0;
98
t_fbe=0;
t_conclude=toc;
t_total=t_init+t_ffe+t_fbe+t_sfe+t_conclude;
formatSpec='Total Time needed is %f seconds\n';
fprintf(formatSpec,t_total)

99
APPENDIX B

4th Order Runge-Kutta Model:

Main Function:

% tic

%%
clc
clear all
close all

%%
%Initial Conditions for Power
Pp(1) = 5.397e-3;
Ps(1) = 11e-6;
global Sase;
Sase(1) = 0;

%%
%Scale Factors
sf_um = 1e6;
sf_mm = 1e3; %step size

%fiber constants
h = 6.636e-34;
ad = 1.75e-6*sf_um;
rhoer = 4.86e24*sf_um^-3;
tau = 9e-3;

%cross-sections
sigpa = 5.8e-25*sf_um^2; %Pump Absorption CS (um^2)
sigpe = 0; %Pump Emission CS (um^2)
sigsa = 2.92309e-25*sf_um^2;%Signal Absorption CS (um^2)
sigse = 3.47566e-25*sf_um^2;%Signal Emission CS (um^2)

%Constants
dzz = 1e-3*sf_um; %step size for Power
L = 12*sf_mm; %length of the fiber (mm)

%constants
lambdap = 980e-9; %Pump wavelength %m
lambdas = 1550e-9; %Signal wavelength %m
c = 3e8; %Speed of light %m
nup = c/lambdap; %Pump frequency %Hz
100
nus = c/lambdas; %Signal frequency %Hz

%%
%bessel
r_step = 0.1e-6*sf_um; %radius step
r = 0:r_step:10e-6*sf_um;
NA = 0.2; %Numerical Aperture
lambda = [0.980e-6; 1.530e-6]*sf_um;%Re-defn for Bessel func.
V1 = (2*pi*ad*NA)/lambda(1); %Fiber number V for L1
V2 = (2*pi*ad*NA)/lambda(2); %Fiber number V for L2

v1 = 1.1428*V1-0.996; %Approximation from

u1 = (V1^2-v1^2)^0.5; %Jeunhomme
v2 = 1.1428*V2-0.996; %Single Mode Fiber Optics
u2 = (V2^2-v2^2)^0.5; %1983

%for r<ad
%choose the final value for i based on the value of the radius of the fiber

for i=1:1:18
m(i) = (1/pi)*((v1/(ad*V1))*...
(besselj(0,(u1*r(i))/ad)/besselj(1,u1)))^2;
n(i) = (1/pi)*((v2/(ad*V2))*...
(besselj(0,(u2*r(i))/ad)/besselj(1,u2)))^2;
end

%for r>ad

for i=19:1:101
m(i) = (1/pi)*((u1/(ad*V1))*...
(besselk(0,(v1*r(i))/ad)/besselk(1,v1)))^2;
n(i) = (1/pi)*((u2/(ad*V2))*...
(besselk(0,(v2*r(i))/ad)/besselk(1,v2)))^2;
end

%flipping the values

m_new = fliplr(m);
n_new = fliplr(n);
m_new = m_new(1:end-1);
n_new = n_new(1:end-1);

%Modes
%concantenation of the two arrays to create the total mode
Ip01 = (1.55/0.878236845791598)*[m_new,m];
Is01 = (1.55/0.592829910946522)*[n_new,n];

%%
%Function Handles

dPp = @(dzz,Pp,Ps)(Gp('E',Pp,Ps,r,Is01,Ip01)-...
Gp('A',Pp,Ps,r,Is01,Ip01))*Pp;

dPs = @(dzz,Pp,Ps)(Gs('E',Pp,Ps,r,Is01,Ip01)-...
101
 Gs('A',Pp,Ps,r,Is01,Ip01))*Ps;

%%
for i=1:1:L

k1Pp = dPp(dzz, Pp(i), Ps(i));

k1Ps = dPs(dzz, Pp(i), Ps(i));

k2Pp = dPp(dzz, Pp(i)+dzz/2*k1Pp, Ps(i)+dzz/2*k1Ps);

k2Ps = dPs(dzz, Pp(i)+dzz/2*k1Pp, Ps(i)+dzz/2*k1Ps);

k3Pp = dPp(dzz, Pp(i)+dzz/2*k2Pp, Ps(i)+dzz/2*k2Ps);

k3Ps = dPs(dzz, Pp(i)+dzz/2*k2Pp, Ps(i)+dzz/2*k2Ps);

k4Pp = dPp(dzz, Pp(i)+dzz*k3Pp, Ps(i)+dzz*k3Ps);

k4Ps = dPs(dzz, Pp(i)+dzz*k3Pp, Ps(i)+dzz*k3Ps);

A(i) = dzz/6 * (k1Pp + 2*k2Pp + 2*k3Pp + k4Pp);

B(i) = dzz/6 * (k1Ps + 2*k2Ps + 2*k3Ps + k4Ps);

Pp(i+1) = Pp(i) + A(i);

Ps(i+1) = Ps(i) + B(i);

end

%%
z = (1:1:L+1)/sf_mm;

for i=1:1:L+1
Gain(i) = 10*log10(Ps(i)/Ps(1));
end

% %%
% figure(1)
% plot (z,Pp,z,Ps)
% legend ('Pp = Pump','Ps = Signal')
% title ('Pump and Signal powers');
% xlabel ('Distance (m)');
% ylabel ('Power (mW)');
%
%%
figure(2)
plot (z,Gain)
legend ('Gain')
title ('Gain vs. Length');
xlabel ('Distance (m)');
ylabel ('Gain (dB)');

%%
Gain_Max = max(Gain);

formatSpec='\nMax Gain is :%fdB\n';

102
fprintf(formatSpec,Gain_Max)

Gain_End = Gain(i);

formatSpec='\nGain at the end of the fiber is :%fdB\n';

fprintf(formatSpec,Gain_End)

% total_t = toc;
%
% formatSpec='\nTime for execution :%fseconds\n';
% fprintf(formatSpec,total_t)

Gs.m

function G = Gs(P,Pp,Ps,r,Is01,Ip01)

sf_um = 1e6;

%fiber constants
h = 6.636e-34;
ad = 1.75e-6*sf_um;
rhoer = 4.86e24*sf_um^-3;
tau = 9e-3;

%constants
lambdap = 980e-9; %Pump wavelength %m
lambdas = 1550e-9; %Signal wavelength %m
c = 3e8; %Speed of light %m
nup = c/lambdap; %Pump frequency %Hz
nus = c/lambdas; %Signal frequency %Hz

%cross-sections
sigpa = 5.8e-25*sf_um^2; %Pump Absorption CS (um^2)
sigpe = 0; %Pump Emission CS (um^2)
sigsa = 2.92309e-25*sf_um^2;%Signal Absorption CS (um^2)
sigse = 3.47566e-25*sf_um^2;%Signal Emission CS (um^2)

r_step = 0.1e-6*sf_um; %radius step

R = length(Ip01);

for i=1:1:R

Rpa(i) = (sigpa*Pp*Ip01(1,i))/(h*nup);
Rpe(i) = (sigpe*Pp*Ip01(1,i))/(h*nup);
Wsa(i) = (sigsa*Ps*Is01(1,i))/(h*nus);
Wse(i) = (sigse*Ps*Is01(1,i))/(h*nus);

103
 N2(i) = rhoer*(Rpa(i)+Wsa(i))/...
(Rpa(i)+Rpe(i)+Wsa(i)+Wse(i)+(1/tau));
N1(i) = rhoer*(Rpe(i)+Wse(i)+(1/tau))/...
(Rpa(i)+Rpe(i)+Wsa(i)+Wse(i)+(1/tau));

end

if(P=='E')
Gs = sigse*2*pi*...
sum(N2(101:121).*Is01(1,101:121).*r(1:21)*r_step);
end

if(P=='A')
Gs = sigsa*2*pi*...
sum(N1(101:121).*Is01(1,101:121).*r(1:21)*r_step);
end

G=Gs;

Gp.m

function G = Gp(P,Pp,Ps,r,Is01,Ip01)

sf_um = 1e6;

%fiber constants
h = 6.636e-34;
ad = 1.75e-6*sf_um;
rhoer = 4.86e24*sf_um^-3;
tau = 9e-3;

%constants
lambdap = 980e-9; %Pump wavelength %m
lambdas = 1550e-9; %Signal wavelength %m
c = 3e8; %Speed of light %m
nup = c/lambdap; %Pump frequency %Hz
nus = c/lambdas; %Signal frequency %Hz

%cross-sections
sigpa = 5.8e-25*sf_um^2; %Pump Absorption CS (um^2)
sigpe = 0; %Pump Emission CS (um^2)
sigsa = 2.92309e-25*sf_um^2;%Signal Absorption CS (um^2)
sigse = 3.47566e-25*sf_um^2;%Signal Emission CS (um^2)

r_step = 0.1e-6*sf_um; %radius step

R = length(Ip01);

104
for i=1:1:R

Rpa(i) = (sigpa*Pp*Ip01(1,i))/(h*nup);
Rpe(i) = (sigpe*Pp*Ip01(1,i))/(h*nup);
Wsa(i) = (sigsa*Ps*Is01(1,i))/(h*nus);
Wse(i) = (sigse*Ps*Is01(1,i))/(h*nus);

N2(i) = rhoer*(Rpa(i)+Wsa(i))/...
(Rpa(i)+Rpe(i)+Wsa(i)+Wse(i)+(1/tau));
N1(i) = rhoer*(Rpe(i)+Wse(i)+(1/tau))/...
(Rpa(i)+Rpe(i)+Wsa(i)+Wse(i)+(1/tau));

end

if(P=='E')
Gp = sigpe*2*pi*...
sum(N2(101:121).*Ip01(1,101:121).*r(1:21)*r_step);
end

if(P=='A')
Gp = sigpa*2*pi*...
sum(N1(101:121).*Ip01(1,101:121).*r(1:21)*r_step);
end

G=Gp;

105
APPENDIX C

106
REFERENCES

[1] C. J. Koester and E. Snitzer, "Amplification in a Fiber Laser," Applied Optics,

vol. 3, no. 10, pp. 1182-1186, 1964.

[2] C. P. Becker, A. N. Olsson and R. J. Simpson, Erbium-Doped Fiber

Amplifiers Fundamentals and Technology, San Diego: Academic Press, 1999.

[3] J. Stone, C. A. Burrus, A. G. Dentai and B. I. Miller, Applied Physics Letters,

vol. 29, no. 37, 1976.

[4] J. Hegarty, M. M. Broer, B. Golding, J. R. Simpson and J. B. MacChesney,

Physics Review Letter, vol. 51, no. 2033, 1983.

[5] S. B. Poole, D. N. Payne, R. J. Mears, M. E. Fermann and R. I. Laming,

Journal of Lightwave Technology, vol. 21, p. 737, 1985.

[6] B. Pedersen, A. Bjarklev, J. H. Povlsen, K. Dybdal and C. C. Larsen, "The

Design of Erbium-Doped Fiber Amplifiers," Journal of Lightwave Technology, vol.

9, no. 9, pp. 1105-1112, 1991.

107
[7] C. R. Giles and E. Desurivire, "Modelling Erbium-Doped Fiber Amplifiers,"

Journal of Lightwave Technology, vol. 9, no. 2, pp. 271-283, 1991.

[8] E. Desurvire, Erbium-Doped Fiber Amplifiers Principles and Applications,

New York: John Wiley and Sons, 1994.

[9] J. Hecht, Understanding Fiber Optics, Upper Saddle River, NJ: Prentice Hall,

2002.

[10] W. L. Barnes, I. R. Laming, E. J. Tarbox and P. R. Morkel, "Absorption and

Emission Cross Section of Er3+ Doped Silica Fibers," IEEE Journal of Quantum

Electronics, vol. 27, no. 4, pp. 1004-1010, 1991.

[11] D. Geguo and C. Guofu, "Gain performances of 980 nm-pumped erbium-

doped fiber amplifiers," Science in China, vol. 42, no. 3, pp. 286-292, 1999.

[12] A. B. Mohammad, "Gain Simulation of Erbium-Doped Fiber Amplifier," in

TENCON Proceedings, Kuala Lumpur, Malaysia, Malaysia, 2000.

[13] B. Pedersen, A. Bjarklev and O. Lumholt, "Detailed Design of Erbium-Doped

Fiber Amplifiers," IEEE Photonics, vol. 3, no. 6, p. 548, 1991.

[14] W. Zheng, O. Hulten and R. Rylander, "Erbium-Doped Fiber Splicing and

Splice Loss Estimation," Journal of Lightwave Technology, vol. 12, no. 3, p. 430,

1994.

108
109