Comparative Analysis of the Performance ofOptical Amplifiers with Pumping and Injection

Ana Margarida Carvalhão Rodrigues

Instituto Superior Técnico Technical University of Lisbon

Lisbon, Portugal amcr@tecnico.ulisboa.pt

Abstract — This dissertation focuses on the optical amplification

process, covering the signal generating process by studying

semiconductor lasers and giving emphasis to the main

amplification technologies in use nowadays which are

Semiconductor Optical Amplifiers (SOA), Erbium-Doped Fibre

Amplifiers (EDFA) and Raman Amplifiers. In the semiconductor

lasers study, when injection current is modulated directly the

influence over time of parameters such as gain, photon and

carrier lifetime in photon and carrier populations inside the

resonant cavity it’s analysed. On the other hand, in the amplifiers

study it’s intended to understand which factors have impact on

their performance. Thus, in SOAs, which operation is similar to

that of the semiconductor lasers, the gain saturation and also the

amplified pulse behaviour by varying the input power and the

confinement factor are evaluated. Then, in EDFAs the relation

between the gain and the optimal fibre length are analysed and

from its spectral characterization is assessed the impact of gain

and noise figure variations in the signal performance. Finally, in

Raman amplifiers in which the signal is amplified due to

stimulated Raman scattering, the gain spectrum and three

limiting factors of their performance: spontaneous Raman

scattering, Rayleigh backscattering and pump-noise transfer are

considered.

Keywords - Optical Amplifiers, Semiconductor Lasers,

Semiconductor Optical Amplifiers, Erbium-Doped Fibre

Amplifiers, Raman Amplifiers.

1. INTRODUCTION

The development of optical amplifiers in the late 1980s had a similar impact as the laser invention in the early 1960s, having these devices contributed to the evolution of communication systems.

A basic optical communication link consists in a transmitter and a receiver with an optical fibre cable which makes the connection between them. Even if signal propagation in optical fibres suffers much less attenuation than in other materials, such as copper, there is a limit of 100 km above the distance that signals can travel until noise level compromise reception [1].

Thus, it’s necessary to restore the signal power before it reaches the receiver through regeneration [2]. However the use of electronic regenerators 3R implies the realization of electro-optical and opto-electrical conversions in order to rescaling, reshaping and retiming. All this process is complex and limit the system once regenerators only worked for the bitrate which they were designed, having high losses, low transmission rates and high costs.

Optical amplifiers (OA) allow direct amplification of the optical signal without using repeaters or regenerators and are used to amplify weak optical signals in order to increase transmitted signals range and also transmission distances. These devices have several advantages over regenerators once they present higher gain in broad bandwidths and by using a single amplifier it’s possible to amplify multiple channels of a WDM system, simultaneously. So, the OAs has become essential components in optical communication systems with high capacity and high performance.

2. SEMICONDUCTOR LASERS

A semiconductor laser is a semiconductor optical cavity containing an active region which operates as an optical amplifier in which the amplification is converted into oscillation through a feedback mechanism and also where the selection of the laser oscillation frequencies occurs.

Generic laser structure consists of a gain medium, an optical feedback and a mechanism to supply energy. To provide optical feedback it’s used an optical cavity formed by two mirrors on either end of it which also confine light. Furthermore, the gain medium allows the amplification of electromagnetic radiation (light) and must be supplied by an external energy usually an electric current. This process is named pumping.

In semiconductor lasers, the transitions between states with different energy levels will be processed in energy bands and there is a conduction band and a valence band. As illustrated in Figure 2.1, three basic processes are possible: absorption, spontaneous emission and stimulated emission.

Figure 2.1. Interaction processes between electrons and photons: (a)

absorption; (b) spontaneous emission; (c) stimulated emission.

For have lasing, the pumping process must produce population inversion, meaning that the electron’s density in the conduction band must be larger than the electron’s density in the valence band. Hence, the emission dominates over absorption and there is an increase of the electrons population in the conduction band as the gaps population in the valence band. If the population inversion happens in the optical cavity, incident light can be amplified by stimulated emission (Figure 2.1c) and if electrons are available in the conduction

band an interaction between them and the incident photon can occurs. This interaction forces a recombination among the electron with a gap from the valence band, resulting in the emission of a photon with the same energy and in phase with the incident photon at the output.

2.1. Rate Equations

It’s assumed that the semiconductor laser operates in single mode, meaning the laser oscillation corresponds to a single longitudinal mode.

Such as all electromagnetic phenomena, the optical fields propagation is governed by Maxwell's equations which with a quantum mechanical approach to induced polarization let deriving the rate equations [3]:

(2.1)

(2.2)

where and are photons and electrons numbers, respectively, is the gain, is the injection current, is the electron’s charge,

is the total spontaneous emission fraction and and are the photon and carrier lifetimes, respectively.

Analysing the continuous wave (CW) operation regime with a constant injection current in which the rate equations are in steady state, i.e.,

(2.3)

Thus, the rate equations are written as

(2.4)

(2.5)

where and represent the photons and electrons numbers in steady state, respectively, is the spontaneous emission factor and

is the spontaneous emission rate in steady state. The gain function, , is a non-linear relation

that depends on the electrons and photons numbers. Considering the linear regime, it’s assumed that the gain

only varies with the electrons number , doesn’t depending on the photons number, .

Under this model, the electrons population is constant, doesn’t depending on the injection current and it’s threshold value is given by

(2.6)

However, equation (2.6) it’s only valid when the laser is emitting, i.e., when and otherwise, if , then,

(2.7)

Having the electrons population value at the threshold of oscillation given by equation (2.7), the threshold current value is given by

(2.8)

And, finally,

(2.9)

2.2. Direct Modulation of Injection Current

When there is a direct modulation of injection current, current signal is written as (2.10)

where is the injection current, is the pulse current and represents the pulse shape and is given by a rectangular pulse such as

(

) (2.11)

If it’s applied a current pulse to a laser, there is an initial turn-on time delay until the carrier density reaches the threshold value to start a coherent emission and produce photons which after the time delay exhibits relaxation oscillation before it stabilizes at its steady-state value as is shown in Figure 2.2.

Figure 2.2. Input pulse response of a semiconductor laser [4].

2.2.1. Injection Current higher than Threshold Current ( and )

In this situation the injection current is given by

(

) (2.12)

and as known , so it’s inferred that laser is transmitting and so the photons population is greater than zero ( ).

In Figure 2.3 photon and carrier lifetimes are changed to analyse the laser behaviour.

Figure 2.3. Temporal evolution of the injection current (red), photons population (blue) on the left scale and carrier population (green) on

the right scale for different values of and to .

Observing Figure 2.3, it’s concluded that for a fixed value of , changes in values induces a variation in photons number such that electrons number increases as the decreases. On the other hand, when it’s fixed, a decrease in value leads to a faster laser response, i.e. laser takes less time to reach a steady value. So, from the nine images in Figure 2.3 the one which has the intended behaviour is that of the lowest photon lifetime ( ) and the lowest carrier lifetime ( ).

Having reached to the parameters that are considered the most appropriate to the laser behaviour, it will be studied the evolution of photons and carrier densities by changing gain as shown in Figure 2.4.

Figure 2.4. Temporal evolution of the injection current (red), photons population (blue) on the left scale and carrier population (green) on

the right scale for and and for different values of .

Analysing Figure 2.4, it’s concluded that to there is an ultra-short pulse but when gain value increases the bigger are relaxation oscillations frequency and amplitude, worsening performance.

2.2.2. Injection Current lower than Threshold Current ( and )

In this situation the injection current is given by

(

) (2.13)

As there isn’t population inversion and the process that governs is spontaneous emission, so since the laser emission condition it’s not verified, laser only emits in a short period of time and the difference to the situation studied in 2.2.1 is due to the initial condition which in this case is .

Figure 2.5. Temporal evolution of the injection current (red), photons population (blue) on the left scale and carrier population (green) on

the right scale for and and for different values of .

Analysing Figure 2.5 it can be seen that the electron density takes more time to reach threshold value and consequently the photons density response it’s slower due to the fact that emission only occurs after electrons density reaches threshold value. It’s also possible to see that when the current pulse ends, electrons density tends to while photons density tends to zero.

3. SEMICONDUCTOR OPTICAL AMPLIFIERS (SOA)

SOA structure is based on the conventional semiconductor laser structure where the output facet reflectivity’s are between 30 and 35% or antireflection. Such amplifiers may be used either in linear or nonlinear operation modes and can be classified into two groups: resonant or Fabry-Perot (FP) amplifiers and traveling wave (TW) amplifiers. These devices are capable of providing high internal gain (15 to 35 dB), low power consumption and their structure makes them particularly suitable for use in single-mode fibre [5].

3.1. Gain Saturation

SOA’s amplification factor is obtained by using the conventional theory of FP interferometer and it’s given by

( √ ) √ [

] (3.1)

where and are the facet’s reflectivities, are the cavity-resonance frequencies and is the longitudinal-mode spacing or the FP cavity free spectral range.

When , is reduced to the single-pass amplification factor that corresponds to that of a TW amplifier when gain saturation is negligible and it’s given by (3.2)

The amplifier bandwidth is determined by for which drops by 3 dB from its peak value reached whenever coincides with one of the cavity-resonance frequencies and it’s written as

[

√

( √ )

] (3.3)

From equation (3.3) it’s seen that the amplifier bandwidth is a fraction of the cavity free spectral range, typically, and and therefore, once it’s narrow, FP amplifiers are unsuitable for most light wave system applications [6].

To estimate an acceptable value of the facet’s reflectivities it’s necessary to consider maximum and minimum values in equation (3.1) near to the resonance cavity. Thus, the ratio between the maximum and the minimum is written as

( √

√ )

(3.4)

In Figure 3.1 reflectivity’s variations as function of gain (assuming ) for four values of gain are shown and it’s observed that gain oscillations increases with increasing gain as well as increasing reflectivity.

Figure 3.1. Gain oscillations as function of reflectivity for four gain

values.

In order to analyse gain saturation, it’s necessary to consider the peak gain and it’s assumed that it increases linearly with the carrier population as

(3.5)

where is the confinement factor, is the differential gain, is the active volume, and is the value of required at transparency.

Power saturation is defined as

(3.6)

where is the carrier lifetime and is the cross sectional area of the waveguide mode.

SOA’s noise figure is larger than the minimum value of 3 dB for several reasons, being the main contribution due to the spontaneous-emission factor obtained by

(3.7)

Another contribution is due to internal losses, such as free-carrier absorption or scattering losses that reduce the available gain from to .

So, the noise figure can be written as

(

) (

) (3.8)

3.2. Pulse Amplifcation

Amplification of optical pulses in SOAs is governed by

(3.9)

| |

(3.10)

where carrier-induced index variations are included through the line width enhancement factor , the saturation energy is defined as and is defined as .

By introducing and √ and from equations (3.9) and (3.10) are written

(3.11)

(3.12)

(3.13)

The amplification factor is given by

(3.14)

where is the unsaturated amplifier gain and the partial energy of the input pulse is defined as ∫

.

The frequency chirp is defined by

(3.15)

where the phase shift is found by integrating equation (3.12) over the amplifier length and is given by

∫

(3.16)

Figure 3.2 shows the chirp profiles for several input pulse energies.

Figure 3.2. Frequency chirp as function of time for four different values of , assuming a Gaussian pulse, and

.

It’s seen that the frequency chirp is larger for the more energetic pulses once gain saturation occurs earlier for those pulses.

In Figure 3.3 is shown the pulse expected shape when a Gaussian pulse of energy is amplified by a SOA.

Figure 3.3. Pulse shape at SOA output for three different values of

and for a input Gaussian pulse with .

It’s observed that for temporal changes dependent on the amplifier gain are quite significant and the pulse shape becomes asymmetrical as the unsaturated gain value increases.

In Figure 3.4 it’s possible to observe that the dominant peak is at the pulse front and has been shifted toward a negative frequency (to the red side), becomes broader than the input spectrum and it’s also may be accompanied by one or more satellite peaks.

Figure 3.4. Input signal Spectrum (blue) and its amplified signal

(red).

By analysing Figure 3.5 it’s concluded if the ratio is increased, satellite peaks are arising on the spectrum and it has a stronger shift to the red side.

Figure 3.5. Amplified Gaussian pulse spectrum for three values of

input power.

The confinement factor is the ratio between the intensity in the active region and total intensity.

Figure 3.6. Amplified Gaussian pulse spectrum for three values of the

confinement factor.

Thus, from Figure 3.6 it’s seen that the variation of the confinement factor influences the pulse spectrum in that, for the same input power, systems with lower confinement factors allow for the same output power better beam quality and higher saturation power than systems with higher confinement factors [7].

4. ERBIUM-DOPED FIBRE AMPLIFIERS (EDFA)

An EDFA is an optical fibre which core is doped with erbium ions ( ) and working in the third window ( ). These erbium ions exhibit a radiative decay in which the excited state lifetime is long enough.

In an EDFA the pumping is done by a continuous optical signal from a laser and it can be unidirectional (forward or backward) or bidirectional (forward and backward simultaneously). In this work, it’s considered unidirectional forward pumping.

The power inserted by the laser at the pumping make erbium ions travels from their energy level for a higher energy level, reaching population inversion and making the fibre behaves as an active medium, i.e. with gain.

4.1. Gain

Gain depends on the amplifier wavelength and on the input power. Thus, at each wavelength, gain isn’t uniform and so its spectrum is an important EDFA feature that if it’s used in a WDM communication system makes different levels of amplification for each channel of WDM signal.

The gain can be written as

(4.1)

On the other hand, the absorption coefficient, is given by (4.2)

where is the absorption cross section. The emission coefficient, , is given by

(4.3)

where is the optical confinement factor, is the emission cross section, is the erbium ions total concentration and .

4.2. Simplified Model for EDFA with optimal length

In this section it’s presented a simplified model to solve a particular case of an EDFA with an optimal length and a single signal, besides pumping.

The optimal EDFA length, , is the value for which gain reaches its maximum value for a given pumping power, such that

|

(4.4)

It’s possible to obtain the optimal length by

( )

(4.5)

Pumping and signal power variations along the EDFA are shown in Figure 4.1, and an optimal length of is obtained to in and in .

Figure 4.1. Pumping and signal power variation along EDFA length

considering and .

By Figure 4.1 it’s seen that pumping power decreases as fibre length increases instead of signal power that shows a sharp increase with the increasing EDFA length. The signal power reaches its maximum value to a fibre length of which represents the optimal length for a signal of .

4.3. Spectral characterization

As mentioned above, the EDFA gain is not uniform, since it depends on the fibre length, on the wavelength and on the input power, and therefore the spectral characterization is one of the EDFAs fundamental aspects.

Figure 4.2. Evolution of gain coefficient as function of fibre length

and wavelength.

Observing Figure 4.2 it’s possible to verify that the EDFA gain shows variations along the amplification which are more significant between fibre lengths of and wavelegths of and also for fibre lengths from and between wavelengths of .

The EDFA gain can be written as ∫

(4.6)

In Figure 4.3 is shown the EDFA gain evolution.

Figure 4.3. Gain spectrum as function of wavelength.

It’s verified that gain increases with wavelength up to and from until reaching its maximum value at about . After reaching this value it decreases gradually in the remaining wavelengths.

Having been study the EDFA gain and obtained the optimal length for a transmission signal at , the output power of a WDM signal for four channels centred at , , and in an EDFA with length and all with the same input power ( ) and a pumping power of at , it’s obtained through the differential equation

(4.7)

with

∑

( ∑

) (4.8)

Figure 4.4. Evolution of a WDM signal output power along EDFA

length.

From Figure 4.4 it’s possible to infer that although input powers are the same, output powers at different wavelengths differ due to the gain variation over wavelength (Figure 4.2). Moreover it’s seen that pumping power decreases gradually along the length while the power of the four channels centred at different wavelengths increases up to the maximum value of each channel and then it starts decreasing.

4.4. Amplified Spontaneous Emission (ASE)

Until this section noise had been neglected, however, it existence due to spontaneous emission (ASE) is one of the EDFAs negative aspects.

The ASE noise average power is given by (4.9)

But, when there is a total population inversion the spontaneous emission factor is equal to one and considering , it’s obtained (4.10)

where is the bandwidth. Thus, it’s concluded that the noise associated with ASE is

minimal when there is a complete population inversion. EDFA’s noise figure is the ratio between the signal-to-noise

ratio (SNR) at the input and the SNR at the output and it can be written as

(4.11)

where is the equivalent noise figure at the input.

Figure 4.5. Gain and noise figure evolution as function of pumping

power.

From Figure 4.5 it’s seen that gain and noise figure are the inverse of each other, i.e., as pumping power value increases, gain shows an increasing growth rather than noise figure that is decreasing. A higher gain value due to total population inversion implies that as gain increases, noise figure tends to its minimum value.

5. RAMAN AMPLIFIERS

The Raman amplification effect is achieved by a nonlinear interaction between the signal and the laser inside the fibre.

Raman amplifiers can be classified into two types: Distributed Raman Amplifiers: amplification is

done throughout the transmission fibre and it has lengths greater than 40 km.

Discrete Raman Amplifiers: amplification occurs in an isolated fibre range and it has lengths around 5 km.

The main Raman amplification advantage is its ability to provide distributed amplification within the fibre, thereby increasing length between amplification and regeneration sites. The Raman amplifiers bandwidth is set by the pumping wavelengths that are used and so amplification can be provided in wider areas than by using other amplifier types which depend on dopants or on the device design to set the amplification window. On the other hand, Raman amplification can be used to improve system noise figure and also to improve gain flatness.

5.1. Stimulated Raman Scattering (SRS)

Raman amplification relies on the stimulated Raman scattering effect which is a fundamental nonlinear process that turns optical fibres into broadband Raman amplifiers.

Spontaneous Raman scattering can transfer a small fraction of power from an optical field to another which frequency is reduced by an amount determined by vibrational modes, in any molecular medium. From a practical perspective, incident light acts as a pump to generate radiation and so incident photons provide part of its energy to photons with a lower energy level, being this process assigned Stokes. On the other hand, an Anti-Stokes process is when photons absorbed energy of the incident photon such that its energy level becomes higher than the energy level of incident photons.

Figure 5.1. Raman scaterring process representation.

The Raman scattering process becomes stimulated if the pumping power exceeds the threshold value and it can occur in both forward and backward directions depending in what direction the pumping is done.

5.2. Raman Gain Spectrum

The most important parameter that characterize Raman amplifiers is the gain coefficient which is related to the spontaneous Raman scattering cross section and describes the stokes power evolution as the pumping power is transferred by the SRS [8].

The non-linear interaction between stokes and pump waves under the continuous wave (CW) conditions gives the Raman amplifier gain and it’s governed by

(5.1)

(5.2)

where e represent the fibre losses for the stokes and pump frequencies, respectively and takes the value depending on the pumping configuration, i.e. for backward pumping and for forward pumping. The

frequency ratio is due to the different energy values of the pumping and the signal photons.

The Raman gain spectrum , where is the frequency shift between stokes and pump waves, is shown in Figure 5.2.

Figure 5.2. Raman gain spectrum as function of frequency.

This spectrum exists over a broad frequency range (up to 40 THz) with a gain peak at around 13 THz.

The amplification factor is given by

(5.3)

where the input pumping power is , the effective length is and is the effective area.

5.3. Performance Limiting Factors

5.3.1. Spontaneous Raman Scattering

Spontaneous Raman scattering contributes to the amplified signal and appears as noise due to the random phases associated with photons generated spontaneously. This noise mechanism is similar to the amplified spontaneous emission which affects EDFA performance but, in Raman it depends on photons population in the vibrational state.

Considering an optical filter with a bandwidth of it’s possible to calculate the total ASE power after the amplifier by

∫

(5.4)

where ∫

and the factor of 2

represents the two fibre polarization modes. The noise figure is written as

∫

(5.5)

with and ∫ ( ) .

5.3.2. Rayleigh Backscattering

Rayleigh backscattering affects Raman amplifiers performance, firstly, because part of the backward-propagating noise may appear in the propagation direction increasing total noise and secondly, because double Rayleigh backscattering creates a crosstalk component in the propagation direction that is amplified by the Raman gain and becomes the main source of power penalty in Raman amplification systems.

Having

(5.6)

[ ] (5.7)

[ ] (5.8)

where and represent the average power levels of the noise components created through single and double Rayleigh backscattering, respectively.

The input power fraction that comes out the signal at the fibre end ( ) is obtain by integrating equations (5.6), (5.7) and (5.8) along the fibre such that

∫

∫

(5.9)

where is the Rayleigh scattering coefficient, is the Raman gain in a distance in the forward-propagating amplifier of length .

5.3.3. Pump-Noise Transfer

All lasers have some intensity fluctuations and in the semiconductor lasers used on Raman amplifiers pumping the energy fluctuations level becomes higher due to their reduced size and also due to their higher spontaneous emission rates. This power is quantified by an amount dependent on frequency and it’s named relative intensity noise (RIN).

On the other hand, knowing Raman amplifier gain changes exponentially with the pumping power, it’s expected that any pumping power fluctuation is amplified and results in an amplified signal with higher power fluctuations.

The pump-noise transfer function represents a signal noise enhancement at a specific frequency and it’s defined as

(5.10)

in which

⟨ ⟩

∫

(5.11)

⟨ ⟩ ∫

(5.12)

6. CONCLUSIONS

In this work it’s attempted to understand several aspects related with signal emission and study the behaviour of three technologies used in signal amplification over a fibre optic link, highlighting their main parameters.

In Chapter 2, it’s carried out a study of semiconductor lasers operation and it was assumed that the semiconductor laser operates in single-mode regime, under CW conditions with a constant injection current and in steady state. In thermal equilibrium it isn’t possible that emission predominates over absorption, so to have population inversion it’s necessary a pumping process through injection current such that emission predominates over absorption. In a first approach, in which the injection current is higher than the threshold current ( ) the laser is emitting and it was concluded that carrier and photons populations have an oscillatory behaviour over time, however, as carrier ( ) and photons ( ) lifetime values are decreasing, oscillations becomes smaller and an almost linear profile it’s obtained for and . It was found that if decreases the laser response becomes faster and therefore the lower are values the less time it takes to stabilize at a steady value and the smaller becomes the photon population delay compared to that of carrier population. On the other hand, as decreases laser response gets less relaxation oscillations, improving its performance. Afterwards, it was concluded that for the pulse is ultra-short and it’s also possible to verify that the higher the gain, the greater are relaxation oscillations’s amplitude and frequency. When the injection current value is lower than threshold current ( ), there isn’t population inversion and therefore the laser only emits in a short period of time. In this case, it was concluded that the carrier population takes longer to reach the threshold value and thus the photons population response is also slower. Furthermore it was found that when the current pulse ends, the carrier density tends to while the photon density tends to zero. Finally, concerning gain variations, it was concluded that the laser behaviour was identical to the one seen for the first case.

In Chapter 3, it’s studied semiconductor optical amplifiers and it’s found that even with antireflection coatings some residual reflectivities that appears are result from gain oscillations, overlapping its spectrum. Moreover, it was found that gain fluctuations increases with increasing gain as well as increasing reflectivity. Chirp profiles for several values of the input pulse energy when a Gaussian pulse is amplified by SOA were analysed and it was seen that the frequency chirp is larger for the more energetic pulses. In the pulse shape it was observed that for temporal changes dependent on the amplifier gain are quite significant and it becomes asymmetrical as the unsaturated gain value increases. Analysing the amplified pulse spectrum, it was found that the dominant peak is at the pulse front and has been shifted toward the red side, becomes broader than the input spectrum and it’s also may be accompanied by one or more satellite peaks. By

applying several values of input power and confinement factor it was concluded that an increasing ratio makes appear satellite peaks in the spectrum and cause a stronger red shift. Moreover, systems with lower confinement factors let for the same output power better beam quality and higher saturation power than systems with higher confinement factors.

In Chapter 4, the EDFA amplification process was described. It was seen than gain isn’t uniform once it depends on the fibre length, on the wavelength and on the amplifier input power and so its spectrum is a fundamental characteristic. The behaviour of the pumping and the signal powers it’s also analysed and it was found that as the fibre length increases the pumping power decreases rather than the signal power that displays a sharp rise. By analysing the gain evolution over the fibre length and the wavelength it was concluded that EDFA gain variations introduced during amplification are more significant for fibre lengths of and wavelegths of and also for fibre lengths from and between wavelengths of , whereas the gain peak value it’s obtained at approximately . Finally, the amplified spontaneous emission influence was studied and it was found that as the pumping power value increases, there is a gain increasing while noise figure is decreasing.

In Chapter 5, the Raman amplification process that is based on the stimulated Raman scattering effect to transfer optical pumping power, in a controlled way, through a given wavelength was studied. Then, Raman gain spectrum was analysed and it was found that the spectrum extends over a broad frequency range (up to 40 THz) with a peak located around 13 THz, being its bandwidth used to compensate fibre losses. It was also concluded that to obtain an optimum performance the frequency shift between the signal and the pump beam must be within the Raman gain spectrum bandwidth (around the peak value, 13 THz).

In Table 6.1 is shown a comparison of the three amplification technologies studied in this work.

SOAs offer certain advantages over EDFAs and Raman amplifiers as lower power consumption, smaller size and lower cost. On the other hand, they have lower gain values and higher noise figure. This type of amplifier is often used in telecommunication systems for optical switching, optical signal processing and wavelength conversion and regeneration.

The main EDFA’s advantage is its high gain bandwidth which allows a single EDFA amplify simultaneously multiple channels of different wavelengths into the active region with high transmission rates. Contrary to SOAs, EDFAs have higher gain and lower noise figure, however, their size is bigger and require higher pumping power. These amplifiers may be used in any configuration either as power amplifiers, preamplifiers or line amplifiers and are more attractive in optical fibre systems once they operate around where fibres have minimum loss values and a reduced noise level [9].

Finally, Raman amplifiers are EDFA competitors once they have a low noise figure and a broad gain bandwidth. However, gain is reduced though they require higher pumping powers. These amplifiers are used both as pre-amplifiers and power amplifiers.

Table 6.1. Comparison of the three amplification technologies [9]. SOA EDFA Raman

Amplification Band

Depends on pump

power Depends on

dopant Depends on pump power

Gain Bandwidth

[nm] per

pump

Pump Wavelength

[nm] Electrical

pump shorter than amplified signal range

Pump Power [mW]

Saturation Power

Depends on current

Depends on dopant and

gain pump power

Noise ASE ASE

Raman Scattering, Rayleigh

backscattering and RIN

Noise Figure [dB]

Direction Uni-directional

Uni-directional Bidirectional

Complexity Simple Complex Simple Cost Low Medium High

BIBLIOGRAFY [1] F. Corporation, Ed.“White Paper: Introduction to Optical

Amplifiers,” Junho 2010. [2] C. R. Paiva, Fibras Amplificadoras Dopadas com Érbio, Instituto

Superior Técnico, 2002. [3] C. R. Paiva, Lasers Semicondutores, Instituto Superior Técnico,

2008. [4] H. K. Hisham, G. A. Mahdiraji, A. F. Abas, M. A. Mahdi e F. R.

M. Adikan, “Characterization of Transient Response in Fiber Grating Fabry–Perot Lasers,” IEEE Photonics Journal, vol. 4, December 2012.

[5] J. M. Senior, Optical Fiber Communications: Principles and

Practice, Third ed., Prentice Hall, 2009. [6] G. P. Agrawal, Fiber-Optic Communications Systems, Third ed.,

John Wiley & Sons, 2002. [7] A. G. a. S. Balle, “Influence of the Confinement Factor on the

Wavelength-Dependent Output Properties of a Tapered Traveling-Wave Semiconductor Amplifier,” IEEE PHOTONICS

TECHNOLOGY LETTERS, vol. 11, n.º 11, Novembro 1999. [8] G. P. Agrawal e C. Headley, Raman Amplification in Fiber

Optical Communication Systems, Elsevier Academic Press, 2005. [9] B. Utreja e H. Singh, “A review paper on comparison of optical

amplifiers in optical communication systems,” Canadian Journal

on Electrical and Electronics Engineering, vol. 2, n.º 11, Novembro 2011.