Erbium Doped Fiber Amplifiers

Part I Gain Flattening EDFA Simulations Part II Designing L-Band Amplifiers in WDM Applications Part III EDFAs with Optical Automatic Control Simulations Part IV- Gain-Clamped EDFAs Simulations

Part I Gain Flattening EDFA Simulations

Outline
Why Gain Flattening - Techniques to flatten the gain bandwidth - Typical configurations Key points in designing Gain Flattening amplifiers - Gain Flattening optimizations - Gain Flattening filters Simulating Gain Flattening Amplifiers Evaluating Gain-Flattened EDFA in a WDM system Conclusions

Why Gain Flattening

Critical: gain bandwidth Enlarge and achieve a flat amplification bandwidth Increase the number of input channels / transmission capacity Operating band: C- and L-band amplifiers Applications: - WDM transmission systems - All-optical self-routed wavelength addressable networks

Techniques to Flatten the Gain Bandwidth

Modifying the material composition in the EDF
- silica with different co-dopants - fluoride - telluride Using optical filters to compensate for the variations in the gain spectrum Gain equalizers - fiber Bragg gratings - fiber acousto-optic tunable filters - Mach-Zehnder filters Series/cascade and parallel configuration

Typical Configurations
Multiple fiber stages Element mid-stage: filter or isolator
Series Configuration

Typical Configurations
Parallel Configuration

Key points in designing Gain Flattening Amplifiers

Gain spectrum of a conventional EDFA is not constant over the bandwidth; Non-uniformity of gain in an EDFA is a critical limitation with multiple channels; If there are more than 16 channels, the gain spectra have to be flattened by external methods; Filters are designated to approximate the inverse profile of the gain spectrum.

Gain Flattening Optimizations

Gain flattened filter allows the user to select two points in the layout with a filter between them. The optimization routine will try to find the optimum filter shape to obtain the gain flattening between these two points. In one simulation the user can select multiple optimizations; the user can also mix parameters iterations and optimizations.

Using Gain Flattening Filters

Filters are designed to approximate the inverse profile of the gain spectrum; The gain flattened filter is usually placed between a multiple-stage amplifier (typically two-stages); Optimization is over at least 3 variables: - stage 1 of EDF; - stage 2 of EDF; - filter shape; Different filter optimizations can be accomplished.

Simulating Gain Flattening Amplifiers

Main considerations

Identifying the basic system parameters; Calculating fiber length and gain shape; Determining influence on system performance; Comparing amplifiers performance with and without the gain flattening filter.

Layout of two-fiber stages EDFA

Details of the Project Signal Input

22 Signal Input Channels

s
1528.79 nm to 1562.25 nm spaced by 1.56 nm -20 dBm/channel

Details of the Project EDF

Details of the Project Gain Flattening Filter

Tolerance specifies how flat the gain will be

Details of the Project Receiver

Simulation Results

Flattened Amplifier Output

ASE Shape and Filter Curve

6 5

Intensity (dB)

Optimized Filter Curve

4 3 2 1 0 1520

1530

1540

1550

1560

1570

Wavelength (nm)

Layout with No Filter

Simulation Results

Signal Output at the Receiver

Comparing Results With and Without Filter

Evaluating Gain Flattened EDFA in a WDM system

Using a black-box model Evaluating a multi-span WDM system Using multiple signal input Gain flattening filter optimization

AIM Black Box Component

Signal Observed Along the System

After Flattening the Gain

WDM Network

Conclusion
Gain Flattened Amplifiers simulations Evaluation of the amplifier performance along the system including gain flattening filter optimization Flexibility to include different types of Erdoped fibers, different configurations

Part II Designing L-band Erbium Doped Fiber Amplifiers in WDM Applications

Outline
Overview of basic characteristics of L-band
EDFAs How to obtain long band amplification Simulating L-band amplifiers Fitting experimental results Facilities to split band selection Conclusions

Overview on Basic Characteristics of L-Band EDFAs

Increasing interest in L-band amplifiers comes from the expanding demand for larger number of WDM channels Designing L-band (1570 nm 1610 nm) fiber amplifiers is a simulation challenge Convergence of the numerical algorithms used to solve rate and propagation equations is slower and less stable than for C-band EDFAs (1520 nm to 1570 nm). Configurations: single and multiple Er-doped fiber stages

How to Obtain Long Band Amplification

Main reason to extend the amplification to longer signal wavelengths: average inversion ~ 40 % Conventional band EDFAs: inversion > 70%

Average Inversion to Different Signal Wavelength

Simulating L-Band Amplifiers

Designing amplifier including elements to be considered in a possible experimental evaluation Passive elements settings Choose to include wavelength dependence or not in the passive components settings One-stage Configuration

Inversion Along the EDF

High Pump Absorption ASE generated in the first 10 m of EDF is observed in this region

Propagating Powers Along the EDF

Signal ASE+

(dBm)

ASE Pump

Calculated Gain and NF

20 6.0 5.5

Gain (dB)

15 5.0 4.5 10 4.0 3.5 5

NF (dB)

One-Stage L-Band Amplifier

3.0 2.5

1580

1585

1590

1595

1600

1605

1610

Wavelength (nm)

Fitting Experimental Results

Adjustments performed to C-band amplifiers typically show an error rate of 5% or less when simulated results are compared with experimental data. If L-band amplifiers are considered, agreement between simulated results and experimental data is high. However, the error rate after comparing L-band simulated results with experimental data can be 3% higher than that observed for C-band amplifiers. Adjustments performed to L-band amplifiers that have component settings including wavelength dependence are critical to better match the experimental data.

Critical Points in the Adjustments

1) Correct evaluation of the return loss of the components in the front end of the erbium fiber is fundamental to attempt correct predictions about the performance of the EDFA 2) Inherent uncertainties due to the method used to measure cross sections can be prohibitive to running simulations with precision better than 5%

Experimental Data Adjustments

L-Band Amplifier Setup in One-EDF Stage
-20 dBm 70 m

1480 nm, 110 mW

Setting Different Return Loss in the Coupler at the Fiber Input*

35 30 25 20 15 10 5 0 1560 1570 1580 1590 1600 1610 1620

Gain (dB)

20 dB 30 dB 41 dB 55 dB Measured

Wavelength (nm)

Nilsson et al., IEEE Photon. Techn. Techn. Lett. Lett. Vol. 10, No. 11, p. 1551, 1998.

Gain versus Coupler Return Loss

30

Gain (dB)

25

20

15

10

1560 nm 1590 nm 1580 nm 1570 nm

5 20

25

30

35

40

45

50

55

Coupler Return Loss (dB)

Changing RL in the 2nd Isolator

25

Return Loss

Gain (dB)

30 dB 40 dB 60 dB

20

15

Last Isolator

1560

1570

1580

1590

1600

1610

1620

Wavelength (nm)

Cross-Section Spectra
1.5x10
-25

6.0x10

-25

Cross Sections (m )

1.0x10

-25

4.0x10

-25

1580

1590

1600

2.0x10

-25

1450

1500

1550

1600

1650

Wavelength (nm)

Testing Small Variations in the Cross-Section Values

32 Original +10 % +5 % +3 % 3 % 5% 10 %

Gain (dB)

28

24

20

Return Loss = 41 dB

1560

1570

1580

1590

1600

1610

Wavelength (nm)

Testing Small Variations in the Cross-Section Values

30
+3% +5% 10% 3% 5% +10%

Gain (dB)

25

20

Original

15

Return Loss = 55 dB
10 1560 1570 1580 1590 1600 1610 1620

Wavelength (nm)

Fitting Experimental Data with Modified Cross-Sections

30

RL 41 dB Gain (dB)
25

measured 41 dB +5% 41 dB 55 dB + 3% 55 dB + 5% 41 dB + 3% 55dB

20

15

RL 55 dB

10 1560 1570 1580 1590 1600 1610 1620

Wavelength (nm)

Including Temperature Dependence

Facilities to Simulate L-Band Amplifiers

Split band selection with optimization of the amplifier performed in two different regions separately in the same project (C-band and LBand for example), Design of component to simulate this particular situation.

Splitting/Selecting C-Band and L-Band

Splitting Band

Conclusions
Long band EDFAs design with a WDM signal source coupled to the amplifier is discussed; Simulating amplifiers with different configurations, considering L-band with clamped, automatic control, L-band with gain-flattening; Including the losses in the simulated layout is critical to adjusting the simulated results with experimental data; Cross-section input files; Temperature dependence; ESA effects.

Part III EDFAs with Optical Automatic Control Simulations

Outline
Overview of Automatic Control Techniques
Automatic Gain Control (AGC) Automatic Power Control (APC) Automatic Peak Power Control (APPC) Conclusions

Overview of Automatic Control Techniques

Functionalities added to the EDFAs in order to achieve stable amplification for actual systems applications; Dynamic compensation of low frequency gain or output power fluctuations; Implementation: fully optical or electronic; Control fast power transients in optical networks.

Basic Configurations for AGC in EDFAs

1. Feedforward with Pump Control
Signal Out

2. Feedforward with Auxiliary Signal Control

Signal Out

Auxiliary Signal

Basic Configurations for AGC in EDFAs

3. Feedback with Pump Control
Signal Out

4. Feedback with Auxiliary Signal Control

Signal Out

Auxiliary Signal V

Basic Methods for Implementation of APC in EDFAs

1. Pump Feedback Loop
Signal Out

V 2. Tandem EDFAs
Signal Out

Basic Methods for Implementation of APC in EDFAs

3. Saturable Feedback
Signal Out

4. Ring Laser

Signal Out

Automatic Gain Control (AGC)

How to accomplish: Pump power varies to maintain the
approximately constant gain

How to simulate AGC with steady-state solution:

Select a parameter and a result in user-defined optimizations; The parameter is changed through a number of iterations to hit the target value of the result the target gain or output power.

Amplifier with Automatic Gain Control

Keep the total gain constant thereby controlling the pump power. The probe calculates the gain and an optimization to determine the pump power necessary to achieve this gain. Changing the pump power and controlling the gain between the two isolators.

What is important to specify when starting calculations

Optimized Pump Power

Total signal input power : 13.98 dBm Gain : 24.84 dB Pump power : 31.875 mW

Optimized pump power

Checking Results
35 30 25 20 20 15 10 -14 -12 -10 -8 15 25

Power (dBm)

Gain (dB)

Pump Ps
out

10 -6

Signal Input Power (dBm)

Automatic Power Control (APC)

Function: Maintains the EDFA output power at a fixed level during signal perturbation

How to accomplish: Pump power varies to maintain approximately constant output power How to simulate APC with steady-state solution: Insert a probe in the layout which calculates the signal output power, and set an optimization to calculate the pump power to achieve the target signal output power

Specifying APC Simulation

Parameter to adjust: Power from Pump Laser;

Result to optimize: Total signal power from Probe; Target value (desired output power): 10 mW; Tolerance: 0.5 mW.

Running Simulations
Total Input Power: Total Output Power: 13.98 dBm 10.08 dBm

Optimized pump power

Calculated Total Output Power

Calculated Total Power

25 25

Power (dBm)

Gain
20 20

Pump Power Output Power

15 15

Gain (dB)

10 10 5 -14 -12 -10 -8 -6 -4

Signal Input Power (dBm)

Automatic Peak Power Control

How to simulate APPC:
Monitor the output power of one signal at a particular wavelength.

Adjustable Parameter:
The output power will be maintained constant thereby controlling the pump power.

Focusing the Peak Power:

The probe tool must export the results to be used by the optimization. In this case we are interested in one signal power.

Setting the Optimization

Adjusting: Power from Pump Laser Selected Result to optimize: First signal power from Probe Target value: 10 mW Tolerance: 0.5 mW

Checking Optimized Power

Signal Input Power: Output Power: Pump Power: 20 dBm 10 dBm 82.5 mW

Optimized pump power

Calculated Output Power

- Signal input power from 20 dBm to

10 dBm; - Trying to maintain 10 dBm of signal output power; - Output power to channels 1, 2, 3 and 4.

APPC Results
150 30

Power (dBm)

25

Gain (dB)

100 20 15 10

Pump P1530 nm
5 -20 -18 -16 -14 -12

10

5 -10

Signal Input Power (dBm)

Conclusions
Simulating EDFAs with optical automatic control; User-defined optimizations allow you to select a parameter and a result; Vary the pump power to maintain the gain or output power at approximately constant values; Different parameter selection is possible.

Part IV Gain-Clamped EDFAs Simulations

Outline
Introduction Gain Clamped Function How to Simulate a Gain-Clamped Function Simulating a Gain-Clamped L-Band Amplifier Conclusions

Introduction
The gain spectrum of optical amplifiers depends on the input power. In WDM networks, the gain of the EDFA must be the same at any input power over a desired range. The EDFA gain-clamped function can be carried out using optical or electrical control. Example: A simple all-optical gain control system in which reflective filters are placed at both ends of the doped fiber.

Gain-Clamped Function
Objective: Make the gain less dependent on the input signal power Application:
Useful in WDM networks, where the gain of EDFA must be the same at any input power, over a desired range

How it works: Reflective filters are placed at both ends of the doped fiber
The created cavity generates a laser oscillation at the reflection wavelength of the filters The stimulated emission clamps the average population inversion of Er atoms The saturation fixed by the lasing effect clamps the gain at any other wavelength

How to Simulate a Gain-Clamped Function

Subramaniam et al. IEEE Photon. Techn. Techn. Lett. Lett. Vol.13, No. 8, 785, 2001

Gain-Clamped Including Two Reflective Filters

Components Settings
Signal Input In most studies of C-band GC-EDFA based on single-channel operation [Kobayashi, Electron. Lett. 27, p.486, 1999; Inoue, IEEE Photon. Techn. Lett. 11, p.1108, 1999]. Or saturation operation [Takushima et al. IEEE Photon. Techn. Lett. 9, p. 271, 1999; Luo et al. IEEE Photon. Techn. Lett. 9, p.1346, 1997]. L-band GC-EDFA single and multiple channel operation [Hsu et al. Opt. Commun., Jul 2001].

Components Settings
Erbium Doped Fiber Silica erbium doped fiber Typical fiber length ~ meters in multiples of ten Single and Multiple fiber stage

Reflective Filter Settings

Center Wavelength defines lasing frequency of the cavity

Bandwidth defines linewidth

Gain Results

9.94 dB ( 4.95 dBm) 9.87 dB ( 34.95 dBm)

x x
G = 0.07 dB

Signal Input [mW]

Laser Oscillation Generated at 1520 nm

Repeating Simulations without Reflective Filters/Clamping

Gain Results with No Clamping

15.11 dB ( 34.95 dBm)

x
13.49 dB ( 4.95 dBm)

G = 1.62 dB

Signal Input [mW]

Simulating a Gain-Clamped L-Band Amplifier

Basic principles: - Develop optimum design of GC L-Band EDFA to achieve constant gain variation under input power variation - Use Single and Multiple-channel schemes - Determine best selection of wavelength lasing which offers large gain-clamping dynamic ranges and low NF characteristics

Configuration of Optically GC L-Band EDFA*

Gain-Clamping Loop

(L) 1.48 m

1.48 m

*S. Hsu et al., Opt. Commun. Issue 5-6, Vol. 19S, Sept. 2001

Non-Clamped Optical Gain and NF

10 35 30 25 20 4 15 10 0 20

Single Channel Operation: 1588 nm

Ps = 15 dBm
8

in

Gain (dB)

NF (dB)

Maximum Gain at L = 50 m

Gain at 50 m: 34.88 dB Noise Figure at 50 m: 4.17 dB

40 60 80

EDF Length (m)

Best Selection of Loop Attenuation L=1565 nm

Single Channel Operation
45 40 20

L = 1565 nm
100 dB 10 dB 5 dB 0 dB

Attenuation

Gain (dB)

35 30 25 20 15 10 5 -35

100 dB 10 dB 5 dB 0 dB

18 16 14 12 10 8 6 4 2

NF (dB)

-30

-25

-20

-15

-10

-5

Input Signal Power (dB)

Best Selection of Loop Attenuation L=1610 nm

10 40

L = 1610 nm
8 100 dB 20 dB 19 dB 18 dB

Gain (dB)

35 30 25 20 15 -35 -30 -25 -20 -15 -10 -5

NF (dB)

100 dB 20 dB 19 dB 18 dB

Input Signal Power (dBm)

Multiple Channel Based Design

The Characteristics of Spectral Gain in the L-Band EDFA
30 28

Gain (dB)

Flatter gain between L = 35 m and L = 40 m

26 24 22 20 18 16 14 1570 1575 1580

Fiber Length

20 m 30 m 40 m 50 m

25 m 35 m 45 m

1585

1590

1595

1600

Wavelength (nm)

Selecting the Flatter Gain Between L = 35 m and L = 40 m

25 24 23 22 21 20 1570 1575 1580 1585 1590 1595 1600 1605

Flatter gain at L=37.5 m (G = 0.9 dB)

Gain (dB)

Fiber Length

35 m 37 m 38 m 40 m

36 m 37.5 m 39 m

Wavelength (nm)

Spectral Gain Observed Considering Different Attenuations

24 22 20 18 16

G < 1 dB

32 Chanels

Gain (dB)

Attenuation
14 12 1570 1575 1580 1585 1590

100 dB 20 dB 15 dB 10 dB 5 dB
1595 1600

Wavelength (nm)

Considering 16 Channels
28 27

Attenuation

Gain (dB)

26 25 24 23 22 21 20 1570 1575 1580 1585 1590 1595

16 Channels

100 dB 20 dB 15 dB

G = 1.7 dB

1600

Wavelength (nm)

Conclusions
Gain-clamped amplifiers simulations considering different configurations; Multiple signal analysis; Important applications in DWDM broadband systems.