Professional Documents
Culture Documents
DOI 10.1007/s12596-015-0289-y
RESEARCH ARTICLE
Received: 30 September 2014 / Accepted: 5 May 2015 / Published online: 2 November 2015
The Optical Society of India 2015
Lutz Rapp
lutz.rapp@coriant.com
1
Introduction
Erbiumdoped fiber amplifiers (EDFAs) have paved the
way to the introduction of wavelength division multiplexing
(WDM) in optical communication systems [1] which has
become the basic technology for static and dynamic wavelength routed networks [2]. Such kind of networks provide
a cost efficient solution for content distribution and allow to
respond rapidly to changing traffic demands. However, nonlinear fiber effects and the nonideal dynamic properties of
EDFAs make their performance sensitive to component failures, fiber breaks, or protection switching since these events
are leading to unpredicted changes in optical power. Resulting power variations can propagate to other sites inducing
optical power fluctuations across the whole network and
possibly to oscillations. As a consequence thereof, surviving channels and even channels that are not transmitted
over affected multiplex sections may suffer from performance degradation at the receivers [3]. The most important
impact stems from signal distortions induced by nonlinear
effects in the transmission fibers, power excursions beyond
the dynamic range of the optical receivers, and deterioration
of the optical signal-to-noise ratio (OSNR) [4].
Gain variations have been shown to accumulate in a
cascade of amplifiers and the speed of the transients can
increase along the chain [5]. Although intelligent wavelength assignment policies [6] limiting the maximum power
excursions have been proposed, requirements in view of
amplifier control are still quite stringent. Thus, efficient
amplifier control techniques precisely stabilizing the inversion and the gain profile of an amplifier are mandatory.
Sophisticated alloptical control techniques have been
developed [7, 8] and even have been combined with traditional control concepts [9, 10]. But fast electronic control
210
architectures are still the most economical solution to stabilize the gain of EDFAs [11]. Typically, feedforward and
feedback control techniques are combined [12]. The fast
feedforward control reacts quickly to the power changes and
prevents large gain deviations, but some permanent variations are unavoidable due to inaccuracies of the used models, aging effects, and intrinsic effects. These deficiencies
are compensated by the slow feedback system cleaning up
for any error in the predetermined adjustment made and thus
helping to recover the original gain of the amplifier over
time. This combination allows to make the feedforward control robust against aging effects and changing environmental
conditions by continuously updating the control parameters during operation [13]. On an again larger time scale,
corrections are done by the link control including the continuously running preemphasis [14]. Some indications on
the required bandwidth of the control circuits can be found
in [15].
If a change of the input power is detected, the feedforward control sets the pump power preferably immediately
to a new power level. In combination with the altered signal
powers, this new power level leads to the same steady
state inversion level existent before the power change. In
order to achieve optimum results, the prediction of the
new pump power level should be as accurate as possible
and almost immediate. But in some cases even this is not
sufficient.
This paper deals with effects affecting the performance
of the feedforward control for EDFAs. Methods and setups
eliminating or at least reducing the impact of these effects
are presented. The aim is to give an overview over available
techniques and to explain key concepts. Details can be found
in the referenced documents. First, tasks of the feedforward control and underlying principles are described, before
disturbing effects are considered. A mathematical model
gives deeper insight into the wavelength dependence of the
optimum pump power setting. The following section provides an overview over techniques taking this dependence
into account, whereas the subsequent section treats of timing issues in different kind of amplifier setups. After some
closing remarks, conclusions are drawn.
211
Conservation of gain
Surviving channels
Gain
Average gain
before drop
Pure gain
control
Wavelength
Compensation of SRS tilt change
Surviving channels
Gain
Power exchange
via SRS before
drop
Tilted spectrum
before drop for
compensation
of SRS tilt
Decrease
of VOA
attenuation
Spectrum after drop with
reduced average inversion
for unchanged VOA setting
Wavelength
Fig. 1 Effect of gain ripples and type of control on channel gain.
The lower diagram shows tilt changes of the gain spectrum required
to compensate for the variation of the SRS induced power exchange
among the channels after channel drop
Functional principles
In principle, the feedforward control can act on different
physical parameters to influence the amplifier gain. Most
implementations alter directly the injection currents of the
pump laser diodes in order to adapt the power supplied to
the EDF to the new situation. However, it is also possible to
vary the gain of a variable optical attenuator (VOA) included
in the optical signal path [21]. Typically, this approach will
be combined with the pump power control. Furthermore,
212
15
10
0.00 ms
0.25 ms
0.50 ms
0.75 ms
1.00 ms
1.25 ms
1.50 ms
1.75 ms
2.00 ms
2.25 ms
2.50 ms
Delay
5
0
0
3
Time [ms]
15
Delay
2
0
0
3
Time [ms]
Gainvariation[dB]
10
0
0
0.5
1
Delay[ms]
1.5
Fig. 4 Peak magnitude of gain variations induced by delayed adaptation of the pump power for several input powers measured before the
drop takes place
213
Sensitivity [dB/mW]
10
10
Gain
10
20
sat
=
psig
10.0 dB gain
15.0 dB gain
20.0 dB gain
25.0 dB gain
30.0 dB gain
15
10
10
15
20
Mathematical model
In order to provide insight into the impact of the signal and
pump wavelengths on the feedforward control parameters,
an analytical expression providing the required pump power
as a function of the input power is derived. Starting point
of the considerations is the model for two involved energy
levels presented by Saleh [33] that provides accurate results
as long as the saturation by amplified spontaneous emission
(ASE) can be neglected. Commonly, it is assumed that this
condition is fulfilled for gain values up to 20 dB [34]. For
the sake of a compact representation, we make use of the
socalled Giles parameters [35] and photon fluxes indicating the number of photons per time unit are used instead of
power levels.
The fundamental finding from the model of Saleh is that
the gain provided by an erbiumdoped fiber (EDF) of length
LEDF to a signal only depends on some fix or wavelength
dependent parameters and the total number of incoming
EDF
e ,
+ sig
(2)
a
sig
pin pout
2nd
N pop
=
EDF LEDF
(3)
= a
EDF
ref
ref
e
ref + ref
(4)
In the following, the amplification of a WDM signal consisting of N sig signals is considered. The required pump
p
energy is supplied by a single pump with photon flux pin .
Since the model only refers to the total number of photons
launched into the fiber, the results are equally applicable to
codirectional and counterdirectional pumping schemes. The
sig
different signals are coupled with a photon flux of pk into
the fiber, and the total number of signal photons provided to
sig, tot
with
the fiber is described by the photon flux pin
sig
sig, tot
pin
N
sig
(5)
pk .
k=1
sig
sk =
pk
sig, tot
(6)
pin
sig, tot
= pin +
N
k=1
sig
pk ,
(7)
214
sig, tot
+ Cff
(8)
Sff =
k=1
c
a + e
1 exp ka LEDF + p
k
k
EDF
c
1 exp pa LEDF + p
pa + pe
EDF
50.0% | 32.0 m
52.5% | 24.9 m
55.0% | 20.4 m
57.5% | 17.3 m
60.0% | 15.0 m
62.5% | 13.2 m
65.0% | 11.8 m
67.5% | 10.7 m
70.0% | 9.8 m
5
lin
Normalized slope S ff /G ref
expressing the output fluxes as a product of the input photon fluxes and the corresponding gain according to this
equation, and applying some basic mathematical operations
leads to the linear equation
sig
sk
4
3
2
1
0
1530
1540
1 exp pa LEDF +
pc
EDF
.
pa + pe
tgt = exp a LEDF + p a + e
Glin
(9)
k
k
k
EDF k
corresponds to the target gain of the channel at wavelength
k and the exponential
function in the denominator reprelin
sents the gain Gpump tgt of the pump at the target operating
point, the equation for the slope can be rewritten as
sig
N
sig
sig
sk Glin
sk Glin
k tgt 1
k tgt
k=1
k=1
.
Sff =
1 Glin
1 Glin
pump tgt
pump tgt
50.0% | 32.0 m
52.5% | 24.9 m
55.0% | 20.4 m
57.5% | 17.3 m
60.0% | 15.0 m
62.5% | 13.2 m
65.0% | 11.8 m
67.5% | 10.7 m
70.0% | 9.8 m
72.5% | 9.0 m
75.0% | 8.3 m
77.5% | 7.8 m
80.0% | 7.3 m
sig
N
(10)
The introduced approximation does not affect the accuracy of the feedforward control noticeably since the typical
gain of an EDFA stage is significantly larger than 1. According to this equation, the required pump power is governed
by the sum of the contributions of the different channels
weighted by the target gain profile of the EDF.
For a target gain of 20 dB at a reference wavelength
of 1550 nm, the wavelength dependence of the slope is
illustrated in Fig. 6. For this purpose, amplification of a single signal with varying wavelength at constant inversion is
assumed. Since the required pump power is in the order of
magnitude of the signal output power, the slope has been
lin
Normalized slope S ff /G ref
Cff =
1550
1560
1570
Wavelength[nm]
and
pc
4
3
2
1
0
1530
1540
1550
1560
1570
Wavelength[nm]
Fig. 6 Relationship between the slope Sff and the wavelength of a
single signal for different inversion levels and lengths of the fiber,
respectively. The gain at the reference wavelength 1550 nm equals
20 dB. The upper plot shows the slope for a pump wavelength of
1480 nm, whereas a pump wavelength of 980 nm is assumed for the
lower plot
215
18
16
Offset Cff
14
Length
12
10
10
12
14
16
18
20
Gain[dB]
Fig. 7 Relationship between the offset Cff and the gain provided by
the amplifying fiber for a pump wavelength of 1480 nm and different
lengths of the doped fiber
the linear target gain decreases for all fiber lengths with
increasing gain.
For the sake of correctness, please note that the final
conclusion neglects the dependence of the photon energy
on channel wavelength. However, this is justified since this
simplification introduces an error smaller than 2.5 %.
Input of stage
Power distribution
Isolator
ATT
EDFA
1st stage
Power
Optimum
noise figure
Monitor
Monitor
Feedback
control
Feedforward
control
all er
sm pow
ut
p
in
EDFA
2nd stage
VOA
attenuation
Fig. 8 Compensation of
wavelength dependence by
adding a spectral filter in front of
the input monitor. The drawing
on the right side illustrates the
power distribution in a two stage
EDFA for a typical amplifier
control designed for optimum
noise figure (dashed curve) and
a control keeping the gain of
each stage at a constant level
(solid curve)
straight forward solution to eliminate the wavelength dependence is to place a filter in front of the input monitor whose
transmission characteristic reproduces the spectral shape of
the target gain of the EDF [36], as shown in Fig. 8. However,
this approach has some disadvantages with respect to amplifier cost. Typically, an unfiltered signal will still be needed
so that three additional optical components are required,
namely a splitter, the filter, and an additional photodiode.
But there are also some drawbacks in view of performance. In most commercial amplifiers, the gain of the
involved amplifier stages is redistributed when changing the
input power, as shown on the right side of this figure. This
does not affect the gain profile noticeably, since it essentially depends on the average population probability of the
metastable level of all gain stages only. In this way, noise
figure degradation at smaller input powers can be avoided.
In contrast, all stages within an EDFA have to be operated a
constant gain irrespective of the input power and the overall
gain when using this technique which finally leads to worse
noise performance at small channel counts. Furthermore,
the sensitivity of the input monitor is degraded since a part
of the signal power in the monitor path has to be branched
off for the second monitor with filter.
An alternative solution makes use of the gain flattening
filter (GFF) that is embedded in the amplifier setup anyway [37]. A suitable amplifier setup comprising two EDF
coils of approximately identical length is depicted in Fig. 9.
Gain flattening and suppression of backward propagating
ASE are achieved by a gain flattening filter and an isolator,
respectively, that are placed between two fiber coils. A single laser diode provides pump power for both EDF coils by
forwarding the residual pump power of the first coil via a
pump bypass to the second coil. Replacing the pump bypass
by a pump splitter [25] is also supported.
Under this approach, a correction term that adds to the
F F calculated according to the standard
pump power Ppump
small er
w
input po
Constant
gain per stage
Position in EDF
Constant
external gain
10.0 m
12.0 m
15.0 m
20.0 m
25.0 m
30.0 m
216
Fig. 9 Setup suitable for taking
wavelength dependence into
account by measuring the
average attenuation of the GFF
Erbiumdoped fiber
Erbiumdoped fiber
Pump bypass
Isolator
Isolator
Gain
flattening
filter
Pump
laser
Monitor
before GFF
FF
FF
lin
Ppump
= Ppump
,
+ pump Pout
(11)
Monitor
after GFF
Output
Monitor
2
pump
1
Variation parameter
Input
Monitor
0
0.0 dBm
2.0 dBm
4.0 dBm
6.0 dBm
8.0 dBm
10.0 dBm
12.0 dBm
14.0 dBm
15.0 dBm
1.5
2
2.5
Attenuation by GFF
3
GFF
217
Fig. 11 Compensation of
wavelength dependence based
on the reaction of the EDF to
changes of the input power
pin
p, before
=pin
sig, after
pout
1 Glin
pump
sig, after
pin
sig, before
pout
sig, before
+ pout
.
(13)
Adding a term to the expression between the curly brackets induces changes of the gain profile that counteract to the
218
Isolator
Erbiumdoped fiber
Delay
Control signal
Input
Monitor
Pump
laser
Fig. 12 The optical fiber placed between the signal monitor and the
first EDF allows to delay the signal so that the pump power can be
adjusted before the power drop becomes effective
increase of the noise figure, this also requires larger amplifier housings and runs counter to the current trend to
smaller amplifier size. In the following, techniques reducing gain deviations by limiting the temporary increase of
the filling level of the reservoir formed by the pump level
are considered that do not require additional optical components and thus avoid this drawback. Techniques for a
single amplifier stage are considered first, whereas control
methods for setups with two cascaded stages are studied
thereafter.
Single stage setup
state absorption (ESA) [40]. Furthermore, it has been verified that this technique can be applied continuously without
starting to diverge.
Timing issues
Delaying the signal before it enters the first EDF helps
to counteract to late reaction of the feedforward control.
Such a holdup of the signal can for example be introduced
by inserting a piece of standard fiber between the coupler of the input monitor and the input of the first EDF,
as proposed in [4] (see Fig. 12). However, the authors of
this document pursued another objective with this proposal.
They conclude that there is no improvement of the transient performance when pumping the EDF at 1480 nm,
whereas adjusting the pump power 2 s before the signal reaches the EDF provides minimum gain excursion
when pumping around 980 nm. In fact, the advance adjustment of the pump power reduces the temporary overshoot
of the population probability of the pump level. But this
technique has a significant drawback, since the additional
fiber has to be at least 400 m long. Besides the small
Glin
sig = exp
a
e
2nd
sig
+ sig
N pop
3rd
a 1 N pop
LEDF .
Signals
EDFA
Signals
Time
Output power
Input power
Drop of
channels
Drop of
channels
Time
Drop of
channels
Zero
period
Time
Population
probability
I13/2
Time
Pump power
Pump
(14)
These two population probabilities are the only parameters in this equation that depend on time t. Thus, their
evolution determines the transient behavior of the amplifier. Unfortunately, both parameters are coupled to the
pump power and cannot be adjusted independently from one
another.
Large input power levels require larger pump powers as
compared with small input power levels in order to keep
the amplifier gain constant. Since large pump powers in
general come along with larger population probabilities of
the pump level, the population probability of the metastable
level will change in case of a drop of channels. Resulting
Population
probability
I11/2
Time
219
2nd
a (surv ) + e (surv ) N pop
3nd
= a (surv )N pop
(15)
is fulfilled, where
2nd
2nd
2nd
(t) = N pop
(t) N pop
N pop
before drop
and
3rd
3rd
3rd
(t) = N pop
(t) N pop
N pop
before drop
stand for the time dependent deviation of the population
probabilities of the metastable level and the pump level,
respectively, from the initial values. In other words, a linear
relation between the variations of the two involved population probabilities is required to keep the gain of a single
surviving channel constant.
When reducing the pump power synchronously to the
drop of the input power, immediately less ions are excited to
the pump level. Furthermore, the rate of transitions from the
metastable level to the ground level also decreases abruptly.
In contrast, the rate of the predominantly nonradiative transitions from the pump level to the metastable level decreases
almost exponentially until the new state of equilibrium is
reached. As a consequence, the population probability of
the metastable level and the gain experience an overshoot.
Minimum overshoots are achieved if the population probability of the pump level reaches as soon as possible its
new steadystage value. These considerations suggest to
interrupt transitions to the pump level completely for some
limited time after the channel drop, which means to turn off
0.2
0.2
0.15
0.1
0
0.1
0.1
0
0.05
10
20
30
40
50
0.05
Zero period
0.1
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
220
points are on the right side of the dashdotted line indicating operating points with constant gain of the surviving
channel. After the population probability of the pump level
has reached its final value, there is some readjustment of
the population probability of the metastable level that does
not affect the pump level. Therefore, there are maximum
overshoots for vanishing zero period, but no undershoots.
2nd becomes
Thanks to the zero period, the change rate of N pop
3rd already during the first
smaller as compared with N pop
microseconds and the operating points are closer to the
dashdotted line. If the pump power is set to the new target
value when the population of the pump level has reached its
final value for the first time (in the example after 11 s at the
upper border of the area labeled tolerance area), the population probability of the pump levels remains at this value and
the population probability of the metastable level goes to its
new steadystate values without any undershoots. As indicated by the size of the dots, the last step takes much longer
as compared to the adjustment of the population probability
of the pump level. The plot also indicates that the control is
quite tolerant to variations of the zero period. For durations
up to 18 s, the trace does not cross the dashed line and thus
221
0
Initial state
Final state
2
Ti
Undershoot
10
Signal input
power: 10.0 dBm
12
14
Standard control: 1.09 dB
With zero period: 0.74 dB
Input delay: 0.43 dB
Combined approach: 0.37 dB
2
Standard control
Time
components. Thus, it seems to be favorable to compensate gain variations of a first stage in a multistage setup
by inducing opposite gain variations in a subsequent amplifier stage. A technique working according to this concept is
illustrated in Fig. 17 for a setup consisting of two individually controlled stages enclosing a dispersion compensating
fiber (DCF).
Due to the memory effect introduced by pumping around
980 nm, the gain of a first amplifier stage will experience
some overshoot after a channel drop even if the pump power
Pump power
Drop of
channels
Drop of
channels
Zero
period
Time
Input delay
Drop of
channels
Delay at
input
Time
10
Combined approach
Pump power
Pump power
16
Drop of
channels
Delay at
input
Zero
period
Time
222
Furthermore, some undershoots arise later on. Nevertheless, this dynamical behavior strongly reduces the impact of
transient events on receiver performance.
As indicated by the additional pump power traces in
Fig. 17, the delayed adjustment of the pump power can
be combined with the technique employing a zero period.
Thus, the maximum magnitude of the gain variations may
be halved and the time shift yielding optimum results is
reduced. A more elaborated waveform of the pump signal for the second stage should allow to further reduce the
remaining gain variations of the complete amplifier.
EDFA
Delay in DCF
Delay of control
1st stage
Input
power
Pump
power
Output
power
2nd stage
Input
power
Pump
power
Output
power
Adaptation of
pump power
Time
Fig. 17 Individually controlled stages: Compensation of gain variations induced in the first stage by adapting the pump power of the
second stage prior to the power changes at the input of this stage. Additional curves show power evolution when applying a zero period to the
pump power
(16)
Thanks to the availability of single pump diodes providing fibercoupled pump powers in excess of 700 mW [44],
pump sharing has become a common technique to reduce
amplifier cost. Typically, the splitting ratio is constant since
a fiber based tunable coupler would obscure the economical
benefits arising from a reduced number of pump modules.
However, the strict coupling of the pump powers launched
into different EDFA stages precludes their independent control and entails significant drawback. These are, on the one
hand, increased steadystate noise figures at smaller input
powers [25], and on the other hand, poor transient performance if the power provided by a common pump is
launched into doped fiber coils placed before and after a
component afflicted with delay [26]. Therefore, commonly
no pump splitting is applied to stages surrounding a dispersion compensating fiber (DCF) module and potential for
cost reduction is wasted.
In the following, a technique achieving acceptable transient performance for two coupled EDFA stages with intermediate delay is explained. Again, a setup consisting of two
stages that are separated by a DCF is considered, but the
pump power is now supplied by a single laser diode via
a passive splitter with fix splitting ratio. With the results
obtained for individually controlled stages, adjusting the
pump power after the transient event has reached the first
EDF coil and before it reaches the second EDF coil appears
to be promising. The evolution of the signal power at the
input and the output of both fiber coils as well as the
pump power launched into the fiber coils is illustrated in
Fig. 20. Due to the delayed adaptation of the pump power,
the power of the surviving channel temporarily increases at
the output of the first stage, whereas the gain of the second stages experiences an undershoot. Since the overshoot
and the undershoot are partly compensated by each other,
a quite stable output power of the surviving channel can be
achieved. In contrast to the previous case with individually
controlled stages, the pump power launched into both stages
is adjusted simultaneously and the delay FF specifies
now the period of time between the pump power adjustment
223
0.5
0.2
0.1
0.4
0
0.1
0.3
0.2
10
0.2
10
20
30
0.1
0.1
0.2
2000
4000
6000
8000
10000
Re
5
0
Ti 5
m
e [ 10
m
s]
15
10
]
5
ms
y[
a
l
De
t . 5
t im
ef0
ra 5
m
e [ 10
m
s]
15
10
]
5
ms
y[
a
l
De
224
3.5
2.5
1.5
10
4.5
12
30.0 %
40.0 %
50.0 %
60.0 %
70.0 %
11
10
9
8
7
0.5
50
Time shift [ s]
100
6
5
50
Time shift [ s]
100
50
100
Time shift [ s]
Fig. 21 Setup with pump power splitting: Maximum gain variation versus delay for a two stage setup based on pump power splitting (delay in
DCF: 100 s) for different splitting ratios and initial input powers
225
Pump power
second stage
II
Pump power
EDF
Noise figure
improvement
Pump power
first stage
Total pump power
Modified control
Standard control
Modified control
Time
Standard control
Modified control
Time
Time
Power
ATT
EDFA
Power
manner only have been studied [51], but the VOA has also
been included in the feedforward and the feedback control scheme [52]. Furthermore, it has also been proposed
to block the input signal by a VOA placed in front of the
first EDF in case traffic is interrupted. This technique eliminates surges that arise when traffic is reestablished and the
pump has not been turned off [53]. But due to the significant
deterioration of the noise figure, this solution is not well
suited for optical data transmission systems. Furthermore,
it does not support the continuous operation of surviving
channels.
In general, the VOA has the task to correct gain variations
that arise although the pump power is adjusted correctly.
Such deviations might result from delayed reaction of the
feedforward control or the population of the third energy
level (pump level). Figure 23 illustrates the principle of
such a hybrid control technique. Gain excursions caused
Pump power
Standard control
Pump power
Pump power
Time
Time
Ideal case
Gain
Power
EDFA
Time
Surviving
channel
Time
Reality
ATT
Power
Attenuation
Time
Time
Surviving
channel
226
0.0 s
0.5 s
1.0 s
2.0 s
5.0 s
2.5
2
1.5
Delay
1
0.5
0
0.5
0
0.1
50
100
150
200
0.0 s
0.5 s
1.0 s
2.0 s
5.0 s
0.05
Delay
Closing remarks
0.05
0
50
100
150
200
Conclusions
227
technique is adaptive and achieves good results at variable gain values. Thus, the solution provides full flexibility with respect to gain redistribution within a multistage
amplifier.
A piece of fiber placed between the input monitor and
the first EDF coil is an efficient means to counteract to a
delayed adaptation of the pump power. Furthermore, this
solution may also counteract to the memory effect introduced by the third energy level of the erbium ions. However,
this approach is not compatible with the request for small
amplifier housing and causes some noise figure degradation.
These disadvantages are avoided by a modified pump power
control turning off the pump for some limited time called
zero period. Timing is also less critical as compared with
delaying the input signal. Unfortunately, the solution is not
equally efficient for all amplifier setups. A detailed analysis
evaluating the temporal changes of the population probabilities of all relevant energy levels illustrates the effect of
this approach on amplifier gain in detail. Further techniques
are possible in cascades of amplifier stages, mainly if two
stages are surrounding an optical device affected by delay.
In brief, the control induces gain undershoots in a latter
stage counteracting to gain overshoots arising in previous
stages. Embodiments for individually controlled stages and
1528.8 nm
1532.7 nm
1536.6 nm
1540.6 nm
1548.5 nm
1552.5 nm
1556.6 nm
1560.6 nm
1563.9 nm
1
2
0.8
4
Initial state
Maximum undershoot
0.6
Final state
0.4
8
0.2
10
0
12
10
12
0.2
1000
2000
3000
4000
228
3rd
pop
[ 10 3 ]
10
12
Standard control: 1.08 dB
With zero period: 0.73 dB
Input delay: 0.43 dB
Modified approach: 0.29 dB
0
10
2nd
pop
[ 10 3 ]
Standard control
With zero period
Input delay
Modified approach
500
1.2
1.08 dB
1
0.8
0.73 dB
400
0.6
0.43 dB
0.4
300
0.29 dB
0.2
200
0
0.2
100
0.28 dB
0.18 dB
0.27 dB
0.4
0.43 dB
0
4 2 0 2
10
15 20
30 50 100 200
4 2 0 2
10
15 20
800
Fig. 27 Modified control of the pump power with periodic pump power adjustment after the zero period. Please note that the time axis has been
compressed at later stages in order to show details. For easier orientation, some pictograms emphasizing the key features of the different pump
power curves have been added
minimum leading to the observed undershoots. As a consequence, the optimum duration of the zero period also
depends on wavelength. However, the tolerance area is
large enough so that a single duration of the zero period
can produce minimum gain variations irrespective of the
wavelength.
References
1. S. Matsuoka, NTT Techn. Review 9(8), 1 (2011)
2. R. Ramaswami, K.N. Sivarajan, Optical Networks A Practical
Perspective (Morgan Kaufmann Publishers, San Francisco, 1998)
3. A.K. Srivastava, Y. Sun, J.L. Zyskind, J.W. Sulhoff, IEEE Photon.
Technol. Lett. 9(3), 386 (1997)
4. H. Nakaji, Y. Nakai, M. Shigematsu, M. Nishimura, Opt. Fiber
Techn. 9(1), 25 (2003)
229
5. Y. Sun, A.K. Srivastava, Proceedings Optical Amplifiers and their
Applications, (Victoria, 1997), pp. 333353
6. K. Ishii, J. Kurumida, S. Namiki, Proceedings Optical Fiber Communication Conference, (San Francisco, 2014), pp. 13. W3E.4
7. D.H. Richards, J.L. Jackel, M.A. Ali, IEEE J. Select. Top. Quant.
Electron. 3(4), 1027 (1997)
8. H. Dai, J.Y. Pan, C. Lin, IEEE. Photon. Technol. Lett. 9(6), 737
(1997)
9. S. Sergeyev, E. Vanin, G. Jacobsen, Proceedings Optical Fiber
Communication Conference, (Anaheim, 2002), pp. 518519
10. J.C.R.F. Oliveira, S.M. Rossi, R.F. Silva, J.B. Rosolem, A.C. Bordonalli, Proceedings International Microwave and Optoelectronics
Conference (IMOC), (Salvador, 2007), pp. 683687
11. H. Feng, E. Patzak, J. Saniter, Phot. Netw. Comm. 4(2), 151 (2002)
12. L. Pavel, Proceedings Conference on Control Application, (Istanbul, 2003), pp. 415422
13. B. Huang, A. Yu, Selfadapting feed forward control apparatus
and method in an optical amplifier. US patent 8,139,286 (2012)
14. A.R. Chraplyvy, J.A. Nagel, R.W. Tkach, IEEE Photon. Technol.
Lett. 4(8), 920 (1992)
15. A. Bianciotto, A. Carena, V. Ferrero, G. Gaudino, IEEE Photon.
Technol. Lett. 15(10), 1351 (2003)
16. N. Sindhu, P.K. Shafeena, Intern. Journ. Emerg. Trends Electr.
Electron. 1(2), 50 (2013)
17. L. Rapp, D. Setti, Journ. Opt. Comm. 28(3), 162 (2007)
18. M. Zirngibl, Electron. Lett. 34(8), 789 (1998)
19. M. Bolshtyansky, N. King, G. Cowle, Proceedings Optical Fiber
Communication Conference, (Anaheim, USA, 2007), pp. 13.
JThA15
20. H. Ono, K. Shimano, M. Fukutoku, S. Kuwano, IEEE J. Lightw.
Technol. 20(8), 1335 (2002)
21. K. Komaki, H. Itou, H. Iizuka, Optical amplifier and control
method therefor. US patent 6,977,770 (2005)
22. F. Shehadeh, R.S. Vodhanel, C. Gibbons, M. Ali, Proceedings Optical Fiber Communication Conference, (San Jose, USA,
1996), pp. 190191
23. A.V. Tran, C.J. Chae, R.S. Tucker, Y.J. Wen, IEEE Photon.
Technol. Lett. 17(1), 226 (2005)
24. M. Karasek, F.W. Willems, IEEE Photon. Technol. Lett. 10(7),
1036 (1998)
25. L. Rapp, Journ. Opt. Comm. 27(4), 194 (2006)
26. L. Rapp, Journ. Opt. Comm. 31(2), 66 (2010)
27. L. Rapp, IEEE J. Lightw. Technol. 25(3), 726 (2007)
28. J. Ye, Y.W. Lu, Y. Cao, Optical amplifier system with
transient control. United States Patent, US 6,414,788 B1
(2002)
29. C. Tian, S. Kinoshita, J.R. Rodriguez, Control system and method
for an optical amplifier. United States Patent, US 6,690,508 B2
(2004)
30. L. Rapp, IEEE Photon. Technol. Lett. 22(16), 1256 (2010)
31. J.M. Ferreira, D. Fonseca, A.N. Pinto, L. Rapp, Internal Conference on Transparent Optical Networks, (Cartagena, Spain, 2013),
pp. 14
32. M. Bolshtyansky, P. Wysocki, N. Conti, IEEE J. Lightw. Technol.
18(11), 1533 (2000)
33. A.A.M. Saleh, R.M. Jopson, J.D. Evankow, J. Aspell, IEEE
Photon. Technol. Lett. 2(10), 714 (1990)
34. P.C. Becker, N.A. Olsson, J.R. Simpson, ErbiumDoped Fiber
Amplifiers Fundamentals and Technology (Academic Press,
San Diego, 1999)
35. C.R. Giles, E. Desurvire, IEEE J. Lightw. Technol. 9(2), 271
(1991)
36. Y. Sun, J. Ye, Optical amplifier system with transient control using
spectrally filtered input. United States Patent, US 6,341,034 B1
(2002)
230
37. L. Rapp, Optical fiber amplifier comprising an embedded filter and a control method with improved feedforward control performance. European Patent, EP 2474110 B1
(2013)
38. L. Rapp, Journ. Opt. Comm. 28(2), 82 (2007)
39. E. Desurvire, ErbiumDoped Fiber Amplifiers Principles and
Applications (Wiley, New York, 1994)
40. L. Rapp, Proceedings of the 5th International Workshop on Design
of Reliable Communication Networks, (Ischia, 2005), pp. 127
134
41. L. Rapp, Journ. Opt. Comm. 31(3), 130 (2010)
42. C.B. Layne, W.H. Lowdermilk, M.J. Weber, Phys. Rev. Lett. B
16(1), 10 (1977)
43. W.J. Miniscalco, IEEE J. Lightw. Technol. 9(2), 234 (1991)
44. B. Schmidt, S. Pawlik, N. Matuschek, J. Muller, T. Pliska,
J. Troger, N. Lichtenstein, A. Wittmann, S. Mohrdieck,
B. Sverdlow, C. Harder, Proceedings Optical Fiber Communication Conference, (Anaheim, 2002), pp. 702
703
45. L. Rapp, Europ. Trans. Telecom. 18(4), 381 (2007)