Вы находитесь на странице: 1из 8

Ferreira et al. VOL. 4, NO. 8/AUGUST 2012/J. OPT. COMMUN. NETW.

595
Weighted Undepleted Pump Model for
Broadband Counter-Pumped Raman Fiber
Ampliers
Joo M. Ferreira, Rogrio Nogueira, Paulo Monteiro, and Armando N. Pinto
AbstractWe develop a low complexity computational
model for the gain prole and amplied spontaneous emission
noise for broadband counter-pumped Raman ber ampliers.
The proposed model is based on two adjustment parameters
used to account for the interactions between the pumps. The
obtained results showa good agreement between experimental
measurements and detailed numerical simulations, for differ-
ent combinations of pump wavelengths and pump powers, with
a processing time several times lower than the time taken by a
detailed numerical model.
Index TermsCounter-pump; Gain; Optical ampliers; Ra-
man ampliers; Raman scattering.
I. INTRODUCTION
A
s optical networks become more complex, new tools and
protocols such as network planning tools or routing
wavelength assignment (RWA) protocols have to be developed
and implemented in order to help operators to plan, optimize
and operate the networks [1]. However, in order to develop
and implement such tools and protocols, it is necessary to
estimate correctly and in a fast way the gures of merit of
network components, like the gain and noise gures of Raman
ber ampliers (RFAs) [2,3]. Nowadays, transmission systems
can have more than 80 signals and tens of other resources
like ampliers and transponders. Therefore, it is necessary to
develop and implement models with low computational effort
in order provide reliable results in a useful time interval.
RFAs are suitable for long and ultra-long haul amplication
due to their low noise gures [4]. When modeling a
transmission link with Raman amplication, it is known that
Manuscript received November 22, 2011; revised June 14, 2012; accepted
June 20, 2012; published July 17, 2012 (Doc. ID 157908).
Joo M. Ferreira (e-mail: joao.2.ferreira@nsn.com) and Paulo Monteiro
are with Nokia Siemens Networks Portugal, S.A. Rua Irmos Siemens,
2720-093 Amadora, Portugal, and are also with the Department of Electronics,
Telecommunications and Informatics, University of Aveiro, Campus de Santiago,
3810-193 Aveiro, Portugal, and are further afliated with the Instituto de
Telecomunicaes, Campus de Santiago, 3810-193 Aveiro, Portugal.
Rogrio Nogueira is with Nokia Siemens Networks Portugal, S.A. Rua
Irmos Siemens, 2720-093 Amadora, Portugal, and is also with the Instituto de
Telecomunicaes, Campus de Santiago, 3810-193 Aveiro, Portugal.
Armando N. Pinto is with the Department of Electronics, Telecommunications
and Informatics, University of Aveiro, Campus de Santiago, 3810-193 Aveiro,
Portugal, and is also with the Instituto de Telecomunicaes, Campus de
Santiago, 3810-193 Aveiro, Portugal.
Digital Object Identier 10.1364/JOCN.4.000595
there are three major interactions [5], i.e., pump-to-pump,
pump-to-signal and signal-to-signal interactions. When consid-
ering network planning tools or RWA protocols, the signal-
to-signal interaction has to be taken into account even if the
transmission link does not have RFAs [6]. For this reason this
interaction is usually accounted for in the transfer function
of the transmission ber. Several models have been developed
to predict and describe the RFA behavior. In [7], the authors
stated that the composite Raman gain can be expressed as the
logarithmic sum of each Raman gain created by each pump
wavelength with a weighting factor. These weighting factors
are found after solving the propagation equations. In [8], a
model suitable for the small signal regime of a counter-pumped
RFA with a single pump was proposed. This approach is known
as the undepleted pump model. An approximated analytical
expression to model the noise gure in dispersion managed
bers was presented in [9]. A model for the gain and noise
gure for a multi-pump RFA without considering the pump-to-
pump interactions was developed in [10]. However, as shown
in [11], the pump-to-pump interactions tend to be relevant as
the number of pumps and the pump power increase. In [11], the
authors also presented an analytical approximated solution
for the gain considering the pump-to-pump interactions by
iteratively solving the propagation equations for the pumps.
These interactions can also be estimated by numerical or semi-
analytical models as shown in [12], where a semi-analytical
algorithm for calculating the gain of counter-pumped RFAs
with multiple pumps and signals was proposed. In this case,
the interactions between the pumps and signals were solved
using the RungeKutta method. In [13], a semi-analytical
model considering the depleted regime of an RFA with a single
pump and with frequency dependent losses was presented.
These analytical or semi-analytical model approaches tend to
imply a signicant number of approximations, which, in most
cases, diminish the generality of the models. On the other
hand, some detailed numerical models have been published
in the literature. In [14], a detailed numerical model based
on the application of the average power analysis technique
to solve the propagation equations of RFAs was presented.
This model shows a similar accuracy and is computationally
more efcient than the fourth-order RungeKutta routine [15].
However, it is still not suitable for treating large networks
due to the computation power required, especially when large
numbers of pumps and signals are considered. In [16,17], the
authors developed a closed integral form of coupled Raman
equations; these equations allowed a reduction of processing
time while maintaining good accuracy. Nevertheless, an
1943-0620/12/080595-08/$15.00 2012 Optical Society of America
596 J. OPT. COMMUN. NETW./VOL. 4, NO. 8/AUGUST 2012 Ferreira et al.
iterative numerical method is still necessary to obtain the
results, and the time needed for this process may preclude their
utilization in some more complex scenarios.
We develop a low complexity computational model for the
gain of the RFA, suitable for use with multiple pumps and
with a large number of signals. To achieve this, we customize
the undepleted pump model with the introduction of two
new parameters, in order to account for the pump-to-pump
interactions. These parameters can be adjusted based on
experimental measurements, simulation results, or analytical
expressions. Using these two parameters, an accurate black
box model for the gain prole is obtained, without increasing
the computational complexity of the undepleted pump model.
Based on this approach, we also present and discuss a method
to estimate the amplier spontaneous emission (ASE) noise
and optical signal-to-noise ratio (OSNR). The obtained results
for the gain prole and ASE noise show a good agreement
with experimental measurements and detailed numerical
simulations, with a computational effort much lower than the
detailed numerical model. With this model it is possible to
estimate the gain and ASE noise of the amplier based on a
simple input/output relation, and to the best of the authors
knowledge it is the rst time that the problem has been
addressed in this way.
This paper is organized as follows. In Section II, we
introduce and explain the proposed model for the gain of
RFAs and the approaches used to adjust the parameters. In
Section III, an extensive comparison with experimental mea-
surements and detailed numerical simulations are presented.
In Section IV, we present a method to estimate the ASE and
the OSNR based on the proposed model. The conclusions are
presented in Section V.
II. WEIGHTED MODEL
To account for the pump-to-signal interaction, let us
consider the undepleted pump model, where a small signal
regime is assumed [8]. The output signal power is given by
P
s
(L) =P
s
(0) exp
_
C
s, p
P
p
(L)L
eff

s
L
_
, (1)
where P
s
(0) is the input signal power, the effective ber length
is given by L
eff
=
1exp(
p
L)

p
, P
p
(L) is the input pump power,

p
is the ber attenuation at the pump wavelength,
s
is
the ber attenuation at the signal wavelength and L is the
transmission length. The C
s, p
is the Raman gain efciency
of the ber, considering the wavelength of the signal and the
pump. Generically, for waves i and j centered at frequencies
i
and
j
, respectively, the Raman gain efciency is given by [18]
C
i, j
=
_

_
g
R
(
j

i
)
2 A
eff
if
j

i
0,

j
g
R
(
i

j
)
2 A
eff
if
j

i
<0,
(2)
where A
eff
is the bers effective area, g
R
() is the Raman
gain coefcient of the ber and is the frequency spacing
between the two waves. The Raman gain coefcient can be
obtained using the technique presented in [18,19].
TABLE I
PUMP WAVELENGTHS AND INPUT PUMP POWERS OF RFA
CONFIGURATIONS
Pumps
1426 nm 1444 nm 1462 nm 1487 nm
Conf. 1 182.7 mW 237.5 mW 88 mW 233 mW
Conf. 2 355.4 mW 256.8 mW
Conf. 3 397.7 mW 279 mW 83.8 mW
Conf. 4 25500 mW 25500 mW
We can extend the undepleted pump model to support
multiple pumps, where the contribution of the pump for the
signal output power in Eq. (1) is replaced by a sum over all the
pumps. In this case, the output signal power is given by
P
s
(L) =P
s
(0) exp
__
N

i=1
C
s,i
P
i
(L)L
eff,i
_

s
L
_
, (3)
where L
eff,i
=
1exp(
i
L)

i
is the effective length for each pump,
P
i
(L) is the input pump power of pump i,
i
is the ber
attenuation at the pump wavelength i and N is the number
of pumps.
In order to assess the accuracy of the undepleted pump
model given by Eq. (3) we perform measurements and detailed
numerical simulations in a transmission system amplied
with a counter-pumped RFA. The detailed numerical model
used in this work is the one developed in [14]. To model
the Raman gain coefcient, g
R
(), we use a set of 14
Gaussian functions with the parameters shown in [18]. The
RFA implemented, conguration 1 in Table I, has four pumps
centered at 1426 nm, 1444 nm, 1462 nm and 1487 nm, with
attenuation coefcients of 0.26 dB/km, 0.25 dB/km, 0.23 dB/km
and 0.23 dB/km, respectively. The pumps P
1
, P
2
, P
3
, P
4
are
coupled to the ber using an optical multiplexer, followed by
an optical coupler. The transmission ber is a single mode
ber (SMF) with 80 km, with an effective area of 80 m
2
and an attenuation coefcient of 0.2 dB/km around 1550 nm.
Since we are interested in measuring the pump-to-signal and
pump-to-pump interactions, a tunable laser operating at 1 mW
is used as the probe to be amplied. We measure the On/Off
gain from 1510 nm to 1610 nm using an optical spectrum
analyzer (OSA). The On/Off gain of the RFA for a wavelength
s, G
dB,s
, in decibels is dened as
G
dB,s
=10log
10
_
P
s,On
(L)
P
s,Off
(L)
_
, (4)
where P
s,On
(L) is the signal output power with amplication
(when the pumps are On) and P
s,Off
(L) is the signal output
power without amplication (when the pumps are Off).
Figure 1 shows experimental measurements (crosses) for
the RFA On/Off gain compared with the undepleted pump
model (dotted line), Eq. (3), and with detailed numerical results
(full line) [14]. We can observe that the detailed numerical
model accurately describes the experimental data. When
comparing the undepleted pump model with the experimental
or numerical model results, we can observe a large deviation.
This is due to the non-linear effect of stimulated Raman
scattering (SRS), which transfers energy between the pumps,
Ferreira et al. VOL. 4, NO. 8/AUGUST 2012/J. OPT. COMMUN. NETW. 597
Fig. 1. On/Off gain measurements and comparison with numerical
models for a counter-pumped RFA with four pumps and a total pump
power of 745 mW: experimental measurements (crosses), detailed
numerical results (full line), undepleted pump model (dotted line),
linear regression of the detailed numerical model (dashed line), linear
regression of the undepleted pump model (dashdot line). The inset
refers to the pump wavelengths and pump powers of the RFA.
known as pump-to-pump interactions. This fact is highlighted
by observing the linear regression of the results obtained with
each model; see Fig. 1. It can be seen that the undepleted pump
model presents a gain with a negative tilt (dashdot line in
Fig. 1), whereas the real gain prole presents a positive tilt
(dashed line in Fig. 1).
A. Adjusting the On/Off Gain Tilt
It is known that the SRS between pumps changes the
pump power distribution within the ber, resulting in an
amplication of the pumps with higher wavelengths and in
the depletion of pumps with lower wavelengths. We propose
to include this effect in the undepleted pump model, Eq. (3), by
adjusting the input pump power, in order to account for the
pump-to-pump interactions and therefore obtain the correct
output signal power
P
s
(L) =P
s
(0) exp
__
N

i=1
C
s,i
P
SRS
i
L
eff,i
_

s
L
_
, (5)
where P
SRS
i
is the corrected input pump power of pump i
to account for the pump-to-pump interactions. In order to
estimate the P
SRS
i
, we use
P
SRS
i
=P
i
(L) exp
_
N

j=1
C
i, j
P
j
(L)
1exp(
j
D
int
)

j
_
, (6)
where the D
int
parameter can be understood as an interaction
length between the pumps, in order to account for the
amplication or depletion of the pumps due to SRS between
the pumps. Note that when j = i, C
i, j
=0. In order to estimate
D
int
, we start by writing the amplier On/Off gain using
G
s
=
P
s,On
(L)
P
s,Off
(L)
=exp
_
N

i=1
C
s,i
P
SRS
i
L
eff,i
_
, (7)
where G
s
is the On/Off gain for a wavelength s. To simplify this
equation let us consider the same attenuation coefcient for all
the pump wavelengths; therefore, L
eff,i
= L
eff
. Using Eq. (6),
we obtain
ln(G
s
) L
eff
N

i=1
C
s,i
P
i
(L) exp
_
L
eff

j=1
C
i, j
P
j
(L)
_
, (8)
where L
eff
=
1exp(D
int
)

. Considering the rst term of the


Taylor expansion of the exponential in Eq. (8), we obtain
ln(G
s
)
L
eff

i=1
C
s,i
P
i
L
eff

i=1
N

j=1
C
s,i
C
i, j
P
i
P
j
. (9)
Expanding the exponentials of L
eff
and L
eff
, we can write
D
int

ln(G
s
) L

N
i=1
C
s,i
P
i
L

N
i=1

N
j=1
C
s,i
C
i, j
P
i
P
j
. (10)
This expression relates D
int
with the On/Off gain for the signal
s. By substituting the On/Off gain, G
s
, with an experimental
or numerical result, it is possible to obtain D
int
. However,
this derivation is made assuming some approximations, the
most relevant one being the assumption that the rst-order
Taylor series is a good approximation for the exponentials. To
overcome this, the chosen wavelength, s, should be the one
which adjusts best the On/Off gain tilt given by Eq. (5), when
compared with the detailed numerical model or experimental
measurements. Therefore, a set of wavelengths has to be
tested, and the one which provides the best description for the
tilt is chosen.
Let us consider the amplier of conguration 1. By
measuring the On/Off gain for different wavelengths in the
amplier region of interest and assessing the gain tilt based
on Eqs. (5), (6) and (10), and comparing with the detailed
numerical model, we verify that the wavelength which best
adjusts the tilt is 1571 nm. Using this wavelength to calculate
the On/Off gain (see Fig. 2), we can see that the gain tilt is well
adjusted. This fact is highlighted by the linear regressions of
both models. However, an offset in the gain is still present.
B. Adjusting the On/Off Gain Offset
We can adjust the offset in the On/Off gain of the RFA by
including another parameter in the model, i.e.,
P
s
(L) =P
s
(0) exp
__
N

i=1
C
s,i
P
SRS
i
L

eff,i
_

s
L
_
, (11)
where L

eff,i
is given by
L

eff,i
=
1exp((
i
+
SRS
)L)

i
+
SRS
, (12)
where
SRS
accounts for a gain reduction due to the depletion
of the pumps.
We can write the On/Off gain difference between the model
with the adjusted tilt, G
t
, obtained with Eq. (5), and the
598 J. OPT. COMMUN. NETW./VOL. 4, NO. 8/AUGUST 2012 Ferreira et al.
(a)
(b)
Fig. 2. On/Off gain measurements and comparison with numerical
models for a counter-pumped RFA with four pumps and a total pump
power of 745 mW: experimental measurements (crosses), detailed
numerical results (full line), weighted undepleted pump model (dotted
line), linear regression of the detailed numerical model (dashed line),
linear regression of the weighted undepleted pump model (dashdot
line), when adjusting (a) the tilt and (b) the tilt and offset.
experimental or numerical result, G
s
, for the wavelength s, as
G
s
= exp
_
N

i=1
C
s,i
P
SRS
i
L
eff,i
_
exp
_
N

i=1
C
s,i
P
SRS
i
L

eff,i
_
. (13)
Solving Eq. (13) in order to L

eff,i
, we obtain
L

eff,i
=
G
s

N
i=1
(C
s,i
P
SRS
i
)
+L
eff,i
. (14)
Making L
eff,i
1/
i
, and L

eff,i
1/(
i
+
SRS
), we obtain

SRS
=
1

G
s

N
i=1
C
s,i
P
SRS
i
+
1

i
. (15)
The results obtained based on the measurement of the
reference wavelength considered previously are presented in
Fig. 2(b). As we can see, the results show that both the tilt and
the offset are well adjusted.
C. Analytical Expressions for D
int
and
SRS
We have shown that it is possible to obtain accurate
results for the amplier gain based on a customization of
the undepleted pump model with two extra parameters.
These parameters are estimated based on a comparison
with experimental or numerical results; we call this method
weighted by comparison (WBC). Nevertheless, it can be
useful to have analytical expressions for D
int
and
SRS
.
Considering that the weighted undepleted pump model
should converge to the undepleted pump model when the
number of pumps is one, we expect that D
int
and
SRS
should approach zero for a single pump conguration. Besides
that, D
int
should be inversely proportional to the total
pump power [20], and should account for the pump-to-pump
interactions. The parameter
SRS
accounts for the pump power
losses due to the SRS effect between pumps. Therefore, we
can derive an expression in dB/km based on the depletion of
the rst pump, which is the most attenuated pump. Based on
the principles stated, and only on known parameters of the
system, we derive analytical expressions which better adjust
the results obtained from simulations considering different
scenarios:
D
int

N1
P
total

N
j=2
|C
1, j
|
, (16)

SRS

10
L
log
10
_
P
SRS
1
P
1
(L)
_
, (17)
where C
1, j
is the Raman gain efciency of the ber between
the rst pump P
1
and the pump j; P
SRS
1
is the remaining
power after accounting for the depletion of the rst pump,
which is the most attenuated pump; and P
1
(L) is the input
pump power of the rst pump. With this approach, we are
able to propose a model for the output signal power of RFAs
with different numbers of pumps and powers, without the need
to compare with experimental measurements or numerical
simulations to adjust the parameters. We call this approach
weighted by analytical expressions (WAE); it is given by
Eqs. (6) and (11), and the parameters D
int
and
SRS
are
obtained with the analytical expressions (16) and (17).
Figure 3 shows the results obtained with the WAE model
for the RFA of conguration 1. As we can see, the results
obtained present a good accuracy; however, it is slightly worse
than the accuracy obtained when D
int
and
SRS
are obtained
through comparison with experimental or detailed numerical
simulation, WBC. Nevertheless, based on WAE the results
are obtained without any comparison with experimental or
numerical data, which makes this method simpler to use.
III. PERFORMANCE ASSESSMENT
In order to assess the proposed model in scenarios with
different numbers of pumps, pump wavelengths and powers,
we experimentally implement counter-pumped RFAs with two,
three and four pumps; see Table I. We measure the On/Off
gain from 1510 nm to 1610 nm when considering the amplier
with conguration 1 and from 1510 nm to 1570 nm for
Ferreira et al. VOL. 4, NO. 8/AUGUST 2012/J. OPT. COMMUN. NETW. 599
Fig. 3. On/Off gain measurements and comparison with numerical
models for a counter-pumped RFA with four pumps and a total pump
power of 745 mW: experimental measurements (crosses); detailed
numerical results (full line); WAE (dotted line); linear regression of the
detailed numerical model (dashed line); linear regression of the WAE
(dashdot line).
congurations 2 and 3, using an OSA. The obtained results for
the considered congurations are shown in Figs. 3 and 4.
To assess the error of the gain prole we measure the mean
absolute error (MAE) between the estimated and measured
gain proles. Figure 5 shows the MAE calculated between
1510 nm and 1610 nm for congurations 1, 2 and 3. As we
can observe, the detailed numerical model describes correctly
the experimental measurements for all the setups considered,
with an MAE below 0.5 dB. When considering the undepleted
pump model, we verify that it presents a higher error than
the detailed numerical model, and this error increases with
the number of pumps and pump powers; it reaches an MAE
of 2.8 dB in conguration 1. When comparing the weighted
undepleted pump model, WBC and WAE with the detailed
numerical model, we verify that for some congurations
the MAE is almost the same as the detailed numerical
model, which highlights the accuracy of the proposed models.
Nevertheless, we verify that when using the WBC, the error
is typically smaller than when using the WAE. Regarding
conguration 2, the WBC presents a slightly higher error
than the undepleted pump model. This happens because the
depletion of the rst pump is overestimated by the model, due
to the fact that, in this conguration, the rst pump has a
higher input power than the second pump and the wavelength
separation between them is small, about 18 nm, which means
that the depletion is negligible.
In Fig. 6, we can observe the MAE as a function of the total
input power, between the gain estimated from the undepleted
pump model and the weighted model using the two methods to
adjust the parameters, and the detailed numerical model for
congurations 1 and 4. The power is increased equally in all
the pumps. Considering the WBC approach, we verify that the
MAE obtained is below 0.4 dB for all cases tested. Regarding
the WAE approach, we conclude that it tends to be suitable for
a total pump power up to 900 mW and a wavelength separation
below 60 nm between the rst and last pumps, considering
an MAE up to 1 dB. When comparing the WBC and WAE
(a)
(b)
Fig. 4. On/Off gain measurements and comparison with numerical
models for a counter-pumped RFA in (a) conguration 2 with a
total pump power of 612 mW and (b) conguration 3 with a
pump power of 760 mW: experimental measurements (crosses);
detailed numerical results (full line); undepleted pump model (dotted
line); weighted undepleted pump model based on comparison with
experimental or detailed numerical simulation, WBC (dashed line);
weighted undepleted pump model based on analytical expressions,
WAE (dashdot line). The inset refers to the pump wavelengths and
pump powers of the RFA.
Fig. 5. MAE of the On/Off gain for the different models for
congurations 1, 2 and 3, when compared with experimental results.
approaches with the undepleted pump model, we verify that
the results are signicantly improved.
600 J. OPT. COMMUN. NETW./VOL. 4, NO. 8/AUGUST 2012 Ferreira et al.
Fig. 6. MAE between the On/Off gain obtained with the undepleted
pump model, WBC and WAE, and the detailed numerical model for
different total input pump powers for congurations 1 and 4; see
Table I.
TABLE II
PROCESSING TIMES (ms) OF THE MODELS IN AN INTEL
T5500 PROCESSOR AT 1.66 GHz WITH 2 GB OF RAM
Model Conf. 1 Conf. 2 Conf. 3
Detailed numerical model 7300 2500 4700
Undepleted pump model 90 35 50
WAE; WBC 90 40 50
We also calculate the MAE as a function of the ber length,
between the gain estimated from the undepleted pump model
and fromthe WBCand WAE, and the detailed numerical model
for conguration 1. When considering the WBC, the MAE is
below 0.3 dB for all the distances considered, up to 200 km.
For the WAE, the results present an MAE below 0.5 dB up to
180 km for conguration 1. The results are similar for all the
other congurations.
The processing time is also an important metric for
assessing the applicability of a simplied model. Table II
shows the processing time of each model. When considering
conguration 1, the processing time to calculate the On/Off
gain of 100 signals, one by one, by the detailed numerical
model, with a numerical resolution of 200 m and a convergence
criterion of 110
6
, is more than 7.3 s while the WAE takes
less than 100 ms. This means that the WAE is more than
70 times faster than the detailed numerical model for this
conguration. When considering congurations 2 and 3, to
calculate the On/Off gain of 60 signals, one by one, the detailed
numerical model takes more than 2.3 s and 4.7 s, respectively.
When considering the WAE it takes less than 40 ms and
50 ms, respectively, to obtain these results. As we can observe,
the WAE is faster than the detailed numerical model even
in simpler congurations. Also, when considering the WAE,
the processing time is almost the same despite the complexity
of the RFA congurations. As expected, when comparing the
WAE with the undepleted pump model, the computational time
is similar. Regarding the WBC and WAE approaches, their
computational times are the same after nding the parameters
D
int
and
SRS
.
IV. ASE NOISE MODEL AND OSNR
Some planning tools and RWA algorithms are based not only
on the amplier gain prole, but also on the signal OSNR
after the transmission links. Therefore, it is also important
to develop and improve the ASE estimation and the OSNR
prediction for the transmission systems.
The ASE noise of the amplier can be related to the
amplier gain [21,22]. Therefore, we can apply the proposed
approach to better estimate the ASE prole of the RFA. Let us
consider a multi-pumped RFA. The net gain, G, is given by
G =exp(
si gnals
L)

i
g
i
, (18)
where i refers to the pump of the conguration imple-
mented. When considering the undepleted pump model, g
i
=
exp
_
C
s,i
P
i
(L)L
eff,i
_
, and when considering the proposed model
we can estimate g
i
using g
i
= exp
_
C
s,i
P
SRS
i
L

eff,i
_
, where
P
SRS
i
is the corrected input pump power given by Eq. (6)
and L

eff,i
is given by Eq. (12). Based on [21] and considering
the proposed model, the ASE noise power at the end of the
amplier, L, can be written as
P
ASE
(L) = hB
0
_
i
E
i
ln(g
i
)

i
ln(g
i
)
_

__
1+
L
eff
exp(L)

i
ln(g
i
)
_
G
_
1+
L
eff

i
ln(g
i
)
__
, (19)
where E
i
is the spontaneous factor, given by
E
i
=1+
1
exp
_
h
i
kT
_
1
, (20)
and h is Plancks constant, k is Boltzmanns constant, T is
the temperature in Kelvin,
i
is the frequency difference
between the i-th pump and the signal and B
0
is the reference
bandwidth.
Figure 7(a) shows experimental measurements for the ASE
noise of conguration 1, from 1520 nm to 1590 nm. The ASE
is measured without signals and the reference bandwidth is
2.5 GHz. These experimental results are compared with results
obtained with the detailed numerical model, and with results
obtained by the ASE model when using the undepleted pump
model and the WAE model to estimate the gain of the amplier.
We verify a good agreement between the experimental data,
the detailed numerical model and the proposed model.
This improved estimation of the amplier gain and ASE
noise allows a better estimation of the OSNR of the
transmission system. To assess this, we measure the output
signal power and the ASE power experimentally with an OSA,
from 1520 nm to 1590 nm. Figure 7(b) shows the experimental
results measured and the comparison with the undepleted
pump model, with the WAE and with the detailed numerical
model. As we can observe, the OSNR is better described by the
proposed approach than the results obtained by the undepleted
pump model.
Ferreira et al. VOL. 4, NO. 8/AUGUST 2012/J. OPT. COMMUN. NETW. 601
(a)
(b)
Fig. 7. (a) ASEnoise measured and estimated based on the undepleted
pump model and on the WAE for the amplier of conguration 1; (b)
OSNR measurements and estimation based on the undepleted pump
model and based on the WAE for conguration 1.
V. CONCLUSION
A novel model to estimate the On/Off gain and ASE
noise of a transmission link with RFA amplication with
multiple pumps is proposed. This model is based on the
customization of the undepleted pump power with two extra
parameters, D
int
and
SRS
, to account for the pump-to-pump
interactions. Two methods are proposed to adjust these
parameters, one based on a comparison with experimental
data or detailed numerical simulations, WBC, and the other
based on analytical expressions, WAE. The second method
is simpler to implement, because it can be used without
comparison with experimental or numerical simulations. When
comparing the WAE with experimental measurements and
detailed numerical simulations, we found that this model
provides accurate results for a total input pump power up to
900 mW, for a wavelength separation between pumps of up to
60 nm and for a transmission distance of up to 180 km.
The proposed model presents an improved computational
performance without losing signicant accuracy for the most
used pump congurations. In fact, it was veried that the
model can be more than 70 times faster than the detailed
numerical model. Therefore, it can be used in network planning
tools, in applications which have to estimate the gain and the
ASE noise or to design RFAs with a limited computational
power.
ACKNOWLEDGMENTS
This work has been partially supported by the Fundao
para a Cincia e a Tecnologia (FCT), under the re-
search grant SFRH/BDE/51095/2010, and the OSP-HNLF-
PTDC/EEA-TEL/105254/2008 and CONTACT-PTDC/EEA-
TEL/114144/2009 projects.
REFERENCES
[1] A. Morea, F. Leplingard, T. Zami, and N. Brogard, New trans-
mission systems enabling transparent network perspectives,
Compt. Rend., vol. 9, no. 910, pp. 9851001, 2008.
[2] F. Leplingard, T. Zami, A. Morea, N. Brogard, and D. Bayart, De-
termination of the impact of a quality of transmission estimator
margin on the dimensioning of an optical network, in Optical
Fiber Communication Conf., 2008, OWA6.
[3] K. Manousakis, K. Christodoulopoulos, E. Kamitsas, I. Tomkos,
and E. Varvarigos, Ofine impairment-aware routing and wave-
length assignment algorithms in translucent WDM optical net-
works, J. Lightwave Technol., vol. 27, no. 12, pp. 18661877,
2009.
[4] J. Bromage, Raman amplication for ber communications sys-
tems, J. Lightwave Technol., vol. 22, no. 1, pp. 7993, 2004.
[5] C. Headley and G. Agrawal, Raman Amplication in Fiber
Optical Communication Systems. Academic Press, 2005.
[6] M. Zirngibl, Analytical model of Raman gain effects in massive
wavelength division multiplexed transmission systems, Elec-
tron. Lett., vol. 34, no. 8, pp. 789790, 1998.
[7] S. Namiki and Y. Emori, Ultrabroad-band Raman ampliers
pumped and gain-equalized by wavelength-division-multiplexed
high-power laser diodes, IEEE J. Sel. Top. Quantum Electron.,
vol. 7, no. 1, pp. 316, 2001.
[8] S. Chinn, Analysis of counter-pumped small-signal bre Raman
ampliers, Electron. Lett., vol. 33, no. 7, pp. 607608, 1997.
[9] A. Kobyakov, M. Vasilyev, S. Tsuda, G. Giudice, and S. Ten,
Analytical model for Raman noise gure in dispersion-managed
bers, IEEE Photon. Technol. Lett., vol. 15, no. 1, pp. 3032,
2003.
[10] X. Zhou, C. Lu, P. Shum, and T. Cheng, A simplied model
and optimal design of a multiwavelength backward-pumped ber
Raman amplier, IEEE Photon. Technol. Lett., vol. 13, no. 9, pp.
945947, 2001.
[11] S. Cani, L. de Calazans Calmon, M. Pontes, M. Ribeiro,
M. Segatto, and A. Cartaxo, An analytical approximated so-
lution for the gain of broadband Raman ampliers with mul-
tiple counter-pumps, J. Lightwave Technol., vol. 27, no. 7, pp.
944951, 2009.
[12] J. Zhou, J. Chen, X. Li, and W. Jiang, A novel algorithm for
backward-pumped Raman amplier, Fiber Integr. Opt., vol. 24,
no. 6, pp. 529535, 2005.
[13] M. Santagiustina, Exact, implicit, integral solution of depletion
and saturation in Raman and Brillouin ber ampliers, in Eu-
ropean Conf. on Lasers and Electro-Optics, and the Int. Quantum
Electronics Conf., CLEOE-IQEC 2007, 2007.
[14] B. Min, W. Lee, and N. Park, Efcient formulation of Raman
amplier propagation equations with average power analysis,
IEEE Photon. Technol. Lett., vol. 12, no. 11, pp. 14861488, 2000.
602 J. OPT. COMMUN. NETW./VOL. 4, NO. 8/AUGUST 2012 Ferreira et al.
[15] M. Karsek, J. Ka nka, P. Honztko, and P. Peterka, Time-
domain simulation of power transients in Raman bre ampli-
ers, Int. J. Numer. Model., vol. 17, no. 2, pp. 165176, 2004.
[16] J. Park, P. Kim, J. Park, H. Lee, and N. Park, Closed integral
form expansion of Raman equation for efcient gain optimiza-
tion process, IEEE Photon. Technol. Lett., vol. 16, no. 7, pp.
16491651, 2004.
[17] J. Park and P. Kim, Gain and noise gure spectrum control
algorithm for ber Raman ampliers, IEEE Photon. Technol.
Lett., vol. 18, no. 10, pp. 11251127, 2006.
[18] M. Fugihara and A. Pinto, Low-cost Raman amplier for CWDM
systems, Microwave Opt. Technol. Lett., vol. 50, no. 2, pp.
297301, 2008.
[19] J. Ferreira, M. Fugihara, and A. Pinto, Transient response and
control of pump-reecting Raman ber ampliers, Fiber Integr.
Opt., vol. 29, no. 1, pp. 4461, 2010.
[20] J. Ferreira, R. Nogueira, P. Monteiro, and A. Pinto, Raman
amplier undepleted pump model customization to include
pump-to-pump interactions, in IEEE Int. Conf. on Computer as
a Tool (EUROCON), 2011.
[21] C. Fludger, V. Handerek, N. Jolley, and R. Mears, Fundamental
noise limits in broadband Raman ampliers, in Optical Fiber
Communication Conf., 2001, MA5.
[22] N. Muga, M. Fugihara, M. Ferreira, and A. Pinto, Non-Gaussian
ASE noise in Raman amplication systems, J. Lightwave Tech-
nol., vol. 27, no. 16, pp. 33893398, 2009.
Joo Melo Ferreira was born in Oliveira de Azemis, Portugal.
He graduated in Physics Engineering from University of Coimbra,
Portugal, and obtained his M.Sc. degree in Physics Engineering from
University of Aveiro, Portugal, in 2006 and 2008, respectively. He
joined Instituto de Telecomunicaes in 2007 where he has been
working in the eld of optical ber ampliers. In 2009 he joined
Nokia Siemens Networks, Amadora, Portugal as a researcher in the
eld of signal quality in transparent optical networks. He is currently
working toward his Ph.D. degree at the Department of Electronics,
Telecommunications and Informatics, University of Aveiro, Portugal.
His main research interests are optical ber ampliers for wavelength
division multiplexing optical networks, and transient effects in
transparent optical networks.
Rogrio Nogueira graduated in Physics Engineering from the
University of Aveiro, Portugal, in 1998. He also obtained a Ph.D. degree
in Physics at the same university in 2005. He is now an Assistant
Researcher at the Instituto de Telecomunicaes where he has been
working in the eld of ber optics since 1999, participating in several
projects nanced by national and EU organizations and private
companies. He has also been a Research Expert at Nokia Siemens
Networks since 2009. Dr. Nogueira is a co-author of more than 250
publications, including international journals, book chapters, patents
and conferences. His research interests include design and production
of optical components and ber optical communication systems.
Paulo Monteiro was born in Coimbra, Portugal, in 1964. He
received his diploma and doctoral (Ph.D.) degrees in Electronics
and Telecommunications from the University of Aveiro and his
M.Sc. degree from the University of Wales, UK. Presently, he is a
sub-division manager of the Network Optimization unit at NWS ON
RD OE Nokia Siemens Networks Portugal. He is also an Associate
Professor at the University of Aveiro and a researcher at the Instituto
de Telecomunicaes. His main research interests include high speed
communications for access and core optical networks and xedmobile
convergence. He has acted as a reviewer for the IEEE Journal of
Lightwave Technology, IEE Electronics Letters, ETRI Journal, OSA
Journal of Optical Networking and SPIE Optical Engineering. He
has participated in several national and European projects and
also the coordination of the large-scale integrating project FUTON
(FP7 ICT-2007-215533). He has authored/co-authored more than 18
patent applications and over 250 refereed papers and conference
contributions.
Armando N. Pinto (M99-SM07) graduated in Electronic and
Telecommunications Engineering in 1994 and received his Ph.D. de-
gree in Electrical Engineering in 1999, both from the University of
Aveiro. In 2000, he became an Assistant Professor in the Department
of Electronics, Telecommunications and Informatics of the University
of Aveiro, and a researcher at the Instituto de Telecomunicaes.
During the academic year of 20062007 he was a Visiting Professor
at the Institute of Optics, University of Rochester. At the present, he
leads a research group at the Instituto de Telecomunicaes focusing
on high speed optical communication systems and networks. He has
published more than 100 scientic papers in international journals
and conferences. He has served on the technical committees of various
scientic international conferences. He is presently a member of the
Editorial Board of the International Journal of Optics. Dr. Pinto is a
member of the Optical Society of America (OSA).

Вам также может понравиться