0% found this document useful (0 votes)
95 views9 pages

Dynamic Stimulated Brillouin Scattering Analysis: A. Djupsjöbacka, G. Jacobsen, and B. Tromborg

Uploaded by

Anonymous KCES2o
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
95 views9 pages

Dynamic Stimulated Brillouin Scattering Analysis: A. Djupsjöbacka, G. Jacobsen, and B. Tromborg

Uploaded by

Anonymous KCES2o
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

416 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO.

3, MARCH 2000

Dynamic Stimulated Brillouin Scattering Analysis


A. Djupsjbacka, G. Jacobsen, and B. Tromborg

AbstractWe present a new simple analysisincluding the ef- to cross-phase modulation (XPM) between the signal chan-
fect of spontaneous emissionof the (dynamic) influence of stim- nels. XPM is especially severe for adjacent channels with
ulated Brillouin scattering (SBS) on the detected receiver eye dia-
gram. It applies in principle for general types of modulation for-
the same dispersion value because this allows long interac-
mats such as the digital formats of amplitude shift keying (ASK), tion lengths between the channels during transmission.
frequency shift keying (FSK), and phase shift keying (PSK). The Stimulated Raman scattering (SRS) may cause transfer of
analysis is formulated for a determination of the signal power de- signal power among the WDM channels over the total WDM
pletion and intersymbol interference (ISI) caused by the combined
effect of fiber dispersion, fiber attenuation and nonlinear fiber ef-
transmission window leading to cross-talk and power deple-
fects such as the effect of self-phase modulation (SPM) and SBS. tion effects. The strongest SRS gain is for channels with
The analysis allows a quantification of the dithering influence on a relatively large frequency separation (12 THz apart from
the SBS threshold. Representative numerical examples are pre- each other). In a WDM system configuration in the 1550-nm
sented for two single-channel ONOFF modulated 10-Gb/s systems
utilizing FranzKeldysh and MachZehnder-type modulators.
wavelength range SRS leads to power depletion of low wave-
length channels and cross-talk induced power enhancement
Index TermsNonlinear optics, numerical analysis, optical fiber of longer wavelength channels. The SRS effect depends on
communication, optical Kerr effect, optical receivers, stimulated
Brillouin scattering (SBS), transmission system performance. the dispersion difference between interacting signals (the ef-
fective interaction length) in a similar way as XPM.
Another dispersion sensitive nonlinear effect may occur
I. INTRODUCTION due to four-wave mixing (FWM) effects between adjacent
(neighboring) WDM channels. This effect is of special
O PTICAL transmission systems using wavelength-divi-
sion-multiplexing (WDM) techniques and operating in
the 1550nm fiber transmission window are currently being
concern for fibers with low dispersion and low dispersion
slope (dispersion shifted fibers in the 1550-nm transmission
prototyped for use in the telecom fiber back-bone network. window) [8]. FWM effects have especially severe system
One state of the art research system has a channel bit rate of quality consequences using WDM channel plans with the
160 Gb/s and 19 signal channels for transmission over 40 km same interchannel frequency separation between all channels
of dispersion shifted fiber [1]. Another example system has 64 as described in the ITU recommendations [9].
signal channels (channel bit-rate 10 Gb/s) and transmits over One last important nonlinear effect that affects the trans-
7200 km of conventional fiber [2]. mission is the stimulated Brillouin scattering (SBS) effect
In the practical design of these WDM systems (and WDM which differs from all the nonlinear effects discussed so far
based optical networks) that operates with unrepeated trans- in that the SBS gain profile has a relatively narrow frequency
mission lengths of several hundreds of kilometers it is im- bandwidth (of the order of 100 MHz for typical fibers)
portant to have a rigorous system model that accounts for meaning that it does not lead to cross coupling effects among
the linear as well as nonlinear fiber transmission properties WDM signal channels. The SBS leads through the interac-
as discussed in tutorial format in [3], [4]. The transmission tion between an optical pump signal and an acoustic wave in
of an ONOFF modulated signal for a single-channel propa- the fiber to a the set-up of a counter propagating stimulated
gation in the fiber at bit-rates beyond the order of 10 Gb/s (Stokes) wave which is downshifted 11 GHz in frequency
is influenced by fiber attenuation and fiber dispersion (in- relative to the pump. The other stimulated nonlinear pro-
cluding the dispersion slope at the signal wavelength) as well cesses (FWM, SRS) are broad-band processes [bandwidths
as self-phase modulation (SPM) through the signal intensity in the range of hundreds of GHz (FWM) to the order of 12
dependent fiber refractive index. This is especially the case THz (SRS)] where the stimulated waves are copropagating
in applications with cascaded optical amplifiers where the with the pump signal(s).
signal power at the amplifier output is above the order of 1 The dynamic properties for the transmitted optical
mW (0 dBm). For WDM applications with several optical signals are in principle described by the time dependent
signal channels separated in frequency by 100200 GHz the Maxwell equations whichusing an equivalent baseband
intensity dependence refractive index may in addition lead formulationleads to modified forms of the Schrdinger
equations [3]. The equations have a particular simple form
Manuscript received May 3, 1999; revised December 1, 1999. (accounting for the lowest order nonlinear influence) when
A. Djupsjbacka and G. Jacobsen are with Ericsson Telecom AB, Stockholm dealing with time scales above the order of 110 psas is
S-126 25, Sweden. of interest in this paperwhereas they are of significantly
B. Tromborg is with the Research Center COM, Technical University of Den-
mark, Lyngby DK-2800, Denmark. more involved form when dealing with femtosecond time
Publisher Item Identifier S 0733-8724(00)02194-0. scales [3]. From numerical solutions of the Schrdinger
07338724/00$10.00 2000 IEEE
DJUPSJBACKA et al.: DYNAMIC STIMULATED BRILLOUIN SCATTERING ANALYSIS 417

equations the time domain properties of a received optical larization state of the pump wave is essentially constant).1 The
signal can be specifiedfor instance by looking at the equations read
resulting eye diagram. A practical evaluation accounting
for effects of SPM, XPM, SRS, and FWM is in principle
straightforward because all optical signals that are injected
into the transmission fiber or are generated in the stimulated (1)
processes are co-propagating and thus interacts on the
fly, i.e., the resulting optical signal amplitude and phase
at a given time and specified fiber position are uniquely
determined by all other signals specified at that time Here the fiber attenuation is described via the parameter and
instant and at that fiber position. The influence of the SBS the resulting SBS process is specified via the gain (measured
process is more complicated because it involves a long relative to the effective fiber core area ) and the spontaneous
SBS interaction time relative to the bit period and because emission factor which accounts for the (Stokes) wave genera-
the counter propagation of the stimulated wave relative to tion occurring throughout the fiber and originating mainly from
the signal which means that a long bit-sequence has to be spontaneous Brillouin scattering from thermally excited sound
considered when evaluating the dynamic influence of the waves. specifies the intensity in the pump signal which
SBS on the detected eye diagram.In this paper, we will is injected into the fiber at . describes the gener-
present a new simplified formulation of the SBS problem ated and counter propagating stimulated wave (Stokes wave)
for a high bit rate application which allows a quantification which is due to the SBS process. Equation (1) (omitting the
of the SBS influence on the detected eye diagram. In order spontaneous emission terms) originates from [10] where it is
to make the current presentation clear and straightforward formulated for light propagation in liquids and is given in [3],
we will focus our attention on the single-channel case in [4] for the optical fiber case. We have corrected the last for-
the mathematical formulation of the modified Schrdinger mulation by including the normalization for the fiber core area.
equation and in our numerical examples. This means that One should note that the spontaneous emission term in the pump
we will not explicitly consider XPM, SRS, and FWM signal equation is insignificant relative to the contribution due
effects. It should however be noted that our formulation to fiber attenuation. The term is included here in order to give
can be modified in a straightforward way to include these a mathematically consistent model. The spontaneous emission
effects in a WDM case. term is more significant as far as the Stokes signal equation is
concerned. Here it is mandatory to include in order to describe
the on-set of the SBS process in a physically correct way.
Note thatalthough this is not explicitly shown in (1)the
II. THEORETICAL OUTLINE
carrier frequency of is downshifted due to conservation of
momentum in the SBS process by the acoustic wave frequency
A. The Continuous-Wave (CW) Case of around 11 GHz.
The form of (1) shows that a direct numerical solution is pos-
sible. This allows a straightforward numerical implementation
1) Our Simplified Model: As a start in our theoretical treat-
in a form which is compatible with our (dynamic) solution of
ment, we will consider a CW-signal case where a coherent pump
the Schrdinger pump signal equation that will be derived in
wave is injected into the transmission fiber with the (cyclic) car-
Section II-B.
rier frequency , at the transmission distance and with
2) Discussion of Alternative CW Models Relative to Our
intensity . In reality the formulation holds for a CW laser
Approach: In [5] a more rigorous formulation of the CW SBS
pump signal which has a linewidth (caused by phase noise) less
problem is given (on an electromagnetic field basis) based
than the SBS bandwidth which is about 100 MHz. Thus, the
upon three coupled first-order differential equations for the
formulation applies for all practical distributed feedback (DFB)
pump signal, the stimulated signal and the acoustic wave field.
laser sources which have typical linewidths in the order of a
This model is mathematically much more involved than our
few MHz. The resulting signal propagation through the fiber
approach. This rigorous formulation includes the effect of spon-
to the transmission distance is determined through two
taneous emission (earlier theoretical work in this area is found
nonlinear coupled differential equations. The precise form of
in [6]) but requiresas our model formulation doesseparate
the equations depends upon the relative polarization states of
input data for the spectral SBS gain profile. The spontaneous
the pump and stimulated (Stokes) waves. In the following we
emission factor is determined from known physical and mate-
present equations for the case where the polarization states are
rial parameters for the fiber. The model includes the influence
the same for the two waves. This represents a worst case de-
scription of the SBS influence. In the situation above the SBS 1For input pump powers around or below the threshold the coupling to the

threshold this description is however expected to closely rep- Stokes wave occurs over longer fiber lengths and the assumption of constant
(relatve) polarization state over time (or distance) is questionable. The effect of
resent the real situation for an installed transmission fiber (i.e., the polarization variation may be taken into account by multiplying in (1) the
with bends where the bending radius is significantly larger than SBS gain g by a constant which varies between 0.5 (random polarization)
the 10-cm range). This is because the major interaction between and 1 (constant polarization) [3]. In our model we have chosen = 1 in all
cases because we will focus on situations of interest for practical optical com-
the pump and the Stokes waves happens close to the entrance of munication system performance (i.e. pump powers close to or above the SBS
the transmission fiber (i.e., within a fiber length where the po- threshold).
418 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 3, MARCH 2000

of polarization effects. The resulting differential equations a CW signal and contains only one wavelength (11 GHz down-
gives the SBS threshold value and specifies the pump depletion shifted relative to the carrier component of ). This leads
and the Stokes wave generation caused by the SBS process. The to
numerical solution of the equations is feasible for maximum
fiber lengths of about 1 km (which is a short length compared
to lengths of interest for the current problem). Longer lengths
may be considered using in the model the approximation of
considering shorter effective lengths combined with a higher
fiber attenuation (corresponding to that of the actual fiber).
Our more approximate formulation does not specify the spon-
taneous emission parameter from first principles and does not (2)
include polarization effects. We end up with a mathematically
much simpler CW model where we need to specify the SBS
threshold and the SBS bandwidth as input parameters to the
calculation (and to adjust the spontaneous emission factor to
match the selected SBS thresholdsee the discussion in Sec-
tion V). Results using our model describe the typical SBS be-
havior (pump depletion and Stokes wave generation) of impor-
tance for optical communication systems remarkably well for where denotes the carrier frequency.
input pump powers below, around and above the SBS threshold. In (2) we have included the effective SBS bandwidth by as-
Furthermore, the analytical (and numerical) formulation is easy suming that the SBS gain only affects the carrier wave.4 The
to evaluate for transmission lengths of practical interest (in the form of the spontaneous emission term in the Stokes signal
order of 100 km) and it allows in a straightforward way the mod- equation in (2) shows that this contribution is most significant
eling of the dynamic SBS problem as it will be shown in Sec- when the Stokes signal is weak, i.e., in the situation where the
tion II-B. spontaneous emission initiates the Stokes signal generation. In
An analytical solution to (1) (in the absence of spontaneous this case the term must be treated with special care in order to
emission) is presented in [3]. In [7] it is pointed out that this solu- avoid numerical instabilities in connection with imposing the
tion is erroneous. An exact solution is given in [7] in terms of an (Neuman type) boundary condition of for ,
integral equation. The solution could not be expressed in closed where denotes the length of the transmission fiber.
form using tabulated standard functions, i.e., it requires a nu- The spontaneous emission is a stochastic process which can
merical evaluation. A major draw back with the CW SBS model be described by a term in the time domain
that does not account for spontaneous Stokes wave generation equation for [5]. The factor represents the
is thatin order to describe the on-set of the SBS processan random fluctuations of density. By assuming to be a
artificial Stokes signal must be injected at the end of the mod- Langevin noise function which satisfies the correlation relation
eled fiber in the opposite direction of the pump.

B. The Dynamic Pump Wave Model


We now proceed to the dynamic model. In order to move on
to a dynamic Schrdinger type of formulation for the signal the equation for is consistent with the Stokes signal
(pump) we transfer (1) to a set differential equations for the equation for the intensity (1). However, for the present problem
pump and Stokes signal amplitudes. We further assumeas it will be very demanding in terms of computation time to sim-
an initial step toward the time dependent formulationthat the ulate a Langevin noise function, and we shall therefore stick to
signal is specified via an average amplitude spectrum the deterministic form (2), which is expected to model the av-
where denotes the (cyclic) optical frequency parameter and erage behavior to a good approximation.
the spectrum is specified over a long bit-sequence. In order to proceed to the general time varying case with a
is in the discussion that follows assumed to have a dominating specified bit-pattern for the signal equation [i.e., the top equa-
carrier component.2 3 Furthermore, we assume that is tion in (2)] we now modify it to read in a general operator form
2It is possible to consider modulation formats that lead to an average spec- (where the signal amplitude may be specified in the time or the
trum A (!; z ) with significant discrete components away from the carrier (this frequency domain according to computational convenience):
happens for instance in the digital frequency shift keying (FSK) case with mod-

ulation index values m close to being integer (m 1; 2; . . .) [11]). The current
formulation may be modified to cover the SBS influence in such cases. This is
discussed in Section III. (3)
3Due to the chosen modulation format the discrete peaks in the average
spectrum A (!; z ) may become broadened (this happens for instance in the
digital frequency shift keying (FSK) case with modulation index values close 4It is worth noting that the influence of SBS is completely determined from

to being half-integer (m 0:5; 1:5; . . .) [11]). In order to account for this the the resulting carrier amplitude. This means, for instance, that using duobinary

effective peak intensity integrated over the ( 100 MHz) bandwidth of the SBS modulation the suppression of the carrier will lead to a significantly enhanced
process accounts for the SBS effectsee the discussion in Section IV dealing SBS threshold (specified in terms of mean transmitted signal power) as it is well
with the effect of dithering a truly discrete carrier component. known [12].
DJUPSJBACKA et al.: DYNAMIC STIMULATED BRILLOUIN SCATTERING ANALYSIS 419

which is a generalization of the classical split-step presentation only. The coupled differential equations that are used initially
of the Schrdinger equation. It is given here in terms of the dis- are as follows:
persion operator , a nonlinear fiber operator and an SBS
operator . Each operator acts upon the signal one by one
in the numerical solution of the equation. The evaluation is con-
veniently done using Fourier transformation techniques by in-
voking in the frequency domain, in the time domain and
in the frequency domain (to specify the depletion of the carrier
signal component). The formal operator definitions are (7)

(4)

Here, the CW solution is only used to calculate


from a given value of the input power
(and for a value of which is determined from the SBS
where calibration proceduresee Section V). The benefit is that the
differential equation can be solved entirely in the frequency
domain which is much faster than a full split-step approach.
When the -value has been calculated, we use the nonlinear
(5) Schrdinger equation for which is given below to
calculate (at the far end of the fiber, i.e., for ,
km in the examples to be discussed in Section VI):

where denotes the fiber dispersion, the dispersion slope,


the nonlinear part of the fiber refractive index ( and
specified at the transmission wavelength ) and the free space (8)
velocity of light.5
In (3), the strength in the SPM process is determined through
the signal intensity. Thus

where it has been taken into account in the formulation that only
(6)
the carrier component of the pump signal contributes to the gen-
eration of the Stokes signal.
where denotes Inverse Fourier transformation. A more rigorous (and ambitious) evaluation approach for the
The amplitude spectrum is specified from the type of modula- model discussed here is to use the nonlinear Schrdinger equa-
tion as well as the bit-sequence under consideration. The general tion for even in the first phase when is calcu-
formulation which is outlined here thus applies for general mod- lated. However, for most realistic solutions the differences be-
ulation formats including the classical digital amplitude shift tween these two methods are quite small as will be shown by an
keying (ASK), frequency shift keying (FSK) and phase shift example in the Section V.
keying (PSK) formats. In the examples to be considered in the
following we focus on ASK generated using a FranzKeldysh C. The Fully Dynamic SBS Model
and MachZehnder type of external modulator [13], [14]. The The transient SBS analysis can be presented in a form which
bit-sequence is specified in periodic form, i.e., using a pseudo- is principally more rigorous than the present in terms of three
random bit-sequence (PRBS) of length bits ( denotes coupled differential equations, one [Schrdinger typesee (8)]
the order of the PRBS sequence). for the pump wave, a similar one for the stimulated wave and an
In the practical numerical solution of the problem, the SBS equation that describes the acoustic wave (see [3, Section 9.2.4],
process is initially solved in the CW domain acting on the carrier [15] and [16]). Thus the pump, stimulated and acoustic wave are
all treated fully dynamically.
A complete numerical solution of the three coupled equations
is extremely difficult. A significant problem arises because the
time resolved Stokes wave that originates at the fiber input is
D
5The split-step dispersion operator  has been given in a form which applies
for normal single mode fibers (where the influence of the parameter D domi-
caused by the pump wave over the total SBS interaction length
nates compared to the influence of D ) as well as for dispersion shifted fibers of the fiber, i.e., over the order of 20 km (corresponding to con-
(with D D
0 i.e. the influence of is important). tributions from about bit-periods of the signal at 10 Gb/s).
420 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 3, MARCH 2000

Our current dynamic formulation simplifies the evaluation TABLE I


significantly by grasping the main physical effect of the PARAMETERS USED IN THE
SIMULATIONS
SBSnamely, the depletion of the signal carrier wavein this
way avoiding the need for an equation for the acoustic wave
altogether as well as avoiding a dynamic and time resolved
evaluation of the stimulated wave evolution.

III. INFLUENCE OF ADDITIONAL DISCRETE SIDEBANDS


In special signal modulation cases the mean amplitude spec-
trum may contain sidebands that each have a power that exceeds
the SBS threshold. In cases where the sidebands are closer to the
carrier than the SBS bandwidth they have to be included in the
effective carrier pump power in order to determine the effective
carrier depletion as discussed in the previous section. In cases
where the sidebands are separated by frequencies that are more
than the SBS bandwidth each sideband is influenced separately
by the SBS and the power depletion can be accounted for by
considering each sideband separately.

IV. DITHERING
It is well known that by making a periodic modulation of the
signal carrier with a spectral width which is larger
than the bandwidth of the stimulated Brillouin process
(but much lower than the bit rate) the SBS effect is reduced. In a
simple conceptual way this type of dithering modulation may be
seen as leading to a Lorentzian type carrier wave with a (cyclic)
specified half width which is determined by the spectral width
parameter . The relative effective signal carrier power which
is influenced by the SBS process may be evaluated as

(9)

The effect of dithering of the carrier wave may now be in-


cluded in the description of the previous sections by multiplying
the SBS gain in (1), (2), (4), (7), and (8) by the quantity spec-
ified in (9).

V. CALIBRATION CONSIDERATIONS
The calibration of the SBS process is performed in the way
it was discussed for the CW case (Section II-A2). We calibrate
the solution relative to an SBS threshold value for a CW-pump
wave. Within the mathematical framework discussed here the
calibration is done considering a total fiber length and re-
quiring that . This, in turn, allows a specification of
the SBS threshold power by adjusting the spontaneous emission
parameter . In this paper, we define the threshold input power
as the one where the resulting output stimulated signal Fig. 1. Pump (full curves) and stimulated power (dashed curves) as a function
power is 10 dB below, i.e., . of transmission distance. Different full curves are for input powers (i.e. power
at zero fiber length) of 10, 7, 3, and 0 dBm. (Stimulated wave powers are larger
In the following examples, we consider a basic 10-Gb/s 0
than 10 dBm only for input powers of 10 (upper dashed curve), 7 and 3 dBm
single-channel system configuration as specified in Table I. The and fiber lengths of less than a few kilometers.)
fiber and SBS parameters , and are
considered as typical for a normal single-mode fiber whereas
has been specified to give an SBS threshold of 0 dBm (1
mW) which we consider a typical value for such fibers. transmission distance as full curves and the stimulated power
In Fig. 1, we showbased on the solution of (7)the pump as dashed curves. The figure is for input CW powers of 10, 7,
power (i.e., the power in the signal carrier) as a function of 3, and 0 dBm. From the figure, it appears that for transmission
DJUPSJBACKA et al.: DYNAMIC STIMULATED BRILLOUIN SCATTERING ANALYSIS 421

lengths larger than about 5 km the pump power decrease is en-


tirely determined by the fiber attenuation and the transmitted
power level saturates due to the SBS generation throughout the
initial fiber length. For shorter distances we have an SBS in-
fluence which leads to a faster pump power decay and which is
more pronounced for the higher input powers as can also be seen
from the stimulated signal power curves. We observe a clear in-
fluence of SBS for the input pump powers of 3 dBm and above.
In order to further investigate the SBS model we show in
Fig. 2 the pump and stimulated power dependences for input
carrier powers of and dBm. The figure shows
results for the total fiber length of 50 km where Fig. 1 only dis-
plays results for the initial 20 km of transmission length (in order
to show more clearly the SBS generation for high input powers Fig. 2. As Fig. 1 for input powers of 10; 0; 010 ; and 020 dBm.
close to the entrance of the fiber). In Fig. 2, it appears that the
stimulated signal power close to the total transmission length of
km is dominated by spontaneous emission (observe the
Stokes signal power decrease between 45 and 50 km of trans-
mission distance) whereas for shorter fiber lengths there is a
balance between SBS gain, fiber attenuation and spontaneous
emission. Close to the fiber entrance there is a sharp increase
in stimulated signal power for cases above or close to the SBS
threshold of 0 dBm.
The discussion of the CW SBS problem is conveniently sum-
marized in Fig. 3 which showsbased upon the previous fig-
uresthe Stokes power and the transmitted pump power as a
function of the input pump power. It is obvious how the Stokes
wave power grows and the transmitted power saturates above
the SBS threshold (0 dBm). Fig. 3. Transmitted pump power, I (L) (full curveleft hand y -axis) and
In the examples to be considered in the following the influ- Stokes power, I (0) (dashed curveright hand y -axis) as a function of input
pump power, I (0) for a fiber with an SBS threshold of 0 dBm.
ence of SBS is treated relative to the CW calibration presented in
Figs. 1, 2, and 3. This is done by first specifying the mean power
injected into the transmission fiber. Secondly the average ampli-
tude spectrum is evaluated and the power in the carrier
component (for ) is used as the power that determines
the SBS influence using (7). This determines the initial condi-
tion [i.e., the value of ] which is used in the full solution
of the Schrdinger Equation (8) when the final eye diagrams are
evaluated.
We have finally compared a solution of the SBS calibration
problem considering the simple carrier description (7) as well
as the full Schrdinger Equation (8) for the case specified in
Table I. In Section VI we show in Fig. 8 a typical resulting de-
tected eye diagram which is evaluated (using the CW calibra-
tion result for the SBS influence that was discussed above) for
an (average) input power of 10 dBm and using a MachZehnder Fig. 4. Difference between the CW-SBS and Schrdinger-SBS solution for
the 10-dBm example of Fig. 8.
amplitude modulator in the transmitter. The figure will be dis-
cussed in detail in Section VI. As a calibration reference, we
VI. NUMERICAL RESULTS
show in Fig. 4 the difference between this eye diagram and the
one which is obtained using the full Schrdinger equation in the In the following, we consider eye diagrams that are ob-
calibration of the SBS threshold. It is obvious that the differ- tained detecting ONOFF modulated digital signals using a
ence in results is less than 0.2% of the eye opening and thus is photo detector with infinite bandwidth and a fourth-order
not significant and we have chosen to utilize the CW calibration BesselThompson filter with a 3-dB bandwidth of 7.5 GHz
throughout our calculations. (We have in our numerical exam- and a noise bandwidth of 7.8 GHz. On the transmitter side a
ples tested the model for input power levels up to 20 dBm and nonfiltered nonreturn-to-zero (NRZ)-signal with raised-cosine
have found that using the CW calibration method gives similar flanks were used as input signal to the optical modulators. The
accuracy as depicted in Fig. 4 in all cases and speeds up our nu- modulators were modeled with realistic chirp, but otherwise
merical calculations by about a factor of 15.) considered as ideal. The bit-sequence used was a PRBS-10
422 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 3, MARCH 2000

Fig. 5. Detected eye diagram after 50-km transmission using a FranzKeldysh Fig. 6. Same as Fig. 5 for an average input power of 10 dBm.
absorption modulator and for an avarage input power of 2 dBm. Black (gray)
curves are without (with) SBS influence.

pattern, i.e., bits. The eye diagrams show the influence


of ISI effects but do not display the influence of additive
noise contributions due to, for instance, amplified spontaneous
emission noise, shot noise and thermal noise. The optimum
noise bandwidth for the receiver in order to yield the best
bit-error ratio performance is around 7.5 GHz (depending
on the spectral shape of the noise limiting filter). Here, we
have chosen a BesselThompson filter according to the ITU
recommendation for eye diagram measurements [9] and scaled
it to a 10-Gb/s (STM-64/OC-192) system. We have done that
in order to display as clearly as possible the influence of SPM
and SBS effects on the detected eye.
Fig. 7. Detected eye diagram after 50-km transmission using a MachZehnder
In Fig. 5, we show the resulting eye diagram after transmis- amplitude modulator and for an avarage input power of 2 dBm. Black (gray)
sion over 50 km of fiber in the case without (black curves) and curves are without (with) SBS influence.
with influence of SBS (grey curves) for an (average) input power
of 2 dBm. The modulation is in this case accomplished using
a typical 10-Gb/s FranzKeldysh type of external modulator
which gives a resulting time dependent signal chirp in combi-
nation with the non ideal amplitude (in reality intensity) mod-
ulation which also has a finite extinction ratio [13]. The eye is
normalized relative to the situation without any SBS influence.
The signal is influenced by SPM as well as SBS. It is obvious
that we have a small effect of the SBS which is to reduce the
effective detected signal amplitude (eye opening) whereas the
shape of the eye does not change in any significant way.
In Fig. 6, we show the resulting eye diagram after transmis-
sion over 50 km of fiber in the case of an input power of 10 dBm.
The distortion of the top part of the eye in the case without SBS
which is seen when comparing to the previous figure is due to
Fig. 8. Same as Fig. 7 for an average input power of 10 dBm.
enhanced SPM influence at 10 dBm input power. It is obvious
that we have a significant effect of SBS which is to reduce the
effective signal amplitude (eye opening) whereas the shape of OFF signal periodsis clearly identified in the bottom of the
the eye does not change very much. eye-diagram.
In Figs. 7 and 8, we consider the case of using a In Fig. 9, we consider for the FranzKeldysh modulator case
MachZehnder type of modulator with a more ideal signal am- the influence of the detected eye under the influence of dithering
plitude modulation performancestill with a finite extinction of the signal with the dithering bandwidth as parameter in the
ratio [14]. Here a distortion of the eye caused by SBS is more situation with 10 dBm of (average) input power. It is obvious
visible as can be seen comparing the 10 dBm cases with and that the SBS threshold increases to above 10 dBm using a mod-
without SBS influence. The distortionwhich is caused when ulation bandwidth of 500 MHz (or more). Similar results are
SBS creates a sign change in the pump wave amplitude during found using the MachZehnder modulator.
DJUPSJBACKA et al.: DYNAMIC STIMULATED BRILLOUIN SCATTERING ANALYSIS 423

Fig. 9. Detected eye after 50-km transmission using a FranzKeldysh Fig. 11. The influence of considering only carrier depletion (black full curves)
absorption modulator and dithering of the signal with average signal power and to consider in addition depletion of the four nearest descrete sidebands
of 10 dBm. The dithering bandwidth isin the order of increasing eye (gray full curves) for sinusoidal modulation and an input power of 10 dBm.
opening0, 100, 200, and 500 MHz. The dashed curves (magnified by a factor of 10) show the difference between
full black and gray curves.

the combined effect of fiber dispersion, fiber attenuation and


nonlinear fiber effects such as the effect of SPM and SBS.
The carrier depletion of a CW signal is in our numerical
examples used to calibrate the dynamic time dependent
Schrdinger equation. This is done by adjusting the sponta-
neous emission parameter in order to obtain a specified value
for the SBS threshold. Using this value for the spontaneous
emission parameter the SBS influence is uniquely given
for input pump powers below, around as well as above the
threshold. The analysis allows a quantification of the dithering
influence on the SBS threshold.
Representative numerical examples are presented for two
Fig. 10. Detected eye after 50 km of transmission for a sinusoidal signal single-channel ONOFF modulated 10-Gb/s systems utilizing
modulation. Black (gray) curves are without (with) SBS influence for an typical FranzKeldysh and MachZehnder-type modulators. It
average input power of 10 dBm. The frame indicates the part of the figure is obvious that in the FranzKeldysh casewhere the signal
which is magnified in Fig. 11.
modulation is imposed upon the optical signal intensitythe
main effect of SBS on the detected eye is a reduction in the eye
In Figs. 10 and 11, we consider as a final example a situation opening. In the MachZehnder casewhere the modulation is
with sinusoidal amplitude modulation at 10 GHz, which results an ideal modulation of the optical signal amplitudeSBS may
in an amplitude spectrum with discrete signal components sep- in addition to the reduction of the eye opening cause distortion
arated by multiples of 10 GHz. We compare the detected eye of the eye shape due to SBS induced sign changes in the optical
accounting for the SBS influence on the carrier alone as well as signal amplitude during the OFF signal periods.
accounting for the influence also on the four sidebands closest
to the carrier on the low as well as on the high frequency side.
It appears that the influence of SBS on the sidebands (that are REFERENCES
at least 9.5 dB below the carrier at the fiber input) only causes [1] S. Kawanishi, H. Takara, K. Uchiyama, I. Shake, and K. Mori, 3 Tbit/s
minor changes to the detected eye. 2
(160 Gbit/s 16 ch) OTDM-WDM transmission experiment, in Proc.
OFC'99, Postdeadline paper PD1.
[2] N. S. Bergano, C. R. Davidson, C. J. Chen, B. Pedersen, M. A. Mills,
VII. CONCLUSION N. Ramanujam, H. D. Kidorf, A. B. Puc, and M. D. Levonas, 640
Gb/s transmission of sixty-four 10 Gb/s WDM channels over 7200 km
We present a new simplified analysisthat includes in a with 0.33 (bits/s/)/Hz spectral efficiency, in Proc. OFC'99, Postdead-
straightforward way the effect of spontaneous emissionof the line paper PD2.
[3] G. P. Agrawal, Nonlinear Fiber Optics, 2nd ed. New York: Academic,
(dynamic) influence of stimulated Brillouin scattering (SBS) 1995.
on the detected receiver eye diagram. The analysis describes [4] R. W. Boyd, Nonlinear Optics. New York: Academic, 1992.
the main influence of SBS on the transmitted signal as that of [5] S. Rae, I. Bennion, and M. J. Carswell, New numerical model of stim-
ulated Brillouin scattering in optical fibers with nonuniformity, Opt.
depleting the signal carrier wave and it applies in principle for Commun., vol. 123, pp. 611616, 1996.
general types of modulation formats such as the digital formats [6] R. W. Boyd, K. Rzazewski, and P. Narum, Noise initiation of stimulated
of ASK, FSK, and PSK. The dynamic formulation allows in Brillouin scattering, Phys. Rev. A, vol. 42, pp. 55145521, 1990.
[7] L. Chen and X. Bao, Analytical and numerical solutions for steady state
a split step numerical evaluation a determination of the signal stimulated Brillouin scattering in a single-mode fiber, Opt. Commun.,
power depletion and intersymbol-interference (ISI) caused by vol. 152, pp. 6570, 1998.
424 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 18, NO. 3, MARCH 2000

[8] R. W. Tkack, A. R. Craplyvy, F. Forghieri, A. H. Gnauck, and R. M. [16] G. Grosso and A. Hk, Generation of short pulses by stimulated Bril-
Drosier, Four-photon mixing and high speed WDM systems, J. Light- louin scattering in optical fibers, J. Opt. Soc. Amer. B, vol. 10, pp.
wave Technol., vol. 13, pp. 841849, 1995. 946951, 1993.
[9] ITU Recommendations G.957, G.692.
[10] R. H. Inns and I. P. Batra, Saturation and depletion in stimulated light
scattering, Phys. Lett., vol. 28A, pp. 591592, 1969.
[11] R. W. Lucky, J. Salz, and E. J. Weldon Jr., Principles of Data Commu-
nication. New York: McGraw-Hill, 1968, ch. 8. A. Djupsjbacka, photograph and biography not available at the time of pub-
[12] K. Yonenaga and S. Kuwano, Dispersion-tolerant optical transmission lication.
system using duobinary transmitter and binary receiver, J. Lightwave
Technol., vol. 15, pp. 15301537, 1997.
[13] O. Sahln, Optimization of DFB lasers integrated with FranzKeldysh
absorption modulators, J. Lightwave Technol., vol. 12, pp. 969976,
1994. G. Jacobsen, photograph and biography not available at the time of publication.
[14] A. Djupsjbacka, Residual chirp in integrated-optic modulators, IEEE
Photon.s Technol. Lett., vol. 4, pp. 4143, 1992.
[15] E. M. Dianov, B. Ya. Zel'dovich, A. Ya. Karasik, and A. N. Pilipetskii,
Feasibility of suppression of steady-state and transient stimulated Bril-
louin scattering, Sov. J. Quantum Electron., vol. 19, pp. 10511053, B. Tromborg, photograph and biography not available at the time of publica-
1989. tion.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy