1/549

Optical Communication Systems (OPT428)

Govind P. Agrawal
Institute of Optics
University of Rochester
Rochester, NY 14627

JJ
II
2007
c G. P. Agrawal
J
I
Back
Close
Chapter 7:
Dispersion Management
306/549

• Dispersion Problem and Its Solution

• Dispersion-Compensating Fibers
• Dispersion-Equalizing Filters
• Fiber Bragg Gratings
• Optical Phase Conjugation
JJ
• Other Techniques II
J
• High-Speed Lightwave Systems I
Back
Close
Dispersion Problem and Its Solution
• Systems built during 1980s used standard fibers with their 307/549
zero-dispersion wavelength near 1.3 µm.
• Standard fibers have large dispersion near 1.55 µm.
• Operation near zero-dispersion wavelength not realistic
for WDM systems.
• Even with DFB lasers, transmission distance is limited to
1 πc
L< = .
16|β2|B2 8λ 2|D|B2
JJ
• L < 35 km at B =10 Gb/s if we use |β2| ≈ 21 ps2/km. II
• Dispersion must be compensated or managed using a suitable J
technique before old systems can be upgraded to 10 Gb/s. I
Back
Close
Basic Idea
• Pulse propagation in the linear case is governed by
308/549
2 3
∂ A iβ2 ∂ A β3 ∂ A
+ − = 0.
∂z 2 ∂t 2 6 ∂t 3
• Using the Fourier-transform method, the solution is
 
1 ∞ i i
Z
A(z,t) = Ã(0, ω) exp β2ω 2z + β3ω 3z − iωt dω.
2π −∞ 2 6
• Fiber acts as an optical filter with the transfer function
H f (z, ω) = exp(iβ2ω 2z/2 + iβ3ω 3z/6).
JJ
• All dispersion-management schemes implement a dispersion com- II
pensating “filter” that cancels this phase factor. J
I
• If H(ω) = H ∗f (L, ω), the output signal can be restored.
Back
Close
Dispersion-Compensating Filters
Dispersion
Compensation
Fiber Link 309/549
Transmitter Receiver

• Optical field after the filter is given by

 
1 ∞ i i
Z
A(L,t) = Ã(0, ω)H(ω) exp β2ω 2L + β3ω 3L − iωt dω.
2π −∞ 2 6
• Expanding the phase of H(ω) in a Taylor series:
H(ω) ≈ |H(ω)| exp[i(φ0 + φ1ω + 12 φ2ω 2 + 16 φ3ω 3)].
• Constant phase φ0 and time delay φ1 can be ignored.
JJ
• Dispersion compensated when φ2 = −β2L and φ3 = −β3L. II
J
• Signal is restored perfectly only if |H(ω)| = 1 and higher-order I
terms in the expansion are negligible. Back
Close
Dispersion-Compensating Fibers
• Optical filters with H(ω) = H ∗f (L, ω) are not easy to design.
310/549
• Simplest solution: Use a fiber as an optical filter because it
automatically has the desired form of the transfer function.
• This solution was suggested as early as 1980.
• It provides an all-optical, fiber-based solution to the
dispersion problem.
• Special dispersion-compensating fibers (DCFs) developed.
• Such fibers are routinely used for upgrading old fiber links.
JJ
• Such a scheme works well even when the nonlinear effects are not II
negligible as long as the average optical power launched into the J
fiber link is optimized properly. I
Back
Close
Conditions for Dispersion Compensation
• After two fibers of lengths L1 and L2, optical field is given by
311/549
1 ∞
Z
A(L1 + L2,t) = Ã(0, ω)H f 1(L1, ω)H f 2(L2, ω) exp(−iωt)dω.
2π −∞

• If second fiber (DCF) is designed such that H f 1(L1, ω)H f 2(L2, ω)

= 1, the pulse will fully recover its original shape.
• Conditions for perfect dispersion compensation are

β21L1 + β22L2 = 0, β31L1 + β32L2 = 0.

• In terms of dispersion parameter D and dispersion slope S JJ

II
D1L1 + D2L2 = 0, S1L1 + S2L2 = 0. J
I
• First condition sufficient if TOD does not affect a bit stream. Back
Close
Requirements for DCFs
• Consider the upgrade problem for fiber links made with standard
312/549
telecommunication fibers.
• Such fibers have D1 ≈ 16 ps/(km-nm) near 1.55-µm.
• The DCF must exhibit normal GVD (D2 < 0).
• For practical reasons, L2 should be as small as possible.
• This is possible only if the DCF has a large negative value of D2.
• As an example, if we assume L1 = 50 km, we need a 10-km-long
DCF when D2 = −80 ps/(km-nm).
• This length can be reduced to 6.7 km if the DCF is designed to JJ
II
have D2 = −120 ps/(km-nm).
J
• DCFs with larger values of |D2| are preferred to minimize extra I
losses incurred inside a DCF. Back
Close
DCFs for WDM Systems
• For a WDM system, the same DCF must compensate dispersion
313/549
over for all channels.
• The slope condition, S1L1 + S2L2 = 0 must be satisfied.
• Reason: both D1 and D2 are wavelength-dependent.
• The condition D1L1 + D2L2 = 0 is replaced with
D1(λn)L1 + D2(λn)L2 = 0 (n = 1, . . . , N),

• Near the ZDWL of a fiber, D j (λn) = Dcj + S j (λn − λc).

• Dispersion slop of the DCF should satisfy
JJ
S2 = −S1(L1/L2) = S1(D2/D1). II
J
• Ratio S/D, called relative dispersion slope should be the same for I
both fibers. Back
Close
Negative-Slope DCFs
• Using D ≈ 16 ps/(km-nm) and S ≈ 0.05 ps/(km-nm2), ratio S/D
314/549
is positive and about 0.003 nm−1 for standard fibers.
• Since D is negative for a DCF, S should also be negative such that
S2/S1 = D2/D1.
• For a DCF with D ≈ −100 ps/(km-nm), dispersion slope S should
be −0.3 ps/(km-nm2).
• The use of negative-slope DCFs offers the simplest solution for
WDM systems with a large number of channels.
• Such DCFs were developed and commercialized during the 1990s. JJ
II
• In 2001, broadband DCFs were used to transmit 101 channels, each
J
operating at 10 Gb/s, over 9,000 km. I
Back
Close
Dispersion Maps
• A fiber link may contain multiple types of fibers with different dis-
315/549
persion characteristics.
• Solution for an arbitrary form of β2(z) is given by
 
1 ∞ i
Z
A(z,t) = Ã(0, ω) exp da(z)ω 2 − iωt dω.
2π −∞ 2
Rz
• Total accumulated dispersion da(z) = 0 β2(z0) dz0.
• Dispersion management requires da(L) = 0 at the end of a fiber
link so that A(L,t) = A(0,t).
JJ
• Three schemes used in practice: (a) precompensation, (b) post- II
compensation, and (c) periodic compensation. J
I
Back
Close
Dispersion Maps
Precompensation

Accumulated Dispersion
316/549
DCF

Distance along Fiber Link

(a)
Accumulated Dispersion

Distance along Fiber Link

Postcompensation

DCF
(b)

JJ
Accumulated Dispersion

Distance along Fiber Link

II
J
Periodic Dispersion Map
I
(c)
Back
Close
Dispersion Maps
• Precompensation: Dispersion accumulated over the entire link is
317/549
compensated at the transmitter end.
• Postcompensation: A DCF of appropriate length is placed at the
receiver end.
• Periodic compensation: Dispersion is compensated in a periodic
fashion all along the link.
• For a truly linear system (no nonlinear effects), all three schemes
are identical.
• Three configurations behave differently when nonlinear effects are JJ
included. II
J
• System performance improved by optimizing dispersion map. I
Back
Close
Single-Mode DCF Design
• In a single-mode design, V parameter is made close to 1.
318/549
• Accomplished in practice by reducing the core size
(diameter 4–5 µm).
• A large fraction of the mode propagates outside the core.
• Waveguiding contribution to dispersion is enhanced, resulting in
large negative values of D.
• Values of D < −100 ps/(km-nm) can be realized.
• Such DCFs suffer from two problems, both resulting from their
relatively narrow core diameter. JJ
II
• Relatively high losses (α = 0.4–0.6 dB/km). J
I
• Nonlinear parameter γ is larger by about a factor of 4.
Back
Close
Improved DCF Design
319/549

• DCF is designed with two concentric cores, separated by a ring-

shaped cladding region.
• Size parameters a, b, and c and refractive indices n1, n2, and n3
optimized to realized desired dispersion characteristics. JJ
II
• D can be as large as −5,000 ps/(km-nm) when a = 1 µm, b =
J
15.2 µm, and c = 22 µm. I
Back
Close
Two-Mode DCF
• Best Solution: Employ a two-mode DCF with V ≈ 2.5.
320/549
• Second mode exhibits large negative values of D.
• A 1-km length can compensate dispersion accumulated over 50 km,
while adding little extra loss or nonlinear degradation.
• The use of a two-mode DCF requires a mode-conversion device.
• Mode converter should be polarization-insensitive and operate over
a broad bandwidth.
• A long-period grating is used for this purpose.
JJ
• Grating period Λ ∼ 100 µm is chosen to match the index difference II
δ n̄ between two modes (Λ = λ /δ n̄). J
I
Back
Close
Two-Mode DCF Design
321/549

(a) (b)

• First grating transfers power to higher-order mode.

• Seconds grating transfers power back into fundamental mode.
• Measured dispersion characteristics of such a 2-km-long DCF show
D = −420 ps/(km-nm) near 1,550 nm. JJ
II
• Such DCFs are polarization-insensitive, exhibit low insertion loss,
J
and offer dispersion compensation over the entire C band. I
Back
Close
Photonic-Crystal Design
322/549

• Photonic-crystal fibers contain a two-dimensional array of air holes

that modify dispersion characteristics.
• D for a PCF is also depends on the core diameter.
• Values as large as −2,000 ps/(km-nm) are possible with a suitable JJ
II
design.
J
• Broadband dispersion compensation can be realized by tailoring size I
and spacing of air holes. Back
Close
Reverse-Dispersion Fibers
• Such fibers are designed such that the signs of both D and S are
323/549
reversed compared to standard fibers.
• Dispersion is compensated using fiber sections of same lengths.
• Lengths of fiber sections are reduced below 10 km so that the map
period Lm becomes a small fraction of amplifier spacing LA.
• This technique is referred to as short-period or dense dispersion
management.
• Length of fiber drawn from a single perform is close to 5 km.
• Fiber cable is made by combining two types of fibers, resulting in a JJ
II
dispersion-free cable.
J
I
Back
Close
Dispersion-Equalizing Filters
324/549

• A shortcoming of DCFs is that a relatively long length (>5 km) is

required.
• Losses encountered within each DCF add considerably to total link
loss.
JJ
• Most dispersion-equalizing filters are relatively compact.
II
• Such a filter can be combined with the amplifier to compensate J
fiber losses and dispersion simultaneously in a periodic fashion. I
Back
Close
Fabry–Perot Filters
• Any interferometer acts as an optical filter because its
325/549
transmission (or reflection) is frequency dependent.
• A simple example is provided by the Fabry–Perot interferometer.
• The only problem is that its transfer function affects both the
amplitude and phase.
• A good dispersion-equalizing filter should affect only the phase of
light propagating through it.
• This problem can be solved by using a Gires–Tournois
interferometer. JJ
II
• It is just a FP interferometer whose back mirror is made
J
100% reflective. I
Back
Close
Gires–Tournois Filters
• Transfer function of a GT filter:
  326/549
−r + exp(iωTr )
HGT(ω) = H0 .
1 − r exp(−iωTr )

• Constant H0 takes into account all losses, |r|2 is front-mirror

reflectivity, and Tr is round-trip time within the cavity.
• If losses are constant over the signal bandwidth, only spectral phase
is modified by such a filter.
• Phase φ (ω) of HGT(ω) is far from ideal.
JJ
• It is a periodic function, peaking at frequencies that correspond to II
longitudinal modes of the cavity. J
• Near each peak, phase variations are nearly quadratic in ω. I
Back
Close
Dispersion of Gires–Tournois Filters
• Group delay τg = dφ (ω)/dω, is also a periodic function.
327/549
• φ2 = dτg/dω is related to the slope of the group delay as
φ2 = 2Tr2r(1 − r)/(1 + r)3.

• For a 2-cm-thick GT filter designed with r = 0.8, φ2 ≈ 2, 200 ps2.

• Such a filter can compensate dispersion acquired over 110 km of
standard fiber.
• A GT filter can compensate dispersion for multiple channels
simultaneously as it exhibits a periodic response.
JJ
• Periodic nature also indicates that φ2 is same for all channels. II
J
• A GT filter cannot compensate for the dispersion slope of
I
transmission fiber without suitable design modifications. Back
Close
Dispersion Slope Compensation
GT1 GT2
328/549

Group Delay
Input Output

(a) (b) Wavelength

• In one approach, two GT filters are cascaded in series.

• Two filters have different cavity lengths and reflectivities, resulting
in slightly shifted peaks and different amplitudes.
JJ
• Figure shows group delay for individual filters and the total group II
delay (gray curve). Dark lines show the slope. J
I
• Different slopes indicate different dispersion near each peak. Back
Close
Mach–Zehnder Interferometer
329/549

• A MZ interferometer constructed by connecting two

3-dB directional couplers in series.
• First coupler splits input signal into two equal parts.
• Different phase shifts acquired in the MZ arms.
• Two fields interfere at the second coupler. JJ
II
• Transfer function for the bar port J
HMZ(ω) = 12 [1 + exp(iωτ)]. I
Back
Close
Mach–Zehnder Chain
330/549

• A cascaded chain of several MZ interferometers used in practice.

JJ
• Fabricated in the form of a planar lightwave circuit using II
silica-on-silicon technology. J
I
• A chromium heater provides thermo-optic control of phase shift.
Back
Close
Mach–Zehnder Chain
• Functioning of the MZ chain can be understood as follows.
331/549
• Higher-frequency components of a pulse propagate in the longer
arm of the MZ interferometers.
• Lower-frequency components take the shorter route.
• Relative delay is just the opposite of that introduced by a standard
fiber exhibiting anomalous dispersion.
• In a 1994 implementation, a MZ chain with only five MZ interfer-
ometers provided a relative delay of 836 ps/nm.
• Such a 5-cm device can compensate dispersion acquired over 50 km. JJ
II
• Main limitations: Relatively narrow bandwidth (∼10 GHz) and sen- J
sitivity to input polarization. I
Back
Close
Dispersion-Slope Compensation
332/549

• A planar lightwave circuit capable of compensating both dispersion

and dispersion slope is used.
• A separate MZ chain is employed for each WDM channel.
• WDM signal demultiplexed and then multiplexed back using arrayed
waveguide gratings (AWGs). JJ
II
• All components can be integrated on a single chip using silica-on- J
silicon technology. I
Back
Close
All-Pass Filters
333/549

(a) (b) (c)

• (a) A simple ring resonator with a built-in phase shifter; cascading

of multiple rings increases the amount of dispersion.
• An asymmetric or symmetric MZ configuration also acts as an all-
pass filter.
• Phase shifters are incorporated using thin-film chromium heaters.
JJ
• Such devices can compensate even the dispersion slope of a fiber. II
• One device exhibited dispersion that varied from −378 to J
−3026 ps/nm depending on the channel wavelength. I
Back
Close
All-Pass Filters
334/549

• (a) A transmissive filter with controllable dispersion for each channel

through optical delay lines and phase shifters.
• (b) A reflective filter with a fixed mirror.
JJ
• (c) A reflective filter with moving mirrors acting as delay lines. II
J
• Such designs, although complicated, provide the most flexibility. I
Back
Close
Fiber Bragg Gratings
• Bragg gratings act as optical filters because of a stop band.
335/549
• Light reflected back if its wavelengths falls within stop band.
• Stop band centered at the Bragg wavelength: λB = 2n̄Λ.
• Grating period Λ ≈ 0.5 µm near 1.55 µm.
• A holographic technique is used for making Bragg gratings.
• Use of gratings for dispersion compensation proposed in the 1980s.
• Their use became practical after 1990.
• Fiber gratings are available commercially and used routinely JJ
for a variety of applications. II
J
I
Back
Close
Coupled-Mode Equations
• Refractive index varies along the length periodically as
336/549
n(z) = n̄ + ng cos(2πz/Λ).
• Index modulation depth ng ∼ 10−4.
• Bragg gratings analyzed using coupled-mode equations
dA f /dz = +iδ A f + iκAb,
dAb/dz = −iδ Ab − iκA f .
πng Γ
• Detuning δ = 2π 2π
λ0 − λB and coupling coefficient κ = λB .

• Transfer function is found to be

JJ
Ab(0) iκ sin(qLg) II
H(ω) = r(ω) = = .
A f (0) q cos(qLg) − iδ sin(qLg) J
I
• Dispersion relation q2 = δ 2 − κ 2 (Lg = grating length). Back
Close
Grating-Induced Dispersion
• Dispersion of the grating is related to the frequency dependence of
337/549
the phase of H(ω).
• Grating-induced dispersion exists mostly outside the stop band.
• In this region (|δ | > κ), dispersion parameters are
sgn(δ )κ 2/v2g 3|δ |κ 2/v3g
β2g =− 2 , β3g = 2 .
(δ − κ 2)3/2 (δ − κ 2)5/2

• Grating dispersion normal (β2g > 0) on the “red” side of the stop
band (used for dispersion compensation).
JJ
• A single 2-cm-long grating can compensate dispersion II
accumulated over 100 km of fiber. J
I
Back
Close
Apodized Gratings
338/549

(a) (b)

• An apodization technique is used to improve grating response.

• Index change ng nonuniform, resulting in a z-dependent κ.
JJ
• Reflectivity spectrum of an apodized 7.5-cm-long grating. II
J
• In some gratings κ is varied linearly over length.
I
Back
Close
Chirped Fiber Gratings
339/549

• Period n̄Λ of a chirped grating nonuniform over its length.

• Bragg wavelength λB = 2n̄Λ also varies along grating length. JJ
II
• Equivalent to multiple cascaded gratings with different λB.
J
• Resulting stop band can become quite wide (>1 nm). I
Back
Close
Dispersion of Chirped Gratings
• Origin of Dispersion: Different spectral components of an optical
340/549
pulse are reflected at different points within the grating where the
Bragg condition is satisfied locally.
• Low-frequency components of a pulse are delayed more if optical
period increases along the grating.
• This situation corresponds to anomalous GVD.
• The same grating can provide normal GVD if it is flipped.
• Optical period n̄Λ of the grating should decrease for it to provide
normal GVD. JJ
II
• Dispersion magnitude determined by the rate at which
J
n̄Λ decreases. I
Back
Close
Dispersion Parameter
• Dispersion parameter Dg of a chirped grating of length Lg is
341/549
determined from the relation TR = DgLg∆λ .
• Here TR is the round-trip time and ∆λ is the difference in the Bragg
wavelengths at the two ends of the grating.
• Since TR = 2n̄Lg/c, grating dispersion is given by Dg = 2n̄/(c∆λ ).
• As an example, Dg ≈ 5 × 107 ps/(km-nm) for a grating bandwidth
∆λ = 0.2 nm.
• Because of such large values of Dg, a 10-cm-long chirped grating
can compensate dispersion acquired over 300 km. JJ
II
• This is remarkable for an optical filter that is only 10 cm long.
J
I
Back
Close
Apodized Chirped Fiber Gratings
342/549

• Fraction F of the grating length over which a chirped grating is

apodized plays an important role.
• (a) Reflectivity and (b) group delay for chirped gratings with 50%
(solid) or 95% (dashed) reflectivity for different values of F. JJ
II
• Group delay should vary with wavelength linearly to produce a con-
J
stant GVD across the signal spectrum. I
• It should be as ripple-free as possible. Back
Close
Mesured Reflectivity Spectrum
343/549

Normalized Reflectivity (dB)

Time Delay (ps)

Wavelength (nm)

• Measured reflectivity and group delay for a linearly chirped fiber

grating with a bandwidth of 0.12 nm.
• In a 1996 experiment, two chirped gratings were cascaded in JJ
series to compensate fiber dispersion over 537 km. II
J
• Chirped gratings work as a reflection filter. An optical circulator is I
used in practice to reduce insertion losses. Back
Close
Chirped Gratings for WDM Systems
RIANT et al.: CHIRPED FBG’S FOR WDM CHROMATIC DISPERSION COMPENSATION 1317

344/549

Fig. 12. Experimental setup for system characterization of CDC’s.

• A chirped grating can have a stop band as wide as 10 nm if it is
made long enough.
• When WDM bandwidth is larger than that, several gratings are
JJ
cascaded in series.
Fig. 13. Eye diagrams at the input, after 80 km without compensation and
with compensation.
II
• By
ensure2000,
operation inthis
the linearapproach was
regime. The receiver consistsapplied
of to a 32-channel WDM system J
a preamplifier, two 0.4- and 1.2-nm bandwidth filters, and a
10-Gb/s electrical receiver.
with 18-nm bandwidth. I
The system characterization starts by launching each wave-
length individually through the whole transmission line (5 80 Back
km) and by using the DCG’s independently. The best BER
measurements are recorded by fine tuning the input wavelength Close
to overcome the problems due to variations in the time delay.
Sampled Gratings
Subgratings Sampling period
345/549

Λs

• A sampled or superstructure grating consists of multiple subgratings

separated from each other by a section of uniform index.
• Each subgrating is a sample, hence the name “sampled” grating.
• Made by blocking certain regions during fabrication such that κ = 0
in the blocked regions.
• It can also be made by etching away parts of an existing grating. JJ
II
• New feature: κ(z) varies periodically along z. J
• This periodicity modifies the stop band. I
Back
Close
Amplitude-Sampled Gratings
• Coupled-mode equations show that a sampled grating exhibits mul-
346/549
tiple periodic stop bands.
• Spacing ∆ν p among reflectivity peaks is set by sample period Λs as
∆ν p = c/(2ngΛs).
• If subgratings are chirped, dispersion of each reflectivity peak is
governed by the local chirp.
• Sampling period Λs shoud be about 1 mm to ensure that ∆ν p is
close to 100 GHz.
• In the simplest kind of grating, sampling function is a “rect” JJ
function such that S(z) = 1 over each subgrating. II
J
I
Back
Close
Design of Sampled Gratings
• Shape of the reflectivity spectrum is governed by the Fourier trans-
347/549
form of S(z).
• For a“rect” function S(z), reflectivity follows a “sinc” function.
• A constant reflectivity for all peaks can be realized using
S(z) = sin(az)/az.
• Dispersion slope can be compensated by introducing a chirp in the
sampling period Λs, in addition to the grating period.
• Figure shows the reflection and dispersion characteristics of a 10-
cm-long grating designed for 8 channels with 100-GHz spacing. JJ
II
J
I
Back
Close
Dispersion of Sampled Gratings
348/549

JJ
II
J
I
Back
Close
Phase-Sampled Gratings
• Amplitude smapling impractical as the number of WDM channels
349/549
increases.
• In phase-sampled gratings S(z) modifies phase of κ, rather than its
amplitude.
• In contrast with the case of amplitude sampling, refractive index is
modulated over the entire grating length.
• Mathematically, index variations are of the form

n(z) = n̄ + ng Re{exp[2iπ(z/Λ0) + iφs(z)]}.

JJ
• Reflectivity, group delay, and dispersion of a phase-sampled II
grating designed for 16 WDM channels are shown in the J
following figure. I
Back
Close
Phase-Sampled Gratings
350/549

JJ
II
J
I
Back
Close
Optical Phase Conjugation
351/549

• Four-wave mixing used to generate phase-conjugated idler field

in the middle of fiber link.
• β2 reversed for the phase-conjugated field:
∂ A iβ2 ∂ 2A ∂ A∗ iβ2 ∂ 2A∗
+ =0 → − = 0.
∂z 2 ∂t 2 ∂z 2 ∂t 2
JJ
• Pulse shape restored at the fiber end. II
• Basic idea patented in 1979. J
I
• First experimental demonstration in 1993. Back
Close
Thory Behind Phase Conjugation
• Pulse spectrum just before the phase conjugator:
352/549
Ã(L/2, ω) = Ã(0, ω) exp(iω 2β2L/4).
• Pulse spectrum just after phase conjugation:
Ã∗(L/2, ω) = Ã∗(0, −ω) exp(−iω 2β2L/4).

• Spectrum inverted because ωc = 2ω p − ω.

• Optical field at the end of fiber link:
   
1 L i
Z ∞
A∗(L,t) = Ã∗ , ω exp β2Lω 2 − iωt dω.
2π −∞ 2 4
JJ
∗
• It is easy to see that A(L,t) = A (0,t). II
J
• Pulse shape restored to its input form irrespective of
I
how much pulse broadened in the first section. Back
Close
SPM Compensation
• Using A(z,t) = B(z,t)p(z), pulse propagation is governed by
353/549

∂ B iβ2 ∂ 2B
+ 2
= iγ p(z)|B|2B.
∂z 2 ∂t
• Signs of both β2 and γ change when B → B∗.
• Both SPM and GVD can be compensated by OPC when p(z) = 1.
• Fiber losses destroy this important property of midspan OPC.
• Physical reason: SPM-induced phase shift is power dependent.
• Much larger phase shifts are induced in the first-half of the link than JJ
the second half. II
J
• Use of an optical amplifier at z = L/2 does not help. I
Back
Close
SPM Compensation
Phase
Conjugator 354/549
Transmitter Receiver

Fiber Link Fiber Link

(a)
Power

Distance
(b)
JJ
• Dashed line shows p(z) required for SPM compensation (p(z) = II
p(L − z)). J
I
• Distributed amplification helps to some extent.
Back
Close
Dispersion-Decreasing Fibers
• Perfect compensation of both GVD and SPM can be realized by
355/549
employing dispersion-decreasing fibers.
• In such fibers |β2| decreases along fiber length.
Rz
• With the transformation ξ = 0 p(z) dz,
∂B i ∂ 2B
+ b(ξ ) 2 = iγ|B|2B.
∂ξ 2 ∂t
• Effective dispersion parameter b(z) = β2(z)/p(z).
• If β2(z) decreases in exactly the same way as p(z), b(z) becomes
independent of z as the ratio remains constant. JJ
II
• Thus, GVD should decrease as e−αz. J
I
• Such fibers can be made by tailoring core radius of the fiber. Back
Close
Experimental Results
356/549

• A long fiber used for OPC in a 1993 experiment.

• Pump wavelength coincided with zero-dispersion wavelength. JJ
II
• Practical issues: Wavelength shift of OPC signal, polarization sen- J
sitivity, insertion losses, higher-order dispersion, etc. I
Back
Close
WDM Systems
357/549
-
.

-
.

• A Periodically poled lithium niobate (PPLN) can also act as a phase

conjugator.
• It was used in 2004 to demonstrate transmission of 16 channels (at
40 Gb/s) over 800 km of standard fiber.
JJ
• A single pump phase-conjugated all 16 WDM channels as it inverted II
the signal spectrum around the pump wavelength. J
I
Back
Close
Prechirp Technique
• Modifies input pulses before they are launched into fiber link.
358/549
• Prechirping of input pulse modifies a Gaussian pulse as
"  2#
1 + iC t
A(0,t) = A0 exp − .
2 T0

• Suitably chirped pulses can propagate over longer distances

before they broaden outside its bit slot.
√
• Assuming broadening by 2 is tolerable,
√
C + 1 + 2C2
L= LD . JJ
1 +C2
II
• Maximize L with respect to the chirp parameter C. J
√ √ I
• L = 2LD for C = 1/ 2 (41% increase). Back
Close
Prechirp Technique (continued)
359/549

• Frequency of DFB laser modulated (FM) through

direct current modulation. JJ
II
• An external modulator modulates envelope (AM). J
I
• Simultaneous AM and FM produces chirped pulses. Back
Close
Prechirp Technique (continued)
• FM optical signal can be written as
360/549
2
E(0,t) = A0 exp(−t /T02) exp[−iω0(1 + δ sin ωmt)t],

• Near pulse center, sin(ωmt) ≈ ωmt, and

"  2#
1 + iC t
E(0,t) ≈ A0 exp − exp(−iω0t).
2 T0

• Effective Chirp parameter C = 2δ ωmω0T02.

• Both the sign and magnitude of C can be controlled by changing
FM parameters δ and ωm. JJ
II
• Phase modulation can also be used: J
E(0,t) = A0 exp(−t 2/T02) exp[−iω0t + iδ cos(ωmt)]. I
Back
Close
FSK Format
361/549

• FSK: 1 and 0 bits transmitted with different carrier wavelengths.

• Two wavelengths travel at slightly different speeds.
• Wavelength shift ∆λ delays 0 bits by ∆T = DL∆λ . JJ
II
• ∆λ chosen such that ∆T = TB = 1/B. J
I
• This scheme is called dispersion-supported transmission.
Back
Close
Duobinary Coding
• Duobinary coding reduces signal bandwidth by 50%.
362/549
• Dispersive effects reduced for a smaller-bandwidth signal.

• Two successive bits in the digital bit stream summed

to form a three-level duobinary code at half the bit rate.

1 + 1 = 2, 0 + 0 = 0, 0 + 1 = 1, 1 + 0 = 1.

• Receiver design quite complicated because of the ambiguity between

0 + 1 and 1 + 0 combinations.
JJ
• Phase information is used to distinguish the two. II
J
I
Back
Close
Nonlinear Prehirping
363/549

• Amplify transmitter output using an SOA.

• Gain saturation leads to time-dependent variations in the carrier
density, and thus in the refractive index.
JJ
• SOA not only amplifies the pulse but also chirps it. II
J
• Input pulse compressed when β2 < 0. I
• 16-Gb/s signal transmitted over 70 km of standard fiber. Back
Close
SPM-Induced Prehirping
• Uses self-phase modulation (SPM) for chirping the pulse.
364/549
• Transmitter output passed through a fiber of suitable length:
p
A(0,t) = P(t) exp[iγLmP(t)].

• In the case of Gaussian pulses

"  2#
√ 1 + iC t
A(0,t) ≈ P0 exp − exp(−iγLmP0).
2 T0

• Effective SPM-induced hirp parameter: C = 2γLmP0. JJ

• Transmission fiber itself can be used for chirping the pulse. II
J
• This is the basic idea behind solitons. I
Back
Close
Postcompensation Techniques
• Employs an electronic technique at the receiver.
365/549
• Relatively easy to implement if a heterodyne receiver is used.
• Heterodyne receivers first convert data into microwave format.
• A microwave bandpass filter cancel the effects of GVD.
• Much harder to solve the GVD problem for direct detection since
all phase information is lost.
• Several nonlinear equalization techniques permit signal recovery.
• They require electronic logic circuits operating at the bit rate. JJ
• Electronic equalization limited to low bit rates and to distances of II
J
only a few dispersion lengths.
I
Back
Close
Tunable Dispersion Compensation
• Not all WDM channels can be compensated perfectly
366/549
by a single DCF.
• Residual dispersion for each channel needs compensation at the
receiver (called postcompensation).

• Precise amount of residual dispersion not known in practice

(dispersion variations along fiber length).
• Dynamic variations can occur because of temperature
fluctuations.
JJ
• Solution: Tunable dispersion compensation II
J
I
Back
Close
Stretched Fiber Gratings
367/549

JJ
• Dispersion tuned by stretching a nonlinearly chirped grating. II
J
• Grating is placed on a mechanical stretcher and a piezoelectric I
transducer is used to stretch it. Back
Close
Role of Nonlinear Chirp
2 Lg
R
• In a chirped grating, group delay τg = c 0 n̄(z) dz. 368/549

• Stress-induced changes in mode index n̄ change the local Bragg

wavelength as λB(z) = 2n̄(z)Λ(z).
• Slope of group delay at a given wavelength does not change when
n̄ is a linear function of z.
• Grating dispersion is given by
Z Lg

dτg 2 d
Dg(λ ) = = n̄(z) dz .
dλ c dλ 0
JJ
• Value of Dg at any wavelength can be altered by changing mode II
index n̄ (through heating or stretching). J
I
Back
Close
Stretched Fiber Gratings
369/549

• Group delay as a function of wavelength at several applied

voltages for a 5-cm-long nonlinearly chirped fiber grating.
• For a fixed channel wavelength, dispersion can be changed JJ
II
from −300 to −1,000 ps/nm by changing voltage.
J
• Tunable compensation for multiple channels possible by using I
a sampled grating with nonlinear chirp. Back
Close
Temperature Tuning
370/549

(a) (b)

• Grating is made with a linear chirp, and a temperature gradient is

used to produce tunable dispersion.
• Distributed heating requires a thin-film heater deposited on the
JJ
outer surface of the fiber grating. II
• (a) Reflection spectrum and (b) total GVD as a function of voltage J
for a fiber grating with temperature gradient. I
Back
Close
Temperature Tuning
371/549
-

• A segmented thin-film heater provides better temperature control.

JJ
• 32 chromium heating elements formed on a silica substrate.
II
• Only a few volts required to change dispersion slope from +100 to J
−300 ps/nm2. I
Back
Close
Experimental Results
372/549

• Solid and dashed curves show power penalties with (filled circles)
and without (empty circles) the dispersion equalizer.
JJ
• Recorded eye diagrams are shown at two data points (arrows). II
• Tolerable dispersion range can be more than doubled. J
I
Back
Close
Virtually Imaged Phased Array
373/549

• A virtually imaged phased array can provide tunable dispersion.

• Signal is focused onto a tilted glass plate with 100% and 98% re-
flecting layers on its front and back surfaces.
• This arrangement creates multiple beams that appear to diverge
from an array of virtual images. JJ
• Interference among these beams produces output at an angle that II
varies with wavelength. J
I
Back
Close
Virtually Imaged Phased Array
374/549

• Light is focused on a mirror that provides controllable wavelength- JJ

dependent group delay by moving the mirror along one axis. II
J
• Dispersion can be varied from −800 to +800 ps/nm. I
Back
Close
Higher-Order Dispersion Management
375/549

• Third-order dispersion requires β31L1 + β32L2 = 0.

• Necessary when short pulses are used at high bit rates.
• Cascaded MZ filters can be used for this purpose.
• Pulse distorted when a 2.1-ps pulse was transmitted over 100 km.
JJ
• Equalizing filter eliminated oscillatory tail and reduces pulse width II
to 2.8 ps. J
I
• Residual increase in the pulse width is due to PMD. Back
Close
Cascaded Chirped Fiber Gratings
376/549

• A nonlinearly chirped fiber grating can compensate TOD.

• Cascading of two chirped gratings ccomensates β3 without
affecting β2.
• One of the chirped grating is flipped so that the combination JJ
II
provides no net GVD.
J
• Their TOD contributions add up to produce a nearly parabolic shape I
for the relative group delay. Back
Close
Cascaded Chirped Fiber Gratings
377/549

• A linearly strain-chirped fiber Bragg grating (LSCFBG) is

cascaded with another that is nonlinearly chirped (NSCFBG).
• Both gratings are mounted on a substrate that could be bent by
moving a block. JJ
II
• It was possible to change only dispersion slope from J
0 to −58 ps/nm2 over a bandwidth of 1.7 nm. I
Back
Close
PMD Problem
378/549

• A PMD-limited system is quantified through outage probability.

• Outage probability depends on data format; performance
better for RZ format with shorter pulses.
JJ
• Outage probability < 10−5 (5 min/year) is required. II
J
• Average DGD should satisfy σT < 0.1/B. I
Back
Close
Need for PMD Compensation
• Average pulse broadening governed by the PMD parameter:
379/549
√
σT = D p L.

• If we use BσT = 0.1, B2L < (10 D p)−2.

of “old” fiber links, B2L < 104 (Gb/s)2-km, if we use
• In the case √
D p = 1 ps/ km as a representative value.
• Such fibers require PMD compensation at B = 10 Gb/s when link
length exceeds even 100 km.
√
• For modern fibers D p < 0.1 ps/ km. As a result, B2L exceeds
JJ
106 (Gb/s)2-km. II
• PMD compensation is not necessary at 10 Gb/s but may be J
I
required at 40 Gb/s if the link length exceeds 600 km.
Back
Close
PMD Compensation Techniques
380/549

• Schematic illustration of (a) optical and (b) electrical PMD

compensators.
• Electrical PMD equalizer corrects for the PMD effects within the
receiver using a transversal filter.
• This filter splits electrical signal x(t) into a number of branches
N−1
using multiple delay lines to form y(t) = ∑m=0 cm x(t − mτ). JJ
II
• Error signal for control electronics is based on closing of the J
“eye” at the receiver. I
Back
Close
Optical PMD Compensation
• PMD-distorted signal is separated into two components along PSPs,
381/549
which are delayed by different amounts before being combined.
• Delay is adjustable in one branch through a variable delay line.
• A feedback loop is used to adjust polarization controller in
response to changes in the fiber PSPs.
• The success of this technique depends on L/LPMD, where
LPMD = (T0/D p)2.
• Considerable improvement expected for L < 4LPMD.
√ JJ
• LPMD ∼ 10,000 km for D p ≈ 0.1 ps/ km and T0 = 10 ps.
II
• Optical PMD compensators can work over transoceanic distances J
for 10-Gb/s systems. I
Back
Close
Tunable PMD Compensation
382/549

(a) (b)

• A birefringent chirped fiber grating can be used.

• Because of birefringence, two components have different Bragg
wavelengths and slightly shifted stop bands.
• Resulting DGD that can compensate for the PMD-induced DGD. JJ
II
• This DGD is wavelength-dependent for a chirped grating. J
• It can be tuned over 5 nm by stretching the grating. I
Back
Close
Chapter 8:
Nonlinearity Management
383/549

• Role of Fiber Nonlinearity

• Solitons in Optical Fibers
• Dispersion-Managed Solitons
• Pseudo-linear Lightwave Systems
• Intrachannel Nonlinear Effects
JJ
• High-Speed Lightwave Systems II
J
I
Back
Close
Role of Fiber Nonlinearity
• In the absence of nonlinear effects, system performance is only lim- 384/549
ited by the SNR degradation induced by amplifier noise.
• Since SNR can be improved by increasing input optical power, link
length can be made arbitrarily long.
• However, nonlinear effects are not negligible for long-haul
systems when power levels exceed a few milliwats.
• Degradation induced by the nonlinear effects depends on the
dispersion map employed.
• Different dispersion maps can lead to different Q factors. JJ
II
• An optimum power level exists at which BER is the lowest J
and the system performs best. I
Back
Close
Nonlinear Schrödinger Equation
• Propagation of an optical bit stream inside a dispersion-managed
385/549
system is governed by the NLS equation:
∂ A β2 ∂ 2A 2 i
i − + γ|A| A = (g0 − α)A.
∂z 2 ∂t 2 2
p
• With the transformation A(z,t) = P0 p(z)U(z,t), this equation
becomes
∂U β2 ∂ 2U 2
i − + γP 0 p(z)|U| U = 0.
∂z 2 ∂t 2
• P0 = input peak power; p(z) = exp ( 0z[g0(z) − α(z)] dz).
R
JJ
• p(zm) = 1, where zm = mLA is amplifier location. II
Rz J
• In the case of lumped amplifiers, p(z) = exp[− 0 α(z) dz]. I
Back
Close
System Design Issues
• Two major design issues exist for a dispersion-managed system:
386/549
? What is the optimum dispersion map?
? Which modulation format provides the best performance?
• Both of them studied by solving the NLS equation numerically.
• Dispersion map: 50 km of standard fiber [D = 16 ps/(km-nm),
α = 0.2 dB/km, and γ = 1.31 W−1/km] followed by 10 km of DCF
[D = −80 ps/(km-nm), α = 0.5 dB/km, and γ = 5.24 W−1/km].
• Optical amplifiers with 6-dB noise figure placed 60 km apart.
JJ
• Maximum transmission distance L calculated at which eye opening II
is reduced by 1 dB for a 40-Gb/s system. J
I
Back
Close
NRZ versus RZ Format
387/549
Distance (km)

Average Input Power (dBm) Average Input Power (dBm)

(a) (b)

• Results for (a) NRZ and (b) RZ formats.

• Without amplifier noise, distance can be increased by decreasing
launched power (empty symbols).
JJ
• When noise is included, an optimum power level exists for which II
link length is maximum. J
I
• This distance is <400 km for the NRZ format. Back
Close
Nonlinear Effects Within DCF
• Reason: RZ-format pulses spread quickly and their peak power is
388/549
reduced considerably.
• This reduction in the peak power lowers the impact of SPM.
• Buildup of nonlinear effects within DCFs also affects system
performance.
• Even for RZ format, maximum distance is <900 km at a power of
−4 dBm because of DCF-induced nonlinear degradation.
• Not only DCFs have a larger nonlinear parameter, pulses are also
compressed inside them, resulting in much higher peak powers. JJ
II
• If the nonlinear effects can be suppressed within DCF, maximum
J
distance can be increased close to 1,500 km. I
Back
Close
Recirculating Fiber Loop
389/549

• Recirculating fiber loop used to demonstrate transmission of

a 10-Gb/s signal over 2,040 km of standard fiber.
• Two 102-km sections of standard fiber and two 20-km DCFs used. JJ
II
• A filter with a 1-nm bandwidth used to reduce ASE noise.
J
• Two acousto-optic switches control the the loop. I
Back
Close
System Design
• Perfect compensation of GVD in each map period is not the best
390/549
solution in the presence of nonlinear effects.
• A numerical approach is used to optimize the design of dispersion-
managed lightwave systems.
• In a 1998 experiment, a 40-Gb/s signal was transmitted over
2,000 km of standard fiber using a novel dispersion map.
• Distance could be increased to 16,500 km at 10 Gb/s by placing
amplifier right after the DCF.
• NRZ format can be used at 10 Gb/s but the RZ format is superior JJ
for lightwave systems operating at 40 Gb/s or more. II
J
I
Back
Close
Semianalytic Approach
• Considerable insight possible by adopting a semianalytic approach
391/549
based on a single Gaussian pulse (an isolated 1 bit).
• Using the moment or variational method, NLS equation is reduced
to two coupled equations:
dT β2(z)C dC β2(z) γ(z)p(z)E0
= , = (1 +C2) 2 + √ .
dz T dz T 2π T
• Details of loss and dispersion managements appear through z de-
pendence of β2, γ, and p.
• For given values of three input pulse parameters (T0, C0, and E0) JJ
these equations can be solved numerically. II
• Pulse energy E0 is related to average power as J
√ I
Pav = 12 BE0 = ( π/2)P0(T0/Tb).
Back
Close
Solution in the Linear Case
• Consider first the linear case by setting γ(z) = 0.
392/549
• E0 plays no role because pulse propagation is independent of input
pulse energy.
• Moment equations can be solved analytically:
z 1 +C02 z
Z Z
2
T (z) = T02 + 2 β2(z)C(z) dz, C(z) = C0 + β2(z) dz.
0 T02 0

• For a two-section dispersion map values of T and C at the end of

the map period z = Lmap are given by
JJ
T1 = T0[(1 +C0d)2 + d 2]1/2, C1 = C0 + (1 +C02)d. II
R Lmap J
β̄2 Lmap
• Parameter d is defined as d = T12 0 β2(z) dz = T02
. I
0
Back
Close
Solution in the Nonlinear Case
150
(a) 393/549
Width (ps) 0.1 mW
100
10 mW
50

0
0 10 20 30 40 50 60
Distance (km)
0

−5 (b)
Chirp

10 mW
−10

−15 0.1 mW

−20
0 10 20 30 40 50 60 JJ
Distance (km)
II
• Nonlinear effects modify both width and chirp but changes are not J
large even for a 10-mW launched power. I
Back
Close
Buildup of Nonlinear Effects
100
Width (ps) (a) Pav = 10 mW 394/549

50 5 mW

1 mW

0
0 500 1000 1500 2000 2500 3000
Distance (km)
8
(b)
6 Pav = 10 mW

4
Chirp

5 mW
2
1 mW
0
−2
0 500 1000 1500 2000 2500 3000
JJ
Distance (km) II
• Even for Pav = 5 mW, pulse width becomes larger than the bit slot J
after a distance of 1,000 km. I
Back
Close
Soliton and Pseudo-linear Regimes
• Management of nonlinear effects is important.
395/549
• Parameters associated with dispersion map can be controlled to
manage the nonlinearity problem.
• Two main techniques have evolved: Systems employing them are
said to operate in the pseudo-linear and soliton regimes.
• It was noted in several experiments that a nonlinear system
performs best when GVD compensation is only 90 to 95% .
• Solitons can form when residual dispersion is anomalous.
• Performance improved if input pulse is initially chirped such that JJ
β̄2C < 0. II
J
• This observation led to the adoption of the chirped RZ (CRZ) I
format used for pseudo-linear systems. Back
Close
Normalized NLS equation
• Consider a lightwave system in which dispersion is compensated
396/549
only at the transmitter and receiver ends.
• Introduce two length scales LD = T02/|β2| and LNL = (γP0)−1.
• Using τ as τ = t/T0, NLS equation becomes
∂U s ∂ 2U LD 2
iLD − + p(z)|U| U = 0.
∂ z 2 ∂ τ 2 LNL

• s = sign(β2) = ±1 depending on the sign of β2.

• For γ = 2 W−1/km, LNL ∼ 100 km for P0 of 2 to 4 mW. JJ
• Dispersion length LD can vary over a wide range (from ∼1 to 10,000 II
km) depending on the bit rate of the system and type of fibers used. J
I
Back
Close
Soliton Regime
• If LD  LNL and L < LD, dispersive effects play a minor role.
397/549
• This is the situation at a bit rate of 2.5 Gb/s or less.
• LD exceeds 1,000 km at B = 2.5 Gb/s even for standard fibers and
can exceed 10,000 km for dispersion-shifted fibers.
• If LD and LNL are comparable, dispersive and nonlinear terms are
equally important in the NLS equation.
• This is the situation for 10-Gb/s systems. The use of solitons is
most beneficial in the regime.
• A soliton-based system confines each pulse tightly to its bit slot JJ
through by a careful balance of GVD and SPM effects. II
J
• Since GVD is used to offset the impact of nonlinear effects, I
dispersion is never fully compensated in soliton-based systems. Back
Close
Pseudo-linear Regime
• If LD  LNL, dispersive effects dominate locally, and nonlinear
398/549
effects can be treated in a perturbative manner.
• This situation is encountered at a bit rate of 40 Gb/s or more.
• If T0 is <10 ps, LD is reduced to below 5 km.
• Input pulses spread quickly over several neighboring bits.
• Extreme broadening reduces their peak power by a large factor.
• Nonlinear effects are reduced considerably because of averaging that
produces a nearly constant total power in all bit slots.
JJ
• Overlapping of neighboring pulses enhances intrachannel II
nonlinear effects. J
I
Back
Close
Soliton in Optical Fibers
• Solitons maintain their shape by balancing the dispersive and
399/549
nonlinear effects.
• GVD broadens optical pulses except when the pulse is initially chirped
such that β2C < 0.
• SPM imposes a chirp on the optical pulse such that C > 0.
• Soliton formation possible only when β2 < 0.
• SPM and GVD can cooperate when input power is adjusted such
that SPM-induced chirp just cancels GVD-induced broadening.
• Nonlinear Schrödinger Equation JJ
II
∂ A β2 ∂ 2 A J
i − 2
+ γ|A|2A = 0. I
∂z 2 ∂t
Back
Close
Properties of Solitons
√
• Introducing ξ = z/Ld , τ = t/T0, and U = A/ P0:
400/549

∂U 1 ∂ 2U
i ± 2
+ N 2|U|2U = 0.
∂ξ 2 ∂τ

• Its solution depends on a single parameter N defined as

N 2 = LD = LD/LNL = γP0T02/|β2|.

T02
• Dispersive and nonlinear lengths: LD = |β2 | , LNL = γP10 .
• The two are balanced when LNL = LD or N = 1.
JJ
• Input pulses of the form u(0, τ) = N sech(τ) evolve in a periodic II
fashion (inverse scattering method). J
I
Back
Close
Soliton Evolution
401/549

JJ
II
• Pulses shape invariant for N = 1 (Fundamental soliton). J
T2
I
• Periodic evolution for N > 1 with period z0 = π2 LD = π2 |β02| . Back
Close
Fundamental Soliton Solution
• For fundamental solitons, NLS equation becomes
402/549
∂ u 1 ∂ 2u
i + 2
+ |u|2u = 0.
∂ξ 2∂τ

d 2V
• If u(ξ , τ) = V (τ) exp[iφ (ξ )], V satisfies dτ 2
= 2V (K −V 2).
• Multiplying by 2 (dV /dτ) and integrating over τ,

(dV /dτ)2 = 2KV 2 −V 4 +C.

• C = 0 from the boundary condition V → 0 as |τ| → ∞.

JJ
1
• Constant K = 2 using V = 1 and dV /dτ = 0 at τ = 0. II
J
• Final Solution: u(ξ , τ) = sech(τ) exp(iξ /2). I
Back
Close
Stability of Fundamental Solitons
403/549

• Evolution of a Gaussian pulse with N = 1.

• Very stable; can be excited using any pulse shape. JJ
II
• Nonlinear index ∆n = n2I(t) larger near the pulse center. J
I
• Solitons is a temporal mode of a SPM-induced waveguide. Back
Close
Loss-Managed Solitons
• Fiber losses destroy the balance needed for solitons.
404/549
• Soliton energy and peak power decrease along the fiber.
• Nonlinear effects become weaker and cannot balance dispersion
completely.
• Pulse width begins to increase.

• Solution: Compensate losses periodically using amplifiers.

• Solitons sustained through periodic amplification are called
loss-managed solitons. JJ
II
• They must be launched with a higher energy. J
I
Back
Close
Path-Averaged Solitons
• The NLS equation with losses included through p(z):
405/549
∂ u 1 ∂ 2u 2
i + + p(z)|u| u = 0.
∂ ξ 2 ∂ τ2

• Rapid variations in p(z) can destroy a soliton if its width changes

rapidly.
• Solitons evolve little over a distance short compared with LD.
• If LA  LD, width of a soliton remains virtually unchanged even if
its peak power varies between two amplifiers.
−1 LA −αz
R JJ
• In effect, replace p(z) by its average value p̄ = LA 0 e dz. II
• Fundamental soliton can be excited if input peak power Ps is larger J
by a factor of 1/ p̄. I
Back
Close
Energy Enhancement Factor
• Energy enhancement factor for loss-managed solitons is given by
406/549
Ps 1 αLA G ln G
fLM = = = = .
P0 p̄ 1 − exp(−αLA) G − 1

• Launched peak power must be larger by a factor fLM for solitons to

survive in lossy fibers.
• As an example, G = 10 and fLM ≈ 2.56 when LA = 50 km and
α = 0.2 dB/km.
• Condition LA  LD must be satisfied for such soliton systems.
JJ
• The moment method can be used to study how fiber losses affect II
evolution of solitons. J
I
Back
Close
Soliton Evolution in Lossy Fibers
• Assume U(z,t) = a sech(t/T ) exp(−iCt 2/T 2 + iφ ).
407/549
• Using the moment method, we obtain:
dT β2C
=
dz T
 
dC 4 2 β2 2γ p(z)E0
= +C + .
dz π2 T2 π 2T
• Losses included through p(z) = exp(−αz).
• If α = 0, both derivatives vanish at z = 0 if
β2 < 0, C = 0 and E0 = 2|β2|/(γT0). JJ
II
• Using E0 = 2P0T0, this occurs for N = LD/LNL = 1. J
I
Back
Close
Soliton Evolution in Lossy Fibers
1.1 75 km 50 km 408/549

Width, T/T0
1 25 km
0.9

0.8
0 500 1000 1500 2000 2500 3000 3500 4000
Distance (km)
0.1

0 50 km 25 km
Chirp, C

−0.1
75 km
−0.2
0 500 1000 1500 2000 2500 3000 3500 4000
Distance (km)

• Evolution of pulse with and chirp for when LD = 100 km.

JJ
• For LA = 25 km, width and chirp remain close to input values.
II
• Width can change by more than 10% when LA = 75 km. J
I
• If LA/LD > 1, pulse width starts to increase exponentially. Back
Close
Numerical Evolution over Long Fiber Links
409/549

(a) (b)

• Evolution of a loss-managed soliton over 10,000 km.

• Amplifier spacing is fixed at LA = 50 km.
• Dispersion length LD is varied by changing T0.
JJ
• When LD = 200 km, soliton is preserved even after 10,000 km. II
J
• If dispersion length is reduced to 25 km, soliton is unable to I
sustain itself. Back
Close
Design of Soliton Systems
p
• Condition LA < LD with LD = T02/|β2| leads to T0 > |β2|LA.
410/549
• T0 must be a small fraction of Tb = 1/B to ensure that neighboring
solitons are well separated.
• This requirement can be used to relate T0 to the bit rate B using
Tb = 2q0T0.
• Typically, q0 exceeds 4 to ensure pulse tails do not overlap.
• Using T0 = (2q0B)−1, we obtain B2LA < (4q20|β2|)−1.
• For β2 = −2 ps2/km, LA = 50 km, and q0 = 5, we obtain T0 > 10 ps
and B < 10 Gb/s. JJ
II
• To operate at 10 Gb/s, one must reduce LA if β2 is kept fixed. J
I
Back
Close
Design of Soliton Systems
• Condition LA  LD can be relaxed considerably by employing
411/549
distributed amplification.
• A distributed-amplification scheme provides a nearly lossless fiber
by compensating losses locally at every point along fiber link.
• Distributed Raman amplification was used by 1985.
• A 1988 experiment transmitted solitons over 4000 km using
periodic Raman amplification.
• This experiment was the first to demonstrate that solitons can be
transmitted over transoceanic distances. JJ
II
• Main drawback is that Raman amplification requires pump lasers
J
emitting more than 500 mW of power near 1.46 µm. I
Back
Close
Dispersion-Managed solitons
• Dispersion management is employed commonly for modern
412/549
WDM systems.
• Solitons can form even when β2 varies along the link but their prop-
erties
are quite different.
• A scheme proposed in 1987 relaxes the restriction LA  LD by
employing a new kind of fiber in which GVD varies along its length.
• Such fibers are called dispersion-decreasing fibers (DDFs).
• They are designed such that the decreasing GVD counteracts the JJ
reduced SPM experienced by solitons weakened from fiber losses. II
J
I
Back
Close
Dispersion-Decreasing Fibers
• In the NLS equation β2 is a function of z.
413/549
T0−2 0z β2(z) dz
R
• Introducing ξ = and τ = t/T0,
∂U 1 ∂ 2U
i + 2
+ N 2(z)|U|2U = 0.
∂ξ 2 ∂τ
• Here, N 2(z) = γP0T02 p(z)/|β2(z)|.
• If |β2(z)| = |β2(0)|p(z), N becomes a constant.
• Fiber losses then have no effect on a soliton.
• LA can exceed LD if GVD decreases between two amplifiers as JJ
|β2(z)| = |β2(0)| exp(−αz). II
J
• Under such conditions, a fundamental soliton maintains its shape
I
and width even in a lossy fiber. Back
Close
Dispersion-Decreasing Fibers
• Fibers with a nearly exponential GVD profile have been
414/549
fabricated.
• A practical technique for making DDFs consists of reducing core
diameter along fiber length during fiber-drawing process.
• Variations in fiber diameter reduce |β2|.
• GVD can be varied by a factor of 10 over a length of 20 to 40 km
with an accuracy better than 0.1 ps2/km.
• Propagation of solitons in DDFs has been observed in several
experiments.
JJ
• Exponential GVD profile can be approximated with a staircase by II
splicing together several constant-dispersion fibers. J
I
• Benefits of DDFs can be realized using just four fiber segments. Back
Close
Periodic Dispersion Maps
• Use of dispersion management forces each soliton to propagate in
415/549
the normal-dispersion regime of a fiber.
• At first sight, such a scheme should not even work because the
normal-GVD fibers do not support solitons.
• It turns out that new kinds of solitons (called dispersion-managed
solitons) can still form.
• Pulses then evolve in a linear fashion over a single map period.
• On a longer length scale, solitons form if SPM effects are
balanced by the average dispersion. JJ
II
• Not only the peak power but also the width and shape of such
J
solitons oscillate in a periodic fashion. I
Back
Close
Input Pulse Parameters
• Moment Equations can be used to study dispersion-managed
416/549
solitons.
• Width and chirp equations should be solved with the periodic
boundary conditions to ensure that a DM soliton recovers its
initial state after each amplifier.
• Periodic boundary conditions fix the initial width T0 and chirp C0
of input pulses at z = 0.
• A new feature of DM solitons is that the input pulse width depends
on the dispersion map and cannot be chosen arbitrarily.
JJ
• In general, input pulses must be chirped appropriately. II
• Pulse parameters depends on the dispersion map used and should J
I
be determined numerically. Back
Close
Input Pulse Parameters
20
4

Input chirp
2
417/549
15
1
Pulse width (ps)

10 0.0 0.5 1.0

Input energy (pJ)

T0
5

Tm

0
0.001 0.010 0.100 1.000

Input energy (pJ)

(a) (b)

• (a) Changes in T0 and Tm for α = 0 (solid lines) and 0.25 dB/km

(dashed lines). Inset shows input chirp C0.
JJ
• (b) Evolution of DM soliton over one map period for E0 = 0.1 pJ. II
Dispersion Map: Two 5-km fiber sections with β2 = ∓4 ps2/km. J
I
• Minimum pulse width Tm occurs in the anomalous-GVD section.
Back
Close
Periodic Width and Chirp Variations
• Both T0 and Tm decrease rapidly as pulse energy is increased.
418/549
• T0 attains its minimum value at a certain pulse energy Ec.
• T0 and Tm differ by a large factor for E0  Ec.
• Pulse width changes considerably in each fiber section when this
regime is approached. (a) E0 = 0.1 pJ; (b) E0 close to Ec.
5 5 5 5

4 4 4 4

Pulse width (ps)

Pulse width (ps)

3 3 3 3

Chirp
Chirp

2 2 2 2

1 1 1 1
JJ
0 0 0 0
II
-1 -1 -1 -1

-2 -2 -2 -2
J
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Distance (km) Distance (km)
I
(a) (b) Back
Close
Soliton System Design
• Many different DM solitons coexist for the same map with
419/549
different values of E0, T0, and C0.
• How should one choose among these multiple solutions?
• Pulse energies much smaller than Ec should be avoided because a
low average power would lead to SNR degradation.
• When E0  Ec, large variations in pulse width induce XPM-induced
interaction between neighboring solitons.
• Region near E0 = Ec is most suited for designing DM soliton
systems. JJ
II
• Numerical solutions of the NLS equation confirm this conclusion.
J
I
Back
Close
Optimum Pulse Width
• Optimum values of T0 can be found from the moment equations:
420/549
s 1/2
1 +C02

|β2nβ2alnla|
T0 = Tmap , Tmap = .
|C0| β2nln − β2ala

• Tmap is a parameter with dimensions of time involving only the map

parameters.
• It provides a time scale associated with an arbitrary
dispersion map.
• Minimum√ value of T0 occurs for |C0| = 1 and is given by JJ
T0min = 2Tmap. II
• Minimum pulse width Tm = Tmap under such conditions. J
I
Back
Close
Map Strength
• It is useful to look for other combinations of map parameters that
421/549
play an important role in designing a DM soliton system.
• Two useful parameters are defined as
β2nln + β2ala β2nln − β2ala
β̄2 = , Smap = 2
.
ln + la TFWHM

• TFWHM ≈ 1.665Tm is the minimum FWHM.

• β̄2 represents average GVD of the entire link.
• Map strength Smap is a measure of how much GVD changes
JJ
abruptly between two fibers in each map period. II
• DM solitons can exist even when average GVD is normal J
provided map strength exceeds a critical value Scr. I
Back
Close
Map Strength
422/549

• Peak power of DM solitons as a function of β̄2/β2a.

• Map strength is zero for the rightmost curve, increases in step of 2
JJ
until 20, and becomes 25 for the leftmost curve.
II
• Periodic solutions in the normal-GVD regime exist if Smap J
exceeds 4.8. I
Back
Close
Experiments with DM Solitons
• In a 1996 experiment, a periodic dispersion map enabled
423/549
transmission of 20-Gb/s soliton bit stream over 5520 km.
• In another 20-Gb/s experiment, solitons were transmitted
over 9,000 km.
• In a 1997 experiment, a 10-Gb/s signal was transmitted over
28,000 km using a fiber loop consisting of 100 km of normal-GVD
fiber and 8 km of anomalous-GVD fiber.
• By 1999, 10-Gb/s DM solitons could be transmitted over 16,000 km
of standard fiber.
JJ
• Solitons system work quite well at 10 Gb/s but their performance II
is less satisfactory at 40 Gb/s. J
I
Back
Close
Timing Jitter
• Timing jitter problem severe for soliton-based systems.
424/549
• In the case of DM solitons, the moment method provides the fol-
lowing expression for it:
SASETm2
σt2 = [NA(1 +C02) + NA(NA − 1)C0d + 61 NA(NA − 1)(2NA − 1)d 2]
E0
R LA
• NA = Number of amplifiers; d = T12 β2(z) dz = Tβ̄22 LA = LLDA .
m 0 m

• For NA  1, jitter is approximately given by

σt2 SASE 3 2 SASELT3 JJ
≈ N d = ,
Tm2 3E0 A 3E0LD2 LA II
J
• LD = Tm2/|β̄2| and NA = LT /LA. I
Back
Close
Timing Jitter
8
425/549
Timing jitter (ps) Standard Soliton
6

2 DM Soliton

0
0 2000 4000 6000 8000
Distance (km)

• ASE-induced timing jitter for a 20-Gb/s system.

• Jitter should be less than 10% of the bit slot (< 5 ps).
JJ
• Dispersion map consists of 10.5 km of anomalous-GVD fiber and II
9.7 km of normal-GVD fiber [D = ±4 ps/(km-nm)]. J
I
Back
Close
Control of Timing Jitter
• Optical filters can reduce timing jitter of solitons.
426/549
• Soliton bit stream passes through the filter but most of ASE is
blocked by it.
• If an optical filter is placed after each amplifier, it improves the SNR
as well as timing jitter.
• Filter technique can be improved by allowing the center frequency
of filters to slide slowly along the link.
• Such sliding-frequency filters avoid accumulation of ASE within the
filter bandwidth.
JJ
• As filter passband shifts, solitons shift their spectrum to minimize II
filter-induced losses. J
I
• ASE noise accumulated over a few amplifiers is filtered out later. Back
Close
Sliding-Frequency Filters
427/549

• Timing jitter with (dotted curves) and without (solid curves)

sliding-frequency filters.
JJ
• Inset shows a Gaussian fit to numerically simulated jitter at II
10,000 km for a 10-Gb/s system. J
I
• Bit-rate dependence is due to contribution of acoustic waves.
Back
Close
Synchronous Modulation
• Soliton jitter can also be controlled using synchronous amplitude
428/549
modulation (implemented using a LiNbO3 modulator).
• Technique works by introducing additional losses for those
solitons that have shifted from their original position.
• Modulator forces solitons to move toward its transmission peak
where the loss is minimum.
• This technique can also be implemented using a phase modulator.
• A frequency shift is associated with all time-dependent phase
variations. JJ
II
• Since a change in soliton frequency is equivalent to a change in the
J
group velocity, phase modulation leads to temporal displacement. I
Back
Close
Postcompensation of Dispersion
• Postcompensation of accumulated dispersion can be used for
429/549
reducing timing jitter.
• Cubic jitter term depends on the accumulated dispersion.
• If accumulated dispersion is compensated using fiber of length Lc
and GVD β2c, jitter becomes

σc2 = NA3d 2Tm2(SASE/E0)(y2 − y + 1/3).

• y = −dc/(NAd) is the fraction by which accumulated dispersion

NAd is compensated.
JJ
• Minimum value occurs for y = 0.5. Timing jitter of solitons can be II
reduced by a factor of 2 by postcompensating accumulated J
dispersion by 50%. I
Back
Close
Pseudo-linear Lightwave Systems
• Local dispersion length is much shorter than nonlinear length in all
430/549
fiber sections of a pseudo-linear system.
• This approach is most suitable for systems operating at bit rates of
40 Gb/s or more.
• Relatively short pulses spread quickly over multiple bits.
• This spreading reduces peak power and lowers the impact of SPM.
• In one design, pulses spread throughout the link and are compressed
back at the receiver end.
• In another, pulses are spread even before they are launched using a JJ
II
DCF (precompensation).
J
I
Back
Close
Design of Pseudo-linear Systems
• It is not essential to compensate dispersion only once at the
431/549
transmitter or the receiver end.
• A periodic dispersion map can also be used.
• It is made such that the pulse broadens by a large factor in the first
section and is compressed back in the second section.
• A small amount of dispersion is left uncompensated in each
map period.
• This residual dispersion per span can be used to control the
impact of intrachannel nonlinear effects. JJ
II
• Combination of pre- and post-compensation is employed to
J
improve further system performance. I
Back
Close
Intrachannel Nonlinear Effects
• Optical pulses spread considerably outside their assigned bit slot in
432/549
all pseudo-linear systems.
• They overlap and interact with each other through the nonlinear
term in the NLS equation.
• Enhanced nonlinear interaction among the 1 bits of the same
channel produces intrachannel nonlinear effects.
• If left uncontrolled, they limit performance of all pseudo-linear
systems.
• Important question is whether pulse spreading helps to lower the JJ
overall impact of fiber nonlinearity. II
J
• The answer to this question turned out to be yes. I
Back
Close
Origin of Intrachannel Effects
• In a numerical approach, NLS equation is solved using a pseudo-
433/549
random bit stream with the input U(0,t) = ∑Mj=1 U j (0,t − t j ).
• Considerable physical insight can be gained with a semi-analytic
approach focusing on three neighboring pulses.
• Writing U = U1 +U2 +U3 in the NLS equations, we obtain
∂U1 β2 ∂ 2U1
i − 2
+ γP0 p(z)[(|U1|2 + 2|U2|2 + 2|U3|2)U1 +U22U3∗] = 0,
∂z 2 ∂t
∂U2 β2 ∂ 2U2
i − 2
+ γP0 p(z)[(|U2|2 + 2|U1|2 + 2|U3|2)U2 + 2U1U2∗U3] = 0,
∂z 2 ∂t
∂U3 β2 ∂ 2U3 2 2 2 2 ∗ JJ
i − + γP0 p(z)[(|U3 | + 2|U1 | + 2|U2 | )U3 +U2 U1 ] = 0.
∂z 2 ∂t 2 II
J
I
• Last nonlinear term corresponds to four-wave mixing. Back
Close
Intrachannel XPM
• Consider two isolated 1 bits by setting U3 = 0:
434/549
∂Un β2 ∂ 2Un
i − 2
+ γP0 p(z)(|Un|2 + 2|U3−n|2)Un = 0.
∂z 2 ∂t
• Over a distance ∆z, XPM shifts the phase by

φn(z,t) = 2γP0 p(z)∆z|U3−n(z,t)|2.

• As this phase shift depends on pulse shape, it produces frequency

chirp
∂ φn ∂
δ ωn ≡ − = −2γP0 p(z)∆z |U3−n(z,t)|2. JJ
∂t ∂t
II
• Similar to an ASE-induced frequency shift, XPM-induced J
frequency shift translates into a timing jitter. I
Back
Close
XPM-Induced Timing Jitter
• If all pulses were to shift in time by the same amount, this effect
435/549
would be harmless.
• Because of XPM, time shift depends on the pattern of bits
surrounding each pulse.
• This shift varies from bit to bit depending on the data transmitted.
• Pulses shift in their respective bit slots by random amounts
(timing jitter).
• XPM also introduces amplitude fluctuations.
• A quantitative analysis of the XPM effects can be carried out with JJ
II
the moment method.
J
• Results of this approach reveal several interesting features. I
Back
Close
XPM-Induced Frequency Shift

Normalized Frequency Shift

Tb 436/549

T0

Normalized Pulse Width, T0 / Tb

• Consider two Gaussian pulses separated by Tb.

• Frequency shift is largest when T0 ≈ Tb.
• Surprisingly, ∆ν is small for wide pulses.
JJ
• Frequency chirp depends on dP/dt. This slope is smaller for II
wider pulses and changes sign, resulting in an averaging effect. J
I
• Stretching of optical pulses over multiple bit slots helps. Back
Close
Frequency and Temporal Shifts
437/549

(a) (b)

• A 100-km link with two 50-km sections (D = ±10 ps/km/nm).

• (a) Frequency shift for two 5-ps pulses separated by 25 ps.
• (b) Change in pulse spacing as a function of launch position. JJ
II
• Pulse position does not shift for a symmetric dispersion map as
J
timing shifts produced in the two sections cancel each other. I
Back
Close
Frequency and Temporal Shifts
438/549

• Frequency and time shifts after 100 km as a function of DCF length JJ

used for chirping input pulses. II
J
• XPM-induced time shift can be cancelled by suitably chirping
I
input pulses. Back
Close
XPM-Induced Degradation
439/549

• 40-Gb/s bit stream in 80-km fiber with D = 4 ps/(km-nm). JJ

II
• Dashed curve shows for comparison the input bit stream. J
I
• Output bit stream exhibits both amplitude and timing jitters. Back
Close
Intrachannel FWM
• Intrachannel FWM is of concern because it transfers energy from
440/549
one pulse to neighboring pulses.
• It can create new pulses in bit slots that represent 0’s and contain
no pulse initially.
• Such FWM-generated pulses (called ghost pulses) are undesirable
because they can lead to additional errors.
• Numerical simulations are often used to predict the impact of such
ghost pulses.
• As an example, consider a 40-Gb/s system designed using 80 km of
JJ
standard fiber with D = 17 ps/(km-nm). II
• 5-ps chirped Gaussian input pulses propagate through the link. J
I
• Bit stream is severely degraded only after 80 km. Back
Close
FWM-Induced Degradation
441/549

JJ
• 40-Gb/s bit stream in 80-km fiber with D = 17 ps/(km-nm). II
J
• Dashed curve shows for comparison the input bit stream. I
• Ghost pulses degrade the eye diagram considerably. Back
Close
Intrachannel FWM

Peak Power (mW)

Peak Power (mW)

442/549

Distance (km) Pulse Separation (ps)

(a) (b)

• Peak power of ghost pulse as a function of (a) link length L and

(b) pulse separation Tb obtained analytically (solid curves).
• Dotted curves show an asymptotic approximation.
• Symbols show the results of numerical simulations. JJ
II
• Total peak power at the end of a link of length L grows as J
Pt (L) = Pg(Lmap)(L/Lmap)2. I
Back
Close
Control of Intrachannel Nonlinear Effects
• Optimization of dispersion map can reduce the impact of
443/549
intrachannel nonlinear effects.
• Two main choices: (i) dispersion accumulates along the link and is
compensated using DCFs at the transmitter and receiver ends.
• (ii) Dispersion is compensated periodically at least partially.
• Both types of dispersion maps have been used for 40-Gb/s
systems.
• In the first case, one has the choice of pre- or post-compensation.
• Next figure shows measured and calculated power penalties as a JJ
II
function of launched power for two choices.
J
I
Back
Close
Cmparison of Pre- and Post-compensation
444/549

• Eye diagrams are simulated numerically. JJ

II
• Much higher powers could be launched in the case of post-
J
compensation, while keeping the penalty below 0.5 dB. I
Back
Close
Role of Amplifier Spacing
445/549

(a) (b)

• This experiment used standard fibers and compensated dispersion

only at the receiver end.
• It employed 2.5-ps pulses at 40-Gb/s with LA = 120 km (left). JJ
II
• For LA = 120 km, system length was limited to 720 km.
J
• Longer distances could be realized by reducing LA to 80 km. I
Back
Close
Optimization of Dispersion maps
• Optimization of a dispersion map is not a trivial task.
446/549
• It involves a large number of design parameters (lengths and dis-
persion of individual fibers used to make the map, the amount of
pre- and post-compensation employed, pulse width, etc.).
• Extensive numerical simulations reveal several interesting features.
• When fiber dispersion is relatively small [D < 4 ps/(km-nm)], soliton
regime works best with an RZ duty cycle near 50%.
• When dispersion is large along most of the link, pseudo-linear regime
is more desirable for designing a 40-Gb/s system. JJ
II
• Pseudo-linear systems are most suitable for old links made with
J
standard fibers. I
Back
Close
Symmetric Dispersion maps

Amplitude

Amplitude
447/549

Time (ps) Time (ps)

(a) (b)

• Can intrachannel nonlinear effects be controlled by optimizing a

dispersion map? Answer: Yes.
• Both amplitude and timing jitter are reduced if dispersion map is
symmetric: da(z) = da(L − z), where da(z) = 0z D(z) dz.
R

• This can be realized by compensating 50% of dispersion at trans-

JJ
mitter and remaining 50% at receiver. II
• Numerical simulations show eye diagrams for 2.5-ps pulses with a J
25-ps bit slot propagated over 1,600 km of standard fiber. I
Back
Close
Optimization of Dispersion maps
• Timing jitter results from XPM-induced frequency shifts that
448/549
cancel for a symmetric map.
• Indeed, timing jitter would vanish in the absence of losses
[p(z) = 1].
• Residual jitter is due to variations in the average power along the
link when lumped amplifiers are used.
• How one one symmetrize the dispersion map?
• If a periodic dispersion map is made with two fiber sections of equal
lengths, reversing two fibers in every alternate map period makes JJ
the map symmetric. II
J
I
Back
Close
Symmetric Dispersion maps
+ + + +
449/549

• Timing and amplitude jitter over 16 spans (each 80 km long) for JJ

II
symmetric (solid) and asymmetric (dashed) links.
J
• Launched powers are 3, 6, and 9 dBm for diamonds, circles, and I
squares, respectively. Back
Close
Phase-Alternation Techniques
450/549

(a) (b)

• Power dependence of Q factor found numerically for a 40-Gb/s

channel at a distance of 1,000 km for four modulation formats.
• In (a) D = 19 ps/(km-nm) for the first and third 30-km sections
but D = −28 ps/(km-nm) for the 40-km-long middle section.
• Map (b) employs 100 km of standard fiber with D = 17 ps/(km-nm) JJ
whose dispersion is compensated using DCFs. II
J
• DPSK and AMI formats provide better performance compared with
I
RZ and CSRZ formats. Back
Close
Growth of Ghost-Pulse Power
451/549

Power (mW)

Distance (km)

• Growth of power with distance for a 40-Gb/s signal (6.25-ps pulses)

and three RZ-type formats.
JJ
• Power of ghost pulses depends on phases of neighboring bits.
II
• AP-RZ format works best because a π/2 phase difference J
minimizes buildup of ghost pulses. I
Back
Close
Polarization Bit Interleaving
• This technique alternates polarization of neighboring bits.
452/549
• Both XPM and FWM processes depend on the state of polarization
of interacting waves.
• If neighboring bits are polarized orthogonally, their impact is reduced
considerably.
• Bit interleaving was first used in 1991 for reducing interaction
between neighboring solitons.
• In a different approach, neighboring channels in a WDM system are
orthogonally polarized to reduce channel crosstalk. JJ
II
• Reduction of Intrachannel nonlinear effects requires that neighbor-
J
ing bits of the same channel be polarized orthogonally. I
Back
Close
Polarization Bit Interleaving
∆τ = Tb
∆τ
453/549

• Two schemes used for for polarization bit interleaving.

• In (a) phase modulator first imposes phase shift on pulse train.
• This train is split into polarization components that are combined
back after one bit delay. A data modulator codes the RZ signal. JJ
II
• In (b) pulse train is first coded with data, then split into its com-
J
ponents that are combined back after a phase modulator imposes I
phase shift on one of the components. Back
Close
Spectra for Four Modulation Formats
454/549

• Spectra of standard (a) RZ and (b) CSRZ signals. JJ

II
• Modified spectra of (c) RZ and (d) CSRZ signals when neighboring J
bits are orthogonally polarized. I
Back
Close
Polarization Bit Interleaving
455/549

• BER at a distance of 2,000 km for the four formats whose spectra

are shown in previous Figure.
• Q2 factor improves by 4.5 dB when neighboring bits are orthogonally JJ
II
polarized.
J
• With polarization alternation, intrachannel nonlinear impairments I
are reduced significantly and lead to a much lower BER. Back
Close
High-Speed Lightwave Systems
• If intrachannel nonlinear effects can be controlled, it is possible to
456/549
increase the bit rate beyond 40 Gb/s.
• Such optical signals cannot be generated electrically because of
limitations imposed by high-speed electronics.
• Time-division multiplexing (TDM) is employed to create bit streams
at data rates higher than 40 Gb/s.
• Optical TDM (OTDM) has been used to transmit data at a single
carrier wavelength at bit rates as high as 1.128 Tb/s.
• Use of OTDM requires new types of transmitters and receivers for JJ
all-optical multiplexing and demultiplexing. II
J
I
Back
Close
OTDM Transmitters
457/549

• A laser emitting a pulse train at bit rate B is used.

• Pulse width Tp shouls satisfy Tp < Tb = (NB)−1 to ensure that each
pulse will fit within its allocated time slot Tb.
• Laser output is split into N branches.
JJ
• Bit stream in the nth branch is delayed by (n − 1)/(NB). II
J
• The output of all branches is combined to form a composite signal. I
Back
Close
OTDM Receivers
458/549

JJ
• Demultiplexing schemes: (a) XPM within a Sagnac interferometer II
and (b) FWM inside a nonlinear medium. J
I
• A semiconductor optical amplifier also used in place of fiber.
Back
Close
Performance of OTDM Systems
• Transmission distance of OTDM systems is limited by fiber
459/549
dispersion because of the use of short optical pulses.
• A 200-Gb/s system is limited to <50 km even even when β2 = 0.
• OTDM systems require simultaneous compensation of both
second- and third-order dispersions.
• Even then, PMD is a limiting factor and its compensation is
necessary.
• Intrachannel nonlinear effects also limit performance.
• By 1999, operation at 3 Tb/s was realized by combining 19 JJ
II
channels operating at 160 Gb/s.
J
I
Back
Close
Performance of OTDM Systems
460/549

• Schematic of a 320-Gb/s OTDM experiment over 200 km. JJ

II
• In 2000, a 1.28-Tb/s ODTM signal was transmitted over 70 km,
J
but it required compensation of fourth-order dispersion. I
Back
Close