Fiber Optical Communication Lecture 3, Slide 1
Lecture 3
• Dispersion in single-mode fibers
– Material dispersion
– Waveguide dispersion
• Limitations from dispersion
– Propagation equations
– Gaussian pulse broadening
– Bit-rate limitations
• Fiber losses
Fiber Optical Communication Lecture 3, Slide 2
Dispersion, qualitatively
• Different wavelengths (frequency components) propagate differently
• A pulse has a certain spectral width and will broaden during propagation
The index of refraction as a function of wavelength
The dispersion in SMF (red) and different dispersion-shifted fibers
Fiber Optical Communication Lecture 3, Slide 3
Each spectral component of a pulse has a specific group velocity
The group delay after a distance L is
The group velocity is related to the mode group index given by
Assuming that Δω is the spectral width, the pulse broadening is governed by
where β2 is known as the GVD parameter (unit is s2/m or ps2/km)
Group delay, group index, and GVD parameter (2.3.1)
22
2
Ld
dL
d
dTT
d
ndn
c
d
ndn
c
n
cv
g
g
d
dL
v
LT
g
d
ndnng
Fiber Optical Communication Lecture 3, Slide 4
The dispersion parameterMeasuring the spectral width in units of wavelength (rather than frequency), we can write the broadening as
ΔT = D Δλ L,
where D [ps/(nm km)] is called the dispersion parameter
D is related to β2 and the effective mode index according to
The dispersion parameter has two contributions:
material dispersion, DM: The index of refraction of the fiber material depends on the frequency
waveguide dispersion, DW: The guided mode has a frequency dependence
2
2
22
1
2
2
2222
212,
2
d
nd
d
nd
vd
dc
d
dv
d
dcD
g
g
Fiber Optical Communication Lecture 3, Slide 5
Material dispersion (2.3.2)The material dispersion is related to the dependence of the cladding material’s group index on the frequency
An approximate relation for the material dispersion in silica is
where DM is given in ps/(nm km)
d
dnD
g
M
2
2
2
MMD 01122
Fiber Optical Communication Lecture 3, Slide 6
Waveguide dispersion (2.3.3)The waveguide dispersion arises from the modes’ dependence on frequency
n2g: the cladding group index
V: the normalized frequency
b: the normalized waveguide index
dV
Vbd
d
dn
dV
VbVd
n
nD
gg
W
2
2
2
2
2
2
2
2
V
22
1
2
2
2
1 anc
nnaV
21
2
nn
nnb
Fiber Optical Communication Lecture 3, Slide 7
Total dispersionThe total dispersion D is the sum of the waveguide and material contributions
D = DW + DM
Note: DW increases the net zero dispersion wavelength
The zero-dispersion wavelength is denoted either λ0 or λZD
An estimate of the dispersion-limited bit-rate is
|D|B Δλ L < 1
where B is the bit-rate, Δλ the spectral width, and L the fiber length
Fiber Optical Communication Lecture 3, Slide 8
The dispersion can have different signs in a standard single-mode fiber (SMF)
D > 0 for λ > 1.31 μm: “anomalous dispersion”, the group velocity of higher frequencies is higher than for lower frequencies
D < 0 for λ < 1.31 μm: “normal dispersion”, the group velocity of higher frequencies is lower than for lower frequency components
Pulses are affected differently by nonlinear effects in these two cases
Anomalous and normal dispersion
Fiber Optical Communication Lecture 3, Slide 9
Different fiber types• The fiber parameters can be tailored to shift the λ0-wavelength from
≈1.3 μm to 1.55 μm, dispersion-shifted fiber (DSF)
• A fiber with small D over a wide spectral range (typically with two λ0-wavelengths), dispersion-flattened fiber (DFF)
• A short fiber with large normal dispersion can compensate the dispersion in a long SMF, dispersion compensating fibers (DCF)
Dispersion compensating fiber
Fiber Optical Communication Lecture 3, Slide 10
Fibers in the labThis dispersion compensating module contains 4 km of DCF...
...and it compensates the dispersion in this 25 km roll of SMF
Fiber Optical Communication Lecture 3, Slide 11
Index profiles of different fiber types• Standard single-mode fiber
(SMF)
• Dispersion-shifted fiber (DSF)
• Dispersion-flattened fiber (DFF)
Fiber Optical Communication Lecture 3, Slide 12
Higher order dispersion (2.3.4)• Near the zero-dispersion wavelength D ≈ 0
– The variation of D with the wavelength must be accounted for
– We have used β2 = 0
• S [ps/(nm2 km)] is called the dispersion slope
– Typical value in SMF is 0.07 ps/(nm2 km)
3
32
2
2
d
dc
d
dDS
Fiber Optical Communication Lecture 3, Slide 13
Basic propagation equationWe will now develop the theory for signal propagation in fibers
The electric field is written as
• The field is polarized in the x-direction
• F(x, y) describes the mode in the transverse directions
• A(z, t) is the complex field envelope
• β0 is the propagation constant corresponding to ω0
Only A(z, t) changes upon propagation (described in the Fourier domain)
Each spectral component of a pulse propagates differently
)exp(),(),(ˆRe),( 00 tizitzAyxFt xrE
dtitzAzA
ziziAzA
)exp(),(),(~
)(exp),0(~
),(~
0
Fiber Optical Communication Lecture 3, Slide 14
The propagation constant• The propagation constant is in general complex
– α is the attenuation
– δnNL is a small nonlinear (= power dependent) change of the refractive index
• Dispersion arises from βL(ω)
– The frequency dependence of βNL and α is small
• We now expand βL(ω) in a Taylor series around ω = ω0 (Δω = ω – ω0)
1/vp 1/vg GVD(rel. to D) dispersion slope(related to S)
0
332210 ...,)(
6)(
2)()(
m
m
mLd
d
2/)()()(2/)()/)](()([)( 00 iicnn NLLNL
Fiber Optical Communication Lecture 3, Slide 15
Substitute β with the Taylor expansion in the expression for the evolution
of A(z, ω), calculate ∂A/∂z, and write in time domain by using Δω↔ i ∂/∂t
The nonlinearity is quantified by using δnNL = n2I where n2 [m2/W] is a
measure of the strength of the nonlinearity, and I is the light intensity
βNL = γ|A|2, where γ = 2πn2/(λ0Aeff) is the nonlinear coefficient
Aeff is the effective mode area and |A|2 is normalized to represent the power
γ is typically 1–20 W–1 km–1
Basic propagation equation (2.4.1)
AAit
A
t
Ai
t
A
z
ANL
262 3
3
3
2
2
21
Fiber Optical Communication Lecture 3, Slide 16
Basic propagation equation• Use a coordinate system that moves with the pulse group velocity!
– This is called retarded time, t’ = t – β1z
– We neglect β3 to get
• This is the nonlinear Schrödinger equation (NLSE)
– The primes are implicit
• The loss reduces the power ⇒ reduces the impact from the nonlinearity
• The average power of the signal during propagation in the fiber is
• Note: α is in m-1 while loss is often expressed in dB/km
AAAit
Ai
z
A
22
2
2
2
2
2/
2/
av
2
av )0(),(1
lim)(
T
T
z
TePdttzA
TzP
Fiber Optical Communication Lecture 3, Slide 17
Chirped Gaussian pulses (2.4.2)• To study dispersion, we neglect nonlinearity and loss
• The formal solution is
• Note: Dispersion acts like an all-pass filter
• We study chirped Gaussian pulses
– A0 is the peak amplitude
– C is the chirp parameter
– T0 is the 1/e half width (power)
02 2
2
2
t
Ai
z
A
ziAzA 22
2exp),0(
~),(
~
2
021
0 )/)(1(exp),0( TtiCAtA
00
2/1
FWHM 665.1)2(ln2 TTT
Fiber Optical Communication Lecture 3, Slide 18
• For a chirped pulse, the frequency of the pulse changes with time
– What does this mean???
• Study a CW (continuous wave)
– A is a constant
• Writing A exp(iβ0z – iω0t) = A exp(iφ), we see that ω0 = –∂φ/∂t
• We define the chirp frequency to be
– We allow φ to have a time dependence
– We get φ from the complex amplitude
• In this way, the chirp frequency can depend on time
– For the Gaussian pulse we get ωc = Ct/T02
Chirp frequency
)exp(),(),(ˆRe),( 00 tizitzAyxFt xrE
ttc /)(
Fiber Optical Communication Lecture 3, Slide 19
A linearly chirped pulseFrequency increases with time Frequency decreases with time
ωc ωc
tt
Fiber Optical Communication Lecture 3, Slide 20
Time-bandwidth productThe Fourier transform of the input Gaussian pulse is
The 1/e spectral half width (intensity) is
The product of the spectral and temporal widths is
If C = 0 then the pulses are chirp-free and said to be transform-limited
as they occupy the smallest possible spectral width
Using the full width at half maximum (FWHM), we get
)1(2exp
1
2),0(
~2
0
22/1
2
00
iC
T
iC
TAA
0
2
0 /1 TC2
00 1 CT
22
FWHMFWHM 144.012ln2
CCT
Fiber Optical Communication Lecture 3, Slide 21
We introduce ξ = z/LD where the dispersion length LD = T02/|β2|
In the time domain the dispersed pulse is
The output width (1/e-intensity point) broadens as
Chirped Gaussian pulses (2.4.2)
A Gaussian pulse remains Gaussian during propagation
The chirp, C1(ξ), evolves as the pulse propagates
If (C β2) is negative, the pulse will initially be compressed
C
i
bT
tiC
b
AtA
ff1
arctan22
)1(exp),(
22
0
2
10
)(sign
)1()(
)1()(
2
2
1
2/122
s
CsCC
sCb f
2/12
2
0
2
2
2
0
2
0
1 1)(
)(
T
z
T
zC
T
zTzb f
Fiber Optical Communication Lecture 3, Slide 22
Broadening of chirp-free Gaussian pulses
Short pulses broaden more quickly than longer pulses
(Compare with diffraction of beams)
2
2
0
2
2
11)(
T
z
L
zzb
D
f
Fiber Optical Communication Lecture 3, Slide 23
Broadening of linearly chirped Gaussian pulses
For (C β2) < 0, pulses initially compress and reaches a minimum at
z = |C|/(1+C2)LD at which C1 = 0 and
Chirped pulses eventually broaden more quickly than unchirped pulses0
2
0min
1
1
1
C
TT
Fiber Optical Communication Lecture 3, Slide 24
Chirped Gaussian pulses in the presence of β3
Higher order dispersion gives rise to oscillations and pulse shape changes
2
3
0
2
3
2
2
0
2
2
2
0
2
2
0
2
24
)1(
221
CLLLC2/00 T
Fiber Optical Communication Lecture 3, Slide 25
Effect from source spectrum widthUsing a light source with a broad spectrum leads to strong dispersive broadening of the signal pulses
In practice, this only needs to be considered when the source spectral width approaches the symbol rate
For a Gaussian-shaped source spectrum with RMS-width σω and with Gaussian pulses, we have
where Vω = 2σωσ0
2
3
0
3222
2
2
0
22
2
2
0
2
2
0
2
24)1(
2)1(
21
LVC
LV
LCp
Vω << 1 when the source spectral width << the signal spectral width
Fiber Optical Communication Lecture 3, Slide 26
If, as for an LED light source, Vω >> 1 we obtain approximately
A common criteria for the bit rate is that
For the Gaussian pulse, this means that 95% of the pulse energy remains within the bit slot
In the limit of large broadening
σλ is the source RMS width in wavelength units
Example: D = 17 ps/(km nm), σλ = 15 nm ⇒ (BL)max ≈ 1 (Gbit/s) km
Limitations on bit rate, incoherent source (2.4.3)
22
0
2
2
2
0
2 )()( DLL
)4/(14/ BTB
14 DBL
Fiber Optical Communication Lecture 3, Slide 27
In the case of operation at λ = λZD, β2 = 0 we have
With the same condition on the pulse broadening, we obtain
The dispersion slope, S, will determine the bit rate-distance product
Example: D = 0, S = 0.08 ps/(km nm2), σλ = 15 nm ⇒ (BL)max ≈ 20 (Gbit/s) km
Limitations on bit rate, incoherent source
22
212
0
22
3212
0
2 )()( SLL
18 2 SBL
Fiber Optical Communication Lecture 3, Slide 28
For most lasers Vω << 1 and can be neglected and the criteria become
Neglecting β3:
The output pulse width is minimized for a certain input pulse width giving
Example: β2 = 20 ps2/km → (B2L)max ≈ 3000 (Gbit/s)2 km
500 km @ 2.5 Gbit/s, 30 km @ 10 Gbit/s
If β2 = 0 (close to λ0):
For an optimal input pulse width, we get
Limitations on bit rate, coherent source (2.4.3)
22
0
2
02
2
0
2 )2/( DL
14 2 LB
22
0
22
03
2
0
2 2/)4/( DL
324.0)( 3/1
3 LB
Fiber Optical Communication Lecture 3, Slide 29
Limitations on bit rate, summary
A coherent source improves the bit rate-distance product
Operation near the zero-dispersion wavelength also is beneficial…
...but may lead to problems with nonlinear signal distortion
Fiber Optical Communication Lecture 3, Slide 30
Dispersion compensation
• Dispersion is a key limiting factor for an optical transmission system
• Several ways to compensate for the dispersion exist
– More about this in a later lecture...
• One way is to periodically insert fiber with opposite sign of D
– This is called dispersion-compensating fiber (DCF)
– Figure shows a system with both SMF and DCF
– The GVD parameters are β21 and β22
• Group-velocity dispersion is perfectly compensated when
β21l1 + β22l2 = 0, which is equivalent to D1l1 + D2l2 = 0
• GVD and PMD can also be compensated in digital signal processing (DSP)
Fiber Optical Communication Lecture 3, Slide 31
Fiber losses (2.5)• Fiber have low loss but the loss grows exponentially with distance
– Approx. 20–25 dB loss over 100 km
– Optical receivers add noise...
– ...and the input power may be too low to obtain sufficient SNR
• The optical power in a fiber decreases exponentially with the propagation distance as Pout = Pin exp(–αz)
– α is the attenuation coefficient (unit m-1)
• Often, attenuation is given in dB/km and its relation to α is
• Typical value in SMF at 1550 nm αdB = 0.2 dB/km ⇒α = 0.046 km-1 = 1/(21.7 km)
343.410log
10
10log
log10log10
110dB
LL e
Le
L
Fiber Optical Communication Lecture 3, Slide 32
Attenuation mechanisms• Material absorption
– Intrinsic absorption: In the SiO2 material
• Electronic transitions (UV absorption)
• Vibrational transitions (IR absorption)
– Extrinsic: Due to impurity atoms
• Metal and OH– ions, dopants
• Rayleigh scattering
– Occurs when waves scatter off small, randomly oriented particles
– (Makes the sky blue!)
– Proportional to λ-4
• Waveguide imperfections
– Core-cladding imperfections on > λ length scales (Mie scattering)
– Micro-bending (bending curvature λ)
– Macro-bending (negligible unless bending curvature < 1–5 mm)
Fiber Optical Communication Lecture 3, Slide 33
Total attenuation• Minimum theoretical loss is 0.15 dB/km at 1550 nm
• Main contributions: Rayleigh scattering and IR absorption
• Left figure: Theoretical curves and measured loss for typical fiber
• Right figure: Loss for sophisticated fiber with negligible loss peak
Fiber Optical Communication Lecture 3, Slide 34
Lecture
• Why/when are nonlinear phenomena important?
• Different types of fiber nonlinearities
• The Kerr effect: SPM, XPM, FWM
Fiber Optical Communication Lecture 3, Slide 35
Nonlinear effects• When is a phenomenon ”nonlinear”?
– Superposition does not apply
– The phenomenon is changed by an amplitude (power) change
• Which is the same, e.g., doubling the amplitude is equivalent to a superposition of a pulse on itself
• In nonlinear optics, light cannot be viewed as a superposition of independently propagating spectral components
– Spectral components interact
– New frequencies can be generated, existing components can lose power
• IR light can become visible (green)
• Fibers nonlinearity is important for moderate powers because
– The fiber core is small, the electric field intensity is high
– A fiber is long, allowing nonlinear distortion to accumulate
Fiber Optical Communication Lecture 3, Slide 36
Why study fiber nonlinearities?• What transmitted power would you choose in a fiber optic link?
– Laser output power is sufficient
– The energy cost is small (typical input power is 1 mW)
• The figure shows that the SNR is proportional to the input power
• Clearly, higher input power is always better!?!
– No, actually it is not...
Fiber Optical Communication Lecture 3, Slide 37
Why study fiber nonlinearities?• What limits the launch power?
• Before 1990: Limited by laser output power to 1 mW
• After 1990: EDFAs enable power levels up to > 100 mW
– Performance is limited by fiber nonlinearities
The nonlinear trade-off:
• Low power: System is limited by noise
• High power: System is limited by nonlinearities
There exist an optimum launch power
A higher power is not always better!
BER for a system with-out nonlinearities
Nonlinear limitation
Noise limitation
Fiber Optical Communication Lecture 3, Slide 38
Nonlinearities in fibersTwo types of important nonlinear effects in fibers:
• Electrostriction
– Intensity modulation in the fiber leads to pressure changes in the density of the medium, which leads to changes of the refractive index
– Responsible for Stimulated Brillouin Scattering (SBS)
• The Kerr effect
– The refractive index is changed in proportion to the optical intensity
– This gives rise to
• Self-phase modulation (SPM)
• Cross-phase modulation (XPM)
• Four-wave mixing (FWM)
• Modulation instability
• Solitons, which propagate without any change of the shape
– The delayed response of the Kerr effect gives rise to a nonlinear frequency downshift called Stimulated Raman Scattering (SRS)
Fiber Optical Communication Lecture 3, Slide 39
Nonlinearities in fibers, scattering processesStimulated Brillouin scattering
• Occurs only in the backward direction
• Light will be backscattered and downshifted 10 GHz
– Remaining photon energy is absorbed as a vibration mode in the fiber
• Requires power levels 10 mW
Stimulated Raman scattering
• Occurs both in the forward and backward direction
• Appears over a wide spectral range (15 THz, 100 nm)
• Photons are downshifted in frequency
– Remaining photon energy is absorbed by the fiber
• Requires power levels of about 0.1–1 W
Fiber Optical Communication Lecture 3, Slide 40
Nonlinearities in fibers, the Kerr effect• The Kerr effect means that the refractive index is intensity dependent
– The propagation constant becomes β(ω) = βlin (ω) + γ|A(t)|2
• The Kerr-effect gives rise to
– Self-phase modulation (SPM)
• Causes spectral broadening
• Can counteract anomalous dispersion
• Can give rise to soliton pulses
– Solitons do not broaden in time or frequency
– Cross-phase modulation (XPM)
• Causes frequency shift of other WDM channels
• Limits WDM systems performance
– Four-wave mixing (FWM)
• Causes power exchange between WDM channels
• Limits WDM system performance
The fundamental phenomenon is SPM
XPM and FWM appear when we “interpret” SPM in a WDM system
Fiber Optical Communication Lecture 3, Slide 41
• Start from the NLSE and eliminate loss term by
– U is the normalized amplitude
• The NLSE for U(z, t) becomes
• The function p(z) varies periodically between 1 and exp(–αLA)
– LA is the amplifier spacing
• Neglecting the impact from dispersion, the NLSE is
– LNL = 1/(γ P0) is the nonlinear length
Self-phase modulation (2.6.2)),()(),( 0 tzUzpPtzA
UUzpPit
Ui
z
U 2
02
2
2 )(2
UUL
zpi
z
U
NL
2)(
The nonlinear length is the propagation distance over which the nonlinear effects become important
Fiber Optical Communication Lecture 3, Slide 42
The solution to the NLSE without dispersion is
The signal phase is changed by the signal itself ⇒ self-phase modulation
We have introduced Leff and φNL
• φNL is the nonlinear phase shift
• Leff is the effective length
– The power decreases during propagation, the nonlinearity becomes weaker
– Therefore, the effective length is shorter than the physical length
We have
where NA is the number of amplified sections of fiber (often called “spans”)
Self-phase modulation
),(exp),0(/),0(exp),0(),( eff
2tLitULLtUitUtLU NLNL
//)exp(1)()(0 0
eff AAA
L L
A NLNdzzpNdzzpLA
Fiber Optical Communication Lecture 3, Slide 43
SPM impact on pulses
• In the absence of dispersion, the pulse shape will not change
• SPM introduces chirp and continually broadens the spectrum
• The chirping depends on pulse shape
– Super-Gaussian different from Gaussian pulse
• Solid line: A Gaussian pulse
• Dashed line: A super-Gaussian pulse with m = 3
(Remember the chirp frequency from last lecture)
eff0effmax / LPLL NL
2eff ),0()( tUtL
L
tt
NL
NL
Fiber Optical Communication Lecture 3, Slide 44
Spectral broadening from SPM• Figures show the spectra
for chirped Gaussian pulses affected by SPM
• Dispersion and loss are neglected
• In this numerical example φmax = 4.5 π
• Spectral broadening will continue if more SPM is introduced
• Chirp on the pulse will change the effect from SPM significantly
• When φmax is large, the spectral broadening is strong
• Dispersion will change this result!
– SPM and GVD acting simultaneously leads to nontrivial phenomena
Fiber Optical Communication Lecture 3, Slide 45
Linear dispersive effects
In the time domain:
• Pulses broaden...
– ...and start to interfere
• A phase shift (chirp) will become an amplitude change
The length scale for dispersion is the dispersion length LD = T0
2/|β2|
In the frequency domain:
• The amplitude is not changed
• Quadratic phase modulation
Fig. shows spectrum for single pulse
L = 1.5LD
L = 0
time (bit slots)
|A|2
|A|2
L = 0
L > 0
frequency (normalized)
|A|2
|A|2 arg(A)
arg(A)
Fiber Optical Communication Lecture 3, Slide 46
Nonlinear propagation, SPM
In the time domain:
• The amplitude is not changed
• A pulse-shaped phase shift is introduced
– Self-phase modulation
In the frequency domain:
• The spectrum is broadened
• Energy is conserved
– Notice: Different y-scales
The length scale for the nonlinearity is the nonlinear length LNL = 1/(γ P0)
L > 0
L = 0
time (bit slots)
L = 0
L > 0
frequency (normalized)
|A|2
|A|2
|A|2
|A|2 arg(A)
arg(A)
Fiber Optical Communication Lecture 3, Slide 47
Cross-phase modulation• Consider (again) the case A = a exp(–iωat) + b exp(–iωbt), insert into the
NLSE, neglect FWM, and split into a coupled system of equations
• The group velocities are different
– This causes walk-off and limits the impact of XPM
• The wave at ωa “notices” the presence of the wave at ωb through the additional nonlinear term
– And vice versa
• XPM is stronger than SPM by a factor of two, but walk-off limits the impact from XPM, i.e., dispersion reduces XPM
• The equation system can be used only for waves well separated in freq.
bbait
bi
t
b
vz
b
abait
ai
t
a
vz
a
bg
ag
22
2
2
2
,
22
2
2
2
,
22
1
22
1
Fiber Optical Communication Lecture 3, Slide 48
Cross-phase modulation in WDM systems• XPM on channel b from channel a gives b → b exp[i2γPa(t)z]
– This changes the absolute phase, but can also...
– ...introduce a chirp that shifts the pulse up or down in frequency
• Figure shows that the sign of the shift depends on the pulse position
– Blue, solid line is the a channel, affected by the red, dashed b channel
– Remember the chirp frequency, ωc = –∂φ(t)/∂t
• The frequency shift depends on the relative position of the pulses
• The frequency shift will, via dispersion, give rise to timing jitter
• Dispersive walk-off will decrease the impact of XPM
frequency upshift no frequency change frequency downshift
Fiber Optical Communication Lecture 3, Slide 49
Four-wave mixing• The waves at three frequencies generate a fourth
– The frequencies can be different or some may be the same
– With N different frequencies, FWM will generate N2(N–1)/2 mixing products
• The strength of each mixing product depends on
– The degeneracy (how many terms that contribute)
– How close the process is to being phase matched
• Phase matching is strongly dependent on the dispersion
– FWM is strong for low dispersion, e.g., near the zero-dispersion wavelength
– At symbol rates > 10 Gbaud, FWM is weak
Figure: Non-degenerate FWMLeft: Measured FWMRight: Original and generated frequencies (dispersion not accounted for)
Fiber Optical Communication Lecture 3, Slide 50
Four-wave mixing in WDM systems• Equal channel spacing ⇒ FWM
components overlap with the data channels
– FWM can be a problem
• Solution:
– Decrease the dispersion length to reduce phase matching
• SMF/DCF better than DSF
• Only SMF is even better
– DSP dispersion compensation
– Use unequal channel spacing
• Not compliant with standard frequency assignment (ITU grid)
• Increases optical bandwidth
Original signal
Equalspacing
Unequal spacing