You are on page 1of 151

Optical Networking Technologies

Tuan Hoang – Senior System Engineer


November 3rd , 2022
Course Objectives

Upon completing this course, you will understand:


• How we transmit light down optical fiber

• The differences in fiber types

• The effect that loss and dispersion have on a signal

• Optical components and how they effect light

• Network components and how they build a network

• How to calculate dispersion penalties

• Non-Linearities that effect optical signals

2
Course Agenda
Module (Lesson)
Optical Transmisson
Transmitting Optical Signals
Optical Fiber
Span Considerations
Non-Linear Effects in Fiber
SMF types
DWDM Components
MUX/DEMUX/OADMs
Dispersion Compensation
EDFA and Raman Amps
ROADMs
Advanced Network Concepts
BER
Polarization
Non-Linearities due to High Power
3
Optical Transmission:
Transmitting Optical Signals
Objectives

Upon completing this lesson, you will be able to:

• Describe the 3 models of light

• Define frequency and wavelength and the relationship between them

• Recognize a signal in the time and wavelength domains

• Understand the construction of lasers and the differences between idealized and real world lasers

• Differentiate between the types of optical receivers

5
What is Light?
① A photon is the basic unit or particle of light. It
Light can be viewed as a photon, a wave or a ray represents a small quantum of energy.

② If we ‘stack’ a bunch of these


photons together in time, their
Different frequencies of photons have different energies.
energies add to create more optical
Two photons of the same frequency are identical.
power.

We can turn the power on and off , creating a regular pattern we call a wave.

③ Finally, these light signals can be viewed as a ray. Rays are most useful when considering beam steering.

Prism

6
Frequency and Wavelength
Frequency is defined as the number of full cycles a wave can go through in one second (in Hertz).

P (dB)
Here we are using a
sinusoidal signal for our
wave. This is one full cycle
of the signal.
t (sec)
The power of the signal 1 sec
is represented by the y- We can see that the frequency of this signal is 4 Hz.
axis
A signal’s wavelength is defined as the distance travelled to complete one full cycle (in meters).

This signal has a From these illustrations, it should


wavelength of 5
mmeters, or 5x10-6
become clear that frequency is
meters inversely proportional to
0.000005 m wavelength
7
Wave Interference
Waves must be ‘in phase’ to interfere with each other.
Two waves of same frequency are out of phase (anti-phase) if their cycles start at
different
360°
points. Anti-phase waves do not ‘effectively’ interfere with one another.

If their amplitudes are in synch, their


powers will add up in ‘constructive
interference’.
If their amplitudes are out of synch, their
180°
powers will subtract in ‘destructive
360° interference’.
Another important property of waves is
interference. When two waves are (i) the
same frequency (wavelength) and (ii)
moving in the same direction, we say
180° they are in phase if their cycles start at
the same point. 8
Wave Interference in 3D
We saw an example of waves moving parallel to each other in 2D, but we also ‘launch’
waves in 3D space. If we consider light coming out of a fiber end, it spreads out in waves, in
the same pattern as a pebbles dropped in a pond:

By adding a second
source, we can create
interference patterns.
Each of these lines
represents a ‘peak’ of
Looking head on
the wave.

9
Looking at a Signal by Time and by l
A typical optical signal is sent at a frequency of 193 teraHertz – that’s 193 x 1012 cycles in
one second!
P (dBm)

t (sec)
While this view gives us useful information about the power of a signal, imagine what
would happen if we tried to view multiple signals at the same time!
It can be helpful to look at P (dBm)
multiple signals with wavelength,
represented by the greek letter
lambda (l), as the x-axis instead
of time. l (nm)
10
How a laser works
A laser is basically a convertor of electricity (charge) into coherent light (photons).
Coherence is defined as light in which all the photons have the same wavelength.
 The cavity fills with + TERMINAL  Only photons with l the same
optical power, held in size as the cavity are allowed to exit
by the reflective end
caps

CAVITY
l
e-  As the optical pressure increases this end cap
- TERMINAL ‘breaks down’ causing the cavity to ‘lase’.
 An electrical charge is applied to the optical cavity
11
EELEDs
Telecom lasers are usually Edge-Emitting. Most are made up of a ‘sandwich’ of materials,
most familiar to us in LEDs, hence the term EE LED.
Fabry-Perot Laser Distributed Feedback Laser (DFB)
+ GaAs layer
AlGaAs Cavity
- GaAs layer

Reflective Coatings Internal Grating

P (dBm) P (dBm)
Side Mode
Suppression
Ratio (SMSR)

l (nm) l (nm)
12
Vertical Surface Cavity Emitting Lasers

VSCELs are:

▪ Much easier to manufacture, since


everything is aligned in the same
direction.
+ terminal
99% Reflector ▪ Able to be tested on the wafer,
rather than individually, once
Cavity
assembled.

99.9 % Reflector ▪ Currently unable to provide the


output powers needed for telecom
- terminal applications.
13
Ideal v. Real World Lasers
In the ideal laser all the photons P (dBm)
produced are the same wavelength
and therefore identical and additive.

l (nm)
IDEAL
Laser
REAL WORLD
P (dBm)
Real world lasers can see changes to their
size due to faulty cooling, vibration
stresses, or many other reasons. Even a
small distortion can produce photons l (nm)
that are of slightly different wavelengths.
This leads to a characteristic ‘bell’ shaped appearance
14
PIN Receivers

PIN receivers are:

▪ Cheap to manufacture and use


e- e- e-
+ relatively low amounts of power
+ terminal
P-type semiconductor
▪ Good sensitivity, typically allows
spans of 80 km

N-type semiconductor
-
- terminal ▪ Deployed in almost every optical
network
15
APD Receivers
Avalanche Photo Diodes (APDs) are a specialized type of PIN diodes. The recovered
electrons are passed through a very high voltage region. The electrons ‘pull off’ more
electrons as they go through this region (much like their namesake mountain avalanches),
making detection of photons more sensitive.
 A photon hits the APD  The electron is drawn to the  As the electron goes
and is converted to an positive terminal of the PIN and through the plates, it
electron through a pair of highly charged draws off electrons causing
plates an ‘avalanche’ effect, thus
+ increasing the receiver’s
sensitivity.

- N e-
I P +
e- e- e- e- e-

- 16
Summary of Lasers

Type Ease of Manufacturing Cost / Unit Output Power Wavelength Range

F-P ✓✓ ✓✓ ✓✓ ✓
DFB ✓ ✓ ✓✓✓ ✓✓✓
VSCEL ✓✓✓ ✓✓✓ ✓ ✓

Summary of Receivers
Type Ease of Manufacturing Cost / Unit Sensitivity

PIN ✓✓✓ ✓✓✓ ✓✓


APD ✓ ✓ ✓✓✓

17
Summary

✓ The 3 models of light at photon (particle), wave and ray

✓ Frequency is the number of cycles/sec a signal goes through

✓ Wavelength is the distance for a signal to go through one cycle, and is inversely proportional to
frequency

✓ Real world lasers have wavelength drift and do not just output one wavelength

✓ APDs are a subset of PIN receivers, using high voltage to get better sensitivity.

18
Optical Transmission:
Optical Fiber
Objectives

Upon completing this lesson, you will be able to:

• Describe the parts of optical fiber and how it steers light

• Explain the difference between reflection and refraction

• Explain how the refractive index affects latency

• Explain how multimode and single mode fiber differ

• List the major causes of loss of signal power in optical transmission

20
What does optical fiber look like?

The three main components of optical fiber are; a buffer or coating, the cladding
and the core, the latter two being made of silica glass, each with slightly different
properties.

250 mm

9 mm 125 mm
CORE

CLADDING

BUFFER

21
Refraction and Reflection
Refraction is used to steer the light into and out of the fiber, reflection is used to keep the light in
the fiber.
Any given ray of light, when approaching a material boundary has two possible
alternatives; to refract or reflect. The outcome is determined by the angle of approach of
the ray and the relative refractive indices (nx) of the two materials.

n1
θ n1 θ

n2 n2

in this case: n1 > n2 in this case: n1 < n2


Refraction Reflection

22
Snell’s Law

Snell’s Law describes the relationship between angle of approach and the
refractive indices of the two materials at the boundary:
When the ray hits the core, n1 𝑛1 sin 𝜃1 = 𝑛2 sin 𝜃2 Here, since n1 ≈ n2 , then  1
= 1 (air) and n2 = 1.5 (silica). ≈  2 - we have Total
There is a maximum angle for Internal Reflection (TIR)
1 before the ray reflects.

𝜃1

23
The NASCAR model of light
If we paint ‘lanes’ on the track, we can let lots of ‘cars’ on it
Imagine if we use stock car racing as a conceptual model for optical transmission. Why
stockcars? Because in stockcar racing all the cars are supposed to have equal performance,
much like every photon travels at the same speed.
The warning tracks on the
In this model the asphalt serves as the core of sides represents the
the optical fiber, carrying all the photons core/cladding boundary

24
n and the speed of light in fiber
This ‘slow down’ of light directly impacts a network’s latency
It has commonly become accepted that the speed of light (c) is constant. However,
more accurately it is constant in any given transmission medium, be it vacuum or
silica glass. While the speed is constant, it does vary according to the refractive index
(n) of the material.
Table of Refractive Indicies
Material Vacuum Air Water Silica
n value 1 ~1 1.3 1.5
The relationship between c and n can be shown by the following equation: c
nm =
Where nm is the material’s refractive index vm
c is the speed of light in a vacuum (3 x 108 m/s)
and vm is the speed of light in the material

Using n = 1.5, we see that the speed of light in fiber is 2 x 108 m/s.

25
Latency: The SAN killer
Providers normally limit storage rings to ~50 km circumferences
Let us assume that in this network the LAN and switches have no latency. Assuming a 2
ms ‘ack’ window, that give us 1 ms for each leg of the communication.

1 ms for signal
optical transport

storage switch
bank network

server farm
1 ms for ‘ack’

We know that vm in fiber is 2 x 108 m/s and that we have 1 x 10-6 seconds for each half of the
signal. This gives us a maximum distance of 200 km on each side of the ring.

26
Loss in fiber

90% of attenuation in fiber is from Rayleigh scattering


As a signal travels down fiber, it loses some of its power over distance. This is called
attenuation or power loss. There are two main causes for attenuation – absorption and
scattering.
This air bubble has n=1 The energy from the photon is absorbed by the
atom and then lost as heat
Types:

▪ Rayleigh
▪ Mie
▪ Geometric
Scattering Absorption
27
Multimode versus Singlemode fiber
BUFFER
CLADDING
Laser 50 mm

multimode fiber

BUFFER
CLADDING
Laser

singlemode fiber
28
Summary of fiber types

Type Multimode Single Mode


Core Size 62.5 + 50 mm 9 mm
Cladding Color Orange Yellow
Operating l 850 nm 1310 and 1550 nm
Max Distance 350 + 500 m 1000’s of km
Distance Limiter Modal Dispersion Attenuation (Loss)

29
Summary

✓The Cladding, Core, and Buffer make up optical fiber and they steer light through total internal
reflection.

✓The refractive index determines the speed of light and hence fiber latency in a network.

✓The main difference between multimode and single mode is core size and hence the angle of
acceptance.

✓Scattering and Absorption are the main causes for loss, with Rayleigh Scattering causing 90% of
the loss.

30
Optical Transmission:
Span Considerations
Objectives

Upon completing this lesson, you will be able to:

• Describe common conversions to logarithmic units and how to use them in simple link budgets.

• Explain the difference between linear and non-linear effects

• Describe attenuation across the wavelength spectrum

• Describe the difference between amplification and regeneration

32
dBs and dBms

dB = 10 log (Pout / Pin) dBm = 10 log (Pin / 1mW)


dB decimal dBm Decimal
10 x 10 30 1W
5 x3 10 10 mW
3 x2 3 2 mW
0 x1 0 1 mW
-3 x 0.5 -3 0.5 mW
-10 x 0.1 -10 100 mW
-20 x 0.01 -20 10 mW
-30 x 0.001 -30 1 mW

The most useful conversions should be memorized


33
What are Linear and Non-Linear effects?
Going back to NASCAR … A linear event is one that effects all participants equally. In the
case of our conceptual model, this can be represented by a yellow caution flag:

A non-linear event is one that effects the participants unequally. In our model, a good
example would be an oil slick on some part of the track:

34
Linear and Non-Linear effects in fiber
Solution: Linear = amplifier, non-linear = regeneration

A linear effect is one in which the signal is affected uniformly across the wavelength
region of interest. A good example is attenuation, or loss, in optical fiber. They are the
easiest of the transmission hurdles to overcome.

Here we see an
idealized ‘brick
wall’ signal.

Non-linear effects are generally random but certainly do not


have uniform consequences across our spectrum of interest.
Generally, non-linearities are very expensive to correct and
should be avoided where at all possible.
35
Attenuation in optical fiber
The most important obstacle to transmission that we must overcome in a fiber network is attenuation,
or loss of optical power. Luckily, attenuation is a linear effect, though it is not constant across
wavelength. We can plot loss (in dB/km) versus wavelength:
The water-absorption peak is
Loss (dB/km) avoided for transmission
0.5

0.4

0.3
This region, centered These regions were opened
0.2 around 1310 nm was the up by optical amplifiers and
original transmission are now preferred.
0.1 window for fiber optic C L
signals. O Band Bands
l (nm)
1200 1300 1400 1500 1600
36
Calculating Link Budgets
Let’s say we have a standard laser operating with 1 miliwatt of input power at a wavelength 1310 nm.
If the signal passes through 30 km of single mode fiber, what is the output power at the receiver?
At 1310 nm single mode fiber has a loss of 0.3
From our table, we know dB/km. We multiply it by 30 km and remember
that 1 mW = 0 dBm that it’s a loss.
1310 nm P I N

0 dBm -9 dB = -9 dBm
If we have a 1310 nm laser operating at 0 dBm, but now know that the minimum sensitivity of the
receiver is -24 dBm, what is the longest span of single mode fiber we can operate on?

We know: Laser Power – Fiber Loss = Receiver Power


Rearranging: Fiber Loss = Laser Power – Receiver Power
Fiber Loss = 0 dBm – (-24) dBm = 24 dB
So our Max Span = Fiber Loss (dB) ÷ Loss/km
x (dB . km-1) = 80 km
37
Modal Dispersion
Loss of peak
Only single mode fiber (SMF) is used for optical networks power

Laser

multimode fiber Pulse broadening


These photons are identical in wavelength, the
differences in colour are to illustrate the
different modes each will take.
Since the difference between the ‘fast’ mode and the slow ‘mode’ increases with
distance, the signal very quickly becomes unrecoverable in multimode fiber.

38
Amplification v. Regeneration

A regenerator takes a signal coming into it and takes it through an O-E-O process, during
which the 3 ‘R’s are performed; reshaping, retiming and re-amplification.
Reshape Retime Reamplify

e-

A fiber amplifier is basically a length of fiber with a pump laser adding energy into it, the
incoming signal ‘takes’ this energy to amplify itself as it travels through.

Laser
39
Calculating Link Budgets II

Look at the following network. If the laser output is the standard 1 mW, and the amplifier’s
gain, or increase in power, is 18 dB, what is the power at the receiver?
36 km 64 km
1550 nm P I N 1550 nm
P I N
Regenerator
AMPLIFIER

Power(L) – FL(1) + Amp – FL(2) = Power(R)


0 dBm – 9 dB + 18 dB – 16 dB = - 7 dBm
What if we used a regenerator instead of the amplifier. What would the power at the
receiver be now? Since we are doing a ‘3R’ on the signal, it’s as if the signal started at the
output of the regen., and we can ignore everything before that point:
0 dBm – 16 dB = - 16dBm

40
Summary

✓dBs are a relative measurement of power while dBms are an absolute measurement tied to 1 mW.

✓Linear effects act the same on all parts of the signal

✓ The loss at 1310 nm is 0.3 dB/km and 0.25 dB/km at 1550 nm.

✓ Amplifiers ‘re-amplify’ the signal while regenerators ‘re-time’, ‘re-shape’ and ‘re-amplify’ the
signal.

41
Optical Transmission:
Non-Linear Effects in Fiber
Objectives

Upon completing this lesson, you will be able to:

• List the three types of dispersion that optical signals must deal with

• Describe the effect that dispersion has on a signal

• Describe the realtionship between optical Group Delay and Chromatic Disperison

43
Can light’s speed be effected by wavelength?
If it’s not the cars, the reason for this has to be the track
Conceptually if you were to run a race as shown, with each colour car representing
a different wavelength of light, thanks to Einstein, what you would expect would
be a dead heat

We can actually run this experiment, using a tunable laser to fire photons at different
wavelengths and a fast stop watch to measure the time taken through a known length of
fiber. The results are surprising:

44
Group Delay and Chromatic Dispersion
We can do the same experiment as in our conceptual model, stepping a tunable
laser in wavelength and plotting the times taken:

Loss (dB/km) Dispersion (D) (ps/nm)


Time Taken (ps)
0.5
x
0.4
x
0.3 x
0.2 x x
x x
x x
0.1 This point of lowest transit time is called
the lambda zero (l0) point
l (nm)
1200 1300 1400 1500 1600
45
Chromatic Dispersion
CD is constant for any given fiber and builds with distance

Remember the real world laser? We noted that not all the photons coming from
it were of exactly the same wavelength.

This means that each of these different photons will travel at a different rate
down the fiber. While the cause of CD is very different from Modal Dispersion,
the effects from all forms of dispersion are always the same:
Loss of peak
power

Pulse broadening
46
Chromatic Dispersion at 10 Gb/s
At relatively low speeds, such as OC-48 (2.5 Gb/s) it will take hundreds of
miles before dispersion has a noticeable effect on the signal. This is because
the spreading is very small in comparison to the bit period:

When we go to OC-192, x4 the bit rate, the amount of dispersion has not changed, but the
bit period is ¼ the size so it’s tolerance to dispersion is much less.
47
What is Polarization?
Polarization divides optical power into different ‘modes’
A lightwave’s state of polarization describes the distribution of the signal’s
optical power (photons) in the x (H) and y (V) axes.
Z Z Z

V V V

H H H

Horizontally Polarized Vertically Polarized Randomly or Un-Polarized

48
PMD’s effect on an optical signal
At 10 Gb/s PMD has 1/50th the impact of CD and can be ignored
A light wave’s state of polarization only becomes of concern due to ‘birefringence in
fiber’, which means that the core of fibers isn’t perfectly homogeneous – there
exists differences in refractive index by axis.
Birefringence means that nx ≠ ny

nX
ny
CORE

CLADDING Since n determines the speed of light in fiber


(remember latency?), each state of polarization travels
BUFFER
at a slightly different velocity.

49
Summary

✓Optical signals must deal with Modal, Chromatic, and Polarization Mode Dispersion

✓All forms of Dispersion cause loss of peak power and pulse broadening

✓Chromatic Dispersion is the rate-of-change of optical Group Delay as you move along the
wavelength range

50
Optical Transmission:
Single Mode Fiber Types
Objectives

Upon completing this lesson, you will be able to:

• Describe loss/km numbers across the wavelength spectrum of interest

• Identify the common types of Single Mode Fiber that you may encounter by loss and dispersion
characteristics

• Explain why the Lambda Zero point of fiber has moved as fiber has evolved over time

52
Single Mode Fiber [l0 = 1310 nm]
Single Mode Fiber (SMF) was first introduced in 1986 and is the most deployed
fiber. An example would be Corning’s SMF-28
Loss (dB/km) D (ps/nm-km)
In the region of most interest SMF
0.5 has ~18 ps/nm-km of D 20

0.4
15
0.3
10
0.2

0.1
l0 5

l (nm)
1200 1300 1400 1500 1600 53
Dispersion-Shifted Fiber [l0 = 1550 nm]
In DSF the l0 was moved to the 1550 nm C-band. This was to allow the signals to
travel longer distances without the need for regeneration.
Loss (dB/km) D (ps/nm-km)
0.5 20

0.4
15
0.3
10
0.2

l0 5
0.1

l (nm)
1200 1300 1400 1500 1600 54
Non-Zero Dispersion-Shifted Fiber [l0 ~ 1480 nm]
NZ-DSF is the fiber of choice for today’s deployments and comes in a variety of
specialized sub-types (reverse slope, extended band, etc …)
Loss (dB/km) D (ps/nm-km)
0.5 20

0.4
In the region of most interest SMF 15
has ~ 6 ps/nm-km of D
0.3
10
0.2

l0 5
0.1

l (nm)
1200 1300 1400 1500 1600 55
Extended Band Fiber [l0 = ~ 1490 nm]
Extended band fiber is similar to NZ-DSF with one exception; the OH- peak has
been removed, allowing a very wide transmission band
Loss (dB/km) D (ps/nm-km)
0.5 20

0.4
15
0.3
10
0.2

l0 5
0.1

l (nm)
1200 1300 1400 1500 1600 56
Summary of fiber types and their uses

Only DSF fiber will cause major design challenges


Loss @ 1550 D @ 1550 Needs dispersion
Fiber Type DWDM?
nm nm compensation?

SMF 0.25 dB/km 18 ps/nm-km ✓ Yes, will require high compensation


values and amplifiers

DSF 0.25 dB/km 0 ps/nm-km  No compensation needed, but only


one laser signal allowed

NZ-DSF 0.25 dB/km 6 ps/nm-km ✓ Yes, but lower amounts and amp
counts.

57
Calculating Dispersion along a link

Let’s look back at our previous example and this time calculate the amount of chromatic
dispersion (D) accumulated by the end of the link:
36 km 64 km
SMF NZ-DSF
1550 nm P I N
AMPLIFIER

D total = ( 36 km x 18 ps.(nm.km)-1) + ( 64 km x 6 ps.(nm.km)-1)


D total = 648 ps/nm + 384 ps/nm = 1032 ps/nm
Receivers are usually given a dispersion tolerance figure. How many kms of NZ-DSF fiber
can a signal go through and still be recovered by a Rx with D tolerance of 3200 ps/nm?

Length = 3200 ps.nm-1 x 6÷ps.(nm.km)-1 = 400 km

58
Summary

✓Fiber has loss characteristics of 0.3 dB/km in the O-Band and 0.25 dB/km in the C-Band

✓The most common types of fiber are SMF, DSF, and NZ-DSF (of different types)

✓ The lambda zero point in fiber has gone from 1310 nm to 1550 when amplifiers were introduced
and slightly off 1550 nm when DWDM was introduced.

59
Optical Transmission:
Other Optical Components
Objectives

Upon completing this lesson, you will be able to:

• Describe the main characteristics of optical filters

• Describe two ways of making optical filters

• Understand the use of isolators, couplers and attenuators

61
Filter Properties
A wideband light source is one that fills all wavelengths with the same level of optical
power. We can study a filter’s properties by putting it in front of the source.
Power(dBm)

0
Peak Power
FWHM
-5
-3 dB
- 10 Channel
Isolation
- 15

- 20 Noise Floor

Center Wavelength
l (nm)
1535.82 1536.61 1537.40 1538.19 1538.98
62
Types of Filter
P (dB)
Also known as low/high pass filters
Uses:
• separating 1310 and 1550 nm windows
Wideband Filter l (nm)
P (dB)
Also known as band/channel pass filters
Uses:
• Dropping a ‘band’ of wavelengths
Notch Filter l (nm) • Cascaded in large multiplexors
P (dB)
Uses
• Single wavelength drops
• Regulating laser performance
1l Filter l (nm)
63
Optical Filters – the network gatekeepers
By using filters we guarantee that one signal will not impact another

One way to think of a filter would be as a tunnel wall, preventing most wavelengths from
continuing while allowing select wavelengths to pass:

The filter will stop some of the


photons. We know that in reality
these photons are reflected, not
destroyed.

The filter creates a virtual ‘lane’ or channel. There


are no lanes painted but only one colour of
photons is allowed through Here we have photons of
different wavelength
64
Thin-Film Filters
In a thin-film filter, alternating layers of coatings with
different refractive indexes are built up upon a silica
glass substrate.

The gas
deposits a very silica glass
uniform and
thin film on the
glass

This is repeated,
typically in l/4
thicknesses
pellet of deposition material

Vacuum Deposition Chamber


Vacuum Deposition Chamber 65
Thin Film Filters 2
Thin-film filters are the most commonly deployed in networks
The process of using vacuum deposition is a
relatively old one, with many common uses,
such as eyeglasses and car rear-view mirrors.

Thin-film filters are:


+ Easy (cheap) to manufacture
Thin-film filters under x12 magnification
+ Low attenuation for ‘pass-through’ light
- Low channel isolation
- Only work on one wavelength range of light (have to use multiples to deal with
multiple signals)
66
Bragg Gratings
Strong UV light permanently changes the properties of
silica (think coke bottle bleached in the desert sun). By
exposing the core and cladding to a source, we can create
changes in the refractive indices of both.

UV SOURCE

UV Mask

Silica sample of a ‘chirped’ FBG


single mode fiber
While the UV light changes the refractive
index of the slice relative to the rest of the fiber, the
relationship between cladding and core remains constant.
67
Bragg Gratings and Circulators
Bragg gratings are seldom used without circulators. Circulators are 3-port devices that always
move light in the same direction. Light entering in Port 1 will exit Port 2, entering Port 2 will
exit Port 3, and entering Port 3 will exit Port 1:

P1

P2
P3
Bragg grating
Bragg gratings are very sensitive to changes in temperature, since that will effect the size
of the ‘slice’, so they are
always hermetically sealed with a
substance with reverse thermal
properties to silica.
68
Isolators, Couplers and Attenuators
Isolators are specialized one-way filters,Attenuators add loss to optical signals, usually by
used to prevent dangerous back- adding an air gap. Variable Optical Attenuators (VOA)
reflections. allow the attenuation to be tuned.
Some of the
power is lost in
the air gap.

Isolator
Depending on the direction of input a Coupler can also be called a splitter. They are
essentially the same device, just connected in different directions.
Splitter Coupler
Pin - 3dB P1in
Pin P1in + P2in
Pin - 3dB P2in
Two common types: Normally couplers are always
▪ ‘3dB’ – 50/50 split of signals 50/50
▪ ‘10dB’ – 90/10 split
69
Peltier Devices
A Peltier device or thermoelectric heat pump is a solid-state active heat pump which
transfers heat from one side of the device to the other. Peltier cooling is also called
thermo-electric cooling (TEC).
COOLER
WARM

e-

WARMER
WARM
Peltier devices are used for devices that need to be temperature stable for linear
operations. The rate of their heating/cooling can be precisely controlled by the voltage
applied. The main drawback is that they fail with power loss.
70
Summary

✓The main characteristics of filters include Channel Isolation, Full Width Half Max bandwidth, and
Center Wavelength.

✓Optical filters can be made of thin-film filters or Bragg gratings.

✓Isolators are one-way devices, couplers put two light waves onto the same fiber, and attenuators
add loss to an optical signal.

71
DWDM Components:
Optical MUX/DEMUX/OADMs
Objectives

Upon completing this lesson, you will be able to:

• Explain the block diagram of a basic Thin-Film Mux

• Explain the workings of an AWG Mux

• Explain the block diagram of a one lambda FOADM

• Recognize the ITU nomenclature for transmission bands.

73
The Optical Filter; Building Block of Muxes
An optical filter basically has the role of allowing some wavelengths of
light to pass through it, while rejecting others, or more precisely it will
refract some wavelengths, while reflecting others.
l l

optical
filter
l l

Since filters are passive devices, basically a piece of silica glass, they can work with any
combination of optical signals; they do not care about bit rate or protocol, only the
wavelength of the signal, and they have no function other than to reject or allow the signal.

74
Building an Optical Mux
Thin-film muxes are passive devices; they draw no power.
We can use 3 thin-film filters to create a 1 x 4 multiplexor; a device that will take
multiple optical signals and put them onto one fiber, separated by wavelength.
We start off with three
mechanical optical filters, much as
optical assembly we’ve seen previously.
assembly
Connecting the
ports on the We then surround the
optical assembly filters with an
to exterior ports assembly to keep
on the them all in a fixed
mechanical distance from each
optical fiber
chassis using other.
fiber, we now OUT IN IN IN IN
have finished
our mux. l l l l l

75
A Matching Twin: The Demux

Since it’s aInpassive


practice,device,
differentadesigns
mux can be used
are used, as principle
but the a demultiplexor
applies simply by
changing the direction of the input:

mechanical
optical assembly
assembly

optical fiber
IN OUT OUT OUT OUT

l l l l l

76
Patching Through Terminal Muxes
At some nodes, not all wavelengths need to be dropped. It is a very common
practice to ‘patch through’ wavelengths across muxes:
Patching through makes sense l
for small counts, but with 88 ls this can be very time
These are the standard symbols for mux / demux
consuming terminal
– a better drop
pairs solution
in network is needed!
diagrams

l l
DEMU

MUX
l l
l l
X
l l
l l

Any ls that need dropping can be


We can patch through any dropped and added back again.
wavelengths that do not drop at
l
this node

77
A 1l Fixed Optical Add Drop Mux (FOADM)
The VOA must be used to match the ADD power to the THRU power
The mux/demux pair that we’ve seen so far are ‘terminal’ components - in
other words all wavelengths are dropped. Often it is useful to drop only
one wavelength without disturbing the others.
l

OUT
l
IN

Variable Optical Attenuator


DROP ADD

l l

78
The ITU Grid (G.694)
The ITU has standardized wavelength spectrum of interest for fiber
transmission by dividing it by band and channel:
With the introduction 20 nm
ITU-T: Standard G.694.2
of ‘allwave’ fibers,
CWDM channels
spaced 20 nm apart
were also defined 1271 1311 1351 … 1671

0.8 nm
ITU-T: Standard G.694.1 The C and L bands have been
divided up by 0.8 nm wide
channels for DWDM
transmission
1552 1554 1556 1558 1560
O Band E Band S Band C L U/XL
1260 - 1360 nm 1360 - 1460 nm 1460 - 1530 nm 1530 – 1565 - 1625 1625 - 1675

l (nm)
1200 1300 1400 1500 1600
79
Multi-l FOADMs
This shortcoming of FOADMs was addressed by the introduction of Arrayed Waveguide (AWG)
Multiplexors
You can increase the numbers of add/drops in an ODAM simply by adding more
filters:

OUT

OUT
IN

IN

DROP DROP DROP ADD ADD ADD

The downside to doing this is that ‘pass through’ signals will take a lot of
additional loss (more filters to go through) without regeneration. This makes
large scale FOADMs impractical in comparison to terminal muxes.

80
Using Interference as a Demux
Irrespective of l count, AWG only have one loss for all channels
Each copy is sent down a
l different path length and
is launched at a different
IN angle.

By calculating the angles


and lengths correctly a
predictable interference
The input signal pattern is created.
is split into OUT OUT OUT OUT
several identical
copies
l l l l

81
Arrayed Waveguides (AWGs)
Here, we see the block diagram of a 1x8 lambda demux:
For higher ls, AWG loss is much less than thin-film muxes

16-ch Arrayed Waveguide Grating Module with 100-GHz Spacing by Tsunetoshi Saito, Toshihiko Ota, Tomoaki Toratani and Yoshimi Ono
Furukawa Review, No. 19. 2000

82
Structure of an AWG mux
Because of the very precise geometry involved, AWGs must be temperature
stabilized using a Peltier device:

If there is a loss of power to the mux, the


Peltier device will stop working and the
AWG will go outside of operating AWG Chip showing a Mux/Demux pair
temperature.
83
Summary of Mux Technologies

Thin–Film Bragg Grating A.W.G.


Manufacturing Ease ✓✓✓ ✓✓✓ ✓
Unit Cost ✓✓✓ ✓✓* ✓
Temp. Sensitivity ✓✓ ✓✓✓* ✓✓✓†
dB Loss (low l mux) ✓✓✓ ✓ ✓✓✓
dB Loss (high l mux) ✓✓ ✓ ✓✓✓
Key: ✓✓✓ = Excellent ✓✓ = Good ✓ = Poor
* - due to cost of hermetic sealing
† - due to use of peltier device
84
Higher Power in Fiber
A warning should be given about one of the unavoidable consequences of
multiplexing optical signals. Optical signals operate in a linear manner in fiber
within a wide range of powers, but multiplexing can quickly increase the total
power in a fiber:
Consider a 1x16 optical mux. Each wavelength in the fiber will be
1
2 added at 1mW of power. This means that while the single
MUX
1x16

channel power is 1mW, the total power in the fiber has increased
to 16mW. Given that amplifiers can easily have 20 dB (x100) of
15
16 gain, total power in fiber can quickly get above 1W.

Conversion Table for Single Channel to Total Power:


1 x ? Mux: 4 8 16 32
Change in input power +6 dB +9 dB +12 dB +15 dB

85
Summary

✓A thin film mux uses a series of filters to drop and add individual channels.

✓Arrayed Wave Guides use interference to split or combine optical signals.

✓In one lambda FOADMs a VOA must be used to balance the powers of each wavelength.

✓The ITU transmission bands of interest are the O,E,S,C,L and U bands.

86
DWDM Components:
Dispersion Compensation
Objectives

Upon completing this lesson, you will be able to:

• Explain what design characteristics of dispersion compensating fiber allow it to mitigate the effects
of CD

• Calculate the amount of dispersion generated by a given length of optical fiber

• Describe how an active dispersion compensator works.

88
Compensating for Chromatic Dispersion

With CD, one advantage we have is that it is linear and constant. Let’s go back
and look at SMF fiber:
Loss (dB/km) D (ps/nm-km)
0.5 Since the region of interest has a 20
GD slope shaped like this,
intuitively if we can find a fiber
0.4
with the reverse shape, their 15
effects should cancel out.
0.3
10
0.2

0.1
l0 5
C L
1530 – 1565 - 1625
l (nm)
89
Dispersion Compensating Fiber
DCFs add extra loss (and latency!) to the optical network
If we can design a fiber to have the reverse characteristics as SMF (or any other fiber type)
running the signal through both should negate any effects of dispersion:
GD D

SMF Fiber

GD D

DCF Fiber

90
FBG based dispersion compensation
Since they are short, FBGs have low latency!
If we can design a fiber to have the reverse characteristics as SMF (or any other fiber type)
running the signal through both should negate any effects of dispersion:

SMF Fiber

P1

P2
P3
‘Chirped’ Fiber Bragg Grating

91
Active Dispersion Compensation
Active dispersion compensation comes into its own in a dynamic network environment and
goes hand-in-hand with ROADMs:
Node A On a fiber break, the path changes and the total D goes
180 ps/nm up to 840 ps/nm of D, so the DCM must also be
10 km changed to reflect the new value.
120 ps/nm Node B
ROADM
20 km
Switch
SMF-28 Ring
720 ps/nm
40 km NZ-DSF Ring
The active path has a
At either node, the flow from ring to 60 km
total of 300 ps/nm of
CPE would be as follows: D, so the DCM is set to
Dispersion Sub-rate -300 ps/nm.
4-l Compensation Customer
FOADM Module Handoff
RING - 300 ps/nm CPE
- 840 ps/nm 92
Tuning a DCG
FBGs provide less total -D than DCFs, but they can be tuned
Since the key to a Bragg grating is the ‘period’, or width, of the grating, you can change the
amount of negative D you get from a FBG based DCM by changing the period. The
simplest way to do this is through the use of a Peltier device:
Fiber Bragg As a current is run through the Peltier device,
Grating the device heats up the FBG. Fiber expands
much more easily along its length than
outwards, causing the Bragg period to increase.

e-
- This same principle holds true for chirped FBG
used in DCMs. By heating and cooling them, you
Peltier Device
+ can create a tunable dispersion compensator.

93
Summary

✓ DCF has a Dispersion slope that is opposite to that of single mode fiber.

✓ Dispersion accumulates by distance according to what fiber type the signal is going over

✓ Active Compensators use a peltier device to increase and decrease the width of a chirped FBG,
causing varying amount of negative dispersion to be added to a network.

94
DWDM Components:
EDFAs and Raman Amplifiers
Objectives

Upon completing this lesson, you will be able to:

• Explain spontaneous and stimulated emission and how it is used to amplify an optical signal

• Understand the key characteristics of EDFAs and how they effect performance

• Understand the process by which Raman amplifiers improve fiber performance

96
Erbium Doped Fiber Amplifiers
EDFAs have four main components:
• Most widely used amplifiers
• Usually works in the C-band (though L-band is available)
• Bitrate and wavelength agnostic (works at the photonic level) The EDF is just single mode fiber
• Generally between 1-35 dB of gain with ions of Erbium mixed, or
‘doped’, into the core.
mechanical
assembly
Coupler

OUT
IN

Isolator

Pump lasers Pump Laser


usually work at a
l of 980 nm or
1480 nm. Erbium Doped Fiber

97
The Photoelectric Effect
If an EDFA has no input signal, all ions naturally return to the fundamental state by
spontaneous emission
Although unlikely, every once in a while, a photon will hit an atom in the
fiber core head on and will disappear, its energy absorbed.
Energy This high energy state is
unstable, and the ions Einstein showed us that the energy of a
The 980 nm ‘spontaneously’ emits a photon, E = h x f
pump laser new photon
photon hits the Where h = plank’s constant
ion and is
High Energy State f = photon frequency
absorbed
so we can say that: E ∝ f

or put another way; the energy in a photon is


This in turn inversely proportional to wavelength.
raises the ion’s
energy state

Fundamental State
Erbium Ions 98
Why Erbium is used in EDFAs

Erbium has an intermediary state of energy, lower than the state it is raised to by
absorption of a photon’s energy. This level is called a ‘metastable’ state.
Energy The ion again is raised to a high
energy state, but stays only briefly
The 980 nm before dropping to a metastable It does this by releasing some of
pump laser state the energy as a photon
photon hits the High Energy State If during the excitation window a
ion and is telecom photon goes by the
absorbed Metastable State metastable ion, it will stimulate
emission of the photon

1550 nm

Fundamental State Er 3+ Er 3+ Er 3+ This is amplification by


Erbium Ions stimulated emission 99
Non-Linearities in EDFAs

We can look at the power output of an EDFA versus wavelength:


This is an EDFA with no input signals to amplify – all
the Erbium ions spontaneously emit their energy. If we add an input into the EDFA,
Please note that the output power is non-linear in we see that a lot of the energy is
Power (dBm) the c-band. now being stimulated into
emission. While this signal is
0 more linear, it still leaves a lot to
be desired.

-10

-20

l (nm)
1510 1530 1550 1570 1590
100
Gain Flattened EDFAs
Since Erbium’s spectral emission is non-linear, we must modify the output of the
EDFA, to make it perform in a linear manner.
The GFF makes EDFAs linear within their operating range
In the EDF, the signal is amplified,
though not linearly.
mechanical
assembly
Coupler
l l

OUT

OUT
l
IN

Isolator Gain Flattening


Pump Laser Filter

Erbium Doped Fiber

The WDM signal, attenuated by the fiber Some amplification is lost in the GFF,
span enters the EDFA but the result is now linear.
101
Two Stage EDFAs
Dual Stage EDFAs have been designed for use with DCF modules.
Almost all EDFAs sold today are dual-stage EDFAs

Coupler GFF Coupler

OUT
IN

Isolator Isolator

980 nm 1480 nm

EDF EDF

Different DCF modules can be By connecting mid-span any


connected mid-span on dual DCF attenuation from the DCF can
stage EDFAs according to the be compensated for by the
network design second EDFA
102
Gain Competition in EDFAs
To avoid this, all input ls must be equal coming into an EDFA
There is one non-linearity in EDFAs that cannot be eliminated by design, but
instead must be avoided at turn-up; Gain Competition.
As a telecom signal enters the EDF, its photons trigger simulated emission. This
creates a cascade like effect causing amplification:

If the input signal is higher (more photons), then more photons will trigger more
simulated emission – causing larger (non-linear) amplification.

103
Optical Signal to Noise Ratio
The only way to correct for a poor OSNR is to regenerate!
EDFAs are not 100% efficient; not all the power of the pump laser becomes
amplification, some is wasted on adding system noise (ASE)
The signal is amplified but
l Amplified Spontaneous
l Emissions (ASE) noise is also
EDFA added.
If we look at what happens when we pass through a series of EDFAs:

l l l

l l l

104
Raman Amplification
Raman amplifiers also work with the similar principle of stimulated Raman scattering (SRS)
to increase the amplitude of a signal, so they also enjoy the same benefits of being bit rate
and wavelength agnostic.
Power
(Vibrational) Energy 980 nm
Now when the atom
The pump ‘spontaneously’ emits a new
photon hits the photon it will be as a different
atom and is wavelength (1550nm)
Coupler
absorbed Vibrational State

Since v. high power is needed, the effect falls


exponentially away from the laser
In this case, some of Raman amps do not use specially doped fiber
the energy is lost in they are attached to the transmission fiber
raising the atom’s and work with the fiber itself as gain medium
energy
Fundamental State (usually in reverse from the signal direction)
Silica Atoms 105
Summary

✓ Stimulated emission is the process which causes amplification while spontaneous emission causes
noise

✓ The key characteristics of EDFAs are Gain, Gain Tilt, Gain Competition, and OSNR

✓ Raman amplifiers use Simulated Raman Scattering to effectively lower the loss of fiber/km

106
DWDM Components:
ROADMs
Objectives

Upon completing this lesson, you will be able to:

• Explain one, two and multi-degree ROADM configurations.

• Step through the block diagram of a ROADM

• List the various WSS technologies deployed in ROADMs

108
Reconfigurable Optical Add Drop Multiplexors (ROADMs)
For anything beyond one degree, multiple WSSs are needed
ROADMs, offer the same functionality as Muxes and FOADMs, but with the ability to change the drop
wavelength or port dynamically. Its core, a Wavelength Selective Switch (WSS) is reconfigurable, acting as
many different types of network component:

The simplest configuration is a 1 x n The functionality of a FOADM can also be


configuration, emulating the function of mimicked by a ROADM. In this case a multi-l
a terminal mux or demux. add/drop node.
This is a two degree configuration
This is a one degree configuration To Ring2
With a true n x n switch, you can
terminate multiple rings on the same To Ring1 To Ring3
ROADM, plus have add/drop ports. To Ring4

These are all called multi-degree


Add/ Drop
configurations

109
ROADM Block Diagram
Regardless of the underlying technology, ROADMs all have similar functionality
and sub-components:
▪ A Wavelength Selective Switch (WSS) – although several technologies
are being touted, the functionality of all WSS are essentially the same:

• The ability to switch an input beam to any of its output ports


• The ability to split (copy) an input beam
• The ability to attenuate input and output ports
▪ A Mux/Demux technology to allow add/dropping of individual ls
▪ Optional components include:
• EDFAs (to help counteract all the losses of the ROADM)
• Tunable dispersion compensation
• Terminal equipment for an Optical Supervisory Channel (OSC) 110
WSS Technology: LCoS

Liquid Crystal over Silicon (LCoS) technology works by changing the phase of a wave.
Unlike a mirror, when light hits a LCoS, the angle of reflection is not fixed, but depends on
the phase change made by changing the pattern on the Liquid Crystal pixels:

+ - + -
CMOS Substrate CMOS Substrate
LC Pixels voltage ‘off’ voltage ‘on’

Liquid Crystal over Silicon (LCoS) has tremendous stability and repeatability though the
trade off is that it generally has greater loss than other designs
111
LCoS Block Diagram

The following illustration represents the block diagram of an LCoS based WSS:

Switches & Attenuates Wavelengths

Separates & Combines Wavelengths

112
WSS Technology: LC
By adding n number of ‘slices’ we can get a 1 x 2n switch
From our discussion of PMD we remember that every optical wave is made up of a vertical
and horizontal states of polarization (SOP). With the LC Cells ‘off’ there is no
The PBS separates the signal into vertically and effect on the signal.
horizontally polarized components

OUT 1
IN
OUT 2
Polarization Polarization
Beam Splitter Beam Combiner

Basically we have a 1x2 switch: LC Cells The LC cells will rotate the phase of the signal
‘Off’
‘On’ 90° when a current goes through it.
LC CELLS

LC CELLS
PBS

PBS

113
WSS Technology: MEMS

Micro-electro-mechanical systems (MEMS) are micro-mirror arrays fabricated in silicon.


With no charge on the
mirror, the beam reflects
to output port1.

By applying a charge
We can similarly to this plate, the
attenuate the beam mirror is moved in the
x-axis

A MEMS mirror MEMS can be designed to


shown deflecting on
the x-axis (x700 move in 2 or 3 dimensions
magnification)
114
Comparison of WSS Technologies

Currently LCoS and LC show the most promise

LCoS LC MEMS
l spacing 0.8, 0.4 nm 0.8 nm 0.8, 0.4 nm
l range c-band c-band 
beam radius 15 mm <60 mm <10 mm
Isolation ✓✓ ✓✓ ✓✓✓
Reliability ✓✓✓ ✓✓✓ ✓
Insertion Loss (low port count) ✓✓ ✓✓✓ ✓✓
Insertion Loss (high port count) ✓✓ ✓ ✓✓

Key: ✓✓✓ = Excellent ✓✓ = Good ✓ = Poor

115
ROADM Block Diagram

Here you see the LCoS


based switch at the
There are two EDFAS in this ROADM,
center of the ROADM one for each fiber direction
This shows the drop ports
coming from the WSS and the
add ports with VOAs built into
their paths

116
ROADMs and Dispersion Compensation

ROADM

In all-optical networks the complexities of circuit paths makes managing dispersion on a per-circuit basis
impossible.
▪ Instead, each span’s dispersion is managed locally. Most often this means bringing D to zero at the
end of the span, but in >10 Gb/s networks some amount of D along the length of the path may be
desirable.
▪ As previously discussed, tunable DCM modules at each ROADM node will allow for changes in D as a
result of a link failure. While this tunability is very useful, it is not dynamic enough to compensate for
PMD.
117
ROADM Limitations
As bandwidth in networks goes up, ROADMs will become generally deployed but cost is currently a
factor

▪ High insertion losses due to free space designs means they generally
are used with EDFA(s).
▪ l switching means that tunable dispersion compensation will also be
necessary.
▪ Some technologies that rely on mechanical switching (i.e. MEMS) have
relatively high failure rates and trouble with sparing ports

▪ Not all WSS technologies may be suitable for 40 and 100 Gb/s
networks

118
Summary

✓ The degree of a ROADM node refers to the number of fiber rings it spans

✓ Most ROADMS are made up of a WSS, multiple EDFAs and multiple VOAs

✓LCoS, LC and MEMS are all types of WSS deployed in ROADM nodes

119
Advanced Network Concepts:
Measuring Bit Error Rate
Objectives

Upon completing this lesson, you will be able to:

• Explain the concept of Bit-Error-Ratio

• Describe how an eye diagram is built

• Explain the types of effect that eye diagrams can highlight

• Explain the role of masks in eye diagrams

121
What is Bit Error Rate?

The Bit Error Rate (BER) of a signal is the ultimate test of its quality. For any given
transmission, the number of bits in error are counted and divided by the total number of
bits sent. The result is usually rounded to a power of 10:

𝐵𝑡𝑠 𝑖𝑛 𝐸𝑟𝑟𝑜𝑟 15 𝑏𝑖𝑡𝑠 𝑖𝑛 𝑒𝑟𝑟𝑜𝑟


BER = e.g.: 32 = 3.5 × 10−9 = 10−9
𝑇𝑜𝑡𝑎𝑙 𝐵𝑖𝑡𝑠 𝑆𝑒𝑛𝑡 2 𝑇𝑜𝑡𝑎𝑙 𝐵𝑖𝑡𝑠 𝑆𝑒𝑛𝑡
Most BER testing is done periodically to ensure the health of the transmission line.
Usually the test set up looks like this:
1 0 1 0 0 1 1 0 1 0 0 1

Signal Generator Digital


Sampling
Scope

122
Eye Diagrams
We talk about an eye diagram being ‘open’ if healthy
One method to quickly check the health of a signal or transmission line is to use the digital sampling
scope to ‘superimpose’ each bit as it is received:

If you think about a 3 bit block of


information, there are only a certain
number of permutations:
0–0–0
1–0–0
0–1–0
0–0–1
1–1–0
1–0–1
0–1–1
1–1–1

123
How Eye Diagrams show BER
Using our previous eye diagram as an example of a ‘healthy’ eye, let’s look at some of the
ways that an eye diagram can show problems in a transmission line:

Eye Diagram showing the


effects of dispersion

Modern digital sampling


Healthy Eye Diagram
scopes use colour to
represent intensity
Eye Diagram showing the
effects of jitter 124
Standards Based Masks
Since BER is so low in modern fiber based systems, even eye diagrams can fail to show the
occasional error. For this purpose masks have been developed to quickly highlight an error.
Masks can be pre-assigned a shape from a transmission standard (e.g. SONET) or can be
user programmed.
No part of the mask If the bitrate of the signal
border should ever be is increased, then the
crossed by a trace mask shape will change
The height of the mask to reflect the decreased
monitors the signal tolerance to bit signal
attenuation non-linearities

The width of the mask User defined masks


OC-48
OC-192
monitors for non- will allow for greater
linearities such as vigilance for specific
dispersion and jitter effects

125
Common BERs in SLAs
Service Level Agreements (SLAs) are the basis of all measurements of system uptime in service
provider contracts. The most common metrics of interest are BER and uptime, which is usually
measured through the use of a BER measurement.

Typical BER metric Type of SLA Example


10-6 Standard (Retail) Offered on most ISP circuits
10-9 Telecom Standard on phone line
10-12 Datacom 10 GbE Standard
Uptime is measured in a percentage and indicates what percentage of a calendar year a network is
up. Providers usually measure uptime in units of error, i.e. errored seconds or errored minutes.

Typical Uptime metric Type of SLA Time down/year


99.9 % Enterprise Internet 8⅔ hrs (526 mins)
99.99 % Leased l SLA 53 mins
99.999 % Bank feeds / custom 5⅓ (315 secs)
126
Summary

✓BER is the ratio of bits sent in error to total bits sent

✓ Eye diagrams are a graphic representation of signal health built by superimposing consecutive
received bits on top of each other

✓Attenuation, Dispersion and Jitter are all network effects shown by eye diagrams

✓Masks enforce the boundaries of performance imposed by standards for any given transmission
protocol.

127
Advanced Network Concepts:
Polarization
Objectives

Upon completing this lesson, you will be able to:

• Define linear polarization

• Understand the workings of an absorptive polarizer

• Understand the workings of an beam-splitting polarizer

• Describe the components of polarization maintaining fiber

129
States of Polarization
Most telecom systems use linearly polarized light for their signals
Polarization is a property of waves that describes the angle of oscillation (usually of
amplitude). Typically in telecom, light signals are polarized (the angle is fixed). This means
that the power varies in the same direction.
It should be noted that the
‘black’ transmission signal is
made up by its x (blue) and y
(red) compenents.

Linearly Polarized Light Circularly Polarized Light

130
Linear Polarizers
To get both SOPs, a splitter and two absorptive polarizers are needed
We know that a light beam is composed of vertically and horizontally polarized light. Linear polarizers
are devices that produce only one of these two States of Polarization (SOP) from their outputs. The
simplest way of doing this is by using an absorptive linear polarizer:
VERTICAL

HORIZONTAL
Absorptive Polarizer

131
Polarization Splitters and Combiners
A combiner is just a PBS used in the reverse direction
Beam-splitting polarizers work on the principle of birefringence..
𝑅𝑒𝑚𝑒𝑚𝑏𝑒𝑟𝑖𝑛𝑔 𝑆𝑛𝑒𝑙𝑙′ 𝑠 𝐿𝑎𝑤: 𝑛1 sin 𝜃1 = 𝑛2 sin 𝜃2
But birefringence means that nV ≠ nH. By treating the vertically and horizontally
polarized components of a beam differently they emerge from a Polarization Beam
Splitter (PBS) at different angles, thus retaining both SOPs

IN OUT

Polarization Polarization
Beam Splitter Combiner

132
Polarization Maintaining Fiber
We are used to the idea of SMF being made up of Core, Cladding and Buffer. To make PMF,
a fourth component is added; stress rods.
Panda fiber keeps previously polarized light from
The rods induce stress in the
core, effectively making the
changing polarization. As such the rotation of the
core oval in shape fiber when coupled to
the laser is critical

CORE

CLADDING

Usually a ‘keyed’ connector


BUFFER is used to ensure correct alignment.
133
Summary

✓Linearly polarized light has its oscillations in amplitude fixed in angle.

✓Absorptive polarizer absorb all but one state of polarization in a beam.

✓Beam-splitting polarizers separate the vertical and horizontal components of the beam by using
birefringence.

✓PMF has stress rods added to the standard Coating, Cladding and Core of optical fiber

134
Advanced Network Concepts:
Non-Linearities Due to High Power
Objectives

Upon completing this lesson, you will be able to:

• List some of the non-linear effects due to high power

• List the components of a modulated signal

• Explain the difference between In-Band and Out-of-Band Cross Talk

• Describe the issues created by Self Phase Modulation and Cross Phase Modulation.

136
Contributors to High Power
As a rule of thumb, SMF should be limited to 1 Watt (30 dBm)
In DWDM systems there are many separate reasons why spans can have too much optical
power concentrated in the fiber. :
▪ EDFAs – Amplifiers provide up to 35 dB of gain to incoming signals. Normally that is divided over
many wavelengths, but in cases where one l fails the extra power for the remaining ls may
cause problems.
▪ EDFAs + DCM – It can be tempting to insert a lot of negative dispersion midstage of an EDFA, but
care must be taken since -D also concentrates the signal and increases the peak power.

▪ High l count systems – In long haul networks the cost of trenching fiber far outweighs the cost
of terminal equipment, therefore there is always the pressure for higher and higher l counts.
Even at 0 dBm, when added in 10s to a fiber, the total power in the fiber can quickly go up.

Much as with Chromatic Dispersion, tolerance for non-linear effects goes down as the bitrate of
the signal goes up. Many of the effects in this lesson can be ignored below 10 Gb/s.

137
Modulation Basics
All modern telecom signals are modulated – they are made up of two signals, modulated
(mixed) together to take advantage of the strengths of both. In fiber optic systems, this is
amplitude modulation (AM).
193 x 1012 Hz Optical
Modulator The Modulation
Envelope carries the
data on the signal

1 x 109 Hz
1550 nm ‘carrier’ signal

1 0 1 0 0
▪ Lowest loss frequency The majority of the optical
▪ Too fast to use for data power is used to get the
signals signal across the span.
10 GbE ‘modulation’ signal
▪ Very easy to recover 1 and 0
▪ Can’t be sent on fiber at
native speed 138
Modulation in the Wavelength Domain
We can also look at the modulation of two signals in the wavelength domain. Let us
consider the example from the previous slide: We can filter out unwanted signals
P (dBm) P (dBm)
Optical
Modulator

l (nm) l (nm)
lc lc-m lc lc+m lm-c lm lm+c
P (dBm)
P (dBm)
overall signal shape

l (nm)
lm
Real world signals have These modulation
sidebands contain the
greater bandwidth than data
these idealized signals: l (nm)
139
Four Wave Mixing
Let’s look at what happens when we have several modulated signals in adjacent channels.
For simplicity, let’s replace the modulated signals with channels.
P (dBm)

CHANNEL 1

CHANNEL 2

CHANNEL 3
l (nm)
l1 l2 l3
140
Four Wave Mixing
This problem can not be avoided and must be solved by design
In DWDM systems, the various wavelengths will modulate with each other, but in regions
of high power this effect will cause noticeable signal degradation.
P (dBm) In this case, all the channels are
Even with only
spaced equally 3 ls the unwanted
But a lot of them do modulations are large:
f

interfere with existing l1 –+ l2 + l3


channels l +l –l
l11 – l22 + l33

CHANNEL 1

CHANNEL 2

CHANNEL 3
l1 – l2 – l3
Some of these spurious 2+1-3 l1 + l2 – l3
modulations are out of
active channels
l1 – l2 – l3
1+2-3 3+1-2
l2 + l1 + l3
l2 – l1 + l3

...
1-2-3 3-1-2 1-2+3 1+2+3
l2 + l1 – l3
l (nm)
l1 l2 l3 l2 – l1 – l3

141
Four Wave Mixing
The ITU grid has been designed to minimize FWM
The solution to this is to space the channels unequally, thereby making their modulation
products fall outside of any defined channel.
P (dBm)
By moving one channel you will Even with only 3 ls, it is
impact multiple products almost impossible to
design a scheme that
moves all modulation
products ‘out of band’ –

CHANNEL 1

CHANNEL 2

CHANNEL 3
you can imagine how
hard it would be to do
1-2+3 for the 10s of ls that
modern DWDM schemes
1+2-3 1+2+3
employ.
l (nm)
l1 l2 l3
142
Example of FWM and solution

Here we see the original DWDM signal and


eye diagram at input into a network

This is the result of using equal channel


spacing on the DWDM Grid

The only difference to the network here is


slightly uneven channel spacings

143
Crosstalk
Imagine a 3l network created using 1x4 mux/demux pairs using a ‘patch through’
configuration as shown below: Each time the signal goes
EDFA through an EDFA all the
channels receive Gain.
l l

l l
What we are interested in is The unused channel also
1x4 1x4
following the l that is not
receives optical power from the
l l EDFA, but since there is no
1x4 being used (and therefore not 1x4 signal in it, it all takes the form
of ASE noise
dropped anywhere) as it goes
around the ring. Since this channel is never dropped, there is
no break in the ring for the optical power to
be dumped out. Soon the power in the
channel breaks down the isolation of filters
l l
and starts to interfere with adjacent, used
EDFA channels
144
In and Out-of-Band Cross Talk
The addition of spurious optical power into a channel is called cross talk. There are two
types, Out-of-band (the easier of the two to prevent) and In-band cross talk.
Out-of-Band (OOB) Cross Talk
OOB Cross Talk was described in the previous slide. In this case the spurious optical power
comes from channels adjacent in wavelength. The easiest way to minimize it is by having an
‘open’ ring – one where all ls in the ring are dropped at one node. As we go to all optical
networks, this becomes less possible.
In-Band Cross Talk
A good example of In-Band Cross Talk would be to consider a 1l FOAMD:
l In the case of cross talk, not all the orange photons would
drop at the first thin-film, but rather would pass through
without being dropped. While the cause is different, the effect
is identical to OOB cross talk, first the orange channel and
then the adjacent red and green channels would go down.
l l

145
Self Phase Modulation (SPM)
In an individual lambda, the signal transmitted is made up of a series of pulses (think when
a ‘0’ goes to a ‘1’). When the total power in a fiber is high enough, these pulses within a
signal will modulate it’s own phase – in other words; the signal mixes with itself:
It is important to note that this effect is different than Chromatic Dispersion, which broadens the
pulse in the time domain.
At high power the signal acts like
two very close signals
The resulting spectrum is modulating one another
broader than the original

l (nm)
146
Cross Phase Modulation (CPM)

To avoid costly regeneration of an optical signal, regions of high power should be avoided through
design and dispersion management

Cross Phase Modulation is very similar to SPM, the main difference being that pulses in adjacent channels will cause
the modulation in phase – CPM is not possible in single channel (non-DWDM) applications.

One strategy to mitigate a lot of non-linear effects is to maintain a certain amount of Chromatic
Dispersion along the span; remember that CD not only causes pulse broadening but also loss of peak
power.
▪ Allowing a controlled amount of D to build along the signal path will spread total power over a larger wavelength
range, avoiding non-linearities.
▪ Dispersion management is well known and the effects of the increased D can be mitigated at span’s end.

▪ Non-Linearities have greater impact at higher speeds and for 40 Gb/s and 100 Gb/s networks, so many are
moving away from NZ-DSF to SMF-28 fiber links since the added D is helpful in avoiding non-linearities.

147
Bandwidth Narrowing in ROADMs
All-optical networks will have circuits than span multiple rings and ROADMs.
ROADM ROADM ROADM

Once this narrowing impacts


The WSS in each the sidebands, the BER will
ROADM also acts degrade.
as a band pass
filter.

l (nm) l (nm) l (nm)


lc lc lc
Any misalignment in the wavelength accuracy of the ROADM start to narrow to
total bandwidth of the signal. Since the signal is never regenerated, this effect is
added to through each successive ROADM 148
Bandwidth Narrowing
To be accurate, the cars in our model can only drive in a straight line. Bandwidth Narrowing
in this case would be represented by the tunnels through our walls not being built exactly to
plumb; at some point, the car will crash!

Bitrate in the model would be


represented by car width – the faster
the signal, the sooner the crash. This is why
this becomes a concern at >10 Gb/s transmission 149
Summary

✓Four Wave Mixing, Cross Talk, and Self-Phase and Cross-Phase Modulation are all non-linear
effects due to high power in fiber.

✓Modulated signals have a central ‘carrier’ wavelength and two modulated ‘sidebands’.

✓In-Band Cross Talk is created by a lack of isolation in ‘drop’ filters while Out-of-Band Cross Talk is
created by a lack of isolation in adjacent-channel filters.

✓Self-Phase and Cross-Phase both create spectral broadening in signals.

150
Thank you

You might also like