Está en la página 1de 10

Erbium-doped fibre

amplifiers:
the latest revolution in
optical communications
by D.A. Chapman
Like many technological advances associated with optical fibre
communication, erbium-doped fibre amplifiers have moved
from the laboratory t o the field with extraordinary speed. They
are capable of low-noise, low-distortion,
high-gain and high-power operation over a broad bandwidth.
This paper describes the physical principles behind their
operation and shows how their unique characteristics are set t o
bring about a qualitative change in the way in which optical
fibre is used.
1

Background

derives from the long distances and high data rates


achievable from this deceptively simple scheme. Long
distances are achievable because optical fibre can now be
manufactured with very low attenuation (say 0.3 dB/km, so
that it takes 10 km for the power to be halved) and because
the attenuation is independent of the signalling rate. The
high data rates are achievable not only because the
attenuation is independent of signalling rate, but also
because systems can be constructed to give very little
dispersion. Consequently, virtually the only degradation to
a signal while propagating along the fibre is the uniform
attenuation, and there is an incentive to develop an optical
amplifier to compensate for the attenuation.

Despite the remarkable changes in telecommunications


already brought about by optical fibre, current uses only
scratch the surface of the possibilities and it has long been
recognised that the full capabilities of single-mode fibre
will only be realised if the signal can be kept optical
throughout an end-to-end link. The main electrical
bottlenecks which need to be eliminated are the electrical
switching and the electrical regeneration. Erbium-doped
fibre amplifiers remove the need for regenerators by
providing wideband, low-distortion, optical amplification
with gains of over 20 dB. In the long t e r n , they could also
contribute to the realisation of practical
optical switching, either merely by
compensating for loss in switching
devices, or by providing an active
component in the switch element itself.2
Conventionally, optical fibre is used
as shown in Fig. 1. A light source
(semiconductor laser diode) is
modulated, usually with essentially onoff digital pulsing, the light is coupled
curmnt
into the fibre and propagates to the
receiver. At the receiver is a
t
semiconductor photodiode which
provides a current output proportional
to the incident optical power. The
I
Fig. 1 Conventional arrangement for optical-fibre communications
success of optical fibre communication
+

ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL APRIL 1994

59

erbium doped
fibre (about 50 m)
input signal
(1550 nm)

wavelength

mEci;i:er

o:oznal

splice

zt

pump light source


(980or 1460 nm)

Fig. 2

Erbium-doped fibre amplifier

Erbium-doped fibre amplifiers have been developed to


the point that they are now available commercially and the
technology is thocght to be sufficiently mature that they
will be used in transatlantic optical-fibre communication
systerns due for construction in 1995.They have all sorts of
possible applications however (they can provide the gain
necessary to support soliton transmission3 and they can be
used in resonant structures to produce fibre-based laser
light sources? to give just two examples) and their full
impact on the way that optical fibre is used is yet to become
clear.
2

Erbium-dopedfibre amplifiers

Erbium-doped fibre amplifiers (EDFAs) operate in the


long-wavelength optical-fibre transmission window at
around 1550 nm. One of their attractions is the simplicity of
the design. At their most basic, all they need is a length of
doped fibre spliced into the transmission path, a pump
source and a means of coupling the light from the pump
into the doped fibre (Fig. 2). Energy is transferred, via the
erbium dopant, from the pump light to the signal, thereby
effecting amplification. The pump source needs to be
reasonably powerful - typically 1-100 mW, depending
upon the required output signal power - but this is now
available from semiconductor laser diodes. Coupling of the
pump light into the fibre can be done at very low loss (say
under 0.2 dB) with a fused-fibre wavelength division
multiplexer. The doped fibre is made from silica glass, and
differs from standard telecommunications fibre mainly by
the presence of erbium (and possibly other dopants) in the
core
region.
Being
based
upon
standard
telecommunications fibre has two advantages: the
manufacture uses established processes enabling the
production of high-quality fibre at low cost, and the

metastablestate

0 8 eV (1 550 nm)

ground state
~

Fig. 3

60

Energy levels of erbium ion electrons

similarity between the doped fibre and the


transmission medium with which it is being
used enables simple joining with low-loss
splices
There arevarious refinements on the design
shown in Fig. 2. In practice, it is usual to
include an isolator at the amplifier output This
is an optical diode, which only allows light to
propagate in one direction. Without the
isolator, reflections from imperfect splices,
couplers, connectors etc could be amplified in
at worst causing instability, at best leading to a
small increase in noise. Fig. 2 shows the pump light
injected so that it propagates in the same direction as the
signal along the doped fibre. This is known as codirectionalpurnping. Counterdirectional pumping,with the
pump injected at the far end of the doped fibre and
propagating in the opposite direction to the signal, is
equally possible. The two options have different
advantages and disadvantages. For example, counterdirectional pumping might allow higher output signal
powers whereas co-directional pumping tends to give
better noise performance. There are all sorts of variations
in possible configurations - such as combinations of
pumping direction and wavelength, and the way in which
isolators are used - aimed at optimising the performance
for particular applications. Furthermore the length of the
doped fibre (typically of the order of tens of metres) and the
dopant concentration and distribution needs to be
optimised.
These details will not be considered further here (see
References 5-8 for examples of some of the variations
considered.) The rest of this Section looks at the
amplification mechanisms and discusses how they
influence the performance characteristics of EDFAs.

Amplification mechanism
EDFAs amplify using stimulated emission. (They are
laser devices in the sense that they use light amplification
by stimulated emission of radiation, but the term laser is
usually reserved for a device in which the amplifying
medium is contained within a resonant cavity in order to
make a light source.)
The electrons associated with the erbium ions can exist
in a number of different energy levels, some of which are
shown (in a much simplified form) in Fig. 3. The first point
to note is the existence of the metastable level at about
0.8 eV (about 1.3 joules) above the ground state. This is
about the same as the energy of the photons of 1550 nm
wavelength light, so that if a 1550nm photon interacts with
an erbium ion in the metastable state, a quantummechanical resonance effect causes the ion to return to the
ground state and an extra photon to be emitted at exactly
the same wavelength as (and in phase with) the incident
wavelength (Fig. 4).
One incident photon has become two photons and we
have the basis of optical amplification.The reverse can also
happen unfortunately: an incident photon at 1550 nm can
be absorbed by exciting an ion from the ground state to the

ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL APRIL 1994

metastable state. However, net amplification is achievable


if, on average, more photons cause resonant emission
rather than absorption. This will happen if incident
photons are more likely to encounter ions in the
metastable state than in the ground state. The condition in
which there are more electrons in the excited, metastable
state than in the ground state is referred to as population
inversion and is achieved by the process known as
pumping. In an EDFA pumping is performed optically by
980 nm or 1480 nm light - photons from the pump excite
electrons from the ground state to the metastable state.
There are two routes for the pumping, depending upon
the pump wavelength (Fig. 5). If the pump wavelength is
around 1480 nm the energy of the pump photons is very
similar to the energy of the signal photons and the
absorption of a pump photon excites an electron from the
ground state directly to the metastable state. (Note that all
the energy levels have a finite width and so correspond to a
small range of energies. See the Panel on Energy level
broadening.) Alternatively, pumping can be at 980 nm, in
which case absorption of a pump photon excites the
electron to a higher energy state. The higher state has a
very short lifetime and the electron very rapidly drops
down to the metastable state, releasing its energy as
mechanical vibrations in the fibre. At the atomic level these
vibrations take the form of waves which, in some ways,
may be viewed as particles (phonons) carrying energy, just
a s photons are viewed a s waves carrying energy. Thus in
the nonradiative decays the transition emits one or more
phonons, whereas in radiative decays the transition emits
photons.

Fig. 4

Photon multiplication by stimulated emission

the probability of an incident photon exciting a stimulated


transition. Doing the calculations gives the result that, in
an ideal amplifier with complete population inversion, the
power spectral density of the output noise would be given
by

where h is Plancks constant and G is the power gain of the


amplifier at frequency v (so hv is the photon energy). The
factor 2 comes from the fact that single-mode fibre
supports two (degenerate) orthogonal polarisation states.
Real amplifierswill always be worse than this, and the noise
output is modelled by
P,,(v)

2 ns,,hv (G- 1)

where nsp is the amplifier excess noise factor. The noise


characteristics of EDFAs are usually described by the
noise figure, which, a s explained in the discussion of preamplifiers below, is equal to 2n,(1010g,,(2n,~) in decibels),

Saturation

Noise

Some photons will return to the ground state


As with any amplifier, as the magnitude of the output
from an EDFA increases, the amplifier eventually starts to
spontaneously, without being stimulated by a signal
saturate. In an EDFA saturation is caused by the reduced
photon. In doing so they emit a photon unrelated to the
signal. This is the only significant source of noise in an
extent of the population inversion (fewer erbium ions in the
EDFA. Photons generated by spontaneous emission can
excited state) leading to a reduction in gain, with
propagate in either direction along the doped fibre.
characteristics of the form shown in Fig. 6. However, there
Whichever way they go they may themselves stimulate
is an important difference between the saturation in
EDFAs and the more familiar saturation in electrical
emission of more photons. The stimulated emission by
amplifiers, or, indeed, some other optical amplifiers.
spontaneously emitted photons causes two problems.
The difference is that the saturation is a long-term effect
Firstly, it amplifies the noise, hence the description
(long, in this context, meaning around 10 ms). Because of
amplified spontaneous emission (ASE) noise. Secondly it
the long lifetime of the metastable state (the time constant
removes some of the electrons from the metastable level.
for spontaneous decay
thereby reducing the
from the metastable state
degree of population
is about 14 ms), and the
inversion. This second
corresponding
low
effect means that even
nonradiative
transition
probability of individual
though
backwardphotons
stimulating
propagating ASE will not
emission from individual
contribute directly to the
ions, high gain is achieved
amplifier output noise, it
1480 nm
by having a large
does degrade the EDFA
reservoir of excited
performance.
erbium ions with which
Theoretical
analysis
b
a
incident photons may
shows that the probability
interact. If the average
of an excited ion
I
signal power is high, the
spontaneously emitting a
Fig. 5 Optical pumping in erbium-doped fibre:
reservoir empties and the
photon is closely related to
(a) a t 1480 n m ( b ) at 980 n m

ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL APRIL 1994

61

1480 nm

1550 nm

Al) .This is known

Is, electrons will tend to b e


Fig. A2
more likely to occupy the lower end of the range. This
explains the success of pumping at 1480 nm (Fig. U).
gain, can b e achieved conveniently spans much of the
the higher energy 1480 nm photons excite electrons
low-loss silica fibre window. Notice also that the broad
energy levels involved also mean that the precise pump
wavelength is not critical. m e higher level, used by
980 pumping, is narrower, so that the exact wavelength
from where they may b e
of 980 nm pumps is important to within a few
top end of the ground state, by
nanometres.)
(lower energy) 1550 nm photoile. The wavelength
If the doped fibre is cooled the thermal broadening
range over which stimulated emission, and therefore
will be reduced, reducing the overlap within the
manifold and giving a level structure more like the
middle of Fig. Al. This is known a s inhomogeneous
broadening. The exact spacing of the levels for any one
erbium ion depends upon the electric fields
experienced by the individual ion. Because of the
amorphous nature of the glass in which the ions are
embedded, the electric field, and therefore the level
structure, are different for each ion. Thus the
inhomogeneously broadened fibre will have levels all
Stark splimng
thermal broadening
the way across the same range as the homogeneously
broadened fibre, but different levels will b e due to
different erbium ions and so will, to some extent, b e
Fig. A I
independent.

- -

gain drops. Short pulses of high power in an otherwise lowpower signal, however, have access to the large reservoir
of excited ions and experience the same gain a s lower
power pulses in the same signal.
There are a number of consequences of this slow
saturation. Firstly, the amplifier is unaffected by
fluctuations in the pump power above about 50 kHz.
Secondly, the amplifier can be operated in the saturation
region without degrading the extinction ratio of the digital
signal, for signalling rates above about 100 kbaud.
Similarly, amplifiers will not severely distort analogue
signals, even when saturating, provided the signal
spectrum is predominantly above 100kHz.
The reduced population inversion of a saturating
amplifier does, however, affect the characteristics in a
number of other ways. A useful feature is that it increases
the efficiency with which pump energy is converted to
signal energy because there is a high probability of all the
pump photons interacting with unexcited erbium ions and
therefore being used, rather than passing through the
doped fibre unabsorbed. Unfortunately however, the
62

reduced population inversion also reduces the noise


figure. This is because a proportion of the signal photons
will pump ground state ions to the metastable state (and be
absorbed in the process). In effect, the degree of
population inversion contributing to spontaneous decay is
then greater than the degree of population inversion
usefully contributing to the amplifier gain.
3

Applications

The most obvious application of an optical amplifier is to


extend the reach of a point-to-point transmission link.
There are three generic configurations (Fig. 7): use a s a
post-amplifier to boost the signal prior to transmission;
a s a preamplifier, to amplify the signal immediately
before detection at the receiver: or as an in-line amplifier
as a nonregenerative repeater (combinations of all three
are also possible). The first two have the advantage of
locating the amplifier only at a terminal, where power
and maintenance access are readily available, but the
third allows the possibility of the longest links

ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL APRIL 1994

(extendible by using more amplifiers). Some insight


into the features and benefits of optical amplifiers is
gained by considering these three configurations in
more detail, so the following subsections look at each in
turn.
The postamplifier
The EDFA is being used here as a power amplifier
following a lower-powered signal source - paralleling the
usual practice for high-powered electrical sources. The
amplifier is operated in saturation (see above), where it is
most efficient in terms of converting pump power to signal
power but has an increased noise figure. Because the noise
is attenuated during transmission with the signal however,
the increased noise figure is unlikely to be significant.
High-power pump sources are needed, and safety
considerations will be important because of the highpowered output signal.
The preamplifier
The performance limit of conventional optical-fibre
systems using pin photodiodes derives from the noise
generated in the receiver, thermal noise in the
photodiode load resistor and amplifier noise in the
following electrical preamplifier.Any means of increasing
the signal power prior to the load resistor and electrical
preamplifier will improve the signal-to-noise ratio for a
given receiver. A conventional approach is to use an
avalanche photodiode (APD) to provide gain within the
photodiode by the avalancheof electrons released by
each incoming photon. As a statistical process, however,
the avalanche gain itself introduces noise. Coherent
detection is an alternative approach, whereby a
(relatively) high-power local optical source is mixed with
the received signal at the photodiode, to provide true
optical amplification. Coherent detection can be very
successful, but requires a complex receiver with phase or
frequency locking of the local oscillator, and polarisation
control (see Reference 9). In addition, both the optical
signal source and the local oscillator lasers need to have
very narrow linewidths.
Using an EDFA as an optical preamplifier is an
alternative which provides better performance than the
use of an avalanche photodiode and is much simpler than
coherent detection.There is no particular constraint on the
source linewidth and there is no need for polarisation
control. It is necessary, however, to consider the effect of
the noise (see the Panel on T h e effect of ASE on an optical
receiver).
A concept which is commonly used to quantify the effect
of amplifier ASE is the noise figure, which is derived by
comparing the signal-to-noise ratio at the output of an ideal
receiver with and without an EDFA used as a preamplifier.
By an ideal receiver is meant one which introduces no
noise itself (no thermal or electrical amplifier noise), so
that without the EDFA the only noise present is the shot
noise (quantum noise) which arises due to the quantum
statistical nature of detection in a photodiode. The signal
power (current squared) out from the photodiode is given
by f = P:p, where P, is the received optical signal power

input power, log scale

Fig. 6

EDFA gain characteristic

and Ris the photodiode responsivity (the ratio of the output


current to incident optical power). R is given by qe/hv,
with q , the quantum efficiency, equal to unity-becausewe
are assuming an ideal receiver and e the charge on an
electron.
The shot noise power (current-squared) is given by
<i,> = 2eIB, where B is the bandwidth of the receiver noise

-k

Tx

Rx

,
pfeamplHier
b

i G R X
E

in-tine amwier
C

9. 7 The three generic configurations for using an


optical amplifier: (a) post-amplifier; (b)preamplifier;
(c) in-line amplifier

ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL APRIL 1994

63

The effect of ASE on an optical receiver


Spectrum of the received opticalsignal

Spectra of the three output noise components


power spectraldensity pmportionalto
mean receivedopticalpower (nB,)

power spectraldensity pmportiinal

The AS spaclrumhas banhrldlh &


Fdlomrq lhn phomdiode,low-pass electrical

llnerinpis irlcdelled86 having a band*ndth &,


In general E- is much greater than W,

The analysis of a receiver in the


presence of amplified spontaneous
emission @SE) noise is qualitatively
different from the analysis of the
receiver in a conventional system.
Whereas all the noise in conventional
systems is electrical noise added
after the photodiode, ASE originates
before the photodiode and results in
three noise components on the
photodiode output. The first
component is the extra shot noise
from the increased mean optical
power. The other two arise due to the
fact that the photodiode output is
proportional to the incident optical
intensity, which is in turn
proportional to the square of the
This
electromagnetic
field.
nonlinearity (squarelaw) of the
mixes different
photodiode
frequencies to generate two sets of
beat frequencies: mixing between
the signal and ASE generates signalspontaneous beat noise; and mixing

-I

2 BWI

The spectrum 01 the SpOnlaneOUS-SpOntatmws


beat noise can be thought 01 as generated by
the trequenwdomam mnvdution 01 lhe
AS -rum
with itsslt, togsther with a DC
term generated by the mean w a r of me
AS. The DC term pmvldns a fixed o W t
which may be negktnd (but generates the
extra shot noise, Shown above)

signal-spontaneous
beat noise

,,

power spectraldensity
pyortional to
slgnal x ASE (ns)

&

power proportional
to nSB-

frequency

6.-

BOD!

The spectrum of the slgnal-spontaneousbeat


n o w can be thought of as generaled by
frequency domain convolution 01 the sgnal
spaarum with the AS spectrum

between
different
frequency
components within the ASE
specbum generates spontaneous
spontaneous beat noise. The
magnitude of the spontaneous
spontaneous beat
noise
is
proportional to the opticalbandwidth

filter, so the signal-to-noiseratio is

of the ASE noise at the receiver input


and can be reduced by putting an
optical filter before the photodiode.
Optical filtering also reduces the
magnitude of the extra shot noise,
but not of the signal-spontaneous
beat noise.

power, GP,, converted to a current by a photodiode with


responsivityR and having a bandwidth of B.)
The signal-twnoiseratio becomes

Putting an EDFA in front of the receiver increases the


optical signal power by the gain G, but also adds noise due
to the ASE. Assuming the use of an optical filter before the
photodiode, it can be shown that the extra noise
component is given by

mD =

GZP:P
+ 4Pn&v (G - 1) GP,B
GPP
2eB(1+ ZnSp(G- 1))

<z?,,~~~,.>
= 4 P n & v ( G - 1)GPJ

(This is the ASE noise, ns#v (G - 1) ,beating with the signal


64

Paralleling the usual definition of noise figure as the ratio


between the signal-to-noiseratio at the input and the signal-

ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL APRIL 1994

to-noise ratio at the output of an amplifier, the noise figure


of an EDFA is defined as the ratio between the signal-tonoise ratio with and without the EDFA, thus

For large gain, this approximates to Zn,. A perfect


amplifier would have n, = 1, giving a noise figure of 2
(3 dB). That is, with an ideal receiver, using a perfect
optical amplifier would halve (degrade) the signal-to-noise
ratio. Real receivers are very much worse than the ideal
receiver because of thermal and other noise, so a real
EDFA (which can approach an ideal EDFA to within a
factor of 2) can give a substantial increase (improvement)
in signal-to-noise ratio.
In-line amplifiers
Using in-line amplifiers is effectively going back to a
technique used in the days of analogue transmission, when
linear amplifiers were used to maintain the signal power
along trunk transmission systems.
In a digital era, when the received wisdom is that
regeneration (detecting and recreating the digital pulse
stream afresh to prevent the accumulation of noise and
distortion) is superior to linear amplification,it might seem
that (linear) optical amplification offers limited benefits.
However, an optical amplifier is much simpler than a
regenerator, and potentially more reliable. Furthermore,
as explained above, the propagation of an optical signal in
fibre can be virtually noise and distortion-free: noise only
appears in the electronics at the receiver. Thus the weakest
link of existing systems is at the receiver, and there is
scope for accepting some in-line signal degradation
(generated by amplifiers) without degrading the overall
performance. In practice, if a series of optical amplifiers are
concatenated the noise (&E) builds-up (the noise power
from successive amplifiers adds) and the receiver noise is
soon dominated by the ASE. Nevertheless, the advantage
of simplicity remains.
There is furthermore the advantage of transparency.
Regenerators have to be designed for a particular system.
They are compatible with a given signalling rate and line
code. Amplifiers, on the other hand, are almost completely
transparent to the conveyed waveform. Thus an installed
system using amplifiers may be changed - upgraded by changing only the terminals, but a regenerated system
requires the replacement of all the regenerators. For
example, the next proposed transatlantic transmission
system, TAT-12, will be installed in 1995 to operate at
5 Gbit/s using 133 amplifiers for the 6000 km link between
Rhode Island and Land's End. It might be possible at some
future date to upgrade to, say, 10 Gbit/s, merely by
changes at the terminals.
The ability to operate EDFAs in saturation without
distorting the conveyed signal can be exploited when
concatenating amplifiers, as it provides a means of passive
automatic gain control. As can be seen from Fig. 8, if the
input power to a saturating amplifier changes, then its gain

changes in the right direction to compensate. Saturating


operation of amplifiers in a transmission system may be
ensured by making the total loss between repeaters less
than the small signal gain of the amplifiers.
Distributed amplification
The gain in an EDFA is inherently distributed over the
length of the doped fibre - typically tens of metres. It is
possible deliberately to distribute the gain over even
longer distances by using a lower erbium doping
concentration in the fibre. In this way it is possible to
arrange for the gain just to compensate for the fibre
attenuation, creating a zero attenuation transmission
medium. If the pump wavelength of 1480 nm is used, then
the silica fibre attenuation of the pump can be kept low
(1480 nm being at the edge, but within, the 1550 nm
transmission window), so a single pump can pump several
kilometres of fibre used in this mode.
The use of distributed amplification to create a zero loss
fibre is particularly attractive for use with soliton
transmission, where it maintains the pulse amplitude
required to generate the nonlinearities necessary for the
stability of solitons.
Wavelength division multiplexing
The transparency of EDFAs across much of the 1550nm
fibre transmission window makes them suitable for use
with systems employing wavelength division multiplexing
(WDM). A number of channels can be simultaneously
amplified in a single EDFA, and, because of the slow gainresponse time (see the discussion of saturation above),
there is negligible crosstalk between channels above about
100kHz. However, there is a problem with using strings of
concatenated amplifiers operated in saturation (the
configuration proposed in the discussion of the in-line
amplifier above). If an amplifier is homogeneously

input power
drops, gain 2
rises to
compensate

input power

L rises. gain

input power

drops to
Compensate

operating
point

Fig. 8

Operation of an EDFA in saturation

ELECTRONICS & COMMUNICATION ENGINEERINGJOURNAL APRIL 1994

65

'dispersion shifted'). This has a


dispersion minimum around 1300 nm
and is highly dispersive at 1550 nm.
EDFAs amplify only in the 1550 nm
window. Research is in progress to try
to find similar amplifier technologies
suitable for 1300 nm (see below), but
nothing as successful as the EDFA
looks likely to emerge in the short term.
A number of techniques for
Fig. 9 A twin-cored fibre can provide automatic gain control for each
overcoming the disDersion limits at
channel of a wavelength division multiplexed signal
1550 nm are being investigated.
broadened (see Panel on 'Energy level broadening'), the
Methods have included equalisation using lengths of fibre
operating point will be determined by the average power
designed to have strong dispersion characteristics in the
across all channels, which will tend to overamplify the
opposite sense to standard fibre15and the exotic technique
more powerful channels and underamplify the weaker
of phase conjugation.16 This lattel: method involves
channels. With concatenated amplifiers the effect buildsinverting the spectrum of the signal half way along a
up along the route and causes initial slight differences in
transmission path, so that the second half of the
transmission 'undoes' all the dispersion done in the first
channel powers to be exaggerated. Even if the channels
half.
are all launched with identical powers, the variation in
Another set of problems brought to light by the use of
amplifier gain across the kequency spectrum, though
amplifiers arises from fibre nonlinearities. At the power
generally small, is sufficient to create, then emphasise,
power differences.
levels and over the repeater spacingsused in the past, it has
A number of solutions to this particular problem are
been sufficient to treat silica fibre a s though its optical
properties (specifically dielectric constant) were purely
currently being investigated. Flattening the gain of
linear. With high powers and longer distances, the
amplifiers,'" or building filters to equalise the amplifier
nonlinear effects can become noticeable. In practice,
gain spectrumlLhelps, assuming that the channels are
however, provided the power levels are kept below about
launched with identical powers. If a return path is available,
+18 dBm, nonlinearities can be kept small such that they
feedback to adjust the channel launch powers on an
individual basis can be used to equalise the received
are not the limiting factor in system perf~rmance.'~
channel powers. Cooling the amplifiers, so that the
4 Fibre a m p l i f i e r s for o p e r a t i o n a t 1300 n m
broadening becomes inhomogeneous rather than
homogeneous (see the Panel on 'Energy level
broadening') is an interesting, but complex a p p r ~ a c h . ' ~ A fibre laser operable at the 1300 nm window is a very
The inhomogeneously broadened amplifier saturates to
desirable goal, because ofthe dispersion zero at 1300 nm in
some extent independently at each of the different
standard singlemode fibre. Praseodymium has the right
energy levels for pumping at 1000 nm to give gain at 1300
wavelengths because the different wavelengths are
amplified by different levels in the inhomogeneous
nm, and amplifiers based on praseodymium doping have
spectrum. This approach is particularly attractive because
been demonstrated.The difficulty is that the lifetime of the
metastable state of praseodymium in silica is very short
it offers the possibility of independent automatic gain
(1 ps), due to nonradiative decays. This means that very
control for each channel.
The same benefit is available from another recently
high pump powers are required to achieve population
inversion. The nonradiative decay occurs because energy
proposed scheme,14 without the need for cooling. This
can be removed by phonons in the silica glass. To some
uses double-cored, doped fibre as the amplifying medium.
extent this can be improved by using a different type of
Light propagating in double-cored fibre moves back and
glass which supports different phonon energies.
forth between the two cores with a cycle length that
Specifically, fluoride glasses have been developed in the
depends upon the light wavelength. Thus different
past for optical fibre transmission at longer wavelengths
channels are present in the cores at different locations
(around 2400 nm), and the energies of phonons in fluoride
(Fig. 9), and some of the gain for each channel derives from
glasses are significantly lower than in silica glasses.
different physical locations in the cores. Saturation at the
Consequently more phonons are needed to carry the
different locations will adjust to the appropriate channel,
energy from a given transition, and so nonradiative decay
again providing the required automatic gain control
is less probable than in silica. Amplifiers for the 1300 nm
separately for each channel.
window based upon praseodymium-doped fluoride glass
fibres are therefore of current interest and show promise.'8
System problems
But since even in fluoride glass the metastable lifetime is
By overcoming the attenuation limits on transmission
only 100 ps (compared to 14 ms for erbium in silica), high
distance, fibre amplifiers bring to the fore other limitations.
pump powers are still required and it is unlikely that
A major problem is that of dispersion. Most of the fibre
praseodymium-doped fluoride fibre amplifiers (PDFFAs)
presently installed in the field (there is a lot of it) is
will ever be as successful a s EDFAs.
standard single-mode fibre ('standard' rather than
I

66

ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL APRIL 1994

Conclusions

Erbium-doped fibre amplifiers provide an almost perfect


technical solution to the need for Optical gain in the 1550
nm optical-fibre transmission window. They have been
developed to the state where the noise performance
approaches theoretical limits to within a few decibels, and
the power handling capacity is more than adequate for any
currently proposed telecommunications application. They
will change the way in which fibre is used in
telecommunication networks by effectively removing the
limitations due to attenuation. As amplifiers come to be
used instead of regenerators, end-to-endlinks will become
transparent to the conveyed signal, opening up the
possibility of widespread use of wavelength division
multiplexing capable of transporting a wide variety of
analogue and/or digital signals.
Further reading

For fiirther reading, a detailed description of the theory of


optical-fibreamplifierswillbe found in a book edited by P.W.
France (Reference 19), but inevitablyin a subject moving as
fast as this is, books tend to date very quickly and the latest
information is only availablekom the technical journals and
conferences. Papers describing the latest work appear in a
number of different physics and engineeringjournals, but in
particular, Electronics Letters (IEE), Photonics Technology
Letters (IEEE) and Journal of Lightwave Technology (IEEE)
have relevant material in every issue. All the major
conferences on optical communication include papers on
EDFAs, but in recent years the Optical Society of America
has held an annual topical meeting on Opticalamplifiers and
their applications;the 1994meeting will be held in the USA
on 3rd-5th August.
References

COCHRANE, P., HEATLEY, D.J.T.. SMYTH, P.P., and


PEARSON, LD.: Optical communications: future prospects,
Electron. & Commun. Eng.J.,August 1993,5, (4), pp.221-232
PANTELL,RH.. DIGONNET, MJ.F., SADOWSKI,RW., and
SHAW, H.J.: Analysis of nonlinear optical switching in an
erbium-doped fibre,Jlightwuue Technol., September 1993,
11, (9),pp.141&1424
TAYLOR J.R: Optical solitons - theory and experiment.
Cambridge Studies in Modem Optics, 10, (Cambridge
University Press, 1992)
TAKARA, H., KAWANISHI, S., and SARUWATARI, M.: 20
GHz, 3.5 ps transform-limited optical pulse generation from a
highly stable, tuneable actively modelocked Er-doped fibre
laser. Cod. on Optical ampliers and their applications,
Yokohama, Japan 4th4th July 1993, Technical Digest
pp.310-313 (Optical Society ofAmerica,Washington DC)
DIGIOVANNI, D.J., WYSOCKI, P.F.. and DAVEY, S.T.:
Tailor fiber design to optimise amplier performance, Laser
Focus World,September 1993,29, (9), pp.95106
MELLIS, J., GARNER, P., MARTIN, J.N., MIDDLEDITCH,
D.J., COLLINS, J.V., WHEATLEY, P., and PARKER BE.:
Modular erbiumdoped fibre ampliers for optical
communications systems, BT Technol. J., October 1991, 9,
(4), pp.12-18

7 W, A, OMAHONY, M.J., and SIDDIQUI, AS.:Analysis of


optical gain enhanced erbiumdoped fiber ampliers using
optical filters, Photon. Technol. Lett., July 1993, 5, (7),
pp.773-775
8 ZERVAS, M.N., LAMMING., RI., and PAYNE. D.N.: Tradeoff and design considerations of the erbiumdoped fibre
amplifier. IEE Colloquium on Optical amplifiers for
communications, London, 20th May 1992, Digest No,
1992/124, Paper 7
9 STANLEY,I.W.: Atutorial review of techniques for coherent
optical fiber transmission systems, IEEE Commun. Mag.,
August 1985,23,(8), pp.37-53
10 DA SILVA, V.L., SILBERBERG, Y., WANG.. J.S..
GOLDSTEIN, E.L., and ANDREJCO. M.: Automatic gain
flattening in Er-doped fiber amplifiers. OFC/IOOC93, San
Jose, California, 21st-26th February 1993, Technical Digest
pp.174-175
11 KASHIWADA, T., NAKAZATO, IC, OHNISHI, M.,
KANAMORI, H., and NISHIMURA. M.: Spectral gain
behaviour of Er-doped fiber with extremely high aluminium
concentration. Conf. on Optical ampliers and their
applications, Yokohama, Japan, 4th4th July 1993,Technical
Digest pp.104-107 (Optical Society of America, Washington
DC)
12 WILLNEQ AE., and HWANG, S.-M.: Passive equalization of
non-uniform EDFA gain by optical filtering for megameter
transmission of 20 WDM channels through a cascade of
EDFAs. Conf. on Optical amplifiers and their applications,
Yokohama, Japan, 4th-6th July 1993, Technical Digest
pp.178-181 (Optical Society of America, Washington DC)
13 GOLDSTEIN, E.L., ESKILDSEN, L., DA SILVA, V.,
ANDREJCO, M., and SILBERBERG, Y.: Suppression of
dynamic cross-saturation in multiwavelength lightwave
networks with inhomogeoneously broadened fiber
amplifiers.Conf. on Optical amplifiers and their applications,
Yokohama, Japan, 4th-6th July 1993, Technical Digest
pp.70-73 (Optical Society of America, Washington DC)
14 LAMING, RI., MINELLY, J.D., DONG, L.. and ZERVAS.
M.N.: Erbium-doped-fiber amplifier with passive spectralgain equalisation. OFC/IOOC93, San Jose, California.
21st-26th February 1993,Techical Digest pp.175-177
15 PAYNE, D.N., LAMING RI.. RICHARDSON, D.J., and
GRUDININ,A: Unleashingthe full capacity of the installed
fibre base. 19th European Conference on Optical
communication (ECOC93). Montreaux, 12th-16th
September 1993,Proceedings Volume 1, pp.92-94 (available
from Swiss Electrotechnical Association (SEV), CH-8034
Zurich)
16 GNAUCK, AH.. JOPSON, RM.. and DEROSIER, RM.: 10
Gb/s 360 km transmission over dispersive fiber using
midsystem spectral inversion, Photon. Technol. Lett., June
1993,5, (6), pp.663-666
17 GOSSET, N., and DUGAN, J.M.: Stretching span distance.
J. Lightwave Technol.,November 1993,.10, (12). pp.5658
18 WHITLEY, T.J., WYATT, R.. SZEBESTA, D.. and DAVEY,
S.T.: Towards a practical 1.3 pm optical fibre amplifier, ET
Technol.J.,April1993.11, (2) pp.115-127
19 FRANCE, P.W.: Optical fibre lasers and amplifiers (Blackie
and Son, Glasgow and London, and CRC Press, USA and
Canada, 1991)
0 IEE 1994
Received 14th December 1993

The author is with the Faculty of Technology, The Open


University, Walton Hall, Milton Keynes, MK7 6AA, UK.

ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL APRIL 1994

67

Schematic Design and Capture


Create your schematics quickly and
etlluently using EASY-PC Professional
Areas of the circuit can be hiahliahted on

our simulation and design programs


,-

Digital and Analogue Simulation


Modify the
configuration
and change
component
values until the
required
performance is
achieved.

Affordable Electronics CAD


/yLI-pC:Entry level P

C
T
EASY- PC Professional: Schematic Capture and
f19500
PCB CAD. Linksto ANALYSER 111 and PULSAR.

PULSAR Professional: Digital Circuit Simulator


- SOLXI ate ca acity

PULSAR: Entry level Digital Circuit Simulator


1500 gate capacity

ANALYSERQIII:Ent: level Linear Analogue Cir/C,UI


Simulator- 130 nodes
-

ANALYSER 111 Professional: Linear Analogue


Circuit Simulator 750 nodes

f98.00
f,95.00
f98,00

f,95,00
t Z-MATCH for Windows: Smith Chad based problem E245,00
solving program for R F Engineers

PCB Design

Enter 002

LETTER TO THE EDITOR

Dear Sir - M.B. Sandlers article in the December 1993


issue of Electron. & Commun. Eng. J. was both excellent
and fascinating. He indicates that the basic concept of
pulse width modulation (PWM), which he describes, is not
particularly new: however, readers may be interested to
hear of its full antiquity. In the late nineteen sixties, at least
two semiconductor manufacturers offered ICs using the
PWM technique. Mullard produced devices for a wide
range of applications. The most advanced device,
TDA2600, appeared in the early seventies. This IC
produced the scanning current required for vertical
deflection of the 30AX range of colour picture tubes. The
circuit was conceived by Brian Attwood at Mullard Central
Applications Laboratory at Mitcham. He, together with
Brian Simpson, brought the circuit to fruition.
Philips used theTDA2600 in their G11 chassis. Kenneth
Smith carried out the field circuit development at Philips
Croydon Works. Naturally, equipment with high-gain VHF
ampliliers had problems of interference from the pulse
switching edges, but careful attention to earth paths solved
these. Also, novel circuitry gave novel mechanisms of noninterlace, but these were discovered and remedied. One
and a half million receivers were made. The specification
called for a life of 10 years and 20000 hours with a failure
fi8

rate below 0.2%/1000 hours. These figures were easily


achieved - a few receivers are known to be in present
regular use - not least that in the shop of the writers
father!
The technique fell into disuse for field scanning in TV
receivers owing to advances in the technology of heat
extraction from IC power transistors. Thus the advantage
of high-efficiency bottomed output devices lost its appeal.
ICs with conventional Class B output stages were simpler
and cheaper.
The authors memory is grateful for help received from
J.C. Warren and J. Douglas.
R. TALKS (F)
Croydon
15th February 1994
Brief letters, preferably typed, on technical and
professional matters are welcomed. They should be sent
to the Staff Editor, Electronics & Communication
Engineering Journal, IEE, Michael Faraday House, Six
Hills Way, Stevenage, Herts SG1 2AY. The Editor
reserves the right to edit letters which are published. The
Editors decision on which letters are published is final.

ELECTRONICS & COMMUNICATION ENGINEERING JOURNAL APRIL 1994

También podría gustarte