Note: Descriptions are shown in the official language in which they were submitted.
CA 02290311 1999-11-24
-2-
MODULAR FILTER FOR EXTRACTING OPTICAL SIGNALS FROM,
AND/OR INSERTING THEM INTO, MULTIPLE-WAVELENGTH
OPTICAL TELECOMMUNICATIONS SYSTEMS
*****
DESCRIPTION
The present invention relates to an optical telecommunications
system, for the transmission of a signal of the wavelength division
multiplexing type, also called WDM, in which this signal can be
extracted from the line in at least one of a plurality of intermediate
stations located along the telecommunications path, in the form of
some of its wavelength components, which can if necessary be
replaced by components having the same wavelengths or can be
processed (amplified, equalized, etc.) and subsequently reinserted into
the line by multiplexing in such a way that the new WDM signal is
regenerated. These operations on the signal can also take place at
terminal stations.
For wavelength division multiplexing, or WDM, transmission, it is
necessary to send a plurality of transmission signals, which are
independent of each other, along the same line, consisting of optical
fibres, by multiplexing in the optical wavelengths domain; the
transmitted signals may be either digital or analog, and are
distinguished from each other in that each of them has a specific
wavelength which is different from those of the other signals.
The implementation of this WDM transmission requires the allocation
of specific wavebands of predetermined width, called "channels" in the
following text, to each of the signals having different wavelengths.
These channels, each identified in the following text by a wavelength
value, called the central wavelength of the channel, carry a signal
having a certain spectral amplitude about the central wavelength value,
which depends, in particular, on the characteristics of the laser source
CA 02290311 1999-11-24
-3-
of the signal and on the modulation imparted to it to associate a data
item with the signal. Typical values of spectral amplitude of the signal
emitted by a laser, in the absence of modulation, are around 10 MHz; in
the presence of an external modulation, at 2.5 Gbit/s for example, the
spectral amplitude is approximately 5 GHz.
For transmitting signals in a large number of channels, making use of
what is called the third transmission window of the optical fibres and of
the pass band of the optical amplifiers, the wavelength separation
between the signals is conveniently of the order of nanometres.
For the purposes of the present description, the WDM or multiple-
wavelength signal is defined as the signal travelling along a single
optical fibre and containing at least two components having different
wavelengths.
In a multiple-wavelength telecommunications network, the said WDM
signals can advantageously be extracted, particularly at the
intermediate stations of the transmission line, for example at the signal
amplification stations, and signals having a different information content
but the same wavelength can be inserted in their place. In this way,
these stations become nodes at which the information passing along
the line can be extracted, modified and/or replaced by terminal
equipment or by a different network which communicates with such
nodes.
In general, in a communications line, a device of this type for
extracting optical channels from the line and inserting them into it,
connected within an intermediate station, causes the formation of a
node in the line in which the data can be redistributed and modified.
A first technique for carrying out the operation of extracting these
signals from the telecommunications line uses WDM signal
demultiplexing devices which separate the various channels so that
each is sent along its own fibre. The extraction of the optical channels
CA 02290311 1999-11-24
-4-
and the insertion by means of optical switches or combining devices
suitable for the purpose takes place in these lengths of fibre.
Patent EP 651528 (Siemens) describes an insertion and extraction
filter comprising a first device for demultiplexing a WDM signal, which
separates the various channels from each other and sends each of
them along one length of optical fibre. In this length, each of the
signals, each of which is contained in one channel, can be extracted
individually by means of optical switching devices and, simultaneously,
a signal at the same wavelength can be reinjected into the same fibre.
Additionally, any signal may simply pass through the length of optical
fibre without undergoing modifications. All the channels reach the
output of the filter, after passing through a multiplexing device which
reconstructs the WDM signal.
The applicant has observed that this type of technology presents
problems caused principally by the difficulty of providing spectral
compatibility between the demultiplexing device and the multiplexing
device following the extraction and insertion of the signals. This is
because the demultiplexed signal generates a series of signals, each at
a specific wavelength, which, after they have been extracted and
replaced by signals at corresponding wavelengths where necessary,
have to be multiplexed to reconstruct the WDM signal. The operation
described is efficient only if the multiplexing device and the
demultiplexing device have perfectly matched wavelength responses.
The matching may be degraded by many factors, for example different
design characteristics, or a differential relative thermal variation
between the two components. In practice, the separation of the two
components which carry out the two functions of multiplexing and
demultiplexing is critical in this configuration.
A second technology integrates both of the said functions in a single
component called a wavelength multiplexer/demultiplexer of the
CA 02290311 1999-11-24
-5-
"Arrayed Waveguide Grating" type, referred to as AWG in the following
text, which enables a multiple-wavelength signal inserted in one of a
plurality of input fibres to be demultiplexed and distributed among a
number of output fibres corresponding to the number of wavelengths of
which the said signal consists. If provided with a plurality of
wavelengths, the device is also capable of multiplexing different
channels and different wavelengths in a single output fibre. The mode
of operation of these devices is described, for example, in the article by
H. Takahashi et al., "Transmission Characteristics of Arrayed
Waveguide NxN Wavelength Multiplexer", J. of Light. Techn., vol. 13,
No. 3, March 1995, pp. 447-455.
The said component has characteristic parameters which define the
nature and number of the wavelengths which the component can
separate; these parameters are the number of inputs and outputs, the
wavelength spacing which must be present between two adjacent
channels so that these can be separated, and the wavelength of the
central channel of the band of channels which the component can
process.
A further parameter of the AWG is the FSR (Free Spectral Range)
which represents the periodicity of an AWG. This is because the device
is capable of extracting, at each individual output, a signal having a
specific wavelength ~, and, also, signals having a wavelength ~,+nFSR,
where n is a positive or negative integer. This takes place at each
output, and consequently the filter has a periodicity defined by the FSR.
A component of this type is available on the market (produced, for
example, by NEL, PIRI, Hitachi, Lucent, etc.). An AWG is described, for
example, in U.S. Patent 5600742. The AWG generally comprises a
number N of input fibres and a number N of output fibres for the
signals. The component is bidirectional and therefore each of the input
fibres may be used as an output fibre, and vice versa.
CA 02290311 1999-11-24
-6-
The principal parameters which determine the performance of this
component are the insertion loss (IL) and two noise contributions which
are defined as "out-of-band crosstalk" and "in-band crosstalk". The
former represents the contribution to an output of all the other
wavelengths which nominally are not coupled to this output. The in-
band crosstalk, on the other hand, is an interference system, having the
same wavelength as the said signal. This may be due, for example, to
the presence of signals using multiple paths within the device.
U.S. Patent 5,414,548 describes an insertion and extraction device
which makes use of the characteristics of an AWG which can be used
to demultiplex a multiple-wavelength signal inserted into one of a
plurality of input fibres and to distribute it to a number of output fibres
corresponding to the number of wavelengths of which the said multiple-
wavelength signal consists.
The said device comprises an AWG having a plurality of inputs and a
plurality of outputs. The WDM signal is inserted into one of the inputs of
the AWG in such a way that it generates a signal having a single
wavelength at each of the corresponding outputs. Each of the said
outputs is connected to a length of optical fibre through which the signal
passes, and in which is located a switch which extracts and reinjects a
signal having nominally the same wavelength in the said length of fibre.
Each of the said lengths of fibre reinjects the signal into an input of the
AWG, thus creating a configuration described as "loopback" because
the output signal of the AWG is reinserted into the input. All the signals
reinjected into the input reconstruct the WDM signal at one of the
outputs of the AWG.
The applicant has observed that a device for the extraction and
insertion of optical channels with a configuration of the loopback type
using an AWG with N inputs is capable of processing N-1 channels
(consequently N-1 loopback channels are present). AWGs do not have
CA 02290311 1999-11-24
the same transfer characteristics for identical signals inserted at
different inputs. This is because the best operating efficiency, in other
words the minimum IL, in an AWG is obtained for signals which have
the central wavelength of the complete range which the component can
process, and are therefore closest to N/2 at the inputs, which are
numbered progressively from 1 to N.
The applicant has observed that the losses due to the attenuation
introduced by the component for signals inserted into these inputs are
smaller than those affecting the signals inserted into the outer inputs.
Consequently, in a configuration of the loopback type, if the signals are
separated and reinjected into corresponding identical outputs and
inputs, the non-uniformity of the paths is doubled. The applicant has
found that, when a configuration of the loopback type is used, it is
advantageous to select the input-output combinations in a suitable way,
in order to maximize the uniformity; this is possible only if FSR = N~0~,,
where N is equal to the number of outputs and ~~, is the spacing
between the wavelengths (channels) which the device can separate.
The combinations usable with low non-uniformity are therefore limited
and may not be those which are best for the purposes of optimizing the
crosstalk performance. The latter parameter, in fact, represents a
considerable limitation on loopback configurations: this structure is
intrinsically highly sensitive to the in-band crosstalk, which is the most
critical factor for the performance of the system.
The diagram in Figure 1, representing an AWG in loopback
configuration according to the prior art, shows the generic path of a
channel ~,k which enters at the port i, together with all the other
channels of the WDM signal, and exits from the port k; this signal is
nominally propagated in the path i-k, and then continues in the branch
k-h in the loopback configuration and subsequently re-enters from the
port h and finally exits from the port j. If a;k,n; denotes the attenuations
CA 02290311 1999-11-24
_$_
affecting the signal through the AWG, where, in particular, a;k is the
attenuation affecting the signal as it enters from the port i and exits from
the port k, and ah~ is the attenuation affecting the signal as it enters from
the port h and exits from the port j, the level of in-band crosstalk
generated by the undesired contribution X;~(~,k), in other words by the
nominally cancelled part of the signal, at the wavelength ~.k, which is
transmitted directly from the port i to the port j, is equal to:
X,; (~k )
arx an;
This is because all the crosstalk contributions due to the residual
signal propagated in the loopback paths p-q are negligible by
comparison with the direct contribution X;~.
Assuming for simplicity that a;k = ak~ = a;~ = a and X;~(~,k) = X, and
bearing in mind that for a commercial device it is possible that (X/a)de =
25 dB and
(a)de = 5 dB, we note that the in-band crosstalk for the configuration in
Figure 1 will be equal to:
(X/a2)de = 20 dB, a value which is insufficient to ensure transmission
with a good signal-to-noise ratio.
The article by O. Ishida et al., published in Photonics Technology
Letters, vol. 6, No. 10, Oct. 1994, pp. 1219-1221, describes a multiple-
selection switch using an AWG in a configuration called "foldback", in
other words one in which the signals from other outputs, which have
therefore already been selected in terms of wavelength, are reinserted
into some outputs of the AWG; this is done by means of branches
consisting of lengths of fibre between which switching can be carried
out in order to provide a
2 x 2 switch operating by wavelength division. Each switch, in one
state, generates a loopback configuration in the branch in which it is
connected, and a foldback configuration in the other branch. The
CA 02290311 1999-11-24
-9-
experiment demonstrated that the performance in the foldback
configuration was better than that of the loopback configuration.
The applicant has tackled the problem of providing a modular filter
for extracting and/or reinserting or processing a desired number of
channels at the node of an optical telecommunications network, in a
configuration which enables the losses to be balanced and the
crosstalk noise to be minimized for all the channels making up the
WDM signal.
In particular, an efficient configuration which uses an NxN AWG and
is of the type described previously as "foldback" has been invented. In
this configuration, in an AWG with N inputs and outputs, which can
therefore separate N signals, each containing at least one wavelength,
Q _< N/2 outputs are connected to the remaining Q outputs by means of
Q branches, each comprising at least one length of fibre. A signal
processor device, comprising at least one filter with a spectral response
such that it transmits the signal at the wavelength or wavelengths which
nominally pass through the said branch and cancels the contributions at
other wavelengths, is connected in at least one of the branches.
The applicant has found that in this way it is possible to obtain a
significant reduction of the crosstalk of the AWG, and in particular of its
in-band crosstalk. An attenuator which equalizes the recombined WDM
signal may also be provided in the branch in question. If desired, the
signal processor comprises a device for inserting and/or extracting
signals in a corresponding branch.
A first aspect of the present invention relates to an optical
telecommunications system comprising:
- at least one transmission station capable of transmitting a multiple-
wavelength optical signal comprising at least two predetermined
wavelengths;
CA 02290311 1999-11-24
- 10-
- a wavelength division multiplexer for sending the said transmission
signals to an optical fibre transmission line;
- at least one station for receiving the said transmission signals;
- the said optical fibre line connecting the said transmission and
receiving stations;
- a modular filter for extracting and/or inserting and/or processing
optical signals, comprising:
- a wavelength multiplexer/demultiplexer device having at least two
first ports for connection to the said line and a number N of
second ports into which the said wavelengths are divided;
- a number Q, less than or equal to N/2, of branches which
interconnect 2Q of the said N second ports into which the said
wavelengths are divided,
characterized in that at least one of the said branches comprises a
signal processor device for at least one of the wavelengths circulating
in the branch.
In particular, the said modular filter introduces an in-band crosstalk of
the said multiple-wavelength optical signals of less than 40 dB.
A further aspect of the present invention relates to a modular I:IIter for
extracting and/or inserting and/or processing optical signals,
comprising:
- a wavelength multiplexer/demultiplexer device having at least two
first ports for a multiple-wavelength signal and a number N of second
ports into which the said wavelengths are divided;
- a number Q, less than or equal to N/2, of branches which
interconnect 2Q of the said N second ports into which the said
wavelengths are divided,
characterized in that it has, in at least one of the said branches, a
signal processor device for at least one of the wavelengths circulating
in the branch.
CA 02290311 1999-11-24
-11-
Preferably, the said signal processor device comprises a
wavelength-selective filter.
Alternatively, the said signal processor device comprises an
attenuator.
In particular, the said signal processor device comprises a switch.
Preferably, the said signal processor device comprises:
- a first signal multiplexer/demultiplexer capable of separating the
wavelengths of the signals from one of the said second ports into at
least two branches;
- a second multiplexer/demultiplexer capable of combining the said
wavelengths of the signals from the two branches at another of the
said second ports,
- a switch connected in at least one of the two branches.
Alternatively, the said signal processor device comprises:
- a first signal divider device capable of dividing the signal injected
into one of the said second ports into a first and a second branch;
- a filter tuned to a first wavelength and a switch connected in series
in the first branch;
- a filter tuned to a second wavelength different from the first,
connected in the second branch;
- a second signal divider device, capable of dividing a signal injected
at another of the said second ports into the two branches, and of
recombining the said signals.
Preferably, the said N/2 branches are connected to the said device in
such a way as to interconnect N/2 remote second ports.
In particular, the said switch comprises two input ports and two
output ports, and is capable of sending the signal passing through each
of the two inputs selectively to either of the two outputs in response to
an operating signal.
CA 02290311 1999-11-24
- 12-
In particular, the said wavelength multiplexer/demultiplexer
comprises an AWG having at least 2 input ports and N output ports.
Alternatively, the said signal processor device comprises an
additional AWG having at least two input ports and N output ports.
A further aspect of the present invention relates to a modular filter for
generating optical signals for multiple-wavelength telecommunications
systems, comprising:
- a wavelength multiplexer/demultiplexer having at least a first port
for receiving a wide-band signal and a number N of second ports in
which the said signal is divided into at least N wavelengths and a
third port for the output of the said optical channels;
- a transmitter having a spectral amplitude greater than that of the
said wavelength multiplexer/demultiplexer,
characterized in that it has
N/2 branches which interconnect the said N second ports into which
the said wavelengths are divided, and which are connected in such a
way as to associate N/2 remote ports with each other.
Preferably, a wavelength-selective filter is connected in at least one
of the said branches.
Alternatively, an attenuator is connected in at least one of the said
branches.
In particular, a radio-frequency modulator is connected in at least
one of the said branches.
In particular, the said modulator is an electro-optical modulator or an
acousto-optical modulator.
A further aspect of the present invention relates to a method for
processing a multiple-wavelength signal, characterized in that it has the
following stages:
- dividing the wavelengths of the said multiple-wavelength signal
present at an input port of a wavelength multiplexer/demultiplexer
CA 02290311 1999-11-24
- 1.3 -
device among a number N of intermediate ports, in such a way that
one or more nominal wavelengths are associated with each port;
- sending the said divided wavelengths to a number of branches less
than or equal to N/2;
- extracting from at least one of the said branches at least one of the
signals circulating in the branch;
- selectively ~Itering in each branch the signals at the nominal
wavelengths circulating in the branch in question;
- reinjecting these signals into a different intermediate port of the said
multiplexer/demultiplexer device;
- re-multiplexing the said signals at an output of the said
multiplexer/demultiplexer device.
In particular, the stage of selectively filtering in each branch
additionally comprises the stage of attenuating the signals circulating in
the branches, by a predetermined amount for each signal.
In particular, the stage of reinjecting these signals comprises
reinjecting these signals into a total of N/2 remote intermediate ports.
A further aspect of the present invention relates to a method for
reducing the crosstalk in an NxN AWG which processes a multiple-
wavelength signal, characterized in that it comprises the following
stages:
- sending a multiple-wavelength signal to an input of the said AWG;
- dividing the components of the WDM signal into a number N of
output ports of the said AWG;
- sending the said divided wavelengths to N/2 branches;
- selectively filtering, in each branch, the signals at the nominal
wavelength circulating in the branch in question;
- reinjecting these signals by means of the said N/2 branches into the
same AWG;
- multiplexing the said signals by means of the same AWG;
CA 02290311 1999-11-24
-14-
- extracting the re-multiplexed multiple-wavelength signal from an
output of the said AWG.
Further details may be obtained from the following description, with
reference to the attached drawings, which show:
in Figure 1, a general circuit of an AWG in a configuration of the
"loopback" type, showing the crosstalk paths, according to the prior art;
in Figure 2, a general circuit of a modular filter for processing the
individual optical channels according to the present invention;
in Figure 3, a signal processor device for extracting and/or inserting
optical channels according to one embodiment of the present invention;
in Figure 4, an operating circuit for the modular ~Iter shown in Figure
2, in which the signal processor device is that illustrated in Figure 3;
in Figure 5, a signal processor device for extracting and/or inserting
optical channels according to a further embodiment of the present
invention;
in Figure 6, a signal processor device for extracting and/or inserting
optical channels according to a further embodiment of the present
invention;
in Figure 7, an operating circuit for the modular filter shown in Figure
2, in which the signal processor device is that illustrated in Figure 5;
in Figure 8, a modular filter for extracting and/or inserting optical
channels with a plurality of stages, according to a further embodiment
of the present invention;
in Figure 9, a general circuit of an AWG in a configuration of the
"foldback" type, showing the crosstalk paths;
in Figure 10, a circuit of an experimental embodiment of a modular
filter for extracting and/or inserting optical channels according to the
present invention;
CA 02290311 1999-11-24
-15-
in Figure 11, a circuit of an experimental embodiment of a signal
processor device for extracting andlor inserting optical channels
according to the circuit in Figure 5;
in Figure 12, a spectral response of the wavelength combining
(dividing) device used in the circuit in Figure 11;
in Figure 13, a transmitter for WDM signals, using a modular filter
according to the present invention.
Figure 2 shows a modular filter comprising an AWG 2 shown, by way
of example, with N inputs and N outputs. To provide an example of the
present invention, the NxN AWG is selected to have an FSR = N~0~,,
where 0~, is the separation between adjacent channels. It should be
noted that the filter according to the present invention can be made with
devices of the AWG type having any number of inputs and outputs. The
number of inputs and outputs is related to the number of wavelengths
which are to be processed by means of the filter. In particular, we shall
consider by way of example a filter based on an NxN AWG and capable
of operating on a WDM signal containing from N/2 to N wavelengths
which are different from each other. Additionally, the filter is
bidirectional, and therefore the input may be used as the output, and
vice versa.
For the purposes of the present invention, the modular filter for
extracting and/or inserting or processing optical channels in multiple-
wavelength optical telecommunications systems comprises a
wavelength multiplexer/demultiplexer device having at least two ports,
one for the input and one for the output of a multiple-wavelength signal,
and a number N of ports into which the said wavelengths are divided.
This operation is carried out by an AWG, but other devices which carry
out the same function can be used in an equivalent way.
CA 02290311 1999-11-24
- 16-
The WDM signal is inserted at one of the inputs A; of this AWG, and
is distributed, according to the wavelengths which it contains, among M
of the N outputs, where M = N/2...N, of the AWG.
Each of the branches comprises an optical path, for example one
made from optical fibres. The optical connections between the
branches B and the AWG are made by known techniques which are
known as "pigtailing" in the case of branches made from optical fibres.
Similarly, the inputs of the AWG are optically connected to optical input
paths such as, for example, the optical fibres A;, A~ connected to the
AWG by means of suitable "pigtailing".
The N outputs U,, U2...UN of the component are interconnected by a
number N/2 of branches B,, B2...BN,2; consequently forming what was
previously defined as a "foldback" configuration. Owing to the
periodicity properties of an NxN AWG with FSR = N~0~,, the selection of
the input A; out of the N possible inputs determines the distribution of
the wavelengths among the outputs. In the configuration in Figure 2,
which is provided by way of example but is not exclusive, the output U,
is connected to the output UN,2" through the branch B,, the output U2 is
connected to the output UN,2+z through the branch B2, and the output
UN,z is connected to the output UN through the branch BN,2. In each of
the said branches there is a signal processor device P~...PN,2 which can
be used to extract and/or insert and/or process individual components
which form the WDM signal. For example, the device comprises a
switch, a filter, a pair of optical fibre terminations, an attenuator or a
further AWG. Examples of such a device will be described in detail in
the following text according to three corresponding embodiments,
shown in Figures 3, 5 and 6.
The WDM signal is globally reconstructed after the double passage
in the AWG at one of the remaining inputs Au.
CA 02290311 1999-11-24
-17-
The selection of the input fibres Ai and Au, and consequently of the
connections between the outputs, may advantageously be made
according to the characteristics of the component, by calculating the
paths inside the AWG of all the signals at individual wavelengths which
make up the WDM signal, in such a way as to optimize the global
attenuation of the filter and in such a way as to distribute this
attenuation equally among all the paths (maximum uniformity). In
particular, it is advantageously possible to interconnect pairs of
equidistant outputs A; A;+N,2. The AWG component (with FSR = Nxo~,,
where N is the number of ports and 0~, is the spacing between the
channels) used as a demultiplexer has a distribution of losses with a
non-uniformity which may be, for example, 2-3 dB; the configuration
with pairs of equidistant outputs indicated above, whether a number of
channels <= N/2 or a number of channels >N/2 is to be processed,
enables the non-uniformity of the insertion losses IL at the output port
to be reduced by at least half with respect to that of the AWG
component used as a demultiplexer solely by selecting the ports Ai and
Au in a suitable way.
Additionally, by using fixed attenuators with suitable IL in the
connecting branches B,, B2...BNI2~ it is also possible to compensate the
insertion loss (up to less than 0.5 dB) in the different paths; by using
devices having a spectral response such that the propagation of only
the wavelengths required for the branch in question is permitted, and
having a calibrated IL, it is possible to suppress the crosstalk
contributions as well as to equalize the losses.
In order to achieve this, use is made of a procedure which
comprises:
- the complete spectral characterization of the AWG (for example, in
the case of an NxN type, N2 spectral responses will be obtained)
CA 02290311 1999-11-24
-18-
and the characterization of the components (e.g. switches,
attenuators, and filters);
the simulation of all the possible configurations which can be
obtained, according to the pair of ports Ai and Au, by connecting
the output ports of the AWG (without any other component present
in the branch), in terms of IL and contributions to the in-band
crosstalk:
the selection of the optimal configurations in terms of IL and
contributions to the crosstalk: the pair of ports Ai and Au is that
which provides a minimum channel-by-channel insertion loss with
the minimum non-uniformity. A check is also made to ensure that
the contributions to the crosstalk do not exceed the values
permitted by the system in which the module is to be connected
(e.g., for the homodyne in-band crosstalk the worst case (sum of
contributions in phase) is < -35 dB);
the insertion of selected components along the loop paths: it is
necessary to determine which channel has the maximum value of
IL, with respect to which the difference of IL in each other channel
is calculated.
If the performance of the modular filter also requires the suppression
of the crosstalk contributions, it is possible to connect filters having a
bandwidth greater than or equal to the channel bandwidth of the AWG
along the branches B,, B2...B~,z. This is because the spectral
characteristics of the filter must be such that the band of the channels
at the output of Au is not reduced with respect to the performance
obtained for the double passage through the AWG; the maximum
bandwidth and the out-of-band profile depend on the required
suppression of the crosstalk contributions.
If, for example, 200 GHz grating interference filters are to be used for
a modular filter with an inter-channel spacing of 100 GHz, for each
CA 02290311 1999-11-24
-19-
wavelength it is necessary to consider the contributions to the crosstalk
which cannot be eliminated (in other words those due to direct crosstalk
from the port Ai to Au and the contribution at the same wavelength
which is propagated along the same branch B; in the opposite direction
to the signal) and the contributions which are propagated along the
adjacent branches B;_, and B;+,, further attenuated by approximately 10
dB. With this selection, the bandwidth of the filter is such that the
adjacent wavelengths spaced 100 GHz apart are attenuated by 9-10 dB
with respect to the central wavelength, while those spaced 200 GHz
apart are reduced by approximately 35 dB. Therefore the contributions
to the in-band crosstalk which are propagated along the remaining
branches are entirely negligible.
A filter as described above can advantageously be connected in a
multiple-wavelength telecommunications system comprising two
terminal stations, one for transmission and one for reception.
In particular, the transmission station comprises N>1 transmitters of
optical signals, each at one wavelength.
The number N of independent wavelengths used for the signals of
each transmission station, corresponding to the number of optical
channels which can be used for transmission, can be selected
according to the characteristics of the telecommunications system.
The optical transmitters included in the transmission stations are
transmitters with direct modulation or with external modulation,
according to the system requirements; these requirements can be
related, in particular, to the chromatic dispersion of the optical fibres of
the system, to their wavelength and to the specified transmission
speed.
The output of each of the transmitters of the transmission stations is
connected to a corresponding multiplexer which combines the
corresponding optical signals and sends them to a single output which
CA 02290311 1999-11-24
-20-
is connected to the input of optical power amplifiers. In general, the
multiplexers are passive optical devices, by means of which the optical
signals transmitted along corresponding optical fibres are
superimposed in a single fibre; devices of this type consist, for
example, of fused fibre, planar optic, micro-optic and similar couplers.
An example of a suitable coupler is the one marketed under the
symbol SMTC2DOOPH210 by E-Tek Dynamics Inc., 1885 Lundy Ave.,
San Jose, CA, USA.
The power amplifiers raise the level of the signals generated by the
transmission stations to a value sufficient to pass through the next
length of optical fibre interposed before the reception station or before
amplifier means maintaining a sufficient power level at the terminal to
guarantee the required transmission quality.
For the purposes of the present invention and for the application
described above, a suitable power amplifier is, for example, a fibre
optical amplifier of the commercial type, having an input power from -
13.5 to -3.5 dBm and an output power of at least 13 dBm.
An example of a suitable model is the TPA/E-MW, marketed by the
applicant, and using an active optical fibre doped with erbium.
Each of the power amplifiers is then connected to a corresponding
length of optical line, usually consisting of a single-mode optical fibre, of
the step index type, inserted in a suitable optical cable, a few tens (or
hundreds) of kilometres in length: for example, with the amplifier means
described below and the indicated power levels, approximately 100
kilometres.
At the termination of each of the said lengths of optical line there is
one or more intermediate stations for amplifying the optical signal, each
station comprising line amplifiers capable of receiving the signal which
has been attenuated in its travel along the fibre, and of amplifying it to a
sufficient level to be fed to a plurality of corresponding subsequent
CA 02290311 1999-11-24
-21-
lengths of optical line, covering the total transmission distance required
until it reaches a preamplifier or a further amplification station, as
appropriate. The term "preamplifier", in the context of the present
invention, denotes an amplifier designed to compensate for the losses
in the last length of optical line and the insertion losses in the
subsequent demultiplexer stages, in such a way that the signal entering
the receiving stations has a power level suitable for the sensitivity of the
device. The preamplifier also has the function of limiting the dynamics
of the signals, by reducing the variation of the power level of the signals
entering the receiver with respect to the variation of the power level of
the signals arriving from the transmission line. A type of preamplifier
suitable for use for the preamplifiers is, for example, an optical amplifier
using an erbium-doped active optical fibre of the commercial type,
having a total input power from -20 to -9 dBm and an output power of 0-
6 dBm.
A suitable model is, for example, the RPA/E-MW, marketed by the
applicant.
The optical signals multiplexed at the output of the preamplifiers
arrive at corresponding demultiplexers which are capable of separating
the signals and sending them to N output optical fibres, according to the
corresponding wavelengths, the signals then being transmitted to the
corresponding N receivers present in the receiving station. An example
of a demultiplexer suitable for use in the present transmission system is
the demultiplexer described in patent application EP854601 filed by the
present applicant.
If the optical signals to be transmitted are generated by signal
sources which have their own transmission characteristics (such as
wavelength, type of modulation, power) which are different from those
specified for the described connection, each transmission station
comprises interface units capable of receiving the optical signals
CA 02290311 1999-11-24
-22-
generated by the transmission stations, detecting them, regenerating
them with new characteristics suitable for the transmission system, and
sending them to the multiplexers.
US Patent 5,267,073 in the name of the present applicant describes
interfacing units comprising, in particular, a transmission adapter
capable of converting an incoming optical signal into a form suitable for
the optical transmission line, and a receiving adapter, capable of
reconverting the transmitted signal into a form suitable for a receiving
unit.
For use in the system, the transmission adapter preferably comprises
an externally modulated laser as the output signal generation source.
The present optical fibre telecommunications system provides, in
addition to the channels intended for the communication signals and
made available to the users, an independent channel for the
transmission of service signals. A system comprising channels intended
for service signals is described in US Patent 5113459 in the name of
the applicant.
These service signals may be of various kinds, for example signals
for the alarm system, for the monitoring or operation of equipment
located along the line, such as repeaters or amplifiers, or for
communications between the maintenance personnel operating at one
point of the line and an intermediate or terminal station of the line.
In these cases, it is therefore necessary to introduce into the
communication line further signals, which may be received and injected
at any intermediate station or at the terminal stations. These service
signals are transmitted at a wavelength significantly different from the
communication wavelength, in other words one which can be separated
by means of a suitable dichroic coupler.
Although the injection of the service signals into the optical line and
their extraction from it are conveniently executed at the terminal
CA 02290311 1999-11-24
-23-
stations of the line and at the line amplifiers, as described above,
dichroic couplers and corresponding service signal receiving and
transmission stations may also be introduced in any other position of
the optical fibre line where they are required.
Optical amplifiers generally comprise at least one active fibre doped
with a rare earth, capable of amplifying the multiple-wavelength
transmission signal in response to a supply of light radiation at a
pumping wavelength.
This pumping wavelength is different from that of the transmission
signals and is produced by at least one pumping source of the said
active fibre, having an optical power which can be controlled by a
control unit of the station within which the amplifier is located; by way of
example, this source may be a laser. The amplifier also comprises a
coupling device for sending the said pumping radiation and the said
transmission signal into the active fibre.
A modular filter according to the present invention may
advantageously be connected along the said line at the intermediate
amplifier stations to permit extraction from and/or insertion into one or
more optical channels.
Figure 3 shows an embodiment of the signal processor device P
which comprises a switch S, having two inputs I, and IZ and two outputs
E, and E2, which, in response to a suitable operating signal, is capable
of sending the signal passing through each of the two inputs I, or 12 to
one of the two outputs E, and E2 selectively. The input I, is suitable for
connection to a source of an optical signal (not shown in the figure)
modulated by a data element, for example a laser diode which may be
associated with an optical modulator. The output E2 is designed to be
connected to an optical receiver capable of converting the optical signal
into an electrical signal carrying the same information. This switch
therefore has two operating states: in one case, determined by a
CA 02290311 1999-11-24
-24-
specified operating signal, a signal at a wavelength ~,; inserted at the
input I, is sent to the output E2 and a signal ~,;' inserted at the input Iz
is
sent to the output E,; in the second case, with a different operating
signal, the same input signals are inverted at the outputs. The switch is
also connected in series to a selective ~Iter 4 having a spectral
response such that the wavelengths necessary for the branch in
question can be propagated and having an attenuation calibrated as
described above for the equalization of the components of the WDM
signal. The filters suitable for this purpose are wavelength-selective
filters, for example those using in-fibre gratings to transmit a particular
wavelength and reflect all the others. Further types of filters suitable for
the purpose are, for example, filters made by micro-optical technology
and interference ~Iters.
These devices may be selected in such a way that the switch with
minimum IL is assigned to the wavelength of the channel which has the
maximum IL: this path provides the insertion loss ILM"~ from Ai to Au.
For each remaining wavelength, a component having an IL as close as
possible to the difference between ILM,,~ and the insertion loss due to
the travel within the AWG will be connected. The switch S is connected
in series with an attenuator A whose value of IL is selected by the
procedure described above, or a filter F selected to have a bandwidth
such that it eliminates the contributions at wavelengths other than that
which nominally circulates in the branch.
In the example described, the switch is advantageously used to carry
out the function of extraction and insertion of a WDM signal from and
into a channel. In this case, the extracted wavelength ~,; must be
nominally equal to the reinserted wavelength ~,;'.
A switch suitable for this function must conform to the specific
characteristics of the system in which the filter is connected, and
therefore the switching rate depends directly on the maximum period of
CA 02290311 1999-11-24
- 25 -
interruption of a signal in the line which is permitted before the system
detects an anomaly. For example, in the SDH communication standard
this period is a maximum of 50 msec, and therefore the switch must
have a significantly higher switching rate. For example, a thermal-
optical switch provides a switching rate of the order of 2 msec and an
insertion loss of 0.5 dB, but its isolation is not very high, at 15 to 27 dB.
An optomechanical switch provides a switching time of approximately
10 ms, an insertion loss of approximately 0.5 dB and an isolation of
more than 50 dB.
Figure 4 shows an operating circuit for a modular filter in an
embodiment which comprises a signal processor device of the type
shown in Fig. 3 in each of the branches.
This filter can process a WDM signal with a maximum of N/2
wavelengths (one channel only for each branch).
By using a different channel processor, it is possible to extend the
use of an NxN AWG in "foldback" configuration to process up to N
channels. It should be noted that, if the number of wavelengths of the
WDM signal is M, where N/2 < M <_ N, there are two channels at
different wavelengths ~,; and ~.;+N,2 propagated in opposite directions in
M-N/2 of the branches Bi which form the "foldback" paths.
Figure 5 shows an embodiment of a signal processor device P which
can advantageously extend the capacity of the modular filter from N/2
to N channels. This processor comprises a first wavelength
multiplexer/demultiplexer M, which separates into two branches R, and
RZ signals at different wavelengths ~.; and ~,;+~,2 from a port H,, and a
second wavelength multiplexer/demultiplexer MZ which recombines the
signals of the two branches R, and RZ at a port Hz. A switch S, of the
type illustrated in Figure 3 and described previously, which can extract
and reinsert a signal at a given wavelength from and into the branch, is
present in each of the two branches R, and R2.
CA 02290311 1999-11-24
-26-
In this device, therefore, two signals, having two corresponding
optical signals at different wavelengths ~,; and ~;+Nrz which travel in
opposite directions and are inserted, in the first case, at the port H, and,
in the second case, at the port Hz, are sent, in the first case (~.;), along
the branch R,, and in the second case (~.;+N,z) along the branch Rz. In
each of the said branches, the signal is extracted and inserted by
means of a corresponding switch S. The processed signals are sent to
the corresponding ports Hz (~,,) and H, (~,;+N,z).
It should be noted that the operation of the signal processor device
in Figure 5 is bidirectional; this ensures that the modular filter using
these processors is also bidirectional, since the AWG is intrinsically a
component having this characteristic. This is because the signals at the
wavelengths ~,; and 7~;+N,z travel in opposite directions to each other; the
first wavelength multiplexer/demultiplexer M, sends the signal at
wavelength ~,; along the branch R,, and suppresses all the other
components which may be present. In the opposite direction, the
second wavelength multiplexer/demultiplexer Mz sends the signal at
wavelength ~,;,N,z along the branch Rz in the opposite direction to 'the
signal ~.; and suppresses all the other components which may be
present.
Figure 6 shows an alternative embodiment of the signal processor
device P in Figure 5, which comprises two signal division/combination
devices D, and Dz which respectively divide the power of the signal
arriving from port H, and the signal arriving in the opposite direction
from the port Hz, along the branches R, and Rz. A filter F,,N,z tuned to a
wavelength ~,;+N,z and a switch S of the type shown in Figure 3 are
connected in series in the branch R,. A switch S of the type shown in
Figure 3 and a filter F; tuned to a wavelength ~,; are connected in series
in the branch Rz. The devices D1 and D2 then operate to recombine the
signals from R, and Rz respectively at the outputs H, and Hz.
CA 02290311 1999-11-24
-27-
Consequently, a signal having a wavelength ~,; inserted at the port H, is
divided into both branches R, and Rz; in the branch R, there is a filter
tuned to the wavelength of the signal ~.;+Nrz which therefore suppresses
the signal ~,; before it reaches the switch S; in the branch Rz, the signal
~,; is processed, as described previously, by the switch S and passes
unchanged through the filter F; and emerges from the output Hz,
through Dz. The signal at ~.;,Nrz is processed in a similar way along the
path R, and is suppressed in the path Rz. The signal
division/combination devices D,, Dz are, for example, fused fibre or
planar optics dividers.
The filters F; and F;;N,z are wavelength-selective, using, for example,
in-fibre gratings to transmit a particular wavelength and reflect all the
others. Further types of filters suitable for the purpose are, for example,
filters made by micro-optical technology and interference filters.
In the preceding embodiments shown in Figure 5 and Figure 6 there
are filters (M,, Mz in Fig. 5 and F;, F;+N,z in Fig. 6) which suppress the
crosstalk according to the present invention. In the embodiment shown
in Figure 5, the wavelength multiplexers/demultiplexers M,, Mz separate
the two wavelengths circulating in this branch, each in its direction of
propagation, and cause the substantial suppression of all the others.
Similarly, in the embodiment shown in Figure 6 the filters F; and F;+N,z
suppress all the signals at wavelengths other than those which
nominally circulate in the branch, namely ~,;,N,z and ~,;, and
consequently substantially reduce the crosstalk and equalize the
remultiplexed signal after insertion into the AWG.
Figure 7 shows an operating circuit for a modular filter in an
embodiment operating on N wavelengths which comprises signal
processor devices of the type in Figure 5.
In this case also, the input and output are selected in such a way
that ~,; is present at the output U;. By using a number j < N/2 of signal
CA 02290311 1999-11-24
-28-
devices of the type shown in Figure 5 and, for example, using N/2-j
other signal processing devices of the type shown in Figure 3, it is
similarly possible to construct a modular filter capable of operating at
N/2+j predetermined wavelengths.
In all the preceding configurations, if the operation of extracting the
channels and reinserting them into the line has not been carried out on
all the channels which travel along the line, and consequently if the
signal processor P is not connected in all the branches B, or if
processors P whose degrees of attenuation differ from each other are
present in some branches, in order to counterbalance in power terms
the losses introduced in the channels which are extracted, it is possible
to connect in the branches not containing the processor device P an
attenuator with its attenuation determined in such a way as to keep the
output channels balanced. A similar attenuator can also be associated
with the extracted channels or with those to be inserted, to adapt their
power in accordance with the subsequent use.
The signal processors P,...PN,2 present in one or more branches,
shown in Figures 5 and 6 and described previously, are capable of
extracting and inserting the signals travelling in both directions in the
branch in question.
The configuration described can also be extended advantageously to
the case in which an NxN AWG is used and a number of channels M,
which is greater than the number of ports N of the AWG, is to be
processed.
This is because an advantageous characteristic of the filter
according to the present invention is its modularity. The "foldback"
configuration makes it possible to use a cascade of AWGs connected in
a multiple-stage configuration. At each stage, the channels are divided
into groups containing a sub-multiple of the channels which enter this
AWG stage, until all the channels separated from each other
CA 02290311 1999-11-24
-29-
individually or in pairs for each foldback path (B; in Figure 2), and
therefore processed by one or more processor units similar to those in
Figure 3, 5 or 6, are obtained in the final stage with the outputs
connected in foldback. The separation takes place progressively from
one stage to the next. In this embodiment, in order to process a WDM
signal having N components at different wavelengths, it is necessary to
use AWGs with a number of ports N; which may be considerably
smaller (<= M/2) than the number of channels M, and this configuration
may therefore be extended to WDM signals with an arbitrarily large
number of channels.
Figure 8 shows a configuration of this type illustrated by way of
example with a WDM signal which comprises 8 different components
~,,...~,8. In this configuration, there are three AWGs with N; = 4 outputs
and four signal processor devices according to Figure 5. In particular, a
first stage contains an AWG 4 with FSR = 4~0,4, where o~,4 is the
spacing between the channels of the WDM signal at the input of the
said AWG. The WDM signal is sent to the input A, of this AWG and,
having been reprocessed and after insertion and extraction, exits from
the port A3. This AWG separates the channels into pairs N/2 channels
apart at each of the outputs U,...U4. Each of the first two pairs of
signals ~,,, ~,5, and ~.3, ~., is sent to an input of a second AWG 6 which
acts on beams whose spacing is twice that of the AWG 4; this second
AWG has an FSR2 = 4x0,6, where o~ = 2~0,4. Similarly, each of the
pairs of signals ~.z, ~,s, and ~,4, ~,8 is sent to an input of a further AWG 8
with spacing characteristics and FSR similar to those of the AWG 6.
The two AWGs 6 and 8 which form the second stage of the filter have,
at their outputs, connections of the "foldback" type along which are
propagated, in opposite directions, two channels which are then
processed by a signal processor device as shown in Figure 5 or Figure
CA 02290311 1999-11-24
-30-
6, which extracts and inserts the channels in a way similar to that
described with reference to the filter in Figure 7.
The configurations of the "foldback" type according to the invention
not only make it possible to obtain filters with a high uniformity of
channel loses, but also offer considerable advantages for the crosstalk
performance by comparison with the preceding configurations of the
"loopback" type.
Figure 9 shows the generic path of a channel in a configuration of
the "foldback" type in which the WDM signal is inserted at the port i and
passes initially into the AWG 2, after which the signal having a
wavelength ~,k exits from the port k and is reinjected into the AWG at
the port h by means of the branch k-h; the reconstructed WDM signal
then exits at the port j and also includes all the noise contributions
generated by the in-band crosstalk.
By comparison with Figure 1, it may be demonstrated that all the
more significant in-band crosstalk contributions are of the second order,
in other words generated by a double passage along paths in which the
signal is nominally cancelled; additionally, by using signal processor
devices of the type shown in Figure 5 or 6, it is possible to select the
spectral response of the wavelength multiplexer/demultiplexers M, and
M2 shown in Figure 5 and of the filters F; and F;,N,2 shown in Figure 6 in
such a way as to inhibit all the contributions X;p,v(~,K) shown in Figure 9,
where X;P,v(~,K) represents the noise contribution at the wavelength ~.k
which is sent from the port i to another generic port p, and which, when
reinserted into the AWG at the port q by means of the branch p-q, is
superimposed at the output port j on the signal at the same wavelength.
Therefore the dominant contributions of in-band crosstalk in a circuit
of the "foldback" type are:
CA 02290311 1999-11-24
-31-
Xih~~'k~'Xkj~/~'k~ D~(/~'k)
Gzik GL'hj aik ahi
The first contribution is defined as being due to what is called the
"inverse path", in other words the path followed by part of the signal at
the wavelength ~,k in the direction opposite the signal at the same
wavelength in the branch B; the second contribution is due to the input-
output directivity, in other words the part of the signal at the wavelength
~,k which is nominally cancelled and is coupled directly from the input
port to the output.
Generally, x;h(~,), D;~(~,) and a;k are substantially uniform regardless of
the variation of the indices and of the wavelength in the working band of
the AWGs. The typical values are indicated as x, D and a respectively.
Given the typical values of D, a and X/a in commercial devices, we
find that both contributions are 30 dB greater, and therefore their effect
is negligible, since the critical limit for the systems is estimated to be
approximately 25 dB.
The modular filter whose configuration is shown in Figure 10 was
constructed for experimental purposes; it is an embodiment of a
combination of the circuits shown in Figure 4 and Figure 7.
This modular ~Iter consists of an 8x8 AWG 2, made by PIRI (USA),
of the narrowband type (band at 3 dB = 40% of the channel spacing),
capable of processing wavelengths with a spacing of 1.6 nm (200
GHz); the FSR of the AWG is 8 x 1.6 = 12.8 nm. The ports selected for
the input and output of the constructed modular filter are 2 and 6
respectively; the other ports of the AWG, located on the same side as
the input and output, are terminated with adapters 3 in such a way as to
prevent undesired reflections.
The WDM signal used for the experiment consisted of 5 channels
spaced 1.6 nm apart; having the following wavelengths in nm:
,, : 1550.92
CA 02290311 1999-11-24
-32-
~,z: 1552.52
~,3: 1554.13
~,4: 1555.75
~,5: 1557.36
With this selection of the input and output, the following connections
are made at the outputs of the AWG, as shown in Figure 10:
U5-U1 (~1~ ~g), Ug'UZ (~2)~ U7 U3 (~3)~ Ug-U4 (~~4)~
Since one of the foldback branches has two channels passing
through it in opposite directions of propagation, a processor device of
the type shown in Figure 5 is used in this branch, while a device of the
type shown in Figure 3 is sufficient for all the other branches.
In particular, the branch 5-1 has the wavelength ~,,, propagated from
the port 5 to the port 1, and the wavelength ~.5, propagated in the
opposite direction. The other wavelengths are processed by the other
branches as shown below:
from U6 to U2
from U, to U3
from U8 to U4
The processor device P' used in the branch having two wavelengths
is constructed as shown in Figure 11. The WDM units 5 and 7
(multiplexers/demultiplexers) which are used are interference devices
marketed by JDS; in particular, Figure 12 shows the transfer function of
the multiplexer/demultiplexer device 5 whose central wavelength is ~,,.
The reflection insertion loss (IL) is indicated by the curve 121 and the
transmission insertion loss is indicated by the curve 122.
The optical switches 9 and 11, like the switches 13 in Figure 10
which are used in the other branches of the modular filter, are of the
2x2 optomechanical type, and are again marketed by JDS. The
measured optical loss of the signal in the pass branch (in Figure 11: R1
for ~,, and R2 for ~,5) is 2.3 dB.
CA 02290311 1999-11-24
-33-
The in-band crosstalk generated by the passage of the signals
through the nominally isolated branches (in Figure 11: R1 for ~,5 and R2
for ~,,) is 40.7 dB.
With reference to Figure 10, the fixed attenuators 15, having a value
of 2 dB and produced by TELE.S (Italy), have been introduced into the
branches having single wavelengths, in series with the switches, in
order to balance the losses in these branches with respect to the
branch containing the processor of the type shown in Figure 5.
The measured performances of the modular filter are summarized in
the following Table 1, where IL is the loss of the optical input-output
path, XTL is the in-band crosstalk contribution of the 2nd order due to
what is called the "inverse path" shown in Figure 9, and DXT is that due
to the directivity between the input and outcut:
IL (dB) XTL (dB) DXT (dB)
12.3 53.2 44.4
11.6 49.4 33.4
12.0 53.6 43.1
12.1 54.1 53.7
12.3 53.1 44.4
Table 1
As will be noted, the losses are uniform over all the channels and the
in-band crosstalk contributions are much better than the critical limit for
the systems.
In an alternative configuration, a modular filter according to the
present invention may have N/2 foldback branches B; on the N outputs,
arranged in such a way that they interconnect the remote ports
positioned N/2 apart, in which branches no signal processor device is
present. These branches return the signal leaving from one port to the
port connected to it by the corresponding branch. When a wide-band
light source is connected to one of the input ports of the AWG, this
CA 02290311 1999-11-24
-34-
source preferably having a spectral width greater than that of the AWG,
a multiple-wavelength signal is generated on the remaining input port,
the components of this signal being determined by the characteristics of
spectral response of the AWG.
This configuration is shown in Figure 13, in which a wide-band
transmitter T is connected to an input port of an NxN AWG 2 whose N
output ports U,...UN are connected in the previously described foldback
configuration, forming N/2 branches B,...BN,z. In each of the branches,
according to the present invention, there is a signal processor device
P,...P,~2 which comprises an attenuator or a filter with a spectral
response such that the wavelengths necessary for the branch in
question can be propagated and with insertion losses calibrated in such
a way that the suppression of the crosstalk contributions is possible in
addition to the equalization of the losses. It is therefore possible to
generate a multiple-wavelength signal by using an AWG in foldback
configuration and a wide-band light source.
Additionally, if a modulator which superimposes a data element at
radio frequency on the signal circulating in the branch is connected in
an advantageous way in each foldback branch B; or alternatively in
some branches only, this will produce a multiple-wavelength signal
transmitter for a telecommunications line. The said modulator is an
external modulator, in other words a modulator in which the information
is superimposed on a pre-existing optical signal, in this case the signal
circulating in the branch. A modulator of a known type suitable for this
purpose is, for example, an electro-optical modulator, in particular a
modulator of the Mach-Zehnder type, or alternatively an electro-
absorption modulator or an acousto-optical modulator.
