U.S. patent application number 16/405773 was filed with the patent office on 2020-11-12 for method and apparatus for loss-aware optical routing.
This patent application is currently assigned to HUAWEI TECHNOLOGIES CO., LTD.. The applicant listed for this patent is Hamid MEHRVAR. Invention is credited to Hamid MEHRVAR.
Application Number | 20200359116 16/405773 |
Document ID | / |
Family ID | 1000004086187 |
Filed Date | 2020-11-12 |
![](/patent/app/20200359116/US20200359116A1-20201112-D00000.png)
![](/patent/app/20200359116/US20200359116A1-20201112-D00001.png)
![](/patent/app/20200359116/US20200359116A1-20201112-D00002.png)
![](/patent/app/20200359116/US20200359116A1-20201112-D00003.png)
![](/patent/app/20200359116/US20200359116A1-20201112-D00004.png)
![](/patent/app/20200359116/US20200359116A1-20201112-D00005.png)
![](/patent/app/20200359116/US20200359116A1-20201112-D00006.png)
![](/patent/app/20200359116/US20200359116A1-20201112-D00007.png)
![](/patent/app/20200359116/US20200359116A1-20201112-D00008.png)
![](/patent/app/20200359116/US20200359116A1-20201112-D00009.png)
![](/patent/app/20200359116/US20200359116A1-20201112-D00010.png)
View All Diagrams
United States Patent
Application |
20200359116 |
Kind Code |
A1 |
MEHRVAR; Hamid |
November 12, 2020 |
METHOD AND APPARATUS FOR LOSS-AWARE OPTICAL ROUTING
Abstract
A method and apparatus for operating an optical switching fabric
or other optical device are provided. The device has multiple input
and output ports to be selectably connected together via optical
paths. For a requested configuration, an optical device
configuration is determined based on a loss metric, which is based
on one or both of: a number of crossings of the optical paths; and
a length of the optical paths. The crossings can be waveguide
crossings within the switching fabric. The configuration can be
obtained by selecting particular intermediate stages of the
switching fabric for carrying particular optical paths. The number
of waveguide crossings, or the variation in the number of waveguide
crossings, can be limited or minimized in the selected
configuration. In one embodiment, an initial solution is
determined, and the intermediate stages of the switch are
re-ordered to obtain an improved solution in terms of the loss
metric.
Inventors: |
MEHRVAR; Hamid; (Ottawa,
CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
MEHRVAR; Hamid |
Ottawa |
|
CA |
|
|
Assignee: |
HUAWEI TECHNOLOGIES CO.,
LTD.
Shenzhen
CN
|
Family ID: |
1000004086187 |
Appl. No.: |
16/405773 |
Filed: |
May 7, 2019 |
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
H04Q 2011/0026 20130101;
H04Q 2011/0073 20130101; H04J 14/0212 20130101; H04Q 11/0062
20130101; H04Q 11/0005 20130101 |
International
Class: |
H04Q 11/00 20060101
H04Q011/00; H04J 14/02 20060101 H04J014/02 |
Claims
1. A method for operating an optical switching fabric having a
plurality of input ports, a plurality of output ports, and a
plurality of photonic elements operable to selectably optically
couple the input ports to the output ports, the method comprising:
receiving an indication to optically couple each designated one of
a set of the input ports to a designated one of a set of the output
ports; and in response to the indication and by operating the
photonic elements, implementing an optical switch configuration
which establishes a set of optical paths accommodating the
indication, the optical switch configuration determined based at
least in part on a loss metric which is based on one or both of: a
number of waveguide crossings in each of the optical paths; and a
length of each of the optical paths.
2. The method of claim 1, wherein the optical switching fabric
comprises an input stage, an output stage, and a plurality of
intermediate stages, the input stage configured to optically couple
selected ones of the input ports to selected ones of the
intermediate stages, and the output stage configured to optically
couple selected ones of the intermediate stages to selected ones of
the output ports.
3. The method of claim 2, wherein the intermediate stages are
interchangeable with one another, and wherein determining the
optical switch configuration comprises: establishing a
representation of the optical switching fabric, the representation
comprising a plurality of logical intermediate stages
representative of the plurality of intermediate stages; determining
a switch configuration solution indicative of a set of paths in the
representation of the optical switching fabric, the set of paths
accommodating the indication; and obtaining the optical switch
configuration from the switch configuration solution by mapping the
logical intermediate stages to the intermediate stages, the mapping
based at least in part on the loss metric, wherein each one of the
intermediate stages is configured in the same manner as a logical
intermediate stage which is associated with said one of the
intermediate stages via said mapping.
4. The method of claim 2, wherein determining the optical switch
configuration comprises: determining two or more alternative
optical switch configurations, each of the alternative optical
switch configurations establishing a corresponding set of optical
paths accommodating the indication, and the alternative optical
switch configurations differing from one another with respect to a
re-ordering of the plurality of intermediate stages; and selecting
one of the alternative optical switch configurations based at least
in part on the loss metric.
5. The method of claim 2, wherein the optical switch comprises more
than two intermediate stages.
6. The method of claim 2, wherein each of the intermediate stages
comprises a further plurality of sub-stages in a nested
configuration.
7. The method of claim 6, further comprising recursively applying
the method to the intermediate stages, the sub-stages, or a
combination thereof.
8. The method of claim 1, wherein the optical switch is configured
to switchably route optical communications signals or optical
signals between components of a computing device.
9. The method of claim 1, wherein the optical switch is a
re-arrangably non-blocking optical switch.
10. The method of claim 1, wherein the loss metric is based on
numbers of waveguide crossings in each of the optical paths and is
independent of said length of each of the optical paths.
11. The method of claim 1, wherein the loss metric is based on one
of: a difference between a maximum value, among all of the optical
paths, of a per-path loss metric and a minimum value, among all of
the optical paths, of the per-path loss metric; an amount of
variation, among all of the optical paths, of the per-path loss
metric; a maximum value, among all of the optical paths, of the
per-path loss metric; and an average value, among all of the
optical paths, of the per-path loss metric, wherein the per-path
loss metric is applied to a given optical path and is an increasing
function of one or both of: a number of waveguide crossings in the
given optical path; and a length of the given optical path.
12. The method of claim 1, wherein the optical switch configuration
is further determined as a non-blocking configuration.
13. An optical switch comprising: an optical switching fabric
having a plurality of input ports, a plurality of output ports, and
a plurality of photonic elements operable to selectably optically
couple the input ports to the output ports; and a controller
configured to switchably operate the plurality of photonic elements
in response to input indicative of a desired state of the optical
switch, the controller configured to: receive an input to optically
couple each designated one of a set of the input ports to a
designated one of a set of the output ports; and in response to the
indication and by operating the photonic elements, implement an
optical switch configuration which establishes a set of optical
paths accommodating the indication, the optical switch
configuration determined based at least in part on a loss metric
which is based on one or both of: a number of waveguide crossings
in each of the optical paths; and a length of each of the optical
paths.
14. The optical switch of claim 13, wherein the optical switching
fabric comprises an input stage, an output stage, and a plurality
of intermediate stages, the input stage configured to optically
couple selected ones of the input ports to selected ones of the
intermediate stages, and the output stage configured to optically
couple selected ones of the intermediate stages to selected ones of
the output ports, wherein the intermediate stages are
interchangeable with one another, and wherein determining the
optical switch configuration comprises: establishing a
representation of the optical switching fabric, the representation
comprising a plurality of logical intermediate stages
representative of the plurality of intermediate stages; determining
a switch configuration solution indicative of a set of paths in the
representation of the optical switching fabric, the set of paths
accommodating the indication; and obtaining the optical switch
configuration from the switch configuration solution by mapping the
logical intermediate stages to the intermediate stages, the mapping
based at least in part on the loss metric, wherein each one of the
intermediate stages is configured in the same manner as a logical
intermediate stage which is associated with said one of the
intermediate stages via said mapping.
15. The optical switch of claim 13, wherein the optical switching
fabric comprises an input stage, an output stage, and a plurality
of intermediate stages, the input stage configured to optically
couple selected ones of the input ports to selected ones of the
intermediate stages, and the output stage configured to optically
couple selected ones of the intermediate stages to selected ones of
the output ports, and wherein determining the optical switch
configuration comprises: determining two or more alternative
optical switch configurations, each of the alternative optical
switch configurations establishing a corresponding set of optical
paths accommodating the indication, and the alternative optical
switch configurations differing from one another with respect to a
re-ordering of the plurality of intermediate stages; and selecting
one of the alternative optical switch configurations based at least
in part on the loss metric.
16. The optical switch of claim 13, wherein the loss metric is
based on numbers of waveguide crossings in each of the optical
paths and is independent of said length of each of the optical
paths.
17. The optical switch of claim 13, wherein the loss metric is
based on one of: a difference between a maximum value, among all of
the optical paths, of a per-path loss metric and a minimum value,
among all of the optical paths, of the per-path loss metric; an
amount of variation, among all of the optical paths, of the
per-path loss metric; a maximum value, among all of the optical
paths, of the per-path loss metric; and an average value, among all
of the optical paths, of the per-path loss metric, wherein the
per-path loss metric is applied to a given optical path and is an
increasing function of one or both of: a number of waveguide
crossings in the given optical path; and a length of the given
optical path.
18. The optical switch of claim 13, wherein the optical switch
configuration is further determined as a non-blocking
configuration.
19. A method for operating an optical device having a plurality of
input ports, a plurality of output ports, and a plurality of
elements operable to selectably optically couple the input ports to
the output ports, the method comprising: receiving an indication to
optically couple each designated one of a set of the input ports to
a designated one of a set of the output ports; and in response to
the indication and by operating the photonic elements, implementing
an optical device configuration which establishes a set of optical
paths accommodating the indication, the optical device
configuration determined based at least in part on a loss metric
which is based on one or both of: a number of crossings or near
crossings in each of the optical paths; and a length of each of the
optical paths.
20. The method of claim 19, wherein the optical device is an
optical switch having a plurality of switching cells, or a
wavelength selective switch.
21. An optical device comprising: an optical path component having
a plurality of input ports, a plurality of output ports, and a
plurality of elements operable to selectably optically couple the
input ports to the output ports; and a controller configured to
switchably operate the plurality of elements in response to input
indicative of a desired state of the optical path component, the
controller configured to: receive an input to optically couple each
designated one of a set of the input ports to a designated one of a
set of the output ports; and in response to the indication and by
operating the photonic elements, implement an optical path
component configuration which establishes a set of optical paths
accommodating the indication, the optical path component
configuration determined based at least in part on a loss metric
which is based on one or both of: a number of crossings or near
crossings in each of the optical paths; and a length of each of the
optical paths.
22. The optical device of claim 21, wherein the optical device is
an optical switch having a plurality of switching cells, or a
wavelength selective switch.
Description
FIELD OF THE INVENTION
[0001] The present invention pertains to the field of optical
devices and in particular to a method and apparatus for optical
routing in a device such as but not necessarily limited to an
optical switching fabric having a re-arrangably non-blocking
characteristic, the routing accounting for loss characteristics
such as number of optical path crossings or optical path
lengths.
BACKGROUND
[0002] An N.times.N silicon photonic switch includes multiple
interconnected switching elements, such as 2.times.2 Mach-Zehnder
switching element, arranged into multiple stages. The elements of
each stage are interconnected to the previous or next stage by
waveguides. Currently, silicon photonic switches are primarily
implemented in a single-layer lithographic structure. As such, the
inter-connecting waveguides must either cross each other in the
same plane or be vertically separated by a small distance. Such
waveguide crossings contribute to insertion loss of the switch.
This is because the loss from an input of the switch to an output
of the switch depends on input/output coupling loss, number of
switching elements in the switch's internal connection path, total
number of waveguide crossings along the connection path, and total
waveguide length of the connection path. Different paths between
the same switch input and output, as well as between different
switch inputs and outputs, can have different total losses. This
leads to variability in the switch's insertion loss. Such
variability is undesirable as it leads to different and possibly
unpredictable loss characteristics for different switch
configurations. This can lead to implementations that factor in a
`worst-case` insertion loss for the switch, which is
sub-optimal.
[0003] It is further considered that other devices, such as
wavelength selective switches (WSS) can also suffer from similar
insertion loss variability, due to confined or unconfined optical
paths crossing closely together or being variable in length.
[0004] Therefore, there is a need for a method and apparatus for
optical path routing, in an optical switching fabric or other
device, that mitigates insertion loss variability, or that
otherwise obviates or mitigates one or more limitations of the
prior art.
[0005] This background information is provided to reveal
information believed by the applicant to be of possible relevance
to the present invention. No admission is necessarily intended, nor
should be construed, that any of the preceding information
constitutes prior art against the present invention.
SUMMARY
[0006] An object of embodiments of the present disclosure is to
provide a method and apparatus for loss-aware optical routing. In
accordance with embodiments of the present disclosure, a method for
operating an optical switching fabric is provided. The optical
switching fabric has a plurality of input ports, a plurality of
output ports, and a plurality of photonic elements operable to
selectably optically couple the input ports to the output ports.
The method includes receiving an indication to optically couple
each designated one of a set of the input ports to a designated one
of a set of the output ports. The method includes, in response to
the indication, implement an optical switch configuration which
establishes a set of optical paths accommodating the indication.
The optical switch configuration is determined based at least in
part on a loss metric. The loss metric in turn is based on one or
both of: a number of waveguide crossings in each of the optical
paths; and a length of each of the optical paths.
[0007] In accordance with embodiments of the present disclosure, a
method for operating an optical device is provided. The optical
device has a plurality of input ports, a plurality of output ports,
and a plurality of elements operable to selectably optically couple
the input ports to the output ports, via confined or unconfined
optical paths. The method includes receiving an indication to
optically couple each designated one of a set of the input ports to
a designated one of a set of the output ports. The method includes,
in response to the indication, implementing an optical device
configuration which establishes a set of optical paths
accommodating the indication. The optical device configuration is
determined based at least in part on a loss metric. The loss metric
in turn is based on one or both of: a number of crossings or near
crossings in each of the optical paths; and a length of each of the
optical paths.
[0008] In accordance with embodiments of the present disclosure, an
optical switch (or at least the controller thereof) is provided.
The optical switch includes an optical switching fabric having a
plurality of input ports, a plurality of output ports, and a
plurality of photonic elements operable to selectably optically
couple the input ports to the output ports. The controller may
include a computer processor executing instructions stored in
memory, or other suitable electronics for determining and
implementing switch states, and switch driving electronics for
driving the photonic elements accordingly. The controller is
configured to switchably operate the plurality of photonic elements
in response to input indicative of a desired state of the optical
switch. The controller is further configured to receive an input to
optically couple each designated one of a set of the input ports to
a designated one of a set of the output ports. The controller is
further configured, in response to the indication, to implement an
optical switch configuration which establishes a set of optical
paths accommodating the indication. The optical switch
configuration is determined, by the controller, based at least in
part on a loss metric. The loss metric is based on one or both of:
a number of waveguide crossings in each of the optical paths; and a
length of each of the optical paths.
[0009] In accordance with embodiments of the present disclosure, an
optical device (or at least the controller thereof) is provided.
The optical device may be an optical switch as above, or a
wavelength selective switch, or another optical device. The optical
device includes an optical path component having a plurality of
input ports, a plurality of output ports, and a plurality of
elements operable to selectably optically couple the input ports to
the output ports. The controller may include a computer processor
executing instructions stored in memory, or other suitable
electronics for determining and implementing device states, and
device driving electronics for driving the device's elements
accordingly. The controller is configured to switchably operate the
plurality of elements in response to input indicative of a desired
state of the optical path component. The controller is further
configured to receive an input to optically couple each designated
one of a set of the input ports to a designated one of a set of the
output ports. The controller is further configured, in response to
the indication, to implement an optical path component
configuration which establishes a set of optical paths
accommodating the indication. The optical path component
configuration is determined, by the controller, based at least in
part on a loss metric. The loss metric is based on one or both of:
a number of crossings or near crossings in each of the optical
paths; and a length of each of the optical paths.
BRIEF DESCRIPTION OF THE FIGURES
[0010] Further features and advantages of the present disclosure
will become apparent from the following detailed description, taken
in combination with the appended drawings, in which:
[0011] FIG. 1A illustrates an example multiplane optical switching
fabric provided according to or operable with embodiments of the
present disclosure.
[0012] FIG. 1B illustrates a representative waveguide crossing of a
switching fabric.
[0013] FIG. 2 illustrates example optical paths through the optical
switching fabric of FIG. 1A.
[0014] FIG. 3 illustrates a method for operating an optical
switching fabric, according to embodiments of the present
disclosure.
[0015] FIG. 4A illustrates a first example configuration for the
optical switching fabric of FIG. 1A.
[0016] FIG. 4B illustrates a second example configuration for the
optical switching fabric of FIG. 1A, satisfying the same
configuration requirements as that of FIG. 4A.
[0017] FIG. 5 illustrates a method for determining an optical
switch configuration by re-ordering intermediate stages, according
to embodiments of the present disclosure.
[0018] FIG. 6 illustrates a method for determining an optical
switch configuration by mapping logical intermediate stages to
physical intermediate stages, according to embodiments of the
present disclosure.
[0019] FIG. 7A illustrates performance improvement for the
switching fabric of FIG. 1A, obtained for embodiments of the
present disclosure via simulation.
[0020] FIG. 7B illustrates performance improvement for the
switching fabric of FIG. 1A for a different loss metric, obtained
for embodiments of the present disclosure via simulation.
[0021] FIG. 8 illustrates an optical switching fabric incorporating
4 intermediate stages each comprising a copy of the switching
fabric of FIG. 1A.
[0022] FIG. 9 illustrates a method for determining optical paths in
a nested or non-nested optical switching fabric, according to an
embodiment of the present disclosure.
[0023] FIG. 10 illustrates another switching fabric provided
according to or operable with embodiments of the present
disclosure.
[0024] FIG. 11 illustrates a photonic system for computing and
datacenter usage, in which an optical switch may be operated
according to embodiments of the present disclosure.
[0025] FIG. 12 illustrates an optical device such as a wavelength
selective switch, which can be provided and operated according to
embodiments of the present disclosure.
[0026] FIG. 13 illustrates an apparatus for controlling an optical
device such as an optical switching fabric, provided in accordance
with embodiments of the present disclosure.
[0027] It will be noted that throughout the appended drawings, like
features are identified by like reference numerals.
DETAILED DESCRIPTION
[0028] Embodiments of the disclosure account for numbers of optical
path crossings and/or optical path lengths during determination of
an optical path routing solution in an optical device. A
requirement to optically couple one or more designated input ports
to one or more corresponding designated output ports can be
satisfied by multiple different routing solutions. Out of the
multiple routing solutions, one selected solution is implemented
which exhibits a limited or minimum contribution of optical loss
due to path crossings, path lengths, or a combination thereof. For
example, the effect of waveguide crossing loss in the connection
insertion loss can be minimized. Thus, routing such as non-blocking
routing in an optical switch is performed with the additional step
of limiting or minimizing a loss metric.
[0029] The route determination process of embodiments of the
present disclosure can account for optical path crossing loss,
optical path length, or both. Furthermore, the route determination
process can attempt to limit or minimize such loss contributions,
as measured by an appropriate loss metric. This tends to cause
reduced variation in these loss contributions among all the
multiple connections of the optical device. The route determination
process can attempt to limit or minimize loss along each path being
implemented, which indirectly tends to limit or minimize loss
variability among all paths. The route determination process can
alternatively attempt to directly limit or minimize loss variation
among all paths, for example by minimizing a difference between
maximum and minimum losses over all paths. The route determination
process can alternatively attempt to minimize a maximum path loss
or an average path loss.
[0030] Embodiments of the disclosure are applicable to a partially
or fully loaded optical switching fabric (connection matrix) or
other optical (photonic) device optically connecting input ports to
output ports. In particular, embodiments of the present disclosure
limit or minimize variations in the optical device's insertion
loss, in particular due to factors such as number of optical
crossings and/or optical path length. Such variation arises
inherently due to the different number of waveguide crossings along
different paths through the optical device.
[0031] Embodiments of the disclosure can be implemented without
impacting the connection blocking performance of the optical
device. Non-blocking routing is performed, and the non-blocking
routes are then re-arranged to minimize or limit certain losses.
Path-finding (routing) time and crosstalk performance may also be
unimpacted or minimally impacted. The method and apparatus are
applicable to a variety of optical switch architectures, such as
multi-plane (e.g. modular quad as in FIG. 1A) and recursively
dual-plane architectures (e.g. Benes, Dilated Benes, Hybrid Dilated
Banyan, and Hybrid Dilated Benes Enhanced (HDBE)). The present
disclosure was purposefully developed so as not to impact or
minimally impact these performance aspects.
[0032] FIG. 1A illustrates an example multiplane switching fabric
100 which may be the subject of certain embodiments of the present
disclosure. This switching fabric is described in more detail in H.
Mehrvar and E. Bernier, "Fast Photonic Switch Architecture for
Interconnect Applications," European Conference on Optical
Communication, September 2018, which is incorporated herein by
reference. Other applicable switching fabrics include dilated
Banyan, Benes, and Hybrid Dilated Benes.
[0033] Multiplane switching fabrics can be characterized in that
they include an input stage, an output stage, and multiple
intermediate stages. The input stage is configured to optically
couple selected ones of the input ports to selected ones of the
intermediate stages, and the output stage is configured to
optically couple selected ones of the intermediate stages to
selected ones of the output ports. The use of multiple intermediate
stages improves the non-blocking characteristic of the switch. It
is noted that the intermediate stages, being coupled to all the
ports of the input and output stages, and being structurally
similar, functionally similar, or both, are substantially
interchangeable. That is, given a routing solution through the
switching fabric, a different routing solution can be obtained by
interchanging the roles of the intermediate stages.
[0034] For example, in a routing solution, a first one or more
optical paths pass through a first intermediate stage, and a second
one or more optical paths pass through a second intermediate stage.
The different routing solution can be obtained by re-ordering the
intermediate stages, for example including causing the first one or
more optical paths to instead pass through the second (or another)
intermediate stage, and causing the second one or more optical
paths to instead pass through the first (or yet another)
intermediate stage.
[0035] The switching fabric 100 of FIG. 1A, as well as a variety of
other switching fabrics, includes a plurality of input ports 105
and a plurality of output ports 140. The input and output ports can
be connected to external optical carriers such as optical fiber or
waveguides. Internally, the switching fabric 100 comprises a
plurality of controllable switching cells 110, which include
two-input, two-output (2.times.2) switching cells, and typically
also one-input, two-output (1.times.2) and two-input, one-output
(2.times.1) switching cells. The switching cells are examples of
photonic elements. Switching cells can be designed and operated
based on Mach-Zehnder interferometry principles, for example, as
would be readily understood by a worker skilled in the art. The
switching cells 110 are interconnected via waveguides 115 in a
particular manner, resulting in a particular switch topology. In
operation, light enters an input port and is routed through a
selected path by operation of the switching cells to an output
port. The path includes a subset of the switching cells 110 and
interconnecting waveguides 115. Thus, input ports are selectably
optically coupled to output ports. Also illustrated are the input
stage 120, the output stage 130, and a plurality of intermediate
stages 135a, 135b, 135c, 135d.
[0036] FIG. 1A (and similar figures disclosed herein) corresponds
to a physical layout of the switching fabric, which is
substantially located in a single plane. Notably, in the optical
switching fabric 100, the waveguides 115 cross one another at
multiple locations as is evident from the figure. Several selected
waveguide crossings 145 are highlighted for clarity.
[0037] FIG. 1B illustrates a representative waveguide crossing 145
in more detail, as it appears in some embodiments. As illustrated,
the waveguide crossing 145 comprises two coplanar waveguides which
cross one another. The crossing can be configured for example as
described in Yangjin Ma et al., "Ultralow loss single layer
submicron silicon waveguide crossing for SOI optical interconnect,"
Opt. Express 21, 29374-29382 (2013). In this example, the region of
the crossing, the waveguides are tapered according to a certain
predetermined shape which mitigates crosstalk. However, crosstalk
is not altogether eliminated in such crossings. As such, the
present invention can be implemented to further mitigate crosstalk.
Other approaches to configuring waveguide crossings can also be
implemented, for example as described in the references in the
above-cited document. In some embodiments, the waveguide crossing
can include two waveguides crossing each other in two planes, one
of which is layered overtop of the other.
[0038] FIG. 2 illustrates, by way of further example, the optical
switch of FIG. 1A along with several possible optical paths 205,
210, 215. The optical paths 205 and 210 are two alternative paths
interconnecting the same input port with the same output port. The
optical path 215 interconnects a different input port with a
different output port. Each optical path exhibits a different
number of waveguide crossings, for example path 205 exhibits about
59 waveguide crossings, path 210 exhibits about 55 waveguide
crossings, and path 215 exhibits about 33 waveguide crossings. Each
optical path 205, 210, 215 also exhibits a different total length,
measured in terms of the length of the interconnecting waveguide
sections along the path.
[0039] Accordingly, it can be seen that the number of waveguide
crossings and total optical path length is significantly variable
both for different optical paths between the same two input/output
ports, and for different optical paths between different
input/output ports. When multiple optical path alternatives are
available, embodiments of the present disclosure tend to select
paths or sets of paths with fewer waveguide crossings and/or with
lower overall path lengths. It is noted that typically multiple
optical paths are routed at once, rather than one path being routed
at a time.
[0040] In view of FIGS. 1A and 2, embodiments of the present
disclosure provide a method 300 for operating an optical switching
fabric as illustrated in FIG. 3. As discussed above, the optical
switching fabric has a plurality of input ports, a plurality of
output ports, and a plurality of photonic elements (e.g.
Mach-Zehnder type switching cells) operable to selectably optically
couple the input ports to the output ports. The method includes
receiving 310 an indication to optically couple each designated one
of a set of the input ports to a designated one of a set of the
output ports. The indication may be provided in the form of a
routing instruction or request, indicating port connection
requirements at a given time. The method further includes, in
response to the indication, implementing 320 an optical switch
configuration which establishes a set of optical paths
accommodating the indication. The optical switch configuration is
determined based at least in part on a loss metric which is based
on one or both of: a number of waveguide crossings in each of the
optical paths; and a length of each of the optical paths. The
method may further comprise determining 315 the optical switch
configuration.
[0041] FIG. 4A illustrates a first example configuration (routing
map) for the optical switching fabric of FIG. 1A. The configuration
illustrates a switch configuration which satisfies a requirement to
couple input ports to output ports according to the first two
columns of Table 1. Each row of Table 1 gives an input port/output
port pair to be coupled together according to the requirement.
Different optical paths in FIGS. 4A and 4B connecting pairs of
input ports to output ports as in Table 1 are illustrated using
differently patterned lines for the waveguides involved. Switching
cells along the same optical path are designated with the same
reference letters. Table 1 also shows, for the configuration of
FIG. 4A, the intermediate stage which the optical path coupling the
corresponding input port/output port pair passes through. Table 1
also shows, for the configuration of FIG. 4A, the number of
waveguide crossings in the optical path coupling the corresponding
input port/output port. The per-path loss metric in this example is
the number of waveguide crossings along the path.
TABLE-US-00001 TABLE 1 Input Output Intermediate Stage Waveguide
Crossings 1 6 135c 57 2 3 135d 81 3 16 135b 47 4 11 135a 39 5 7
135d 60 6 14 135c 44 7 8 135b 43 8 5 135a 33 9 15 135d 24 10 1 135c
53 11 2 135b 41 12 4 135a 42 13 13 135a 72 14 9 135c 41 15 10 135d
21 16 12 135b 56
[0042] As can be seen from Table 1, the highest number of waveguide
crossings is 81, while the lowest number of waveguide crossings is
21. This gives 81-21=60 as the difference between the maximum
number of waveguide crossings (among all of the optical paths) and
the minimum number of waveguide crossings (among all of the optical
paths).
[0043] FIG. 4B illustrates a different example configuration
satisfying the same configuration requirements as that of FIG. 4A.
Table 2 shows, for the configuration of FIG. 4B, the intermediate
stage which the optical path coupling the corresponding input
port/output port pair passes through. Table 2 also shows, for the
configuration of FIG. 4B, the number of waveguide crossings in the
optical path coupling the corresponding input port/output port.
TABLE-US-00002 TABLE 2 Input Output Intermediate Stage Waveguide
Crossings 1 6 135a 15 2 3 135c 59 3 16 135b 47 4 11 135d 51 5 7
135c 52 6 14 135a 54 7 8 135b 43 8 5 135d 57 9 15 135c 40 10 1 135a
27 11 2 135b 41 12 4 135d 48 13 13 135d 18 14 9 135a 63 15 10 135c
39 16 12 135b 56
[0044] As can be seen from Table 2, the highest number of waveguide
crossings is 63, while the lowest number of waveguide crossings is
15. This gives 63-15=48 as the difference between the maximum
number of waveguide crossings (among all of the optical paths) and
the minimum number of waveguide crossings (among all of the optical
paths). Furthermore, the maximum number of waveguide crossings has
been reduced by 18 relative to Table 1. Thus, the configuration of
FIG. 4B satisfies the same input/output port coupling requirement
but with a lower overall loss metric, which is taken in the present
example to be the difference between the maximum number of
waveguide crossings and the minimum number of waveguide crossings,
or which can alternatively be taken to be the maximum number of
waveguide crossings among all optical paths. Therefore, according
to embodiments of the present disclosure, the configuration of FIG.
4B is preferred over that of FIG. 4A. It is noted that blocking and
crosstalk performance are substantially the same for FIGS. 4A and
4B. The configuration of FIG. 4B may also result when the overall
loss metric is the maximum number of waveguide crossings taken over
all paths in the switching fabric.
[0045] In fact, when the per-path loss metric is taken as the
number of waveguide crossings along a given optical path, and the
overall loss metric is taken as the difference between the maximum
value, among all of the optical paths, of the per-path loss metric
and the minimum value, among all of the optical paths, of the
per-path loss metric, and when the loss is considered as the worst
case loss of a connection, it can be shown that FIG. 4B illustrates
the optical switch configuration with the smallest overall loss
metric that satisfies the given input port/output port coupling
requirement. As such, this configuration is the most preferred
alternative and can be implemented in response to an indication to
optically couple input ports to output ports according to the given
input port/output port coupling requirement.
[0046] The configuration of FIG. 4B can be obtained from the
configuration of FIG. 4A by re-ordering (interchanging) the
intermediate stages. In particular, to obtain the FIG. 4B
configuration from that of FIG. 4A, stage 135a of FIG. 4B is
configured as stage 135c of FIG. 4A, stage 135b of FIG. 4B is
configured as stage 135b of FIG. 4A, stage 135c of FIG. 4B is
configured as stage 135d of FIG. 4A, and stage 135d of FIG. 4B is
configured as stage 135a of FIG. 4A. For clarity, "configuring a
first (intermediate) stage as a second (intermediate) stage"
comprises internally configuring the switching elements of the
first intermediate stage in the same manner as those of the second
intermediate stage, in order to create the same set of optical path
(internally to the intermediate stage). It additionally comprises
configuring the input and output stages so that, whenever an input
port and an output port of the switch were routed to a given input
and output of the second intermediate stage, those same input and
output ports are instead routed to the same corresponding input and
output of the first intermediate stage.
[0047] FIG. 5 generally illustrates a corresponding method 500 for
determining an optical switch configuration by re-ordering
intermediate stages. The re-ordering can be done in a
representation of the optical switching fabric stored in computer
memory before implementation, or the re-ordering can be done by
actually physically reconfiguring the optical switching fabric,
although this second option may result in unnecessary additional
switching. The method 500 includes determining 510 two or more
alternative optical switch configurations. Each of the alternative
optical switch configurations establishes a corresponding set of
optical paths accommodating the indication. In various embodiments,
the alternative optical switch configurations differ from one
another with respect to a re-ordering of the plurality of
intermediate stages of the optical switch. This may be the sole
difference between alternative configurations. Some or all possible
such alternative configurations can be determined, for example by
generating an initial seed configuration and obtaining alternative
configurations by performing a re-ordering. The method further
includes selecting 520 one of the alternative optical switch
configurations based at least in part on the loss metric. For
example, a comparison of the loss metric applied to each of the
alternative optical switch configurations may be made, and the
configuration with the most favorable (e.g. lowest-valued) loss
metric may be selected. The loss metric may be determined for all
alternative configurations or for selected ones of the alternative
configurations. Various exhaustive or non-exhaustive computational
approaches can be used to select a globally optimal, locally
optimal, near-optimal, or adequately optimal switch configuration.
The method further includes implementing 530 the selected
configuration by driving the optical switch appropriately.
[0048] In some embodiments, the use of logical intermediate stages
(also referred to as logical planes) is introduced instead of, or
to facilitate, the process of re-ordering intermediate stages. The
intermediate stages are presumed for ease of exposition to be
substantially interchangeable, for example due to similar or
identical structure or function. As shown in a corresponding method
600 illustrated in FIG. 6, a representation of the optical
switching fabric is established 610, for example as a model of the
switch held by a computer, such as a computer integrated into or
otherwise coupled to the switch controller. The representation
includes a plurality of logical intermediate stages, representative
of, and equal in number to, the intermediate stages of the
switching fabric.
[0049] The computer then determines 620 a switch configuration
solution indicative of a set of paths in the representation of the
optical switching fabric. The set of paths accommodates a given
input port/output port coupling requirement. The coupling
requirement is also referred to as an indication to optically
couple each designated one of a set of the input ports to a
designated one of a set of the output ports. The switch
configuration solution can be indicative of the set of paths, a
configuration of switching cells resulting in the set of paths, or
both. The switch configuration solution can represent a set of
paths that satisfies other criteria. For example, the switch
configuration solution can represent a set of non-blocking paths
through the switching fabric, which exhibit limited or minimal
crosstalk.
[0050] Next, the optical switch configuration is obtained 630 from
the switch configuration solution. This includes mapping 632 the
logical intermediate stages to the actual intermediate stages of
the switch. The mapping is based at least in part on the loss
metric, for example by performing the mapping in a manner which
substantially minimizes or limits the loss metric. The method
further includes implementing 640 the optical switch configuration
by driving the optical switch appropriately. This may include
controlling the switching cells of the switching fabric to
implement desired optical paths.
[0051] The mapping process is described in more detail as follows.
P logical intermediate stages can be associated with P physical
intermediate stages in P factorial different ways. In any given one
of these associations, also referred to as a mapping, the n.sup.th
logical stage (for n ranging from 1 to P) is associated with a
k.sup.th physical stage (for k also ranging from 1 to P), and all
logical and physical stages are so mapped in a one-to-one manner.
According to the mapping, each k.sup.th physical stages is
internally configured in the same manner as its associated n.sup.th
logical stage. The internal configuration refers to the
configuration of switching cells of the physical/logical stage.
Also according to the mapping, whenever a given input port in the
(logical) switch configuration solution is coupled to a given input
of a logical stage, the corresponding input port in the physical
switching fabric is coupled to the same input of the associated
physical stage. Similarly, whenever a given output port in the
(logical) switch configuration solution is coupled to a given
output of a logical stage, the corresponding output port in the
physical switching fabric is coupled to the same output of the
associated physical stage.
[0052] The above can be more clearly understood, by way of example,
by viewing the configuration of FIG. 4A as a (logical) switch
configuration solution and viewing the configuration of FIG. 4B as
an associated physical switch configuration. In FIG. 4A,
intermediate stages 135a, 135b, 135c, 135d are thus viewed as
logical stages, while in FIG. 4A, they are viewed as physical
stages. For purposes of discussion, the intermediate stages in FIG.
4A are labeled as followed: 135a is L1, 135b is L2, 135c is L3, and
135d is L4. The intermediate stages in FIG. 4B are labeled as
followed: 135a is P1, 135b is P2, 135c is P3, and 135d is P4. Now,
according to the mapping of the switch configuration solution of
FIG. 4A to the physical switch configuration of FIG. 4B, L1 is
associated with P4, L2 is associated with P2, L3 is associated with
P1, and L4 is associated with P3. Accordingly, L1 and P4 have the
same internal configuration and have the same inputs and outputs
coupled to the same input ports and output ports, respectively.
Similarly, L2 and P2 have the same internal configuration and
input/output couplings, as do L3 and P1, and L4 and P3.
[0053] As such, a given switch configuration solution can be
physically implemented in N factorial different ways. According to
the present disclosure, one of these different ways is selected
according to a loss metric. In particular, an implementation which
substantially minimizes or otherwise limits the loss metric is
selected. Embodiments of the disclosure may therefore implement a
pathfinding procedure including three phases. First, a set of
logical intermediate stages (planes) representing physical
intermediate stages of the optical switch are defined. Then,
non-blocking routes with least (or limited) crosstalk through the
logical planes are then determined. Then, the logical intermediate
stages are mapped to the physical intermediate stages in a manner
that minimizes or limits the loss metric (e.g. number of waveguide
crossings and/or variations therein). Determining the switch
configuration solution may comprise finding a solution comprising
non-blocking paths with minimum crosstalk, for example. Because the
pathfinding method stays substantially the same (except for a
re-ordering of intermediate planes), there is no impact on blocking
performance of the pathfinding procedure. The main difference in
the present disclosure is that, after routes are allocated (e.g. to
logical planes), an additional step maps the routes (or logical
planes) to physical planes in a manner that limits the loss
metric.
[0054] In some embodiments, for a multiplane switching fabric
having P>2 intermediate stages (P=4 in FIG. 1A) and N
connections to be routed, determining a (logical) switch
configuration solution, or indeed determining a (physical) optical
switch configuration, proceeds as follows. The value N can be equal
to the number of input ports or output ports of a switch (N=16 in
FIG. 1A), and represents the requested number of connections to be
made between input ports and output ports via routed optical paths.
First, a first set of N/P connections that can be routed through a
first (logical or physical) intermediate stage are determined.
Next, for the remaining unrouted connections, another set of N/P
connections that can be routed through a second (logical or
physical) intermediate stage are determined. This process continues
P times until all N connections are routed through all P planes.
Once this is done, further optimization of the obtained routing
solution in view of the loss metric can be performed as already
described above.
[0055] The above description of FIGS. 4A and 4B shows a performance
improvement for a single connection map. FIG. 7A illustrates
comparable performance improvement for the same switch for two
million connection maps, obtained through simulation. The top graph
710 is a histogram illustrating the differences (delta) between the
maximum and minimum number of waveguide crossings for all of the
two million connection maps, without the step of minimizing the
number of waveguide crossings. The bottom graph 720 is a histogram
illustrating the differences (delta) between the maximum and
minimum number of waveguide crossings for all of the two million
connection maps, when the step of minimizing the number of
waveguide crossings is included. As illustrated, the delta is
significantly reduced in the bottom graph, thus illustrating a
performance improvement.
[0056] FIG. 7B illustrates performance improvement for the same
switch, over one million simulated connection maps, when the loss
metric is taken as the maximum number of waveguide crossings. The
top graph 730 is a histogram illustrating the number of waveguide
crossings for all one million connection maps, without the step of
minimizing the number of waveguide crossings. The bottom graph 740
is a histogram illustrating the number of waveguide crossings for
all one million connection maps, when the step of minimizing the
number of waveguide crossings is included. As illustrated, the
maximum number of waveguide crossings in the bottom graph 740 is
limited to a maximum of about 75, versus about 90 in the top graph
730.
[0057] Although the above description refers primarily to the
switching fabric of FIG. 1A by way of example, it should be readily
understood that the present disclosure is not limited to any
particular switching fabric architecture. In particular, switches
with more than 16 input an output ports, and more than 4
intermediate planes can be accommodated. One approach to creating
larger-scale switching fabrics is to build them by combining
multiple copies of smaller-scale switching fabrics. In this and
other cases, embodiments of the disclosure can include recursively
applying the above-described pathfinding routine for example in a
nested optimization approach.
[0058] FIG. 8 illustrates an optical switching fabric 800 having 64
input ports, 64 output ports, and incorporating 4 intermediate
stages 835a, 835b, 835c, 835d each comprising a copy of the
switching fabric of FIG. 1A. Each intermediate stage includes its
own input stage, output stage, and multiple intermediate stages, in
a nested or recursive architecture. This switching fabric, and a
32.times.32 alternative version, are also described in H. Mehrvar
and E. Bernier, "Fast Photonic Switch Architecture for Interconnect
Applications," European Conference on Optical Communication,
September 2018. The switching fabric 800 is again formed using a
plurality of switching cells coupled together using waveguides. The
switching fabric 800 includes an input stage 820 having 64 input
ports 805 and an output stage 830 having 64 output ports 840. The
input stage is operative to couple each input port to any selected
one of the intermediate stages. The output stage is operative to
couple each output port selected one of the intermediate
stages.
[0059] For the switching fabric 800, or for other switching fabrics
employing a recursive architecture, embodiments of the present
disclosure can employ optical pathfinding and optimization in a
correspondingly recursive manner.
[0060] For example, a representation of the optical switching
fabric 800 can be defined, including outer logical intermediate
stages representative of the intermediate stages 835a, 835b, 835c,
835d. The representation of the optical switching fabric further
includes inner logical intermediate stages. Each inner logical
intermediate stage is located within an outer logical intermediate
stage. That is, each of the outer logical intermediate stages
includes four inner logical intermediate stages, each having 16
inputs and 16 outputs. The inner logical intermediate stages may
reflect the same structure as the intermediate stages of FIG. 1A.
Inner logical intermediate stages 837a, 837b, 837c, 837d are
labelled, by way of example, as four of 16 such stages all
belonging to the same outer intermediate stage 835a.
[0061] Given this logical representation, which may be generated a
priori, pathfinding can proceed as follows. For each outer logical
intermediate stage, an outer-level switch configuration solution is
determined which indicates a set of paths for some (e.g. 16) of the
requested connections. The outer-level switch configuration
solution comprises four inner-level switch configuration solutions,
one for each inner logical intermediate stage, which are also
determined at this point. The inner-level switch configuration
solutions each indicate a set of paths for some (e.g. 4)
accommodating the connections required through corresponding the
outer logical intermediate stage. Furthermore, the inner logical
intermediate stages of a given outer logical intermediate stage are
re-arranged (within that outer logical intermediate stage) based at
least in part on the loss metric. After pathfinding (i.e.
determining the inner-level switch configuration solutions and
their order/arrangement) has been performed for all outer logical
intermediate stages these outer logical intermediate stages can be
mapped to the physical intermediate stages 835a, 835b, 835c, 835d
based at least in part on the loss metric. Thus, for a two-stage
recursive switch architecture such as that of switching fabric 800,
intermediate stage mapping or re-ordering based on the loss metric
may be performed twice.
[0062] In cases, such as above, where all intermediate stages (at
multiple levels) are substantially identical (and potentially also
in other cases), recursive pathfinding can proceed as follows. It
is assumed that there are multiple outermost intermediate stages,
and that each intermediate stage contains multiple further
intermediate stages, to a depth of L>1 levels. Proceeding inward
from the outermost level, switch configuration solutions can be
determined indicating sets of optical paths through a corresponding
set of intermediate stages. Furthermore, an ordering of the
intermediate stages, at one, some or all levels, can be determined
based on the loss metric. The ordering gives the mapping of logical
to physical intermediate stages. In some embodiments, for each
applicable level, the determination of ordering (based on the loss
metric) can be performed upon determining the switch configuration
solution for that level. In other embodiments, the determination of
ordering for each applicable level can be performed at a later
time, for example after switch configuration solutions for all
levels have been determined.
[0063] The pathfinding procedure can be repeated at each nested
level stage until the optical paths are fully mapped. At each
recursive level, the ordering of the intermediate stages can be
determined using a re-ordering technique, or a logical intermediate
plane technique, in which the paths are determined in view of the
predetermined loss metric.
[0064] FIG. 9 illustrates a method 900 for determining optical
paths in a nested or non-nested optical switching fabric, according
to an embodiment of the present disclosure. The method includes
starting 905 at a first (e.g. outermost) level of the switching
fabric. The method further includes routing 910 connections through
the current level of switching fabric using a non-blocking method.
For N inputs and outputs, and P intermediate stages, the
connections can be routed through N/P logical intermediate stages.
The method further includes examining 915 up to (N/P) factorial
alternative cases for mapping logical intermediate stages to
physical stages. For each alternative case, a loss metric (e.g.
waveguide crossing loss) can be determined 920. The method further
includes selecting 925 one of the alternative cases for
implementation. The above process, including steps 910, 915, 920,
925 can be repeated 930 progressively for all or selected ones of
the further nested levels of the switching fabric.
[0065] FIG. 10 illustrates another switching fabric 1000 which can
be involved in embodiments of the present disclosure. The
architecture of the switching fabric 1000 is a Benes, Dilated Benes
or Hybrid Dilated Benes Enhanced (HDBE) architecture, depending on
the configuration of the centermost 2.times.2 switching cells 1010.
In particular, when the switching cells 1010 are simple 2.times.2
(e.g. interferometric) switching cells constructed similarly to the
other 2.times.2 switching cells of the fabric, the architecture is
referred to as Dilated Benes. When the switching cells 1010 are
themselves formed from four cross-connected simple 2.times.2
switching cells, the architecture is referred to as HDBE. The
architecture is an example of a recursive dual-plane architecture.
The switching fabric 1000 exhibits at least three levels (layers)
of recursion. Three, four, or more layers of recursion may be
present. There are two outermost intermediate stages (or
sub-planes) 1035a, 1035b, each of which has two further
intermediate stages (e.g. 1037a, 1037b), each of which has two even
further intermediate stages (e.g. 1039a, 1039b), and so on. The
switching fabric 1000 has an input stage 1020 having N=32 input
ports and an output stage 1030 having N=32 output ports. More or
fewer input ports and output ports can be accommodated, for example
in a comparable architecture having more or fewer levels of
recursion, respectively.
[0066] Pathfinding in the switching fabric 1000, and in similar
architectures, can proceed generally as already described above, or
as follows. First, assuming all N input ports are to be routed to
all N output ports in a given correspondence, two sets of N/2
connections that result in a minimum for the loss metric are
determined. This may be done for example by minimizing total
optical path distance or waveguide crossing distance between pairs
of input and output ports.
[0067] For example, number the input ports and output ports from 1
to N, with each input port having the same number as its
diametrically opposite output port. If, in a given requirement to
couple a particular input port to a particular output port, the
number value for an input port plus the number value for an output
port is less than 2N, then the coupling is routed through the upper
plane, otherwise it is routed through the lower plane. In this
context, "upper plane" refers to the intermediate stage that is
physically closer to the output and output ports having the lowest
number value, and "lower plane" refers to the other of the two
intermediate stages.
[0068] Next, for each set of N/2 connections, two sets of N/4
connections that result in a minimum for the loss metric are
determined in a similar manner. This process is repeated
recursively until a full set of optical paths satisfying the given
mapping between input ports and output ports is determined.
[0069] Pathfinding for the dual-plane recursive architectures is
performed on two planes (at the same level) at a time, with
pathfinding applied to the outermost two planes first, followed by
pathfinding applied to the two planes nested within each of the
outermost planes, and so on.
[0070] In some embodiments and similarly to the above, pathfinding
may involve the following. For the N input ports and N output
ports, two sets of N/2 input ports and N/2 output ports are
determined such that the routing method can route each of the two
sets using one of the upper plane and the lower plane. Next, each
of the two sets are evaluated with respect to both to the upper
layer and the lower layer to determine the number of waveguide
crossings that would be required to route a particular set using a
particular layer. (In some embodiments this may refer only to
waveguide crossings from the ports to the layers, rather than
waveguide crossings within the layers.) This yields two alternative
values: a first value representing the total number of waveguide
crossings when the upper layer is used to route the first set and
the lower layer is used to route the second set; and a second value
representing the total number of waveguide crossings when the lower
layer is used to route the first set and the lower upper is used to
route the second set. The set-to-layer assignment yielding the
lower of the first and second values is selected. This process is
then repeated recursively for each of the upper plane and the lower
plane, viewed as its own separate switching fabric. Other routing
metrics can be substituted in the above.
[0071] The example embodiments presented above consider the number
of waveguide crossings as the per-path loss metric. However, it is
considered that the per-path loss metric can additionally or
alternatively reflect other factors. For example, the per-path loss
metric can be based at least in part on the total length of
waveguide sections (optical path length) used to couple the
switching fabric's input port and the output port for an optical
path. Crosstalk, number of switching cells, and similar
considerations can also be factored into the per-path loss metric.
In some embodiments, the per-path loss metric can be based on a
combination of factors. For example, the per-path loss metric can
be a weighted combination of the number of waveguide crossings and
the optical path length, or another increasing function of one or
both of these variables.
[0072] It is also considered that the per-path loss metric can be
different for embodiments which are applied to optical devices
other than an optical switching fabric. For example, the total
optical path length can be considered as part of the per-path loss
metric even when the optical signals are not confined within a
particular waveguide, for example in the case of certain wavelength
selective switches (WSS) or other devices in which optical signals
propagate in an unconfined medium. In such cases, the per-path loss
metric can additionally or alternatively account for the number of
crossings or near-crossings of (potentially unconfined) optical
paths, when such crossings or near-crossings result in signal
degradation.
[0073] The example embodiments presented above consider the overall
loss metric, to be minimized, as the difference between the maximum
number of waveguide crossings among all paths through the switch
and the minimum number of waveguide crossings among all paths
through the switch. However, it is considered that other overall
loss metrics may also be employed. For example, the overall loss
metric may be a difference between a maximum value, among all of
the optical paths, of a particular per-path loss metric and a
minimum value, among all of the optical paths, of the per-path loss
metric. As another example, the overall loss metric may be another
measure of variation, among all of the optical paths, of the
per-path loss metric. The measure of variation may be based for
example on a statistical characteristic such as variance or
standard deviation, or a sum of differences of per-path loss
metrics from an average value. As another example, the overall loss
metric may be a maximum value, among all of the optical paths, of
the per-path loss metric. As another example, the overall loss
metric may be an average value, among all of the optical paths, of
the per-path loss metric. Other overall loss metrics, determined
based on a combination of per-path loss metrics, can also be
employed.
[0074] The optical switch configuration is determined based on
substantially minimizing the overall loss metric. This can include
computing a global minimum for the overall loss metric, or
computing a local minimum. This can also include computing a
configuration which has a loss metric that falls below a
predetermined threshold, or a best of M candidate configurations.
The computation can be exhaustive by considering all alternative
configurations. Alternatively the computation can consider a subset
of all alternative configurations, selected randomly or according
to a given routine or heuristic. Various statistical or exhaustive
minimization routines can be employed as would be readily
understood by a worker skilled in the art. Non-exhaustive methods
may be particularly useful when the switching fabric is large.
[0075] Embodiments of the present disclosure can be used to reduce
insertion loss, variation in insertion loss, or both, for a given
optical switch or other optical device. Insertion loss can be
reduced by reducing optical path or waveguide crossings. For a
16.times.16 switching fabric, insertion loss improvement may be 0.5
dB to 1 dB in some embodiments. This can depend on the amount of
waveguide crossing loss (e.g. 0.03 dB to 0.06 dB per crossing).
Total optical path length may also be reduced. For the illustrated
64.times.64 switching fabric, insertion loss improvement may be 2
dB to 3 dB in some embodiments. This may be due to about 55 fewer
waveguide crossings (40 at a first level and 15 at a second level),
with an estimated maximum number of waveguide crossings being 210
in some embodiments.
[0076] Embodiments of the present disclosure particularly provide
for loss uniformity i.e. limiting of insertion loss variations, in
an optical device. This can improve signal uniformity and reduce
error rates in operations, particularly as traffic switches from
one port to another.
[0077] FIG. 11 illustrates a photonic system for computing and
datacenter usage. The photonic switch 1110 is used to route optical
paths between devices 1120, such as packet cards, CPU devices, and
memory devices. In particular, connection routing can proceed as
described herein. The connection routing is determined by a
controller 1130 and implemented using switch drivers 1140.
[0078] FIG. 12 illustrates an optical device 1200 such as a
wavelength selective switch, which can be provided and operated
according to embodiments of the present disclosure. The optical
device has one or multiple input ports 1210 and multiple output
ports 1220, and elements 1230 operable to selectably optically
couple the input ports to the output ports. The elements can
include wavelength selective switches, optical devices such as
lenses, reflectors, filters, and splitters, or other optical
components. Interferometric devices such as 2.times.2 switching
elements can additionally or alternatively be used. Optical paths
may be confined to media such as waveguides or optical fibers, or
they may be substantially unconfined.
[0079] In the illustrated optical device 1200, by way of example,
there are one or a plurality of input ports 1205 and a plurality of
output ports 1240, for receiving and providing optical signals,
respectively. The optical device 1200 further includes elements
1230 such as diffraction gratings, lenses, beam steering arrays,
polarizers, collimators, and controllable reflective devices such
as MEMS devices. Optical signals of the same or different
wavelengths can be transmitted via physically proximate paths which
potentially interfere with each other. The optical signals can be
propagated at least partially in free space within the optical
device. Two different signals can be routed via paths which
partially overlap, for example before reaching an optical
splitter.
[0080] The optical device 1200 (or an optical switching fabric) may
be operated, by a suitable method or controller, as follows. An
indication is received to optically couple one, some or all of the
designated input ports to corresponding output ports, for example
but not necessarily in a one-to-one manner. In response to the
indication, an optical device configuration is implemented which
establishes a set of optical paths accommodating the indication.
The optical device configuration is determined based at least in
part on a loss metric which is based on one or both of: a number of
crossings or near crossings in each of the optical paths; and a
length of each of the optical paths. The configuration can be
determined as described herein, by selecting one configuration out
of a plurality of alternatives, based on a loss metric. In some
embodiments, this includes determining an initial seed
configuration, and then using symmetries and similarities in the
optical device internal structure (e.g. a plurality of intermediate
stages) to generate one or more alternative configurations. The
seed and alternative configurations are then evaluated and one with
a more (or most) favourable loss metric is selected and
implemented.
[0081] FIG. 13 illustrates an apparatus 1300 provided in accordance
with embodiments of the present disclosure. The apparatus includes
or is operatively coupled to an optical device 1350 such as an
optical switching fabric. The apparatus is configured to control
the optical device, for example so as to cause optical paths
between input ports and output ports to traverse a selected route.
The apparatus includes a controller 1310, such as a computer
processor 1315 executing program instructions stored in memory
1320, or other suitable electronics such as an application specific
integrated circuit, and a set of one or more drivers 1325 such as
optical switching cell drivers configured to put the optical device
into a desired state. The drivers may be electrical devices
configured to generate selectable voltages and/or currents suitable
for disposing components of the optical switch into corresponding
selectable configurations, for example. In one embodiment, the
drivers include a digital-to-analog converter. The apparatus
includes an interface 1330 configured to receive indications to
optically couple designated input ports to designated output ports
in the optical device. The interface may be a digital communication
interface, for example. Such indications can be received
periodically at certain switching times, for example. The apparatus
is configured to determine and implement optical device states in
the manner described herein, in response to received indications.
The apparatus, particularly its internal logic and modelling
parameters, can be customized to the optical device
architecture.
[0082] The controller 1310 can include various functional
components such as a path determiner 1322, a loss evaluator 1324
and a selector 1326. The path determiner can determine a plurality
of sets of potential optical paths satisfying requirements received
at the interface 1330. At least some of the sets of optical paths
can be determined from an initial set of optical paths by
re-arranging intermediate planes of an optical switching fabric.
The initial set of optical paths can be determined using a
non-blocking path determination algorithm. The loss evaluator 1324
can evaluate the loss metric as applied to some or all of the sets
of potential optical paths, in order to select, by the selector
1326, one set of paths having an adequately low loss metric.
[0083] Although the present invention has been described with
reference to specific features and embodiments thereof, it is
evident that various modifications and combinations can be made
thereto without departing from the invention. The specification and
drawings are, accordingly, to be regarded simply as an illustration
of the invention as defined by the appended claims, and are
contemplated to cover any and all modifications, variations,
combinations or equivalents that fall within the scope of the
present invention.
* * * * *