Receivers

Customers require the use of either:

(i) one slow (ms) tunable wavelength receiver (tunable over the entire wavelength spectrum) and one fixed wavelength receiver to monitor the status of the common control channel; or

(ii) one fast (ps) tunable wavelength receiver (tunable over the entire wavelength spectrum) which can flip back and forth between the control channel wavelength and the wavelength used to receive information on.

The first method is preferable, cost permitting. A number of tunable wavelength receivers have been proposed which are suitable for use in LamdaPON, the most promising being the Acousto-Optical Tunable Filter (AOTF)[lll to 113]. It is interesting to compare surveys of tunable optical filters made five years apart [114 to 115], and observe that most of the designs which were earlier considered promising fell by the wayside in 5 years and were replaced by better techniques.

9.4 DYNAMICALLY RECONFIGURABLE WAVELENGTH FILTERS

This is the key component required for the LambdaPON. These need to be capable of providing individual and simultaneous control over either one wavelength or a group of wavelengths in time, operate at a similar rate as that taken to set up a call, i.e. of the order 1 to 3 ms, and be capable of resolving between 60 to 100 wavelengths each operating in the region 2 Mbit/s to 620 Mbit/s.

The need to block out groups of wavelengths in time ehminates the use of straight forward DFB laser/filter technology. However, there are several filters being designed in research labs which can be applied to the problem. Typically these fall into three main categories

(a) bulk optics grating based devices (which have the capability of resolving up to

600 wavelengths).

(b) micro-optic devices (capable of resolving up to 300 wavelengths)

(c) integrated devices

The cost and manufacturing constraints eliminate the use of bulk optic and micro- optic devices whilst favouring integrated devices. An example of one promising component which could be used in the future is a waveguide grating device based on InP as shown in Figures 9.3 and 9.4. Light entering the waveguide from the left of the device and is spread out and then diffracted by the vertically walled grating before being reflected back onto an array of output waveguides. By switching on and off the controllable optical amplifier gates, required wavelengths can be either amplified or blocked. At present only 16 wavelength based devices with channel spacings of

0.8nm have been made, however plans are underway to extend this to cover a greater

number of wavelengths with similar channel spacings.

These devices do suffer from polarisation sensitivity, high chip loss (about -11 to -14

dB) and high coupling loss (total of -6dB). The use of a large number of wavelength

channels (> 100) each with around 10 mW of power will causes on-chip heat dissipation problems. Whether these problem will be overcome is unknown at present.

Other devices which also have promising prospects include the Phased Array Waveguides (the resolution of 4 wavelengths with channels spacings of 195 GHz has been possible, however the devices designed suffer from weak electrical isolation causing side-lobes), the Mach-Zehnder Interferometer (spacings of between 5 and 10 GHz have been demonstrated on 128 wavelength channels, but the device sizes are very large (about 15 inches in length)) and the Multi-Grating Tunable Notch Filter (but only a 4 channel device has so far been tested). Brief details of these components can be found in Appendix 7.

Controllable Optical Amplifier

Gates

Slab w aveguide region - input field diffracts to fill grating mirror

Output Input

Focussing Grating M irror - input field is reflected, diffracted and foccussed onto output waveguides planar silica

W aveguide arrays for input and output

Figure 9.2 Schematic Diagram of a Monolithic Wavelength Demultiplexer

Etched Grating Etched Mirror Etched Mirror A m p lifier

Laser Bar _{E lectrical}

C ontrols to filter X,, to Input and _ Output Signal Curved W a v eg u id es R eflector Laser Bar

Figure 9.3 Active Filter Version of the Wavelength Demultiplexer

9.5 RELIABILITY AND RESILIENCE ISSUES IN BRIEF

In designing the LambdaPON network, a large number of additional areas need to be investigated; too large to be covered in this thesis. However, the area of reliability and network survivability is felt to be of sufficient importance to warrant a brief mention. The loss of services in high capacity fibre systems could be devastating and could result in significant revenue loss [124]. In the access network the failure of a link or node is not as great a concern as a similar failure in the long haul trunk network. However, for certain services (e.g. the transmission of large amounts of data and Private Circuits) customers are willing to pay extra in order to guarantee service protection.

In general, bi-directional rings are considered reliable [84] [85] since they offer the ability to route signals in both the clockwise and counter-clockwise directions around the ring; thereby providing two separate paths between each node pair. Protection against link failure [124] is provided by eitherl

(i) The use of self-healing rings, or

(ii) Utilising dynamically path rearrangeable mesh architectures, or

(iii) Diverse path routing where the traffic is routed around the ring in both directions simultaneously.

The first two techniques require the use of intelligent network nodes capable of processing the signals to detect the occurrence of a fault; and hence are not suitable for the LambdaPON network which employs transparent nodes. The alternative solution is to utilise diverse path routing. However, this will inevitable effect the performance of the network and require the use of different routing and wavelength allocation algorithms. This is an area of further study not covered in this thesis.

One other failure of concern, which requires further investigation, is that which would occur should a dynamic tunable filter fail. This would cause all the wavelengths traversing through the filter to be either

(i) all totally blocked, or

(ii) all the wavelengths could pass through the filter transparently, or (iii) there could be partial blocking of the wavelengths.

In the first case, connections being routed through the filter at fault would all be broken. Whilst is not necessarily a serious problem for telephony based services, it is for large data transfers. Once again, diverse path routing techniques can be used to ensure the successful transmission of information. A failure resulting in the effects mentioned in (ii) and (iii) would result in an overlap of the information transmitted on the same wavelength pairs. So long as the interference between the signals results in white noise at the customers' receivers, confidentially can be maintained. Further work on this is required to establish the exact consequence of a filter failing

For any of the above mentioned failures, the risk is that after a failure occurs and connections in progress broken; all the users effected by the failure will transmit control packets to the Network Routing Manager at roughly the same time to reinitiate their calls. This would results in a burst of traffic all contending for the NRM, a situation best avoided. A solution needs to be built into the contention

resolution protocol to deal with this likelihood, such as the automatic reallocation of wavelength pairs to re-establish the connections lost. However, this raises the question "How does the NRM know a failure has occurred since all the routing is performed transparently" ? A possible method is to designate one wavelength which the NRM uses to reflect signals off a few selected end point terminals in the network to determine the status of the nodes and links in the network.

The Routing Manager is in essence the heart of the network. It stores all the information on the state of the network and the wavelength pairs in use. Should this part of the network fail, calls in progress could not be terminated and no new calls can be established. Hence, in a fully operable system, duplication or dual parenting of the Routing Manager would be advisable.

9.6 CONCLUSION

This chapter has presented possible designs for the Passive Routing Node, and briefly discussed the components required at the customer premises and the current development of wavelength selective filters which could eventually perform the functions required. The majority of components required by the LambdaPON either already exist in the open market or as prototypes in research laboratories. The exception to this is the dynamic tunable chaimel dropping filters. Several integrated devices capable of performing the required filter functions have been suggested, one of which has been demonstrated with greater than 100 wavelength channels (namely the Mach-Zehnder Interferometer), however its size and power loss does not make it the ideal component. One design technique based on waveguide gratings seems to offer hope.

It should be noted that the design of tunable lasers and filters is a very rapidly moving field with new innovations every few months. At present the design of WDM filters capable of selectively filtering 100 channels or greater has been hampered because the need for them has not yet been conclusively proven. Instead, the emphasis has been primarily in designing components for use in WDM system requiring only a few wavelength charmels (4 to 12 wavelengths) each operating at high bit rates used for long haul transmission.

CHAPTER 10