1. Introduction

Unicast and multicast transmission capabilities on access networks constitute a key issue for the development of a future-proof network platform being able to offer service integration. The delivery of what is called triple play service (video + voice + data) has become a requirement for nowadays access networks. However, new services like High Definition TV (HDTV) and Video on Demand (VoD) will require higher transmission capabilities that will be difficult to implement on the present infrastructure [1].

Optical Access networks present the perfect solution due to their virtually unlimited bandwidth and their high transmission range. Cost effectiveness is one of the key parameters when designing this kind of networks so the sharing of costly equipment is very important to reduce both CAPEX and OPEX. Also, bandwidth managing is worth to be considered in order not to over dimension the infrastructure and equipment [2].

We propose a passive optical network (PON) architecture that uses wavelength multiplexing (WDM) for routing and time multiplexing (TDM) for laser sharing [3, 4]. At the same time, multicast and unicast traffic are wavelength multiplexed using the free spectral range (FSR) periodicity of the Arrayed Waveguide Gratings (AWGs) that are used to connect the Exchange Center (Optical Layer Termination - OLT) with the user equipment (Optical Network Unit - ONU) [5, 6]. Two AWG stages are used to share equipment more efficiently [7]. The cost of the ONU is also critical because of the large number of them that are required and the fact that are a dedicated device for each end user. Therefore, a design based on simplicity and WDM transparency is presented, using a reflective modulator.

The main novelties of this topology are the use of the FSR to transmit multicast data and the tunable laser stack located at the OLT to send unicast data to the ONUs on TDM basis. Also, the utilization of the proposed routing technique using two cascaded AWGs to share multicast transmitters is innovative.

2. Network architecture

The network architecture is presented in Fig. 1. It is based on two cascaded AWGs. The first stage is an NxN AWG that interconnects the optical transmission equipment with N remote nodes. These remote nodes are 1xN AWGs that route 2·N wavelengths to each of the output ports, using the main pass band and the adjacent AWG FSR. The ONU has two receivers. The multicast receiver is a simple photo detector. The unicast receiver needs to also be able to modulate upstream data. Any device that can remotely modulate the carrier sent from the OLT is suitable to be used, like a Reflective Semiconductor Amplifier (RSOA) [8]. As there is no light generation at the ONU, these are wavelength independent. However, an unmodulated optical carrier needs to be sent from the OLT.

 

Fig. 1. Network Topology

Download Full Size | PDF

The OLT optical sources are a tunable laser stack and a fixed laser stack. The tunable laser stack performs unicast transmission and is shared on TDM basis among N ONUs, which are connected to one of the N remote nodes. This architecture allows dynamic bandwidth allocation as time slots can be dynamically assigned depending on the transmission requirements. The fixed laser stack is composed of N fixed lasers with wavelengths compatible with the AWG routing table, not using the main pass band but the adjacent FSR. The optical sources are coupled and modulated with a single modulator and then split to the N input ports of the NxN central AWG. Optical Amplification after the modulator is required.

This architecture shares each laser by a factor of N. Each tunable laser serves N users on TDM basis while the N lasers from the multicast laser stack serve the entire network, to which NxN users can be connected.

2.1Hybrid WDM/TDM multiplexing

Wavelength and time multiplexing techniques are used on the proposed topology. Wavelength multiplexing is used for routing purposes and to combine unicast and multicast traffic (see Fig. 2). Routing more than one single wavelength to each outside ports is possible using the FSR of AWGs. We propose the use of this technique to route unicast and multicast traffic.

The laser needs to tune the correct wavelength to reach each of the ONUs. The protocol to have access to the laser can be static or dynamic [9]. In any case, tuning speed is critical. Examples of tunable lasers switching in the range of nanoseconds have already been reported [10]. Also, there is a compromise between tuning speed, time slot wideness and network latency. A wider time slot increments network performance as fewer times the laser requires to be tuned. However, wide time slots increment network latency, which is an important parameter for real-time applications. A dynamical time assignation protocol is the best choice to balance these three parameters. We have done a numerical analysis and have found that to transmit at speeds from 2.5 to 10Gbps, tuning times should be in the range of 1 microsecond to 100 nanoseconds. Time slots of 10,000 to 100,000 bits @ 2.5Gbps (the first for low latency applications, the second for high bandwidth constant bit rate ones) offer a good balance between network throughput and latency.

 

Fig. 2. Wavelength and time multiplexing techniques

Download Full Size | PDF

2.2 Comparison with pure TDM and pure WDM PON solutions

The presented topology finds a compromise between pure TDM and pure WDM PON approaches which are described below. Note that the logical comparison in terms of costs is to compare our network with N TDM or N WDM PONs.

Optical transmission capabilities of our topology are very similar to a pure WDM PON. It offers the advantage of lower losses, when routing the different wavelengths, in comparison with the use of optical splitters. An average 40-port AWG offer insertion losses in the range of 5 to 7 dB, while the equivalent 32-port splitter has insertion losses of 16 to 17 dB. This fact is very relevant as we are using reflective ONUs without light sources and therefore optical power budget is critical. The 10 dB extra power budget turns into 20 dB if we take into account down and upstream transmission.

Another important benefit of our topology is security. As each ONU received a single and dedicated wavelength, the rest of the ONUs can not gain access to the communication between another ONU and the OLT. This is not the case in TDM PONs, as on those networks, the ONUs listen to a broadcast channel and receive the information that it is sent to all the rest of the ONUs.

In terms of network performance, the first thing to mention is that while on a pure WDM PON there is a permanent point to point connection between the OLT each ONU, on both, our hybrid network and TDM PON, transmission resources are shared among a number of users. Therefore, the network solution which offers the best network performance is the pure WDM solution, where there is a dedicated laser for each ONU.

When we compare the pure TDM solution and the hybrid network that we propose, it is important to note that network performance mainly depends on the logical protocol that it is chosen to assign the time slots to each ONU. Implementation of this protocol however, is much simpler on our proposed network as the algorithm needs just be run on the OLT. The ONUs are completely transparent to the protocol. Another advantage of our topology is bandwidth scalability. We can easily add more lasers to increase network performance while on the pure TDM PON just one laser is used for the transmission. Therefore, while on a TDM PON the theoretical bandwidth per user is BW/N, on our proposed network it is L · BW/N, where L is the number of lasers dedicated to one network segment. Also, dynamic assignation of resources is more efficient on our hybrid topology as we can concentrate bandwidth on specific network segments depending on the user needs by assigning wider time slots to the ONUs connected on a specific network segment. We call this concept Geographical Dynamic Bandwidth Allocation.

Finally, talking about costs, the solution which is more cost effective is the pure TDM approach. On the other hand, the most costly one is the WDM approach as one dedicated laser is needed for each ONU. Also, the WDM equipment located at the outside plant (AWGs) is more expensive than optical splitters. Our solution represents an intermediate point between TDM and WDM in terms of costs. The outside plant is similar to a WDM PON but on the OLT side, there is no need to have one laser for each ONU. It is true that the cost of the tunable lasers that we need for the transmission is higher than fixed WDM ones but on the other hand, the number of devices that we need is much lower.

In conclusion, when compared to N pure WDM and TDM PONs, our topology presents the following advantages: more secure, scalable and flexible than TDM PONs also offering higher network performance. At the same time, it is less costly than a pure WDM PON implementation, whereas offering good transmission capabilities. Finally, it also features the novelty of geographical dynamic bandwidth allocation.

3. Experiments and results

In order to verify the correct connectivity and performance of the proposed architecture, optical transmission tests were implemented. The network test bed can be seen in Fig. 3. An 18×18 AWG (Agilent Router Module s/n SG 03 20 B 002 W219-C1) was used as a central router and a 1×18 AWG (Agilent Router Module s/n SG 03 20 B 002 W223-C4) was used as a remote node. Both devices had channel spacing of 100GHz. Insertion losses of the 18×18 AWG were of 6dBs, while the 1×18 AWG losses were between 4 and 5dBs depending on the output port. The wavelengths used for unicast transmission were in the range between 1546.92 nm and 1560.61 nm corresponding to the main band pass of the AWG. For multicast transmission 1561.42 nm to 1575.37 nm were used. This corresponds to the first FSR. Both, the remote and central AWG had the same wavelength routing specifications.

 

Fig. 3. Optical transmission test bed

Download Full Size | PDF

Two laser sources were used to simulate unicast and multicast traffic. Those light sources were Grating-assisted codirectional Couplers with rear Sampled Reflectors (GCSR). Both were tunable, although on a real implementation, the multicast laser stack would be implemented using fixed laser sources. An additional optical splitter was added after the multicast modulator to simulate the splitting losses when distributing multicast data to all the central AWG input ports. This reduced model perfectly matches the concepts of the topology that has been presented so results can be extrapolated to a real network implementation.

 

Fig. 4. Left: BER/Pin using 30km SMF for unicast and multicast transmission. Right: 18×18 AWG FSR

Download Full Size | PDF

Both unicast and multicast data was modulated at 2.5Gbps with PRBS 2^23-1, using two external LiNbO3 modulators. Sequences were decorrelated using an electrical delay line. Both lasers offered nominal output power of 3dBm. 30Km of SMF connected the central AWG to the remote node and an optical attenuator was used after the remote node to obtain BER against Pin curves.

The ONU receiver was composed by an APD photo detector and a 2.5GHz RF amplifier. An optical Coarse WDM (C-WDM) demultiplexer was also used to recover unicast and multicast data independently.

Results for unicast and multicast reception at the ONU are depicted in Fig. 4. Sensitivity in the range of -33dBm was obtained for unicast traffic. Different wavelengths were tested to verify AWG routing consistency. Difference among wavelengths was less than 0.5dB. A sensitivity penalty of 1dB was measured when transmitting multicast traffic. The reason for this decrease on sensitivity is filtering effects at the central AWG.

4. Conclusion

A novel network architecture to transmit unicast and multicast data using the inherent free spectral range periodicity of two cascaded AWGs is presented. A shared tunable laser stack provides unicast transmission capabilities while a fixed laser stack is used for multicast transmission. Optical tests at 2.5Gbps over 30km of SMF confirm the feasibility of the proposed network platform.