Physical Level Constraints

Up until now, none of the models have incorporated the physical layer requirements of PONs. In what concerns the physical layout of the PON, the ones which have to be considered are the optical power budget, the maximum dispersion penalty and the maximum differential distance. The incorporation of these requirements into the procedure is elaborated in this section.

4.1.1 Power Budget and Dispersion

In any telecommunications link, the power budget is the ubiquitous reference as to whether or not the link meets the minimum connectivity requirements. In the end, the link must ensure that the power arriving at the receiver is sufficiently above the noise level to keep the decoding error rate below a given threshold. For PONs, it is no different. Given a transmitter power coupled to the fibre, yw, we need to ensure that the power at the receiver, y_z ,is such that it guarantees the Bit Error Rate (BER) does not exceed a certain limit (ITU-T specifies the reference budget values for a BER of 1E-10). This is accomplished by evaluating all the power losses and penalties applied in the link between the OLT and the ONT (in both directions). Figure 4.1 reviews the loss elements present in a link.

Figure 4.1 - Power budget in PON link. Source: [38].

{ is the fibre attenuation coefficient in dB/km. l is the total distance in km between the ONT and OLT.

e_|}~ is the attenuation caused by the optical splitter. It is divided in two terms. One caused by the actual division of optical power and another by losses in the coupling stages of the splitter. Each stage divides

the power by two, so if the attenuation of each stage is e_, the total attenuation introduced by the splitter is:

e_|}~ . ? 75€DB e_? 75€_cDB (4.1)

is the split ratio of the splitter. Usually e also depends on the splitter type. Splitters with larger split ratios have a lower attenuation per stage (see APPENDIX C).

To calculate the power budget, we must also consider the losses in connectors, splices and couplers.

These were grouped into a parameter called eG. In the calculations performed, a standard link model was considered, present in Figure 4.2.

Figure 4.2 - Standard link model.

With this model we know the amount of connectors that e_G has to account for. A coupler for inserting RF video into the PON is also included because it is a prevalent option in Portuguese PON deployments.

Couplers also exist in the OLT and ONT to separate the wavelengths of the upstream and downstream directions. The splices in the feeder and distribution networks take place at every 1 000 m and 70 m of fibre respectively (an average from PON projects analyzed). For power budget calculations, naturally the drop cables from the NAP to the customer’s house had to be included, as well as the connector to the outlet and the ONT. The drop section length from the NAP to the customer’s house was set to 50 m for all connections as a worst case scenario. If it meets the requirements for the worst-case (usually the highest floor in the building) it will meet them for the remaining ones. The budget must be calculated for both directions, resetting the fibre loss coefficients according to the wavelength used.

To be safeguarded against possible degradations in the link, operators add a system margin, w, so that the PON does not operate with a receiver power too close to the sensitivity limit of the detector.

Inquiries with operators revealed that 3 dB is the most commonly used value. The expression for the power budget can now be defined as:

_wS y_wV y_zV { ? l V e_|}~V uyƒDg_‚ „lB V e (4.2)

uyƒDg‚ „lB is the power penalty associated with dispersion related effects.

In the simulations performed, the values used were based on an industry best practise for GPON found in [39], which relates to Class B+ GPON systems. For further details on the specifications, please refer to APPENDIX C.

The model developed by [17] directly accounts for power budget constraints in the ILP model (and also differential distance). In ILP Model 2, with a fixed architecture, (the one in the standard model), the only real variable in expression (4.2) is distance, so if we define a maximum distance for the feeder and distribution fibre runs, the constraint is easily added to the model. But to limit the overall OLT-ONT distance, one would need additional variables to relate the ONT-splitter and splitter-OLT runs. As such, [17] requires additional constraints to check the budget feasibility of every possible feeder and distribution connection (like the multi-stage formulation did for the capacity restrictions). In this work, a simpler approach was adopted. Because we know beforehand the lengths of all the connection paths, the power budget can be computed for every feeder+distribution combination before the ILP is executed. With the standard link model presented, this is very fast even for large networks. In this way, we find out which ONTs cannot be connected to which splitter locations. If the power budget is not met for a given connection, we simply set the appropriate . in the ILP. This avoids adding complexity to the model and still achieves the same goals.

Another budget related issue that needs to be addressed is the Overload Margin, _:. The power reaching the ONT or the OLT cannot exceed a given threshold, otherwise it risks damaging the equipment. Similar calculations to the power budget must be made, only this time assuming “best-case”

scenarios, that is, the ones where the most power reaches the terminals. This means not considering power penalties or the system margin. The power reaching a detector must be below its Overload Power.

y:, by at least the value of the Overload Margin. This condition is expressed by:

y: ywV { ? l V e|}~V e… : (4.3)

If the maximum receiver power criteria are not met, there is an option of using optical attenuators.

However that was not considered in this framework.

The power penalties play an important role in the link budget. They account for degradation in the Signal-to-Noise Ratio (SNR) caused by physical effects like chromatic dispersion, non-ideal extinction ratios, frequency chirping, etc... In PONs, the optical links are relatively short as are the bit rates when compared to transport links, so these effects are much more negligible [38]. However, because we will look at PONs operating at 10 Gbps in Section 4.2, it is important to make sure these effects do not become more noticeable. There were two penalties considered here. One is the burst-mode penalty. In a TDM-PON, the OLT receives information in bursts, as each ONT transmits in different time periods.

Because each ONT is at a different distance from the OLT, the phase and amplitude of each signal will be different, which forces the receiver to quickly adjust. For this purpose power levelling is used to smooth

the differences in the receiving amplitudes. Because the ITU-T parameters used already factor in a 3 dB penalty to the OLT receiver sensitivity because of burst-mode operation, this penalty is already implied in the model. The other factor is chromatic dispersion. This arises from the fact that the several wavelengths in the optical signal travel at different speeds, which causes the signal to become increasingly distorted in time leading to inter-symbol interference. This is accounted for by setting a power penalty associated with the parameters that influence dispersion that cannot be exceeded. For details, again refer to APPENDIX C. The dispersion effects are mentioned here because they were dealt with in the same way as the power budget. The only changing variable between the various links is distance, so a pre-processing routine was made that checks all ONT-splitter-OLT path combinations and sets the ONT-splitter connections that do not meet the requirements to zero.

4.1.2 Maximum Differential Distance

The maximum differential distance is a constraint on the range of distances that ONTs can have to the OLT. This is important because the further apart two ONTs are relative to the OLT, the harder it is for the latter to perform ranging, power levelling and synchronization procedures. Recommendation G.984.2 defines the maximum differential logical reach at 20 km and the maximum differential optical path loss at 15 dB. Again, [17] included this information (for the logical reach) in the ILP itself. Here, because we included a splitting stage at the CO, the “members” of the same PON are not clearly defined. That was in fact the purpose, because an operator wants to have that kind of flexibility. For this purpose however, each specific ONT is only allocated to a Junction Box, not to any PON. This means we cannot know the differential distances before service is requested. A somewhat naive solution was implemented in this case, by post-processing the solution. When the ILP is executed and the solution obtained, the PONs are defined in the following way: each PON will successively spread its second level splitters by as many Junction Boxes as possible. Since we cannot know when customers will require service and therefore which splitter they will be connected to, the calculations are made for all the PONs that the ONTs have access to, meaning the ones that have splitters in Junction Boxes where the ONTs are connected.

Basically we create virtual super-PONs that include every ONT that can possibly be connected to it and check the differential distance and path loss. The final conclusion however was that the maximum differential distance is too large to impact the scenarios studied (without any reach extension) and the optical path loss is never an issue because the loss of a 1:64 splitting ratio (18 dB) alone is only 10 dB short of the total 28 dB maximum path loss, so there cannot be a 15 dB difference when PONs use 1:64 ratios.