DESIGN OF HYBRID METRO/ACCESS LONG-REACH PONS

1

DESIGN OF HYBRID METRO/ACCESS LONG-REACH PONS

João Carlos Bernardo

Corresponding author: [email protected]

1. INTRODUCTION Nowadays, metro and access are two completely different networks in technology,

protocol and, in most of cases, in the transmission medium. Therefore, its maintenance becomes too complex and expensive. The increasing needs for bandwidth has driven operators to improve the infrastructure of their networks by extending the fibre to the vicinity of subscribers. In this sense, it became clear that the optical fibre, dominant in the core and metro networks due to its high transmission capacity, will prevail as the medium of choice in the final segment of the network. Basically, the implementation of PONs (Passive Optical Networks) and other FTTH (Fibre To The Home) approaches will provide the gradual metro/access convergence, in terms of the transmission medium. PONs have been gaining popularity, because they allow the sharing of costs inherent to an optical access network by several users, enabling the progression of the fibre and providing higher bandwidths. From the solutions already proposed, only two variants of TDM-PONs (Time Division Multiplexing-PONs) are currently massified: ITU-T G.984 GPON or Gigabit PON and widely implemented in North America and Europe; IEEE 802.3ah EPON or Ethernet PON and installed on a large scale in Japan, China and Korea (see Table 1.1) [1].

Table 1.1: GPON and EPON parameters [13]-[14].

GPON EPON Standard ITU-T G.984 IEEE 802.3ah Downstream rate [Mbps] 1244.16, 2488.32 1250 Upstream rate [Mbps] 155.52, 622.08, 1244.16, 2488.32 1250 Wavelength D/U [nm] 1490±10/1310±50 1490±10/1310±50 RF overlay [nm] 1550±5 1550±5 Average efficiency ≈93% ≈65% Maximum splitting ratio 1:64 1:16/1:32 Maximum range [km] 10/20 (physic); 60 (logic) 10/20 (physic) Maximum budget [dB] 15/20/25 15/20 Relative cost/ONU 100% 78%

ABSTRACT This work aims to design a hybrid optical access/metro network based on long-reach PONs that reduces investment and operating costs. Its base contains a ring, along which are distributed the interconnection nodes with the passive optical access components (PONs). Communications schemes, which include hybrid Wavelength Division Multiplexing/Time Division Multiplexing (WDM/TDM) strategies and wavelength conversion techniques, are studied. It is also presented a protection scheme. In order to assess the network’s physical feasibility, the communication schemes are physically modelled and it is developed an optimization process, which uses genetic algorithms to determine the optimal amplification schemes. The physical layer results show that the only scheme that is not practicable for 10 Gbps is the one that uses converters based on cross gain modulation (XGM) in semiconductor optical amplifiers (SOAs). Given this, it is suggested, for future work, the study of the converter based on cross phase modulation (XPM) in SOAs. KEYWORDS Long-reach PON; communication scheme; wavelength conversion; amplification scheme; genetic algorithms.

2

The TDM-PON structure is traditionally based on a tree topology (see Figure 1.1), where the central equipment OLT (Optical Line Terminal) connects to the terminals: ONUs (Optical Network Units) by IEEE or ONTs (Optical Network Terminals) according to ITU-T. In the ODN (Optical Distribution Network), a supply fibre is divided by a certain number of distribution fibres arriving at the ONUs. All active elements are eliminated, gaining in simplicity, reliability, operating and maintenance costs. The communication between OLT and ONUs is full-duplex and uses CWDM (Coarse Wavelength Division Multiplexing) to separate the traffic directions. Besides that, they use MAC (Media Access Control) protocols built with TDM technologies.

Figure 1.1: PON traditional structure [2].

It is commonly accepted that these TDM-PONs (GPON and EPON) will not be able to cope with the future needs of the telecommunications world. Therefore, the community is already thinking about the next generation PON. As an example, we have the work started in September 2006 for the standardization of 10GEPON (10 Gigabit EPON) and that ended in 2009 with the establishment of IEEE 802.3av. 10GEPON represents an extension of the EPON standard and intend to increase the TDM-PON rates up to 10 Gbps [3]. The WDM (Wavelength Division Multiplexing) technology has been regarded as the ideal solution to enhance the capacity of optical networks without drastically changing the available fibre infrastructure. It allows a single fibre to simultaneously carry multiple wavelengths, eventually making the connection to each one of the end users virtually point to point. Thus, in addition to the increasing of the available bandwidth, the tasks of management, protection and security are simplified. A PON where each ONU is served by a single wavelength is called WDM-PON. The proliferation of PONs, coupled with the development of various broadband services and applications, will require the increase of the metro’s capacity, which is traditionally solved with the installation of expensive SDH (Synchronous Digital Hierarchy) and WDM equipments. However, sharing the same transmission medium will also lead to the possibility of metro/access convergence, under the use of similar technologies, starting to benefit from economies of scale and lower maintenance costs. One promising strategy is to improve the scalability of PONs, without necessarily keep them passive, supporting with a single structure a large number of subscribers, which were previously spread across multiple access networks. In publications [2] and [4] is demonstrated the feasibility of an architecture based on 10GEPON that due to the large coverage, high number of ONUs supported and resilience against failures, provides reliable metro/access convergence. The long-reach 10GEPON uses a ring structure, which constitute the metro’s segment, to interconnect several 10GEPONs with one OLT. This work is based on the architecture of this long-reach 10GEPON. Given the evolving needs of users and future services, it is expected the increase of its transmission capacity by the introduction of WDM in the metro’s segment. However, being aware of the prohibitive costs involved in assigning a dedicated wavelength to each user, it is essential to maintain the TDM in the access. This work aims to design a hybrid optical access/metro network based on long-reach PONs that reduces investment and operating costs. The work is divided in five chapters. In chapter 2, is proposed a structure for the network and are studied and compared the various communication schemes which are targeted for possible implementation. In chapters 3 and 4, the solutions identified are physical modelled in order to assess the viability of the network. Chapter 5 presents the conclusions.

3

2. CHARACTERIZATION OF A HYBRID OPTICAL ACCESS/METR O NETWORK BASED ON LONG-REACH PONS

2.1. Network Architecture This paper proposes a long-reach PON that contains two unidirectional rings with opposite traffic directions (see Figure 2.1). The downstream service (DS) and upstream protection (UP) travel across the outer ring and the upstream service and downstream protection circulate in the inner ring. The service and protection wavelengths are identical. The OLT and the remote nodes (RNs) from which derive the access components are distributed along this structure. Besides that, the RNs are responsible for amplification, splitting, combination and switching protection. The access is performed under a totally passive structure, like the traditional TDM-PON.

Figure 2.1: Hybrid optical access/metro network bas ed on long-reach PONs.

In a network of this magnitude it is essential to develop a strategy for the communication between the central office and its subscribers. This strategy can be defined as a transmission scheme, which covers a multiple access method for the upstream and a multiplexing method for the downstream. The most promising methods are those based on time domain (TDM and TDMA) and in the wavelength domain (WDM and WDMA) [5]. 2.2. TDM DL + TDMA UL 1 The first scheme, TDM DL + TDMA UL, uses TDM in the downstream and TDMA in the upstream. PONs that implement this strategy are nicknamed TDM-PON and represent nowadays the most popular Fibre To The Premise (FTTP) approach. The voice and data transmission is done using the 1490 nm in the downstream and the 1310 nm in the upstream (see Table 1.1). Figure 2.2a presents a possible structure for the OLT. The downstream signal transmission is performed by the transmitter (Tx) and the splitter (S), which divides the signal by the rings. The upstream signal reception is made by the combiner (C), the filter and the burst-mode receiver (Rx BM). The combiner collects the signals from both rings and the filter centred on 1310 nm filters out the noise and blocks the downstream signal components. The ONU, observable in Figure 2.2b, consists of a WDM coupler (WDM), a burst-mode transmitter (Tx BM) and one optical receiver (Rx). MAC blocks are responsible for all control and management functions. Figure 2.2c provides the structure proposed in [4] for the RN. It includes splitters that carry the downstream signal towards the access and combiners that put the upstream signal in the ring. Semiconductor Optical Amplifiers (SOAs) are used for power losses compensation. In the downstream is used a filter (DF), centred on 1490 nm, to filter out the noise and to block upstream signals. The link between the RN and the access is performed by a WDM coupler. The switches are an essential part of the protection mechanism and should be operated through a system that monitors the signal level. This approach promotes an economically efficient solution, since the OLT is composed by a single transmitter/receiver set and because all ONUs are identical. However, it critically restrains the bit rate offered to the customers because the network’s transmission capacity is shared by all subscribers (see Table 2.2). 1 DL and UL respectively stands for Downstream Line and Upstream Line.

4

(a) (b)

(c)

Figure 2.2: TDM DL + TDMA UL structures: (a) OLT, ( b) ONU, (c) RN. 2.3. Hybrid WDM DL + WDMA UL In a WDM-PON are overcome many of the typical problems of TDM-PON, both at management and security level as at transmission capacity. However, it is known that the costs involved in the attribution of a wavelength per subscriber are still prohibitive. One way of alleviating this economical problem is to adopt a hybrid WDM/TDM strategy. Terminals belonging to the same access component start operating on a single and unique wavelength, but keeping the TDMA in the upstream, which mean that transmitters must remain burst-mode (see Figure 2.3c). Moreover, now there is an optical source (besides one burst-mode receiver) per RN in the OLT, which implies the existence of a mux/demux scheme (see Figure 2.3a). Thus, in the downstream, each transmitter emits a single wavelength and in which are time multiplexed the data belonging to the terminals of a certain component. Each of the access components can be managed independently and through a MAC protocol similar to the one which operates under the traditional TDM-PON.

(a) (b)

(c)

Figure 2.3: Hybrid WDM DL + WDMA UL structures: (a ) OLT, (b) RN, (c) ONU. Table 2.1 indicates the wavelengths used (4 RNs) and that were selected taking into account the amplifier’s bandwidth (30 nm). The use of alternative wavelengths implies the changing of optical devices, particularly the optical filters belonging to the OLT and RNs (see Figure 2.3). The optical receivers of the OLT and ONUs are respectively maintained centred on 1310 and 1490 nm, because their bandwidths are large enough.

5

Table 2.1: Hybrid WDM DL + WDMA UL wavelengths. Transmission Direction # of PON � �� ����

Downstream (Service or Protection)

1 1 1482.5 2 2 1487.5 3 3 1492.5 4 4 1497.5

Upstream (Service or Protection)

1 5 1302.5 2 6 1307.5 3 7 1312.5 4 8 1317.5

The adoption of the hybrid strategy increases (on average) the transmission capacity per terminal by � times, where � is the number of access components (see Table 2.2). However, its implementation requires maintaining a diverse inventory, because the characteristics of the terminal change according to the access component, which is reflected in substantially higher costs. Thus, although the network is physically conceivable, the costs that falls under the operator and, ultimately, on the subscriber may not be admissible. 2.4. WDM/TDM DL + TDMA UL The previous schemes provide a symmetrical nature to the network. However, most of today’s networks are asymmetric with a greater bit rate in the downstream. The development of digital video services coupled with the growing demand for more content and better quality will give direction to the maintenance of asymmetry in the access. WDM/TDM + TDMA UL provides this asymmetric nature. In the downstream it adopts the WDM DL + WDMA UL method and in the upstream it uses the TDM DL + TDMA UL method (such as bitrates granted and wavelengths used), which favours the downstream. Figure 2.4 presents the OLT structure, whose transmission section comprises four emitters, just like Hybrid WDM DL + WDMA UL, and whose reception is equal to the one from TDM DL + TDMA UL. As the access components are not independent of each other, the MAC block should handle them like the TDM-PON. However, at its output should exist a demux process. The structures shown in Figures 2.2c and 2.3c remain respectively valid for the RN and the ONU.

Figure 2.4: WDM/TDM DL + TDMA UL structure for the OLT.

Given all that was said, this strategy appears to be a good and balanced solution, not only in terms of transmission capacity as in economic. This scheme combines the high downstream capacity of the Hybrid WDM DL + WDMA UL with the moderate monetary burden of the TDM-PON. The only limiting factor of this strategy appears in the upstream, when the number of subscribers is quite high and or the total bit rate is low. 2.5. Hybrid WDM DL + WDMA UL With UWC 2 The biggest problem in the Hybrid WDM DL + WDMA UL is economical and is related to the fact that the terminal’s characteristics are not common to all access components. The previous scheme solves this problem, but critically restrains the upstream bit rates. The idea is to guarantee an upstream bit rate as high as the downstream, but ensuring an acceptable economic efficiency. One way to circumvent this problem is to insert a wavelength converter in the upstream on each RN. ONUs, belonging to each access component, emit at the wavelength specified for the TDM-PON and, in the RN, that wavelength is converted to another, unique on the network. This allows the maintenance of the WDMA in the ring, ensuring an upstream

2 UWC stands for Upstream Wavelength Conversion.

6

bandwidth equal to the Hybrid WDM DL + WDMA UL scheme. In the downstream, all transactions are conducted normally. It is also possible to implement wavelength conversion in this direction, but this proves to be unnecessary. For the long-reach PON analysis, it is selected an all-optical converter based on Cross Gain Modulation (XGM) in SOA, due to its simplicity [7]-[9]. Figure 2.5 presents the RN structure, already with the converter (WC). For the rest, the structure remains the same. A possible OLT structure is the one shown on Figure 2.3b. Regarding the ONU, one possibility is the structure described in the previous section.

Figure 2.5: Hybrid WDM DL + WDMA UL with UWC struct ure for the th RN .

2.6. Protection Mechanism

In a network of this magnitude, a single failure may interrupt several communications, making it necessary to increase the network’s survivability. The protection mechanism used is similar to the Optical Channel Dedicated Protection Ring (OChDPRING), with some modifications [2] in relation to the traditional scheme [10]. In the reference scheme, upstream and downstream replicas travel across both rings simultaneously, but in different directions. However, in the proposed scheme, each of the RNs (and its switches) only forwards the upstream signal up to the inner (service) or to the outer (protection) and never for both. On the downstream, the operation is identical to the classic. The implemented architecture is quite simple, especially due to the autonomy provided to the RNs. They can switch to protection without having to wait for any response from the OLT. In other words, signalling is not necessary to coordinate the switching. However, this architecture does not promote bandwidth efficiency.

Tab. 2.2: Transmission schemes characteristics. 3

Scheme Characteristic

TDM DL +

TDMA UL

WDM DL +

WDMA UL

WDM DL +

WDMA UL

WDM/TDM DL +

TDMA UL

Hybrid WDM DL

+ WDMA UL with UWC

Multiplexing Method TDM WDM WDM/TDM WDM/TDM WDM/TDM Multiple Access Method TDMA WDMA WDMA/TDMA TDMA WDMA/TDMA

Upstream Bit Rate

.bD

N K bD bD

K

bDK

bD

K

Downstream Bit Rate

.bD

N K bD bD

K

.bD

N K

bDK

MAC Protocol At the

network level.

Not necessary.

At the access component

level.

At the network level.

At the access component

level.

Security Mechanism is necessary. Guaranteed.

Mechanism is necessary.

Mechanism is necessary.

Mechanism is necessary.

Cost Low High Medium Low Low

3 is the number of ONUs per RN.

7

3. PHYSICAL LAYER ASSESSMENT: LONG-REACH TDM-PON AN D WDM/TDM-PON

3.1. Theoretical Concepts The average optical power at the OLT’s receiver (��

���) for the �th PON (1 � � � �, where � is the total number of RNs), in the upstream, can be written as:

( )1

,nOLT ONU

R E ONU PON RN L i OLT BM fi

P P A A A A G A P P M=

= − − − + − − − ∆ − ∆ −∑ (3.1)

where ������ denotes the average optical transmission power of a given ONU’s emitter, ���� is

the loss inside the ONU, ���� is the power loss in the access section, � � is the loss in the RN, �� is the loss in the ring link between two RNs, �� is the �th amplifier’s gain, ���� is the loss inside the OLT, ∆� is the chromatic dispersion penalty, ∆��� is the burst-mode penalty and � is the safety margin. The average optical power at a given ONU’s receiver (��

���) for the �th PON, in the downstream, is equal to:

( )1

,nONU OLT

R E OLT RN L i PON ONU fi

P P A A A G A A P M=

= − − + − − − − ∆ −∑ (3.2)

where ������ is the average optical transmission power of the OLT. The referred losses are given

by (see Figures 2.2 and 2.3):

,ONU WDM CA A A= + (3.3)

( )

( )1:

1:

.

.

, if DL,

, if ULSwitch WDM K f F D

f F D K WD

FilterPON

M Switch

A A S L L

L L S A A

AA

αα

++

+ ++ +

+ +

+=

(3.4)

1:22. ,RNA S= (3.5)

. ,L fA Lα= (3.6)

1:2

1:2

, if DL,

, if ULC

OLTC Filter

S AA

A S A

+= + +

(3.7)

where �!"� is the loss in the WDM coupler, �# is the loss in a connector, �$%�&'( is the switch power loss, �)�*&+, is the optical filter insertion loss, -.:0 is the loss in the access section PSC (Passive Splitters-Combiners), is the number of ONUs per access section, 1 is the attenuation coefficient of the optical fibre, 2) is the feeder fibre length, 2" is the distribution fibre length, -.:3 is the power loss verified in a PSC of two entrance-output ports and 2 is the ring fibre (link between RNs) length. All the equations listed remain valid for the WDM/TDM-PON, except for (3.7):

1:4 1:2

1:2 1:4

, if DL,

, if ULC

OLTC Filter

S S AA

A S S A

+ += + + +

(3.8)

where -.:4 is the power loss verified in a PSC of four entrance-output ports. The power loss in a 1: 5 PSC is obtained by -.:6 7 10 log.< 5 = 0.5 log3 5 [11]. Taking into account the noise sources internal to the receiver and the ASE noise originated in amplifiers, the electrical current (�) generated by the photodetector is given by [6]: ,s th shot sm s sp sp spi i i i i i i− −= + + + + + (3.9)

where �@ the signal current, �&( is the thermal noise current, �@(A& is the shot noise current, �@B is the shot multiplication noise current (only in APD), �@C@D is the noise current from the signal-ASE beating and �@DC@D is the noise current from the ASE-ASE beating. Assuming that the total noise current (�E) represents a stationary random process that can be approximated by a zero-mean Gaussian distribution, the total noise power (FE

3) can be estimated by [6]:

2 2 2 2 2 2 ,n th shot sm s sp sp spσ σ σ σ σ σ− −= + + + + (3.10)

where the power of each component is calculated according to [6] and [12]. Notice that, in their way towards the receiver, the ASE noise components are attenuated and eventually amplified. The bit error probability is calculated using [11]:

0 1

,0 ,1

1P . ,

4 . 2 . 2D D

e

n n

I I I Ierfc erfc

σ σ

− −= +

(3.11)

8

where FE,6 and H6 are, respectively, the square root of the total noise power and the electrical signal current associated with the symbol 5 (‘0’ or ‘1’ with equal probability). This current can be calculated according to H6 7 �. IJ. K6 , where � and IJ represent, respectively, the avalanche gain and the photodetector responsitivity and K6 the optical power associated with 5. The optical power of both symbols can be related through L 7 K< K.⁄ , where L denotes the extinction ratio. The noise powers associated with the symbols ‘0’ and ‘1’ are quite unequal, because of the ASE noise. Therefore, the decision threshold current should be calculated using [6]:

2 2

,10 1

,0,0 ,1

ln .. 2 . 2

nD D

nn n

I I I I σσσ σ

− −= +

(3.12)

3.2. Optimal Amplification Schemes The long-reach PON is properly characterized with two different amplification schemes, one for the inner ring and another for the outer. An amplification scheme represents the set of amplifier parameters (gain and saturation power), serving a particular ring. In order to optimize the network performance, we search for the optimal amplification schemes through a class of heuristics called genetic algorithms [16]-[20]. The algorithm, which covers only the service condition, begins with a set (population) of possible solutions (chromosomes) randomly generated (gain lower than 35 dB and saturation power lower than 20 dBm). Figure 3.1 shows the chromosomes structures, whose parameters follow a real representation: the gains for weak signals �<,6 in dB, the saturation powers �@N&,6 in dBm of the amplifiers, with 5 an integer varying from 1 until 4; the average optical transmission power ��� in dBm (of the ONU in the upstream or of the OLT in the downstream); the gain of the converter’s amplifier �<

�!# in dB and the respective saturation power �@N&�!# in dBm, whose values are

identical to all.

�<,. �@N&,. �<,3 �@N&,3 �<,O �@N&,O �<,4 �@N&,4 ��� (a)

�<�!# �@N&

�!# ������

(b) Figure 3.1: Chromosome structure: (a) TDM-PON and W DM/TDM-PON, (b) WDM/TDM-PON with UWC.

The algorithm’s evolution starts from this population and is accomplished through generations. In each generation, the adaptation of each solution is evaluated by an objective function that returns the number of pathways between the central and the terminal, for which the bit error probability is equal or less than 10-12. After evaluation, the selected chromosomes are recombined and or mutated to form a new population. The new population is then used as input for the next iteration. The above process is manually terminated, but not before the 250th iteration. Table 3.1 shows the amplification schemes obtained.

Table 3.1: Optimal amplification schemes.

(a) TDM DL + TDMA UL, Hybrid WDM DL + WDMA UL and WDM/TDM DL + TDMA UL.

Ring Direction PQ,R PQ,S PQ,T PQ,U VWXY,R VWXY,S VWXY,T VWXY,U

Outer Donwstream Service (DS) 34 34 34 34 18 18 18 18

Inner Upstream Service (US) 34 14.5 14.5 14.5 18 17 17 17

(b) Hybrid WDM DL + WDMA UL with UWC.

Ring Direction PQ,R PQ,S PQ,T PQ,U VWXY,R VWXY,S VWXY,T VWXY,U PQZ[\ VWXY

Z[\

Outer DS 34 34 34 34 18 18 18 18 34 0

Inner US 34 14.5 14.5 14.5 18 17 17 17

3.3. Results For The Service Condition Tables 3.2a, b and c show, respectively, the main results at 1.25, 2.5 and 10 Gbps for the long-reach PON physical assessment, among which are the preferred photodetector, the average optical transmission powers necessary to ensure bit error probabilities of 10-12 and 10-3

9

(assumes FEC) [15] and finally the decision over the physical feasibility (average maximum transmission power placed at 7 dBm). The main network parameters are represented in Table 4.1. It’s indifferent to use a PIN or an APD in the upstream, because the thermal and shot noises are negligible compared to the ASE noise. On the other hand, the APD clearly leads to a better performance in the downstream. Any one of the amplification schemes is physically feasible, but at 10 Gbps is necessary to apply a FEC on the upstream signal, which allows the earning of 6-7 dB. Notice that, for the same reason that it is irrelevant the photodetector used in the upstream, the insertion of an optical pre-amplifier in the OLT does not improve system performance. The upstream is more critical than the downstream. In the upstream, only the signal is attenuated in the access section and, in addition, anyone of the upstream signals suffers the interference of all ASE noise components. At first glance, the WDM/TDM-PON performance should be significantly worse than that of TDM-PON. Notice that, in the ring we transmit a signal, whose power is (four times) higher than the TDM-PON signal and that should lead to a saturation effect much more intense. However, this is not true and the amplifiers gains practically don’t change. Now, it is interesting to try to understand why, in the upstream, the WDM/TDM-PON performance is similar to the TDM-PON’s and that, in the downstream, is worse. The reason for this is in the mux/demux scheme, placed at the OLT of the WDM/TDM-PON. In the upstream, both the signal and ASE noise are attenuated by the demux, which allows keeping the same signal to noise ratio value, while, in the downstream, only the signal is attenuate by de mux. If with the introduction of such attenuation, which did not exist in the TDM-PON, only the signal is attenuated, it is logical to think that the performance is degraded sharply.

Tab. 3.2: Main results for the service condition. (a) 1.25 Gbps.

Transmission Scheme Direction Photodetector

V]^,,_`a �bc�� Physical Feasibility Vd 7 RQCRS Vd 7 RQCT

TDM DL + TDMA UL

UL Indifferent 4.0 -1.0

Yes

DL APD -20.2 -25.4 Hybrid WDM DL +

WDMA UL UL Indifferent 5.1 0.4 DL APD -13.1 -18.2

WDM/TDM DL + TDMA UL

UL Indifferent 4.0 -1.0 DL APD -13.1 -18.2

(b) 2.5 Gbps.

Transmission Scheme Direction Photodetector

V]^,,_`a �bc�� Physical Feasibility Vd 7 RQCRS Vd 7 RQCT

TDM DL + TDMA UL

UL Indifferent 6.4 1.0

Yes

DL APD -18.0 -23.5 Hybrid WDM DL +

WDMA UL UL Indifferent 7.3 2.3 DL APD -10.8 -16.4

WDM/TDM DL + TDMA UL

UL Indifferent 6.4 1.0 DL APD -10.8 -16.4

(c) 10 Gbps.

Transmission Scheme Direction Photodetector

V]^,,_`a �bc�� Physical Feasibility Vd 7 RQCRS Vd 7 RQCT

TDM DL + TDMA UL

UL Indifferent 11.6 5.4

Yes

DL APD -12.9 -19.2 Hybrid WDM DL +

WDMA UL UL Indifferent 12.2 6.4 DL APD -5.4 -12.0

WDM/TDM DL + TDMA UL

UL Indifferent 11.6 5.4 DL APD -5.4 -12.0

3.4. Results For The Protection Condition The long-reach PON doesn’t have an acceptable performance in the protection condition, if we use the amplification schemes presented on Table 3.1. To solve this problem, we take advantage of the downstream amplification scheme flexibility and we modify their values by making the scheme that acts on the upstream protection signal equal to the upstream service amplification scheme. Figure 3.1 presents the final amplification schemes and in Table 3.3 the main results.

10

(a) (b) Figure 3.1: Final amplification schemes: (a) servic e and (b) protection conditions.

An acceptable performance is only assured with the use of APDs in the ONUs and the

application of FEC codes in both directions. However, even resolving this situation, the system will fail when it leaves the service condition. To make the system work properly, it is necessary that the amplifiers have a bandwidth of about 200 nm, which is impossible.

Table 3.3: Main results for the long-reach WDM/TDM- PON in the protection condition.

ef�PfgW�

Direction Photodetector

V]^,,_`a �bc�� Physical Feasibility Vd 7 RQCRS Vd 7 RQCT

1.25 UL Indifferent 5.1 0.4

Yes

DL APD 1.8 -4.0

2.5 UL Indifferent 7.3 2.3 DL APD 4.2 -1.7

10 UL Indifferent 12.2 6.4 DL APD 9.5 3.1

In order to overcome this problem, our prospects are the modification of the central wavelength of the upstream signal or of the RN internal structure. Figure 3.2 presents an alternative structure for the RN, which uses a greater number of switches to assure that the amplifiers always act over the same direction.

Fig. 3.2: New RN structure.

4. PHYSICAL LAYER ASSESSMENT: LONG-REACH WDM/TDM-PO N WITH UWC

4.1. Theoretical Concepts Figure 4.1 shows the wavelength converter inserted in the �th RN, which is based on XGM in SOA. The average optical power at the �th converter (���

!##E) is given by:

#

( ),WC n ONUi E ONU PON SwitchP P A A A= − − − (4.1)

where ���� and ���� can be calculated, respectively, from Eq. (3.3) and (3.4). The signal at 1310 nm is combined with a continuous signal at i�Ej4� (see Table 2.1). It is considered that the power of the signal emitted by the converter’s laser is equal to �&

!##E, which is about one order of magnitude lower than the input signal power.

11

Figure 4.1: �th wavelength converter structure.

After combination, it is performed the gain modulation process. When the input signal power is maximal, i.e. when the signal reaches the logical level ‘1’, the amplifier saturates and the signal is amplified by a very small gain kl�� 7 ′1′m. When the input power is minimal, i.e. when the signal goes to ‘0’, the signal is amplified by a high gain kl�� 7 ′0′m. The amplifier gain variation is then mapped in a continuous signal causing the appearance of a signal at the target wavelength. This signal has the same digital information as the input signal, but with a reverse polarization. The �th WCF aims to filter the signal at i�Ej4� and to decrease the ASE noise power generated by the converter. From the power of symbols ‘0’, KA,′<′

!##E 7 K&!##E. kl�� 7 ′1′m. 10C�$n:ojpqrstuv� .<⁄ , and ‘1’,

KA,′.′!##E 7 K&

!##E. kl�� 7 ′0′m. 10C�$n:ojpqrstuv� .<⁄ , it is inferred the extinction ratio expression LA

!##E 7 kw�� 7′ 1′x kw�� 7′ 0′xy . The average optical power at the OLT can be estimated by:

( )#

1

.nOLT WC n

R o Switch RN L i OLT BM fi

P P A A A G A P P M=

= − − + − − − ∆ − ∆ −∑ (4.2)

The equations for the noise components and for the bit error probability remain valid, being only necessary to add the ASE noise component generated by the converter’s amplifier.

Table 4.1: Main parameters.

Parameter Symbol (Unit) Value Number of RNs � 4 Number of ONUs per RN 32 Ring fibre (link between RNs) length 2 �z{� 20 Feeder fibre length 2) �z{� 20 Distribution fibre length 2" �z{� 20 Amplifier noise factor |E,pBD �}~� 8 Amplifier bandwidth ~<,pBD ��{� 30 Filter bandwidth ~< ��{� 1

Fibre loss 1 0.24 in downstream 0.35 in upstream

Responsitivity IJ 1 at 1490 nm 1 at 1310 nm

Data rate �� ���K�� 1.25, 2.5 and 10 Electrical bandwidth ~+,E ���� 0.80 � �f Polarization resistance I� �z� 1 Receiver noise factor |E �}~� 2 Temperature � �� 290 Ionization coefficient zp 0.35

APD optimal gain (� � 7.0345) ��D& 20.0 at 1.25 Gbps 16.8 at 2.5 Gbps 11.9 at 10 Gbps

Insertion loss in the connectors �' �}~� 0.3 Insertion loss in the filters �)�*&+,�}~� 1 Insertion loss in the WDM coupler �!"��}~� 1 Insertion loss in the switch �$%�&'(�}~� 1 Chromatic dispersion penalty ∆���}~� 2 Burst-mode penalty ∆�����}~� 0.3 Safety margin � �}~� 2

12

4.2. Results Table 4.2 presents the main results for the long-reach WDM/TDM-PON with UWC physical assessment (service condition), where it is confirmed its impracticality. The use of a converter based on XGM in SOA degrades the extinction ratio and generates ASE noise that significantly deteriorates the upstream system performance. Even applying FEC coding, the system requires an inadmissible value of transmission power. Therefore, it is concluded that the type of converter studied is not suitable for this network. An alternative is the converter based on Cross-Phase Modulation (XPM) in SOA since it seems not to degrade to such an extent the extinction ratio and the signal.

Table 4.2: Main results for the long-reach WDM/TDM- PON with UWC (service condition).

ef�PfgW� Direction Photodetector V]^,,_`a �bc��

Physical Feasibility Vd 7 RQCRS Vd 7 RQCT

1.25 UL Indifferent 18.8 14.1

No

DL APD -13.1 -18.2

2.5 UL Indifferent 21.1 15.9 DL APD -10.8 -16.4

10 UL Indifferent 26.4 20.2 DL APD -5.4 -12.0

5. CONCLUSIONS

This work focuses on a hybrid metro/access long-reach PON. The transmission schemes that are best suited to this network are identified and studied and it is performed its physical assessment, particularizing to a network with four RNs and 32 ONUs per RN. The transmission schemes are called: TDM DL + TDMA UL, Hybrid WDM DL + WDMA UL, WDM/TDM DL + TDMA UL and Hybrid WDM DL + WDMA UL with UWC. In order to access the network’s physical feasibility, the transmission schemes are physically modelled and it is developed an optimization process which aims to identify the optimal amplification schemes. By analysing its physical layer, it is concluded that all schemes, except the Hybrid WDM DL + WDMA UL with UWC, are physically viable in service. However, in the upstream and at 10 Gbps, it is necessary to use FEC coding techniques. For the impracticable scheme, it should be studied another type of converter, such as the one that is based on XPM in SOA. In order to increase the network’s survivability, it is presented a very simple protection mechanism based on OChDPRING. However, its implementation proves to be incompatible with the RN’s internal structure. To solve this problem, the upstream central wavelength and or the RN structure must be modified. Even following these instructions, there is a need to change the amplification schemes. Ended this, the physical feasibility is attained, not only in service as in protection, and to this end the ONUs need to use APDs and FEC codes must be applied in both directions.

REFERENCES [1] K. Grobe, J.-P. Elbers, “PON Evolution from TDMA to WDM-PON”, Optical Fiber communication/National Fiber Optic Engineers Conference, 2008 (OFC/NFOEC 2008), 24-28 Feb. 2008, NthD6 pp. 1-7. [2] J. Santos, O. Quadros, “Análise e Planeamento de uma Rede de Acesso baseada na Tecnologia PON”, Departamento de Engenharia Electrotécnica e de Computadores, Instituto Superior Técnico, Oct 2008. [3] R. Lin, “Next Generation PON in Emerging Networks”, Optical Fiber communication/National Fiber Optic Engineers Conference, 2008 (OFC/NFOEC 2008), 24-28 Feb. 2008, OWH1 pp. 1-3. [4] J. Santos, J. Pedro, J. Pires, “ Long-Reach 10 Gbps Ethernet Passive Optical Network based on a Protected Ring Architecture” , Optical Fiber communication/National Fiber Optic Engineers Conference, 2008 (OFC/NFOEC 2008), 24-28 Feb. 2008, OTuI2 pp. 1-3. [5] C.-H. Lee, W. V. Sorinm, B. Y. Kim, “Fiber to the Home Using a PON Infrastructure”, Journal of Lightwave Technology, Vol. 24, No. 12, Dec. 2006, pp. 4568-4583. [6] G. P. Agrawal, “Fiber-Optic Communication Systems”, John Wiley & Sons, 1992. [7] D. Wolfson, T. Fjelde, A. Kloch, “Technologies for all-optical wavelength conversion in DWDM networks”, The 4th Pacific Rim Conference on Lasers and Electro-Optics 2001 (CLEO/Pacific Rim 2001), Vol. 2, 15-19 July 2001, pp. 574-575. [8] S. J. B. Yoo, “Wavelength Conversion Technologies for WDM Network Applications”, Journal of Lightwave Technology, Vol. 14, No. 6, June 1996, pp. 955-966.

13

[9] M. M. de la Corte, J. M. H. Elrnirgharni, “Accurate noise characterization of wavelength converters based on XGM in SOAs”, Journal of Lightwave Technology, Vol. 21, No. 1, Jan. 2003, pp. 182-197. [10] M.-J. Li, M. J. Soulliere, D. J. Tebben, L. Nederlof, M. D. Vaughn, R. E. Wagner, “Transparent Optical Protection Ring Architectures and Applications”, Journal of Lightwave Technology, Vol. 23, No. 10, Oct. 2005, pp. 3388-3403. [11] A. Cartaxo, “Transmissão por Fibra Óptica”, Departamento de Engenharia Electrotécnica e de Computadores, Instituto Superior Técnico, Março 2004. [12] D. Mestdagh, C. Martin, “The Super-PON Concept and its Technical Challenges”, Broadband Communications, 1996, pp. 333-344. [13] ITU-T Recommendations G.984 Series, 2003. [14] IEEE Standards 802.3ahTM, 2004. [15] F. Chang, “10G EPON Optical Budget Considerations”, Vitesse, July 2006. [16] htttp://pt.wikipedia.org/wiki/Heur%C%ADstica. [17] http://en.wikipedia.org/wiki/Genetic_algorithm. [18] http://www.obtiko.com/tutorials/genetic-algorithms/portuguese/index.php. [19] http://www.dca.ufrn.br/~estefane/metaheuristicas/index.html. [20] www.ica.ele.puc-rio.br