Evolution of Optical Networks

Since the “optical telegraph” was invented by Claude Chappe in the 1790s, optical communication systems have made significant progress. After centuries of development and in particular the invention of the laser and optical fibres in the 1960s, optical communication became very popular. In 1970 researchers designed single-mode fibres whose loss was below 20 dB/km at 633 nm [18]. Due to high attenuation, such fibres were not used to operate the telephone network. In the 1980s when the attenuation of optical fibres became as low as 0.5 dB/km, long-distance telephone networks and national backbone networks based on optical fibres started to see field deployment [18]. After the opening of the 1310nm and 1550 nm optical communication windows (through the design of optimised fibres, sources and detectors) optical fibre loss was further reduced to 0.2 ~ 0.3 dB/km. Recently

photonic crystal fibres have become attractive in terms of their characteristics which cannot be provided by ordinary optical fibres. Compared to copper cable, optical fibre provides lower error rate and higher capacity. Furthermore copper cables have been replaced by optical fibres. All the improvements in optical fibres motivated the development of optical networks.

In the first generation optical networks, optical devices were only used for information transmission. All switching and other intelligence functions in networks were implemented by electronic devices. A typical example is the Synchronous Optical Network (SONET), which commonly multiplexes channels of 64 kbit/s into data frames with data rates between 155 Mbit/s and 2.5 Gbit/s [19]. In such kinds of optical networks, both the traffic passing by and ending at a node is converted from the optical domain to the electrical domain and switched electronically to an output port (including a port that can drop traffic locally). Following electronic switching, the traffic passing by a node is converted back to the optical domain before departing from the node. With the increase in data transmission rate, electrical switching and optical-electrical-optical (OEO) conversion result in a significant growth in complexity and cost for electronic devices [20]. Therefore, reducing the burden placed on the underlying electronic devices in a node and removing electronic switching for traffic passing by a node became key factors in the development of second generation optical networks.

In order to improve capacity in second generation optical networks different multiplexing strategies were proposed. In optical networks there are two basic multiplexing technologies: Optical Time Division Multiplexing (OTDM) and WDM. In OTDM, lower bit rate optical streams are assigned to different time slots on the multiplexed channel [21]. This is in contrast to Time Division Multiplexing (TDM) which is implemented in electrical domain. OTDM however requires specific optoelectronic devices such as pulsed semiconductor lasers and optical switches. The optical pulses used in OTDM transmission should satisfy several requirements. First the pulse width of the optical signal pulse, which determines the upper bound on the achievable bit rate, must be less than the time slot associated with the desired bit rate. Secondly, the spectral width of the optical signal should be extremely narrow for a given pulse width as the spectral width of the pulse determines the impact of the fibre chromatic dispersion [22]. Currently, using Femtosecond pulses,

OTDM transmission can reach data rates of 1.28 Tb/s over 70 km [23], however single high data rate channels are not very desirable as flexibility/granularity is lost specially that typical individual user rates currently are below 40 Gbit/s.

Fig 2 - 1: A point-to-point WDM Transmission Configuration [20]

WDM technology is very similar to Frequency Division Multiplexing (FDM), which can allow multiple non-overlapping wavelength channels to transmit in the same optical fibre link. Each of these channels can operate at a different data rate with typical rates being 10 Gbit/s, 40 Gbit/s and 100 Gbit/s currently [24]. Fig 2-1 shows a basic point-to-point WDM transmission configuration. At the transmitter, different wavelengths (λ1~ λN) are multiplexed in the same fibre [20]. At the receiver these wavelengths are split back out, or demultiplexed, into separate fibres. Essentially, the bandwidth capacity of the optical fibre is multiplied by the number of wavelengths multiplexed onto it. Each wavelength being an independent channel can transmit data at a different rate. In addition to increasing the bandwidth, consequently reducing the optical fibre cable cost and use of equipment, WDM also gives the great advantage of transparency which allows different network technologies (SONET, IP, Asynchronous Transfer Mode (ATM), etc) to use the same physical transmission infrastructure [20].

Commonly, WDM is divided into different wavelength density classes: Coarse WDM (CWDM) and Dense WDM (DWDM) [25]. The first generation of WDM technology utilised two wavelengths, 1310 nm and 1550 nm. It was implemented using off-the-shelf optical transmitters (meant for single channel operation at the 1310nm and 1550nm windows) without tight control of wavelengths. Because of the simple implementation and lower number of multiplexed wavelengths, the optical multiplexers and demultiplexers are low cost and low in insertion loss [25]. To further increase the capacity of optical fibres, in 2003 the International Telecommunication Union (ITU) standardised a channel grid for use with CWDM wavelengths, using 18 CWDM wavelengths from 1271 nm to 1611 nm with a

channel spacing of 20 nm [26]. The wavelengths in DWDM, are spaced closer than in CWDM, therefore DWDM offers higher capacity. Currently, DWDM can allow up to 100 wavelengths per fibre (DWDM ITU grid) offering a single fibre capacity beyond 4 Tb/s [27]. However, the requirement of tight control over the wavelengths under all operating temperature conditions result in higher cost for DWDM equipment compared to CWDM.

Combining WDM technology with the adoption of many different mature technologies allowed second generation optical networks to support wavelength- routing. The key components of the second generation networks included: Optical Line Terminals (OLT) [28], Optical Add/drop Multiplexers (OADM) [29], Erbium Doped Fibre Amplifiers (EDFA) [30] and Optical Crossconnects (OXC) [31]. Commonly, OLTs are installed at the end of point-point WDM links. OADMs can drop or add some wavelengths at intermediate nodes without any OEO conversion processing. Similarly to OADMs, OXCs can switch a large number of wavelengths, from tens to thousands. All of the above mentioned components can eliminate OEO conversion and provide lightpaths across the network. Therefore, the second generation optical networks provide much higher bandwidths, larger capacities and also improve the flexibility and transparency of networks.

The third generation of optical networks were developed based on the all-IP and all-optical based approach, i.e. retain the IP layer for applications and the optical layer for capacity and eliminate all the intermediate layers such as SONET, ATM etc. However, packet services, rather than conventional WDM systems that provide fixed bandwidths are needed [32]. Therefore, appropriate optical switching techniques, such as, Optical Packet Switching (OPS) and Optical Burst Switching (OBS) were proposed, although OPS and OBS are not part of the third generation IP over WDM networks and may appear in future generations. In the next section, the main optical switching techniques will be reviewed.

The last section of the network, closest to the users, is the access network. Digital Subscriber Line (DSL) was the first widely used broadband technology to provide Internet access by transmitting digital data over the local telephone network twisted wire pairs. Because the bandwidth of the cable is limited, DSL can provide data rates up to 30 Mbit/s normally [33]. Even current DSL technologies, such as the

latest “NodeScale Vectoring” technology (which enables 100 Mbit/s and higher DSL access speed capability [34]) may not be able to handle the future growth in the number of end users and requirements in terms of network services. To replace DSL technologies, the higher bandwidth Fiber-to-the-Home (FTTH) solution was proposed. It has recently surpassed 75 million users and is continuing to grow at a rapid rate [35]. To realise various FTTH solutions, Passive Optical Networsk (PONs) have been considered the most promising technology. Developing from initial Broadband PON (BPON, 155 Mbit/s to 622 Mbit/s) and Ethernet PON (EPON, 1.2 Gbit/s), next generation Gigabit-Capable PON based on the International Telecommunication Union-Telecommunication Sector (ITU-T) G.987.1 standard is expected to deliver downstream speed at up to 10 Gbit/s [36, 37].