Optical CDMA transmission system simulations

Abstract
First of all, the behaviour of optical codes in multiple access scenarios will be studied. Different simulations will be performed modifying various system parameters and configurations of the network elements using software from VPI. New generation Passive Optical Network architectures will be used, being capable of transmitting information without any light source. Finally, the main goal of this study is to apply techniques for Rayleigh scattering penalties reducing in bidirectional transmissions where optical codes are used.

LIST OF FIGURES

FIGURE NO

TITLE

PAGE NO.
5

Internet users per 100 inhabitants between 1997 and 2007 OCDMA-PON architecture from the encoding point of view Behaviour of light travelling through the imperfections of the fiber glass

2 3 4 5 6 7

11

13 14 18 19 20

Rayleigh scattering losses versus wavelength. Source: CISCO FTTH PON architecture Simple scheme of the Rayleigh Backscattering contributions Fig. 3-6: Basic bloc diagram of an OCDMA architecture Fig.3-6:spectrum showing SPM effect on optical pulse

21

Fig.3-5:2-Dimensiona Time-Wavelength Optical Code Matrix, also denominated by Wavelength Hopping Time Spreading (WHTS) 22

10

Fig. 3-6: Series structure of an Encoder/Decoder using FBGs

25

11

Fig. 3-7: Parallel structure of an Encoder/Decoder using FBGs

26

12

Chapter 1
Introduction
Consumer bandwidth demands are growing at an enormously high rate, and are projected to grow for years to come. This growth rate applies not only to the entire Internet, but to a large range of individual institutions.59 % to 64 % of the downstream traffic is web media-related which is mainly because of photo and video communication and real-time streaming. Peerto-peer (P2P) traffic covers over one fifth of downstream and over 60 % of upstream traffic. Services alternative to P2P like file hosting and remote storage are gaining more interest Also the growing interest in voice, video and data delivery on the same infrastructure (tripleplay) has changed the common way of network usage, a necessity of running many applications on several devices connected simultaneously to a single access point has arisen Furthermore, when mature high-definition TV (HDTV) products become available the bandwidth demand will get even higher. Changing from the standard definition TV (SDTV) to HDTV, 1-3 M bit/s to 8 M bit/s per channel respectively, even not including services like video-on-demand (VoD), P2P and online gaming will exceed the capabilities of currently most popular digital subscriber lines (DSLs) drastically. Higher quality HDTV channels, where the information is less compressed, require a much higher bandwidth than the 8 Mbit/s.

Although the capacity offered by the asymmetric DSL (ADSL) technology has improved significantly from 512 Kbit/s (2001) up to 20 Mbit/s (2006), the most sophisticated protocols ADSL2+ and very high speed DSL 2 (VDSL2) are not able to satisfy the next-generation users' bandwidth hunger. This forces the efforts towards developing new generation networks that can offer enough bandwidth to satisfy the increasing needs of the users. Note that VDSL2 already needs the DSLAM to be closer to the user since the capacity decreases rapidly versus increasing distance. This bandwidth-hungry scenario created by both content providers and consumers stimulates the development of novel components and network architectures which should not only be capable of transmitting data at high bit-rates but should also be cost efficient. Latter is a necessity to make them particularly attractive to deploy for network operators and service providers. The physical layer of such network has to be capable of providing bandwidth on-demand, and, since the destination of the traffic load may change in time, the provision of the bandwidth should be made reconfigurable.

Fig. 1-1: Internet users per 100 inhabitants between 1997 and 2007

Also the international scenario is changing rapidly. The global economy doesnt stop growing pushed by the developing countries. These countries are experiencing a similar growth of internet users as the developed countries at the end of the last century.

Passive Optical Networks Architectures

The great majority of current copper-based access installations can not guarantee the increasing bandwidth demand of future users and need to be changed into advanced fiber architectures. Installing optical fiber in the access area is a great challenge in terms of capital expenses (CAPEX) and operational expenses (OPEX). There are many risks that have to be evaluated, it is important to choose the right technologies to guaranty the success of the new network. The decisions on the architecture should be made taking a low cost and low complexity point-of-view otherwise a large-scale commercial deployment may be less interesting. A single star architecture or point-to-point (PTP) provides a direct connection between the central office (CO) and the remote node (RN). Each connection has an optical line termination at the CO and at the RN. With this kind of architecture, the bandwidth provided to each user can easily be upgraded but its deployment costs are higher than other solutions like a point-to-multi point (PTMP) architecture. The cost efficiency of a PTMP architecture becomes more profound when the length and/or the amount of RNs increases.

For PTMP there is additional intelligence in the active node, such as optical amplification or wavelength routing abilities, where the signals are split and distributed to the RNs. This

increases the complexity thus the system costs because the active node requires powering and possibly housing of the components. PON has a passive optical distribution node where the signals are broadcasted to the RNs via power or wavelength splitting. Consequently, PON is a more robust and cost-effective option than the active double star because no active elements are in the field and all remaining active components as well as the intelligence are located at the CO.

Multiple access techniques

A multiple access mechanism has to be employed on the PON because the (feeder) fiber from the central office to the aggregation node is shared and, therefore, if the access is not controlled collisions occur at that node between information transmitted by different users in the upstream direction. The three more important techniques to provide access to the shared fiber link are based on assigning a time slot to each user (TDMA), a wavelength (WDMA) or an optical code (OCDMA). Currently, TDMA is widely implemented in broadband access networks and employing WDMA, either to multiplex TDMA streams or to provide multiple access, is considered as a powerful and future-proof upgrade. OCDMA has many attractive features that make it a promising technology for the (future) access networks. On TDMA system a central clock is present in the system to which all have to synchronize. When implementing TDMA on PON, synchronization difficulties arise and ranging techniques have to be employed in order to properly schedule the transmitted data. Additionally, the system bandwidth increases significantly if a high throughput for each channel is required. TDMA may also be applied on top of a WDM scheme to provide a very high transmission capacity. A WDMA system divides the frequency spectrum into discrete slots whereby each slot is assigned to a user. The resolution of such frequency slots may range from coarse to dense depending on the amount of simultaneously active nodes allowed on the network.

Commonly, in order to accommodate a large amount of active wavelengths, high quality optical sources are installed because of narrow wavelength spacing. In an OCDMA system, a unique and (pseudo-)orthogonal optical code (OC) signs each data bit transmitted by a user which allows the optical carrier to be asynchronously shared with other users on the network. The OCs are designed in such a way that, on the one hand, if the encoded data matches with the proper key at the receiver the autocorrelation is maximal and, on the other hand, if there is no match the cross-correlation approaches zero.

1.1 Goals and motivation

In this project, the performance of PON-OCDMA architecture will be studied. This study will focus on a specific design of these networks where the communication between the Central Office (CO) and the user is made over a single fiber. On this scheme, the bidirectional transmissions are taken place at two different wavelengths for the- 6 -upstream and the downstream data using colorless and sourceless Optical Network Units(ONU). The technique used for this architecture is an adaptation of the Code Division Multiple Access (CDMA) widely used on wireless communications. This Optical CDMA (OCDMA) is based on the transmission of optical codes formed by pulses. In this study we will deepen on the particular case of two-dimensional codes, constructed on time and wavelength dimensions. The light source is located at the CO, for the downstream and upstream. The ONU has a Reflective Semiconductor Optical Amplifier (RSOA) that has the role of modulating this input signal from the CO by reflective modulations, so the information will be sent back to the CO with the upstream data. One of the problems of this kind of bidirectional transmissions are the optical impairments produced by the fiber. One of these intrinsic properties of the material forming the fiber produces the Rayleigh Backscattering (RBS)

effect. This effect produces interferences and in-band crosstalk, causing distortion on signals propagating in opposite directions. It eventually causes penalties on the transmission. This is one of the most important issues to be solved on this kind of schemes. These optical impairments produced by RBS are studied on this thesis. Techniques to reduce the transmission penalties produced by this effect are proposed and its performance evaluated.

Chapter 2 State of the Art

In this chapter, the state of the art of the topics to be studied will be presented. OCDMA technology will be described and explained. It is also explained in this chapter how RBS is produced and what impairments can it induce to optical signals traveling through a fiber. Finally, there will be a brief description of how SPM can be induced and how an optical pulse is distorted when it is self-phase modulated.

2.1 Optical Code Division Multiple Access

Recent improvements on technologies like Wave Division Multiplexing (WDM) or Time Division Multiplexing (TDM) have boosted the development of optical networks for Metropolitan Area Networks (MAN) and Wide Area Network (WAN) environments. However these technologies have known limitations on protocols and hardware that add unnecessary complexity and costs. The limits on the number of users or access latencies of these technologies can be resolved using OCDMA. The optical networks based on OCDMA are a clear example of the extrapolation to the optical domain of a technology widely used on actual wireless networks. OCDMA is capable

of providing a gigabit- or even multi gigabit-per-second for each user both in the up- and downlinks, and OCDMA over WDM PON could be one of the most promising system architectures that can break through the last/first mile bottleneck. It effectively provides an extra dimension on top of each wavelength, meaning that each wavelength is shared by many users and thereby increasing the total available capacity. There are clear advantages that make OCDMA a suitable access technology for passive optical networks with large number of users that randomly use the connection. Some of these advantages are: - There is no need of high quality lasers - There is no centralized network control because of the simplicity of the protocol - Self-routing by code sequence and characterized by a high security in the network - Effective utilization of bandwidth and high tolerance to noise - Low-cost devices - Random and simultaneous access protocol. No need for the strict timing Synchronization

Depending on how the codes are represented in the optical domain, some advantages may not apply. The multiplexing procedure by which each user will encode its data is distinguished by a specific optical code rather than a wavelength or time-slot. An encoding operation optically transforms each data bit before transmission. At the receiver, the reverse decoding operation is required to recover the original data. OCDMA uses this procedure to establish the access among multiple network users because the codes are (pseudo-)orthogonal. OCDMA offers flexibility with coding, large cardinality, and the code conversion. OCDMA coding creates a soft limit in terms of a number of users, allowing the network to operate at different levels of BER performance. It has been shown that this is an important advantage of OCDMA systems; the network can easily be adapted to various networking requirements or load conditions, which might not be feasible in more rigid WDM based networks. Dynamic networks can therefore benefit from the OCDMA routing. A large offered cardinality is also an advantage of the OCDMA: powerful 2-D coding can be used to create a large pool of codes while using few wavelengths. This increased granularity eliminates many routing impairments.

In this study we will consider the use of two dimensional codes. They are constituted by a sequence of short pulses, arranged in time at different wavelengths. On-Off Keying modulation (OOK) is performed, that is, if a logical 1 or 0 arrives the code will be transmitted or not. At the receiver side, with a mechanism of delay lines, the opposite process will be performed, reconstructing the pulses in time. It is clear that a bit slot is divided into smaller slots which are referred to as the chips of an optical code. The different pulses of the code will add up in intensity at the receiver side. If a correct code arrives, the receiver will have a high intensity input while a low intensity input will be received for incorrect codes. This forces the need of thresholds to define where is the limit between correct and incorrect codes. This kind of encoding is cost efficient, because all operations can be done in the optical domain, using multiple lasers and delay lines, reducing access time and increasing bandwidth limits.

Fig. 2-1: OCDMA-PON architecture from the encoding point of view OCDMA systems may be classified according to their coding principle (incoherent or coherent) and their coding domain (time, wavelength, or time and wavelength). The properties of these codes determine the characteristics and capabilities of the network. The more users the networks has, more complex the codes will be and more resources will use, otherwise the performance of the network will deteriorate. For several reasons, an active user may cause a crosstalk to all the other active users. This is referred to as Multiple User Interference (MUI). The MUI is not the focus of this study. With one set of codes access to the network can be given to 40 users, having the bit error rate below one erroneous bit per million.

2.2 Rayleigh Backscattering

The Rayleigh scattering is the effect responsible of reflections and refractions of light when passing through the glass inside the fiber. The intrinsic properties of the material with which the fiber is made and the non-ideal manufacturing process making the glass imperfect. The fiber is characterized by microscopic variations in the material density, compositional fluctuations, and from structural in homogeneities or defects. The fiber manufacturing is a complex process and many different chemicals are involved that need to be mixed as uniformly as possible. This structure contains regions in which the molecular density is either higher or lower than the average density of the glass. These effects change the refractive-index of the glass along the fiber, causing deviations on the lights path,

losses and even interferences given that the light can be reflected and travel into the opposite direction. The refractive index is a way of measuring the speed of light in a material. Light travels fastest in a vacuum, such as in outer space. The actual speed of light in a vacuum is about 300,000 kilometers (186 thousand miles) per second. Index of refraction is calculated by dividing the speed of light in a vacuum by the speed of light in some other medium. The index of refraction of a vacuum is therefore 1, by definition. The typical value for the cladding of an optical fiber is 1.46. The core value is typically 1.48. The larger the index of refraction, the slower light travels in that medium. Therefore, nonlinear refractive indexes cause the light to travel at different speeds, clearly degrading the signal.

Fig. 2-2: Behaviour of light travelling through the imperfections of the fiber glass This effect degrades the quality of the signals travelling through the glass produces an appreciable attenuation of the light. These losses are called Rayleigh scattering loses and have to be considered when designing optical networks where information is transmitted in both directions using a single fiber. Structural in homogeneities and defects created can also cause scattering of light out of the fiber. These defects may be in the form of trapped gas bubbles, un reacted starting materials, and crystallized regions inside the glass. In general, the perform manufacturing methods have evolved and these extrinsic effects are minimized to the point where they are negligible compared to the Rayleigh scattering losses. Scattering directly depends on the size of the particles relative to the wavelength of light. The closer the wavelength is to the particle size, the more scattering. This is why the amount of scattering increases rapidly as the wavelength decreases. This produces that the system performance will directly depend on the laser line-width. In this study we will deepen into the

particular case of Rayleigh backscattering losses. The light passing through the imperfections of the fiber is reflected in many directions, in the case of backward reflections we will have an optical signal at the same wavelength of the original towards the emitter.

Fig. 2-3: Rayleigh scattering losses versus wavelength. Source: CISCO This is an important issue when we talk about bidirectional transmissions. If we are sending information in both directions at the same wavelength, the Rayleigh backscattering effect will not only insert losses but also get in-band interferences of the signal coming from the opposite direction. The degradation of the signal depends on the power ratio between the copropagating signals, so called Attenuation to Crosstalk Ratio(ACR). In systems with narrow line width optical sources, crosstalk penalties due to inter ferometric conversion of laser phase noise to intensity noise of the backscattered light can far exceed the crosstalk penalties observed in bidirectional systems using incoherent sources. Thus, Rayleigh backscattering is a limiting factor for bidirectional transmission. Rayleigh backscattering noise power grows with the fiber length to saturation at a specific value of the launched signal power multiplied by the Rayleigh backscattering coefficient. This means that the signal will also be deteriorated for long fibers because the receiving signal power of the counter-propagating signals will go down due to fiber attenuation, while Rayleigh backscattering noise power will remain constant.

2.3 Self Phase Modulation

Self Phase Modulation (SPM) is an effect caused by the intensity dependence of the refractive index in nonlinear optical devices. This effect causes spectral broadening of optical pulses. SPM can be observed on solids and glasses by using high intensity and short pulses. The physic fundamentals of self phase modulation on optical fibers are explained by the Kerr effect, also called the quadratic electro-optic effect discovered in 1875 by John Kerr, a Scottish physicist. This is the special case in which a slowly varying external electric field is applied by, for instance, a voltage on electrodes across the fiber. Under this influence, the glass becomes birefringent with the different indices of refraction for light polarized parallel to or perpendicular to the applied field. This means that the optical ray will be decomposed into two rays caused by the different refractive indexes. This difference of the refractive index, n , is given by n KE^2 Where is the optical wavelength, K is the Kerr constant, and E^2 is the pump intensity. Therefore, when the light travels through crystal or glasses, the refractive index will be modified proportionally with the intensity of the pulse. In the case of light generated by semiconductor lasers or amplified by semiconductor optical amplifiers (SOAs), a high intensity signal reduces carrier densities, which in turn leads to a modification of the refractive index and thus a phase change per unit length during propagation. The self phase modulation in SOAs is an effect produced as a result of index nonlinearities induced by gain saturation. This is an important issue when talking about short and powerful pulses arriving to a semiconductor optical amplifier. The amplifier can be saturated by the pulse edge and the gain available for the trailing edge will be reduced. This will induce gain saturation modifying the linearity of the refractive index and pulse distortion. The chirp of an optical pulse is usually understood as the time dependence of its instantaneous frequency. Specifically, an up-chirp (down-chirp) means that the instantaneous frequency rises (decreases) with time. When an un-chirped pulse arrives to the amplifier the intensity is highest at the peak of the pulse and therefore, that part of the pulse experiences the highest refractive index and as a result, propagates slower than the leading and trailing parts of the pulse. This causes the carrier wave to stretch out on the leading part of the pulse and to pile up on the trailing part of the pulse. The resulting spectrum will be then modulated and broadened.

When an electromagnetic wave interacts with the bound electrons of a dielectric, the medium response, in general, depends on the optical frequency w. This property, referred to as chromatic dispersion, manifests through the frequency dependence of the refractive index n(w). Fiber dispersion plays a critical role in propagation of short optical pulses because different spectral components associated with the pulse travel at different speeds given by c/n(w). Even when the nonlinear effects are not important, dispersion induced pulse broadening can be detrimental for optical communication systems. If the effects of SPM and chromatic dispersion are combined, the result is amplitude distortion when the pulse is in the anomalous dispersion region. The leading edge of the chirp induced by the SPM will be shifted to the lower frequencies (red-shift) and the trailing edge to the higher frequencies (blue-shift). In the presence of chromaticdispersion, the red-shifted and blue-shifted parts of the pulse travel at a different speed, causing amplitude distortions. The chirp caused by the self-phase modulation cannot be eliminated as this would require a material with a negative Kerr effect. The impact can be reduced by choosing a positive chromatic dispersion coefficient, so the chirp will be cancelled partially.

Chapter 3 System setup

3.1 Architecture description
In this chapter a basic OCDMA architecture will be described, its components analyzed and its performance evaluated. The objective is to know the structure, the possibilities and performance of this kind of new generation Passive Optical Network architectures. The particularity of this kind of schemes is the transmission of pulses between the ONU and the Optical Line Termination (OLT) at the CO instead of continuous signals using OOK modulations. These pulses are transmitted such that together they will form a unique code identifying each user of the network. These codes can be constructed using 1 (time) or 2 (time, wavelength) dimensions and can have different lengths. The length and the dimensions used to form the code will define the characteristics and performance of the system. In order to implement the multipoint-to-multipoint communications, the core networks usually employ the ring and mesh network topologies and the local area networks commonly exploit the star, ring and bus network topologies. The access network is a type of service distribution networks, which may utilize the point-to multipoint tree architecture to form a passive optical network (PON), as shown in Figure3-1. This structure facilitates low cost and high bandwidth utilization of fiber-to-home(FTTH)/premises (FTTP), fiber-to-

building/business (FTTB) and fiber-to-cabinet (FTT Cab). The upstream and downstream links can employ different optical fibers to achieve space division or different wavelengths to realize wavelength division.

Fig. 3-1: FTTH PON architecture

The considered scheme of this study is composed of a colorless reflective ONU with no optical source that can work at any wavelength of the same window, situated at the user-end. On the other side of the network there is an OLT that generates the light remotely, being this component the only active of the system. The CO has the task of sending the (modulated) information downstream and the (unmodulated) optical source signals to the users at the different wavelengths. This optical source will be in form of a train of pulses that will be used by the ONUs to encode their information and send it back to the CO, forming the upstream data. In other kind of bi-directional PONs the light source is a constant signal, being used at the user side to modulate the upstream data. The ONU has two major tasks. Firstly, it terminates the optical path and converts the downstream data to the electrical domain and, secondly, it transmits the upstream data after converting it from the electrical domain to the optical domain. The transmission process of upstreaming data is carried by a reflective modulator. This device uses the optical pulses received from the CO to send the digital signal using optical code and OOK modulation. This module has no optical source and can work at any wavelength considering that it works like a mirror with amplification.The advantage of this network approach is that wavelength referencing and control is provided by CO and thus the more complex devices (high

performance lasers and processing units) are less in number and as a result the costs are reduced dramatically, the devices at the end-user are the simplest possible. Also passive optical components alleviate maintenance requirements in the access network. The node in between the CO and ONU of Fig. 3-1 is a passive coupler. All this transmissions can be done using two fibers, one for the upstream data (and the pulses sent from the CO) and the other for the downstream data. In this study a single fiber transmission will be considered. These bidirectional transmissions have counter-effects like in-band crosstalk caused by reflections inside the fiber and Rayleigh Backscattering .

Fig. 3-2: Simple scheme of the Rayleigh Backscattering contributions

As studied in, RBS is an important issue on bidirectional transmissions. Figure 3-2 shows the two contributions to the Rayleigh Backscattering effects in our scheme, interfering with the upstream signal sent to the CO. The first contribution is generated by the carrier being delivered to the ONU and the second contribution by the modulated upstream signal at the output of the ONU. The backscattered light re-enters the ONU where it is re-modulated and reflected towards the Rx. The relatively low power backscattered light is also amplified by the reflective modulator, which usually provides net gain.

3.2 Optical CDMA architecture principles and components

The system to be described is designed to connect 4 users simultaneously at a distance of 25.4 km from the CO. The system transmission speed is set to 1.25 Gbps due to restrictions on the bandwidth of the RSOA. Figure 3-3 shows the setup to be described on this chapter. The distance consists of 20 km of Standard Single Mode Fiber (SSMF) with 5.4 km of Dispersion Compensating Fiber (DCF) to compensate for the dispersion.

Fig. 3-3: Basic bloc diagram of an OCDMA architecture

The scheme shown in figure 3-3 describes a basic OCDMA setup. The CO side is divided in three parts: the optical pulse source (to be used by the ONU), a second pulse source that will be encoded and modulated (for the downstream data), and the receiver with the decoder (upstream data). After the CO the fiber link is placed and splitters will connect several users to the same CO. The ONU is a simpler device, composed of an RSOA and a encoder that will

send the upstream information, and a receiver with the decoder that will receive the downstream data.

3.2.1 Pulse amplification

The optical pulses used in this architecture are generated at the CO. These pulses are designed to travel through 25.4 km of fiber and be reflected and amplified at the user side by an RSOA. Since pulses behave differently from continuous wave transmissions, the nonlinear effects of pulse amplification have to be considered. Amplification of pulses using semiconductor amplifiers may introduce nonlinear effects and distort optical pulses when being amplified. Dispersive effects are not important for SOAs because of small material dispersion and a short amplifier length (<1 mm in most cases). These effects occur when the pulse width is short and optical power high, causing saturation to the amplifier and modifying the pulse shape. Seen the importance of pulse width and power , the pulse can be analyzed by using its partial energy.

Fig. 3-4: spectrum showing SPM effect on optical pulse

The frequency chirp is larger for more energetic pulses simply because gain saturation sets in earlier for such pulses. Self-phase modulation and the associated frequency chirp can affect light wave systems considerably. The spectrum of the amplified pulse becomes considerably

broad and contains several peaks of different amplitudes. The dominant peak is shifted toward the red side and is broader than the input spectrum. It is also accompanied by one or more satellite peaks. The temporal and spectral changes depend on amplifier gain and are quite significant for G0 = 30 dB. The experiments performed by using pico-second pulses from mode-locked semiconductor lasers confirm this behaviour. In particular, the spectrum of amplified pulses is found to be shifted toward the red side by 50100 GHz, depending on the amplifier gain. Spectral distortion in combination with the frequency chirp would affect the transmission characteristics when amplified pulses are propagated through optical amplifiers.

3.2.2 Optical Codes

The optical codes used by OCDMA systems can be divided into two branches: coherent and incoherent codes. Coherent codes are formed via phase coding, modulating the optical signal using phase modulations. Incoherent codes that use standard techniques of intensity modulations with direct detection. The most common approaches to incoherent OCDMA are based on spectral amplitude coding, spatial coding, temporal spreading , and 2-D wavelength hopping time-spreading (WHTS) coding. Coding in multiple dimensions, such as- spectral amplitude and WHTS, adds more flexibility while increasing the capacity and performance. Incoherent OCDMA systems are preferred because it presents robustness to environmental conditions, nonlinearity, and coherent effects and utilizes commercial off-the-shelf components. Coherent codes have worse performance when talking about signals propagating through multiple hops .

Fig. 3-5: 2-Dimensional Time-Wavelength Optical Code Matrix, also denominated by Wavelength Hopping Time Spreading (WHTS) In OCDMA systems, the pulses are encoded in both the time domain and the wavelength domain simultaneously: the codes can be represented as matrices with time and wavelength on the two axes. The wavelength domain is divided into n channels of bandwidth n and the time domain into m time chips of duration t. A code consists of K short pulses of different

wavelengths, where K is called the weight of the code, within the m available time chips.The biggest merits of 2-D codes are the significant increase of the cardinality, which makes the numbers of users and simultaneous subscribers in a network increase enormously, and the reduction of the code length, which results in the data rate of single user to boost largely. Most recent implementations of 2-D OCDMA utilize multi-wavelength sources to avoid the need for sources that can rapidly wavelength hop. The pulses can be generated from an array of mode-locked lasers, which is limited in scalability due to the higher complexity in controlling a large number of lasers or from a broadband source. The encoder essentially creates a combination of two patterns: a wavelength-hopping pattern and a time-spreading pattern. Some implementations perform the two patterns independently like array waveguide gratings (AWG's) or thin-film filters (TFF's), or simultaneously as fiber Bragg gratings (FBG's), holographic Bragg reflectors (HBR's) , or chirped Moire gratings (CMG's) . The position and length of the delay lines will determine the codes that will be created. The position of the wavelengths within the bit interval can be changed in order to create a different code from the same code family. We define a code family y as the total number of valid codes, which is also called the cardinality, that can be obtained with the same n wavelengths and m chips. A valid code has to satisfy specific values of cross and autocorrelation. The cross correlation of a code refers to the interference caused to other users which is an additive effect in the network. In that sense, as more codes are active, the more mutual interference is experienced. Orthogonal codes have a crosscorrelation of 0; however, many code-sets have relaxed requirements and are designed to be pseudoorthogonal. Physical imperfections also have their impact on the orthogonality of the codeset. While an encoder spreads the wavelengths in time over the bit period, on the decoder side there is a system of delay lines that are ordered in the way that the code will be reconstructed. The pulses, from the different wavelengths are realigned in time and summed up to create a peak of height K. Then a threshold will determine weather or not the code is right, whence the data was sent to that user.

3.2.3 Encoder/Decoder
As discussed previously, there are different kinds of encoders/decoders in OCDMA systems. In this section two kinds of encoders/decoders will be described, using TFFs and BFGs.

Optical Encoders and Decoders using TFFs A broadband light source or a multi-wavelength light source outputs short pulse trains with the repetition rate equal to the bit rate, which are modulated by the input data in an optical modulator. For the on-off keying modulation, the transmitter only outputs an optical pulse when a transmitted data bit is 1. Then, the narrow optical pulse corresponding to the data bit 1 is fed into a 2-D incoherent WHTS optical encoder to perform encoding. The fixed 2-D incoherent WHTS optical encoder consists of a 1:W wavelength division demultiplexer, W fixed fiber-optic delay lines and a W:1 wavelength-division multiplexer, with W the number of wavelengths. The short optical pulse corresponding to data bit 1 is therefore firstly separated into W optical pulses. These pulses will have a defined spectral width and frequency distance, which depends on the quality of the filters and the transmission characteristics of the network that we will be working on. These W optical pulses are finally combined and output by the wavelength division multiplexer. With this procedure, the 2-D WHTS optical encoding of data is achieved.

Optical Encoders and Decoders using FBGs

A two-dimensional incoherent WHTS optical encoder and decoder can be composed by fiber Bragg gratings (FBG's) to reflect the optical signals with different wavelengths, which adopt a series [18] or a parallel structure [19]. A fixed optical encoder and decoder using a series structure consist of an optical circulator with three ports, a circulator, W FBGs and W fiberoptic delay lines. A short optical pulse coming from a broadband light source is firstly modulated by data in a modulator and then the modulated narrow optical pulse is fed into port 1 of the optical circulator. The optical signals output from port 2 of the circulator are reflected back by FBGs with different wavelengths. Meanwhile, two adjacent FBGs are connected with the delay lines with different lengths in order to implement the desired delays in terms of the requirement of a two-dimensional codeword. The returning optical pulse signals with different wavelengths and different delays exit the encoder at port 3.

Fig. 3-6: Series structure of an Encoder/Decoder using FBGs

The structure and operational principle of the optical decoder is similar to its corresponding encoder. The necessary variations are to put these FBGs in reverse order and to change the delays of fiber-optic delay lines between two adjacent FBGs in order to make their delay values be complementary values of those in its corresponding encoder. The structure of a tunable optical encoder/ decoder using serial FBGs is the same as that of its fixed encoder/decoder except that the fixed delay lines are changed into tunable delay lines. However, this would result in large optical power loss. Hence, such a structure is only suitable for the implementation of fixed encoder/decoder. A tunable optical encoder and decoder using parallel FBGs, shown in Figure 8, consists of a circulator with three ports, an arrayed waveguide grating (AWG) wavelength-division demultiplexer, W FBGs and W fiber-optic delay lines. After the short optical pulse from a broadband light source is modulated by data, it is fed into port 1 of a circulator. The optical signal output from port 2 is decomposed into W parallel output by the AWG. They pass through the fiber-optic delay lines with different lengths and are reflected by FBGs with different wavelengths, and return into the AWG again. The resulting optical pulse trains multiplexed by the AWG enter the optical circulator from its port 2 and then output from port

3 of the circulator. For the decoder, the same functional structure is preceded for the decoding.

Fig. 3-7: Parallel structure of an Encoder/Decoder using FBGs .

Amplifier gain
The optical gain, in general, depends not only on the frequency (or wavelength) of the incident signal, but also on the local beam intensity at any point inside the amplifier. Details of the frequency and intensity dependence of the optical gain depend on the amplifier medium. The gain is maximum when the incident frequency w coincides with the atomic transition frequency w0. The gain reduction for w=w0 is governed by a Lorentzian profile that is a characteristic of homogeneously broadened two-level systems [4]. Amplifiers with a relatively large bandwidth are preferred for optical communication systems because the gain is then nearly constant over the entire bandwidth of even a multichannel signal. Gain saturation Ps, depends on gain-medium parameters such as the fluorescence time and the transition cross section. Since g is reduced when P becomes comparable to Ps, the amplification factor G decreases with an increase in the signal power. This means that there is a limit on the input optical power that the amplifier can amplify. This limit is determined by the SOAs carrier pumping capacity. When the input optical power reachesthis limit the amplifier gain will decrease.

3.3 Studied architecture

In this thesis, a simpler version of an Optical CDMA architecture will be studied in order to simplify the influences of the transmitted signal and focus on pulse propagation and RBS reduction. Therefore, information will not be encoded and just a train of optical pulses will be transmitted to the user. It is expected that by studying a single pulse transmission, the RBS effects at the larger OCDMA system can be understood.

Fig. 3-8: Scheme used for VPI simulations (simplified)

Figure 13 shows the schematic used in the following chapters for performance simulations. There are some OCDMA characteristic components used in this schematic are: 1. Data generator, this device has the role of creating a random sequence of data that will be transmitted simulating the users upstream transmission. 2. RF signal, explained in following chapters it implements a RBS reduction technique based on bias dithering.

3. Delay line, which purpose is to synchronize users data with the optical pulses sent from the ONU side. It has a fixed value and its optimization is explained on following chapters. 4. Reflective Semiconductor Optical Amplifier, used for the reflective On-Off 5. Filter, optimized to filter any noise out of the pulse bandwidth, specially ASE noise produced by the RSOA. 6. Fiber link, composed by a Single Mode Fiber (SMF) and a Dispersion Compensating Fiber.

Chapter 4
System parameters
The system to be studied is based on OCDMA architectures and therefore is characterized by the transmission of optical pulses. These pulses will have different behaviours when passing through the components and elements of the system. A detailed description of all these aspects of the pulse transmission will be performed in this chapter. The pulses to be transmitted from the CO will be first analyzed. Eye diagrams and spectrums will be explained after passing through the fiber. It will be shown how the RSOA behaves under different bias conditions and how the signal to be amplified is affected in terms of noise and distortions showing how filtering at the output of the RSOA may improve the quality of the signal. The importance of optimizing data delay will be emphasized. Finally, the quality of the signal arriving to the receiver will be analyzed.

4.1 Pulses
The CO, as discussed in previous chapters, has the role of sending an unmodulated and continuous train of Gaussian pulses towards the user, in order to enable the reflections of these pulses and modulation by the user for upstream data transmission. A reduced scheme will be studied where no codes are used and only one user will perform data transmission at a single wavelength. Pulses will be sent from the CO at a rate equal to the user data bit rate, and therefore each pulse will transport one bit when reflected at the ONU side.

Fig. 4-1: Pulse train transmitted by the CO.

The pulses shown in figure 4-1 are a time domain representation of three pulses at the output of the pulse laser at the CO. They are first-order Gaussian pulses, separated by 0.8 nano seconds, a duration (or Full Width at Half Maximum FWHM) of 19 pico seconds, and a peak

power of 1 mW. These pulses are transmitted at a frequency of 193.1 THz, using a 1550.92 nm wavelength, and have a bandwidth of 50 GHz. In this system no chirping will be induced to the pulses sent from the CO.

4.2 Fiber
The fiber used in this system is composed of a Single Mode Fiber (SMF) preceded by a Dispersion Compensating Fiber (DCF) used to suppress dispersion impairments induced by the SMF. Dispersion is a phenomenon in which the phase velocity of a wave depends on its frequency. Dispersion is sometimes called chromatic dispersion to emphasize its wavelengthdependent nature. A single optical signal will travel at different speeds, depending on the geometry of the fiber and the lights spectrum width. The signal distortion will depend on the length of the fiber mainly because the varying delay in arrival time between different components of the signal produces distortion in time. To decrease this degradation a dispersion compensating fiber is placed before the main fiber. Its role is to pre-compensate the distortion that the SMF fiber will introduce in the downstream direction. In the other direction, the DCF will compensate for the dispersion. For this, it is necessary to calculate the total distortion that the fiber will introduce so that the DCF has the right length and the arriving signal is undistorted. In this study a 20 Km SMF is chosen, having a dispersion factor of 17*10-6. The DCF is therefore chosen to be 5,397 Km of length having a distortion of -63*10-6 which equals the D*L of the SMF. The pulse becomes wider and delayed in time in order to compensate the dispersion that the SMF will introduce. After the SMF, the pulse goes returns to its original shape but attenuated.

4.4 Data delay Data delay is an important parameter to be set for performance optimization. As the basis of the system is pulse transmission, data and pulses have to be synchronized. Pulses can arrive at any time to the RSOA because of different system delays. It is optimum to have the RSOA fully charged by electrons when the optical pulse has to be amplified, and as the data signal is the carrier of the bias current, it is determinant to synchronize data with pulses. It will be assumed that the system does not loose synchronism and therefore the delay will be a constant parameter. This parameter will depend on separation between pulses, and therefore in the range of 0 to 0.8 nanoseconds.

Fig. 4-2: System performance versus data delay

The importance of the data delay is explained by the level of charge that the cavity of the amplifier has. It is optimum to have the amplifier full of electrons ready to recombine and start light emission right when the pulse inputs the device, as shown in figure 4-14. If the arrival of the pulse concurs with the data, the pulse will saturate the amplifier and gain may

be negative. Even attenuation or distortion of the pulse may occur when the delay is not optimized. Figure 4-12 shows the importance, in terms of system performance, of the data delay. No optimization of this parameter may introduce a full error transmission, disabling data transmission. This oscillation of the system performance when modifying the delay can go from almost error free rate to a maximum error rate. In case of no delay, the data and pulses are synchronized to the internal clock of the simulation environment. The objective is to have the pulses corresponding to the 0-level as attenuated as possible and the pulses corresponding to the 1-level as powerful as possible, so that the difference in power between them is the maximum.

Fig. 4-3: Representation of pulses and data arriving to the RSOA without data delay

4.5 Receiver
After travelling through the fiber and be reflected and amplified by the RSOA the pulses arrive to the receiver and are converted into the electrical domain. In this process a photodetector, filters and operational amplifiers are involved in order to have an optimal electrical signal and calculate system performance. Photo-detectors generally have associated operational amplifiers because the voltage that outputs is too low and noise can distort the

signal when going through the decoding circuits. Before this amplifier a filter is introduced to eliminate everything out of the data bandwidth, in our case 12 GHz. The minimal required filter bandwidth can b calculated by considering the Time-Bandwidth Product (TBP) for chirp-free (transform limited) 1st-order Gaussian pulses. The TBP is a multiplication of the pulse duration and the bandwidth at Full-Width at Half-Maximum (FWHM). In this case, the TBP is greater or equal than 0.44 so the bandwidth is equal to 23.15 GHz. The Nyquist theorem then stipulates that the minimum receiver bandwidth should be equal to 0.5*23.15 = 11.6 GHz. This filter is especially important when applying spectral broadening techniques for RBS reduction. These techniques will introduce power out of the pulse bandwidth and it is preferred to be eliminated at the receiver side using low pass filters. It can be seen in figure 415 that the shape of the pulses remains almost unchanged. The photo-detector is optimized for this particular pulses and the optical to electrical transition is achieved linearly. The figure also shows that the pulses voltage is in the order of micro-volts, which would be low for data treatment after being decoded. The amplifier introduces a sensible gain to the pulses, passing from micro-volts to hundreds of micro-volts. In the other hand, thermal noise is added reducing the quality of the signal. There is also a pulse broadening because the pulse exceeds the filter bandwidth and cuts high frequencies. For all simulations and system analysis, a receiver attenuation has been firstestablished in order to reference all changes in performance to a 10-9 bit error rate.

Fig. 4-4: Bit Error Rate versus Receiver attenuation

The system performance has a linear behaviour when increasing the receiver attenuation. On a real system, this attenuation would be the power margin at which the system is guaranteed to work under error-free conditions. Thus, in the following system simulations, the receiver attenuation will be set to 21 dB.

Chapter 5
System Performance
The objective of this study was to apply two different spectral broadening techniques in order to reduce RBS impairments in bidirectional OCDMA transmission systems. The functionality of these techniques has been shown in previous chapters. In this chapter, the receiver behaviour when applying the techniques will also be analyzed Finally the behaviour of the system will be presented in terms of bit error rate for each of the techniques. In this chapter, a comparison of RBS performance will be done. It is known that RBS impairments directly depend on the optical power that travels through the fiber. There actually may be a power threshold that determines at which power level RBS starts influencing the system performance. For the BER performance simulations shown in this chapter, a comparison at different power levels will be done so that this RBS power threshold influence can be also shown.

5.1 Original setup A characteristic of RBS impairments is its power dependence. The more power istransmitted through the fiber, the more power is reflected by the fiber nonlinearities andimperfections. This causes attenuation of the transmitted signal and induces crosstalk tothe signal travelling in the opposite direction.On the original setup the RSOA is driven by a data modulated bias current set to55 mA. The power output by the RSOA is low enough to not induce any RBS effect as itcan be seen on the BER curves of figure 6-1. The difference of performance between curves when no RBS effects are turned on and all RBS impairments is almost none. The pulse power threshold where RBS effects are introduced into the signal is about 3dB higher than the pulses pumped by 55mA of bias current from the RSOA. Setting the bias current to 70mA, this threshold is over passed. The penalty introduced by RBS is less than 0.5 dB. It is a low impairment but also low power transmission.

Fig.5-1: Original setup system performance when data amplitude is 55mA

Fig.5-2: Original setup system performance when data amplitude is 70mA

5.2 Bias Dithering

The main effect produced by RSOA bias dithering, at the optimum settings, is an improvement on the amplifier gain. Therefore, the use of this technique will increment the output power of the amplifier, if comparing it with a no dithered RSOA. It will be expected to have a RBS influence on the system performance at lower data amplitude values compared with the original no dithered system.When using a system with a data amplitude of 70mA a 2 dB penalty can be seen, while using a 55mA system a 1.5 dB penalty is shown. Its worth mentioning that the bias dithering technique introduces a power gain in comparison with the original system. Even though the RBS penalty is bigger, the distance that a pulse can travel through the fiber is longer, possibly making it feasible for specific networks.

Chapter 6
Conclusions and further work

After presenting the system, describing the different components composing the system, explaining the techniques and analyzing the system performance a final conclusion can be made. Comparing both techniques with the original system setup, it is clearly seen that the RBS penalty is lower when the optical power is low. On the other hand, higher pulse powers guarantee longer transmission distances. OCDMA networks are characterized by tree topologies, where many splitters and long fiber links compose the architecture. When talking about a reduced number of users, it may be useful to reduce the amount of power sent into the fiber in order to avoid RBS impairments. When this number increases, the splitting factor and the distance covered also increase forcing the OLT to transmit powerful pulses, making RBS impairments inevitable. When working with large amount of users, a solution to these impairments must be chosen. Bias dithering does not clearly improve RBS penalties but its effects increase the amplification capacity of the RSOA. It is also complex to make it work in an optimum way because of the elements that are involved. If we maintain the system BER under 10-9, an improvement of 1 dB at the received power is achieved when modulating the bias current at 70mA. The penalties produced to the overall system performance when introducing the bias dithering should also be taken into account. For the case of induced SPM, there is a clear difference in terms of transmitted power. The laser at the CO sends a train of 6 power-peak pulses, arriving with a 8 dB attenuation to an RSOA fed by 70mA of bias current. RSOA outputs pulses with a broader spectrum, which explains the good behaviour under RBS conditions. RBS penalty through a chirped pulse (produced by SPM) is in the same penalty orders as the original system but with a 2 dB decrease on the power needed to guarantee a BER under 10-9. It can finally be concluded that inducing SPM at the RSOA, the system performance improves. Pulses are transmitted with more power, which means further travel distances or higher splitting ratios are possible, and pulse spectrum is broader reducing the influence of RBS impairments. All this is achieved by only tuning the power sent by the laser at the CO, to the point where it saturates the RSOA enough to broaden the spectrum but without introducing a big pulse distortion.

Further work
The application of these techniques to scheme closer to a real OCDMA setup maybe a challenge but it is the first thing to do. It would be interesting to see the performance on a setup of more than one user, where splitters and longer fibers are involved. The encoding and higher attenuation may be a challenge when applying bias dithering and SPM for RBS reduction. A further study of SPM effect may be helpful in order to understand its behavior so the elements of the system can be optimized. The introduction of different filters with determined work bands, on different parts of the scheme may improve the pulse quality. Also inducing SPM or finding the way of chirping the pulses from the begging of the trip, at the laser side or through the fiber can also be studied, so RBS will be reduced on the downstream direction. The performance of introducing bias dithering may improve if a real half wave rectifier is used, so the amplification of logical zero pulses reduces. This way the pulse eye will be bigger, enabling a better transmission. Also, a further optimization of the dithering frequency may be a viable option. The RSOA is a complex device, and the overall system performance could be improved adapting the dithering to the limits of the amplifier. Spectral broadening should be achieved by choosing the right RF parameters or pulse shapes.

A. References