Digital implementation of OFDM
The FFT-based OFDM uses IFFT and FFT algorithms to generate and demodulate the OFDM signal. Indeed, as in the case of the analog OFDM, the incoming data stream is first partitioned into blocks of N symbols. The inverse fast Fourier transform (IFFT) is then applied to each block of N symbols to generate a digital OFDM symbol. Considering the orthogonality of subcarriers (i.e. fk = Tk , ∀k ∈ [0, N − 1]), the nth sample of x(t) sampled at NT can be expressed as
N −1 N −1
x(n) =X(k)exp j2π k nT = X(k)exp j2π kn (1.4)
T N N
Note that x(n) is complex. Contrary to FDM transmission systems, where both analog and digital modulation schemes can be used for each of carriers, OFDM transmission can use only digital modulation schemes. Quadrature Amplitude Modulation (QAM) and Phase Shift Keying (PSK) are commonly employed. Two Digital to Analog Converters (DACs) are next used to convert the real and imaginary parts of x(n) to the analog domain, before being sent to an I/Q modulator.At the receiver side, the received OFDM signal is demodulated using a fast Fourier transform (FFT) as follows,
N −1 kn
Y (k) = y(n)exp −j2π (1.5)
In DFT-based OFDM systems, each subcarrier is windowed by a rectangular window of duration T . The rectangular pulse shape in the time domain leads to a sin(x)/x frequency response in the frequency domain. Since the subcarriers are spaced by T1 , each subcarrier peak coincides with the nulls of all the other subcarriers, hence, orthogonality between subcarriers is maintained. This is illustrated in Figure 1.2.
Cyclic prefix for OFDM
Radio channels are not ideal but band-limited channels. The dispersive nature of band-limited channels causes the signal to spread in time, thus introducing the well-known Intersymbol Interference (ISI) distortion. This phenomenon is illustrated in Figure 1.3. In the frequency domain, time-dispersive channels lead to a loss of orthogonality between subcarriers, resulting in Intercarrier Interference (ICI). One straightforward technique to combat ISI and ICI in OFDM systems is the so-called cyclic prefix insertion .
OFDM for optical communications
The explosive growth in the use of broadband applications has drastically increased the IP traﬃc. Figure 1.3 shows Cisco’s forecast for the global IP traﬃc in Exabyte (EB) per month . According to Cisco’s visual networking index (VNI) forecast, the average monthly global traﬃc is expected to reach 121 EB by 2017, which is equivalent to 37 million DVDs per hour. The global traﬃc will continue to be dominated by video (business and consumer) applications. Cisco forecasts that global network users will generate 3 trillion Internet video minutes per month.
To meet the huge growth in IP traﬃc and the continuous growth in demand for broadband applications, service providers have to increase the capacity of their networks to support extended data rates. Optical communication systems have emerged as an eﬃcient alternative of copper-based systems that had reached a saturation in capacity and reach . Indeed, the total available bandwidth of standard optical fibers is enormous (i.e. about 20THz), thus providing huge capacities of 100Tb/s and beyond. Table 1.1 summarizes the advantages and requirements for using optics in long and short distance communications.
However, to take full advantage from the enormous capacity of lightwave systems, the innovation of novel schemes of optical amplification, modulation format and fiber design is required . Two modulation formats can be used in optical communications: single carrier or multicarrier modulations. Optical OFDM has been shown to have several advantages as compared to single-carrier modulation format , [10, 11]. These advantages include:
• Ease of signal processing: straightforward channel estimation and phase estimation can be performed in OFDM transmission, by inserting pilot symbols withing the OFDM signal. In single-carrier modulations, the channel estimation generally relies on algorithms which are prone to error propagation. Furthermore, an eﬃcient one-tap frequency equalization can be performed in OFDM instead the time equalization commonly used in single-carrier modulations.
• Bit and power allocation: Information rate and power are independently adjusted on each subcarrier in accordance with channel conditions, permitting to achieve the desired date rate with the minimal power at the smallest error rate.
• Computation complexity: OFDM has less computation complexity than single-carrier systems due to the use of the eﬃcient algorithm FFT/IFFT, reducing the chip design complexity.
• Adaptability to time-varying channels: The transmitter updates the transmission parameters according to time-varying channel conditions.
Further the features mentioned above, OFDM has other advantages as compared to single-carrier systems including: spectral eﬃciency, high order modulation, flexibility and scalability. Two flavors of OFDM transmission have been investigated for optical communications, namely Coherent Optical OFDM (CO-OFDM) [12, 13] and direct detection optical OFDM -. Coherent optical OFDM provides high spectral eﬃciency, high resistance to fiber chromatic dispersion and receiver sensitivity but requires a complex implementation, making it more suitable for long haul communications. Direct detection OFDM is advantageous for short reach applications because it provides a simple low cost realization. Indeed, the challenge of optical-based short reach communications is to develop cost eﬀective solutions based on the use of fast and low-cost components such as VCSELs and Multimode Fiber (MMF). For cost considerations, IM/DD systems are commonly used in fiber-based short-reach networks such as data centers and Local Area Networks (LANs). For such cost-sensitive networks, CO-OFDM can not be employed because it requires the use of complex architectures and high cost components such as Mach-Zehnder Modulator (MZM) and optical coherent detectors. The use of multimode fiber in short reach networks provides a cost eﬀective solution but it comes at the expense of decreased bandwidth. To overcome the bandwidth limitation due to the multimode fiber intermodal dispersion, an approach is to employ optical OFDM to increase the spectral eﬃciency of short-reach optical communications.The requirements for optical OFDM in next generation of short reach networks are discussed in the following sections.
OFDM in short-range optical fiber communications
Passive optical network
Due to the increasing demand for bandwidth-hungry broadband applications, such as VOIP (Voice over IP), VOD (Video on Demand) and streaming video, Passive Optical Networks (PONs) have been adopted as a cost eﬀective solution to support higher bandwidth capability and provide high speed services to distant users. Copper wire based technologies such as xDSL (x Digital Subscriber Line) can provide a data rate of 100Mb/s over a maximum transmission distance of 300m. By deploying large bandwidth capacity optical fiber transmission lines, passive optical networks can achieve higher data rate (up to 10Gb/s) with larger coverage area (up to 20km). A basic fiber based PON consists of three main components: a) the Optical Line Terminal (OLT) located in the Central Oﬃce (CO) where the assignment of traﬃc to subscribers is managed, b) Optical Network Units (ONUs) located at the subscribers premises c) optical fibers that connect the CO to ONUs. Contrary to Active Optical Networks (AON) where active network elements are used to connect the CO to ONUs, passive optical networks do not need any expensive active components to support services to users, thus reducing the cost of deployment and maintenance. The OLT can be interconnected to ONUs using either a point-to-point (P2P) or pointto-multipoint (P2MP) topology. These two topologies are shown in Figure 1.5. In the point-to-point architecture, a dedicated fiber is used to connect each subscriber (ONU) to the central oﬃce, with diﬀerent wavelengths. In the point-to-multipoint topology, the diﬀerent wavelengths are multiplexed and transmitted over a shared fiber to a passive optical splitter (POS), also called remote node (RN), that splits the optical signal and broadcasts data to all ONUs over separate fibers. Using only one shared fiber, the point-to-multipoint topology of PONs results in additional savings for service providers, making passive optical network one of the dominant broadband access technologies in the access market.
network (10G-EPON) standardized by the institute of electrical and electronics engineers (IEEE 802.3av ). XG-PON1 can provide an asymmetric 10Gb/s downstream and 2.5Gb/s upstream data rates up to 60km physical reach using reach extenders (RE), while 10GE-PON can provide up to 10Gb/s for both downstream and upstream directions with a maximum coverage area of 20km.
OFDM for NGPON2
The increasing demand for modern applications such as high definition TV, multimedia conferencing, multiplayer online gaming and next-generation 3D TV, has pushed operators worldwide to seek for increased-capacity optical access solutions. Next-generation passive optical networks (NG-PONs) are intensively investigated and emerging as a cost-eﬀective solution to satisfy the growing bandwidth demands fueled by both residential and business applications. A gradual migration strategy from the currently deployed PONs to the next-generation solution is adopted in order to distribute the investment over longer periods of time. The first step of migration to the next generation solution referred to as NG-PON1 has to support coexistence with the deployed Gigabit-capable PONs and reuse the outside plant. In the long term, the second phase of migration referred to as NG-PON2 is required to deliver 40Gb/s downstream and 10Gb/s upstream up to 60km. The optimal technology permitting to reach NG-PON2 specifications is under discussion. An overview of the candidate technologies for NG-PON2 is given [24, 25]. These technologies include time division multiple access TDMA-PON, wavelength division multiplexed WDM-PON, OFDM access OFDMA-PON, orthogonal code division access OCDM-PON. A combination of two or more of the aforementioned technologies referred to as hybrid technologies can be considered to increase the aggregated capacity. OOFDM is widely considered as one of the most promising technologies for NG-PON2 due to its particular advantages such as immunity to channel impairments, high spectral eﬃciency, high transmission bit rate and great system scalability. The OFDM-based PON concept was first proposed in . The OFDM-based PON architecture is shown in Figure 1.7.
The OFDMA-based PON (OFDMA-PON) is essentially a hybrid technique, that combines OFDM and Time Division Multiple Access (TDMA). Indeed, in OFDMA-PON, the high-bandwidth is divided to N sub-bands containing each the specific data to each ONU. The OLT broadcasts and transmits the whole sub-bands to ONUs via the passive power splitter, and each ONU selects its specific sub-band with the specific data. One of the distinguishing features of OFDMA-PON is that the number of subcarriers per ONU can be dynamically assigned in diﬀerent time slots depending on the specific requirements of each user, thus allowing a time and frequency two-dimensional bandwidth allocation. Table 1.2 compares the diﬀerent candidate technologies for next generation passive optical networks. As can be seen in table 1.2, OFDMA-PON provides the highest upstream and downstream data rates with the highest passive reach, widely outperforming the other technologies. However, the cost and the power consumption of OFDMA-PON are still high, especially in the ONU which is the most cost-sensitive part of the PON being located at the user side.
The cost-eﬀectiveness of OFDMA-PON can be enabled by combining the exploitation of advanced silicon technologies and sophisticated DSP algorithms tailored to the PON environment:
• The silicon photonic integration, i.e. the integration of chip-scale photonics and electronics in the same semiconductor wafer, is emerging as a promising solution to meet the NG-PON2 requirements in terms of cost, complexity, performance and manufacturing volume . It was shown in , that the FDM/FDMA ONU can be integrated in silicon providing low cost and highly manufacturable devices, which allows the use of very low-cost transceivers at the ONU side.
• For practical deployment of cost-eﬀective OFMDA-PONs, IM/DD systems with Directly Modulated Laser (DML) are highly preferred [29, 30]. In addition, VCSELs are a promising solution for IM/DD systems due to several features such as cost-eﬀectiveness, low power consumption and easy packaging and testing. The main drawbacks of VCSELs as compared to single mode lasers are the limited modulation bandwidth, the limited output optical power and the non-linear characteristic . For the upstream direction of PONs, the required modulation bandwidth and output optical power are quite moderate. Thus, the use of VCSELs as intensity modulators at the ONU side, allows the desired transceiver cost target to be achieved without compromising the NG-PON2 requirements. However, the VCSEL nonlinearity can drastically aﬀect system performance. This issue has to be addressed by researches in order to enable the feasibility of employing VCSELs in OFDMA-PON.
• The aforementioned features allow the mass production of ONU transceivers to meet the mass deployment of NG-PON2. The exploitation of mass-market technologies allows to significantly reduce the cost per unit and achieve the desired cost eﬃciency .
The power consumption of ONUs is still a bottleneck for OFDMA-PON deployment. The high power consumption stems essentially from the high computational power of the DSP to generate OFDM waveforms. The exploitation of advances in complementary metal-oxide-semiconductor (CMOS) can be exploited to reduce the power consumption. However, the use of a very advanced CMOS process can increase the cost per unit . Instead, the exploitation of sophisticated DSP algorithms can significantly reduce the power consumption without considerably increasing the cost of ONU transceivers.
Local area network
A local area network (LAN) is a data communication system which interconnects a number of independent data devices such as computers, mass storage devices, workstations and printers, in a moderate sized geographic area such as university campus, oﬃce buildings, industrial plants and hospitals . Initially, local area networks (LANs) were only based on copper wire links. With the increased demand in bandwidth and data rate, the limited transmission distance of copper cable has become a real shortcoming of copper-based LANs. Optical fibers have then emerged as an extremely eﬀective solution for high-capacity transmission LANs due to their several advantages over copper wire. These advantages include: large bandwidth, light weight and small diameter, long distance transmission, high security and immunity to electromagnetic interference (EMI) [33, 34]. The LAN architecture may have physical and logical topologies. The physical topology refers to the manner of which the network devices are arranged, while the logical topology refers to how the data is assigned. The three basic topologies are: star, ring and bus. Various combinations of these topologies are possible. The advantages and drawbacks of each combination are detailed in .
History and challenges for optical LANs
With the increasing demand in bandwidth due to the growing number of end user applications, optical LANs had to be upgraded from 1Gb/s to 10Gb/s. 1Gb/s LANs employed some millions kilometers of legacy multi-mode fibers such as Fiber Distributed Data Interface (FDDI), OM1 and OM2. Although the diversity in the already installed fiber types and the diﬀerence in their performance and characteristics, the reuse of the existing fiber plant was necessary for a cost-eﬀective upgrade to 10Gb/s. Optical fibers are classified according to two main parameters: the core/cladding diameters and the modal bandwidth that refers to the amount of data the fiber is able to transmit over a certain distance. It is expressed in MHz*km. Table 1.3 lists various multimode fibers standardized by the Telecommunication Industry Association (TIA), the International Electrotechnical Commission (IEC) and the International Standardization Organization (ISO).
Table of contents :
1 Optical OFDM in future optical networks
1.1 OFDM principles
1.1.1 Mathematical description
1.1.2 Digital implementation of OFDM
126.96.36.199 System model
188.8.131.52 Cyclic prefix for OFDM
1.2 OFDM for optical communications
1.3 OFDM in short-range optical fiber communications
1.3.1 Passive optical network
184.108.40.206 OFDM for NGPON2
1.3.2 Local area network
220.127.116.11 History and challenges for optical LANs
18.104.22.168 Multimode solutions for low-cost networks
22.214.171.124 Light sources for low-cost networks
126.96.36.199 OFDM for next generation LANs
1.4 Discrete multitone modulation
1.5 Challenges for OFDM in optical networks
2 Fundamentals of optical OFDM theory
2.1 DMT-based IM/DD system model
2.2 Unipolar OFDM schemes for IM/DD systems
2.2.4 BER performance
2.2.5 Optical power efficiency
3 Hermitian symmetry free optical OFDM
3.1 Hermitian symmetry in optical OFDM
3.2 DHT-based optical OFDM
3.4 Hermitian symmetry free OFDM
3.4.1 System model
3.4.2 PAPR and BER performance
188.8.131.52 Signal to noise ratio
184.108.40.206 Bit error rate
3.4.3 Channel equalization
3.4.4 Computational complexity
3.5 Hermitian symmetry free Flip-OFDM
3.5.1 System model
3.5.2 Bit error rate
4 Asymmetrically Companded DCO-OFDM
4.1 Overview of optical power efficient DMT systems
4.1.1 Asymmetrically clipped DC biased optical OFDM
220.127.116.11 Transmitter model
18.104.22.168 Receiver model
4.1.2 Hybrid asymmetrically clipped optical OFDM
4.1.3 Pilot-assisted modulation
4.1.4 Discussion and conclusion
4.2 The companding concept in OFDM systems
4.3 Asymmetrically companded DCO-OFDM
4.3.1 System model
4.3.2 Companding function
4.4 Simulation results
4.4.1 Clipping noise reduction
4.4.2 BER as a function of Eb(elec)/N0
4.4.3 Optical power efficiency
5 Experimental investigation of DMT for cost-sensitive networks
5.1 VCSEL characterization and modeling
5.1.1 Static characteristic
5.1.2 Quasi-static characteristic
5.1.3 VCSEL nonlinearity modeling
5.2 Simulation results
5.2.1 Optical link model
5.2.2 Optical noise components
5.2.3 Nonlinearity and clipping distortions
5.3.1 Back to back measurements
22.214.171.124 Inverse sinc compensation
126.96.36.199 Channel equalization
5.3.2 Optical link characterization
5.3.3 Experimental validation of VCSEL non-linear model
5.3.4 Companded DCO-OFDM
Appendix A The AWG impact on the DMT signal
A.1 AWG without interleave function
A.2 AWG with interleave function
List of Publications
French thesis summary
5.3 Fondamentaux de l’OFDM optique
5.3.1 Syst`eme IM/DD
5.3.4 Comparaison des deux techniques
188.8.131.52 Puissance optique
5.4 L’OFDM optique sans sym´etrie hermitienne
5.4.1 La sym´etrie hermitienne dans les syst´emes OFDM optiques
5.5 DCO-OFDM asym´etriquement compress´ee
5.5.2 La fonction de companding
5.5.3 R´esultats de simulations
184.108.40.206 Puissance optique
5.6 Exploration exp´erimentale de la DMT pour les r´eseaux `a bas coˆut
5.6.1 L’impact du VCSEL sur la modulation DMT
220.127.116.11 La caract´erisation quasi-statique du VCSEL
18.104.22.168 R´esultats de simulation
22.214.171.124 Validation exp´erimentale
5.6.2 DCO-OFDM compress´ee