Telecommunication operators are investing significant resources in developing passive optical networks (PONs) to meet the increasing capacity requirements. Therefore, wireless transmission has become the bottleneck for the wireless broadband internet access due to the spectrum saturation. This issue can be solved taking advantage of the huge portions of unused spectrum at high-microwave / millimeter-wave (mm-wave) bands, but their generation is power consuming. Radio over fiber (RoF) is a cost-efficient solution for the distribution of high-frequency broadband signals to remote base stations. We present a novel photonic PON-to-RoF bridge based on heterodyning a PON signal with an unmodulated tone generated by an independent laser. The proposed scheme is transparent to modulation format and can generate RF signals in the entire microwave band. The feasibility of the bridge is experimentally shown converting a 2-Gbps orthogonal frequency division multiplexing PON signal using inexepensive distributed feedback lasers, whose phase noise is cancelled employing an envelope detection based mobile terminal.
© 2013 Optical Society of America
In recent years, passive optical networks (PONs) have become ubiquitous all over the world, replacing previously deployed hybrid fiber-coaxial (HFC) distribution networks and the copper local loop, reaching progressively closer to the user terminal . In the first deployed PON systems, the medium access control was based on time division multiple access (TDMA). However, as the bitrate increases, the synchronization requirements are more difficult to meet. Wavelength division multiplexing (WDM) PONs were proposed as a solution to relax the exigent synchronization mechanism. In these systems, a wavelength is dedicated to a single user or to a relatively small number of them, which can access the assigned wavelength using TDMA. In addition to the easier synchronization, WDM-PONs improve the security and allow higher split sizes, since some splitters in TDMA-PONs are substituted by demultiplexers .
WDM-PONs offer high bitrates but they lack the mobility that wireless communications offer. This issue has been addressed for in-door scenarios by deploying wireless local area networks (WLANs). In conventional WLANs, the base station (BS), generally denominated modem, receives downlink data in baseband and generates the RF signal with the desired frequency and modulation scheme. Therefore, both the modulation and RF generation tasks are performed at the modem. This technology has proven to be successful in WLANs operating at 2.4 and 5 GHz with bitrates below 150 Mbps (802.11n). Recently, the distribution of WLAN and 2G/3G cellular signals over fiber network has gained attention . Such systems allow the deployment of high-capacity distributed antenna systems (DAS), where the signal is not radiated by a single antenna in the BS, but by multiple antennas scattered through the area to be covered. DAS improves the aggregated capacity  but, given the narrow allowable bandwidth and the high number of systems operating at these bands, the provision of future multi-Gbps wireless access services is a challenge. An idea that has been echoing in conference halls for the past few years is to increase carrier frequency up to the unused and unlicensed millimiter-wave (mm-wave) bands.
Given the few applications operating at higher frequencies and the broad slices of available spectrum, the use of higher frequency bands has been proposed as a solution to the spectrum scarcity and avoid the wireless capacity bottleneck . At millimeter-wave (mm-wave) frequencies, the high absorption of conventional building materials and high free-space losses limit the maximum distance of the wireless link, reducing the BS coverage area. Consequently, in such systems with numerous BSs, it is critical to keep BSs as simple and power-efficient as possible. In this context, radio over fiber (RoF) has emerged as a cost-efficient and low-consumption solution for delivering high-frequency broad-bandwidth signals to multiple BSs . In RoF systems, all the modulation and upconversion tasks are concentrated at a central station (CS), distributed over an optical network, and converted back to electrical domain at the BSs. Typically, an RoF signal generated at the CS has two tones, being one or both modulated with the downlink data (generally, one of the tones is considered as a reference). The frequency separation between the two components is set to match the desired RF frequency. Hence, the beating at the photodiode (PD) between the two components results in the desired modulated signal at the RF frequency.
In conventional PON access, WLAN services are implemented and controlled by user equipment. In contrast, the wireless link in RoF is controlled by the CS (telecommunication operator equipment). Consequently, the coexistence of PON and RoF requires particular attention. Different approaches have been proposed:
- Approach #1: Both PON and RoF signals are generated at the OLT/CS and transmitted to the ONUs/BSs sharing the distribution network, Fig. 1(a). This scheme has two main advantages: (i) it concentrates all the generation components for the downlink in a single site and (ii) since PON and RoF share a great part of the distribution network, there is not network duplication. Therefore, it has attracted considerable attention in recent years [7, 9–14]. Nevertheless, this scheme presents two drawbacks: (i) the quality of the generated RF signal is severely affected by the fiber losses, since both the reference and modulated components are attenuated, limiting the maximum reach of the network . (ii) The wavelength multiplexing of different RoF signals on the same fiber requires special frequency planning  that may not be compatible with already deployed wavelength-division multiplexing (WDM)-PON grids.
- Approach #2: Alternatively, a conventional PON can be used to transmit data between the OLT and the ONUs. Afterwards, some of the ONUs are connected to a local area network (LAN) where the WLAN BS is substituted by a CS that generates the RoF signal, as shown in Fig. 1(b). In this case, since PON and RoF signals do not share any part of the network, the frequency grid is not an issue. In comparison to Approach #1 the range is increased because the RoF signal is regenerated at the CS. Nevertheless, as the PON signal is converted to the electrical domain and then converted back to optical domain, a lot of components are required, increasing both the cost and power consumption of the conversion.
- Approach #3: A photonic conversion from PON to RoF has been proposed in  where the PON signal is optically remodulated in an intermediate point using a Mach-Zehnder modulator biased at zero-intensity transmission point and fed by a tone at half the desired RF frequency. In this way, two modulation sidebands appear, separated by the RF frequency, Fig. 1(c). The main inconvenience of this scheme is the required broad-bandwidth external modulator as well as the high frequency electronics, which increases its cost.
- Approach #4: Another photonic technique was presented in , where an optically modulated tone (PON signal) is combined with an unmodulated tone at the base station (BS) as shown in Fig. 1(d). In this technique, the frequency of the generated signal can be controlled by tuning the frequency difference between the modulated and the unmodulated tones. Given the temperature tuning of the lasers, the maximum frequency that can be generated using this technique is limited by the bandwidth of the PD, but it requires an independent laser for each BS, resulting in an expensive implementation.
- Approach #5: In this paper, we propose a novel architecture where the conversion from PON to RoF is performed at an intermediate point, as shown in Fig. 1(e). This architecture is based on a photonic PON-to-RoF bridge where a single-wavelength PON signal is combined with an unmodulated reference tone generated by an independent laser, which is referred as local oscillator (LO). The generated RF frequency at the PD can be easily tuned by electrical or temperature tuning of the laser at the PON-to-RoF bridge to cover the whole microwave band. In comparison with Approach #1, Approach #5 is transparent to frequency grid. Besides, in Approach #1, both the PON and LO signals are attenuated by the fiber and, therefore, the generated RF power is proportional to exp(−2αl) where α denotes the fiber attenuation power coefficient and l is the fiber length. In contrast, in Approach #5, the fiber span between the OLT and the bridge only affects the PON signal. Considering that the bridge is much closer to the BSs than to the OLT, the fiber length between the OLT and the bridge can be assumed to be close to l. Therefore, the generated RF power in Approach #5 is proportional to exp(−αl) and, consequently, the maximum reach is extended compared to Approach #1.
Side by side with Approach #2, the conversion requires much less optical and electrical components. Unlike in Approach #3, the proposed bridge does not require a broad-bandwidth external modulator and the maximum generated frequency can be increased up to the bandwidth of the PD. Finally, in our approach, a single laser is used to generate the RoF signals for multiple BSs, in contrast to Approach #4, where each BS requires its own laser.
The main impairments of the proposed PON-to-RoF bridge are the frequency stability and the phase noise of the generated RF signal. While the frequency stability can be enhanced by improving the electrical and temperature controls, the phase noise is difficult to reduce. In mobile terminals (MTs) based on heterodyne detection, where the downconversion from RF frequency to intermediate frequency (IF) is performed mixing the received RF signal and a tone generated by an electrical oscillator, the RF phase noise is converted to amplitude noise due to the phase mismatch between the RF and the oscillator tone. Consequently, narrow-linewidth lasers, such as external cavity lasers (ECLs), would be required in order to reduce the phase noise of the RF signal, increasing the cost of the system . Alternatively, MT schemes based on phase-insensitive downconversion can be employed in order to relax the laser linewidth requirements. The phase-insensitivity of MTs based on self-heterodyning (SH) has been experimentally demonstrated [8, 16]. On the other hand, MTs based on envelope detection (ED) are also insensitive to phase noise . Compared to heterodyning MTs, SH and ED based MTs have worse sensitivity but are simpler and less power consuming, which make them an attractive solution for short range applications.
The paper is organized as follows: Section 2 explains in more detail the PON-to-RoF bridge and its use in WDM-PON networks, in Section 3 the experimental setup is presented and results are discussed. Finally, Section 4 concludes the paper.
2. PON-to-RoF bridge
In this section we discuss the compatibility with PON systems and we present a simple mathematical to show the effect of the laser linewidth on the generated RF signal. Afterwards, the effect of phase noise on different MTs is analyzed to show that in contrast to heterodyning MTs, downconversion using SH or ED-MTs is phase insensitive.
2.1. Compatibility with WDM-PON networks
As shown in Fig. 1(a), the major part of the WDM-PON distribution network remains unaltered, only after the bridge, the RoF distribution network has to be redesigned. Additionally, since the bridge operates after the wavelength segregation, it is independent of the employed frequency grid. This is an important point because WDM-PON networks are evolving to narrower frequency separation grids to accommodate a higher number of channels . Another tendency in PONs is reconfigurability, which adds flexibility and makes a more efficient use of resources. In reconfigurable PONs, the wavelength assigned to a user or group of users can be changed according to a resource management algorithm. If this occurs, the LO laser can be temperature tuned to keep the desired frequency separation between the PON and the unmodulated tone. In , the tuning of a DFB laser over 500 GHz covering 20 channels with a 25 GHz separation is reported. The transparency to modulation format is another important issue. It is expected that future PON networks will use flexible and spectrally efficient modulation formats, such as m-ary quadrature amplitude modulation (mQAM) and multicarrier modulation formats, mainly orthogonal frequency division multiplexing (OFDM) . In this sense, the only requirement for the modulated signal in order to be properly converted by the bridge and detected using SH or ED based MT is that it has to be accompanied by a carrier. This is the case of the signals in PONs with direct detection, which is much more cost efficient than heterodyne PON systems. In addition, due to the sensitivity to the fiber chromatic dispersion and poor spectral efficiency of double-sideband (DSB) modulation, the use of single-sideband (SSB) modulation has been proposed for future PONs. Our scheme can convert both the SSB and DSB signals to the de- sired frequency. To sum up, the PON-to-RoF bridge is capable to deal not only with actual PON systems but also with future deployments.
2.2. Mathematical model for the PON and RF signals in the proposed architecture
The input signal to the bridge, Ein(t), can be written in the form of:
Considering the fiber attenuation, α in Np/m, and neglecting the effect of chromatic dispersion, the optical field at a distance z can be expressed as:
The field at the input of the PD, Erec(t), after a fiber span of length lfiber is:
The photogenerated current, iPD(t), is proportional to the instantaneous received optical power, Prec(t), and can be expressed as :22]. The upper term in Eq. (6) represents the low-frequency components of the generated signal, which can be detected using a conventional low-bandwidth ONU. On the other hand, the lower term in Eq. (6) corresponds to a passband signal centered at ωRF. After segregating the low and high-frequency components we get the expressions for the received PON signal, iPON(t), and the components around ωRF, iRF(t), which will be radiated:
2.3. Downconversion of RF signals
Given the high ωRF, the first stage of the MT front-end is downconversion either to baseband or to an ωIF. Here, we consider four different MT front-end schemes, shown in Fig. 2 [15, 23]. They can be divided in MTs equipped with an electrical oscillator (homodyning and heterodyning MTs) and MTs without electrical oscillator (SH-MTs and ED-MTs). The former do not require the carrier to be transmitted by the BS, while the latter do. In comparison with homo-dyning and heterodyning MTs, SH and ED based MTs present poorer sensitivity since they lack the mixing gain that homodyning and heterodyning MTs have. Additionally, in systems with SH and ED based MTs, the transmission of the amplitude modulated carrier leads to two disadvantages: on the one hand, the BS is less power efficient since the carrier consumes a great part of the radiated power. On the other hand, typically a frequency guard is left between the carrier and the modulation sideband, increasing the occupied bandwidth and reducing the spectral efficiency. However, SH and ED based MTs are insensitive to phase noise and their cost is lower than that of homodyning and heterodyning MTs. SH and ED based MT are, then, attractive for low-cost picocell-based applications.
2.3.1. Suppressed carrier RF signals
Homodyning MTs are equipped with an electrical oscillator operating at the same frequency as the RF signal and, therefore, the downconversion is straight to baseband, while in heterodyning MTs, the RF signal is mixed with a tone at ωRF ± ωIF and the RF signal is downconverted to ωIF in an intermediate step. If a homodyning/heterodyning receiver is employed, single-sideband with suppressed carrier (SSB-SC) can be used to improve both the energy and spectral efficiencies. SSB-SC can be generated at the BS by electrical filtering. For the sake of clarity, we will consider that when homodyning/heterodyning MTs are used, the received RF signal, sSSB(t), is SSB-SC which is obtained filtering iRF(t) in Eq. (7):
2.3.2. Carrier assisted RF signals
In contrast to heterodyning/homodyning MTs, in SH-MTs an RF carrier is required to bias the mixer while in ED-MTs, the carrier avoids the zero-crossing distortion. Both SH-MTs and EDMTs can operate in SSB or DSB mode while ensuring that the power of the RF carrier is high enough. For simplicity we will consider the transmission of carrier assisted DSB signals and, therefore, the received signal, sDSB(t) is expressed as:
2.3.3. Effect of the phase noise on the downconverted signal
As mentioned in the Introduction and shown in Subsection 2.2., the main impairment of the PON-to-RoF bridge is the high phase noise of the generated RF. In this subsection we present different MT implementations and we study the impact of the phase noise of the RF signal when using each of them.
- Homodyning MT: Figure 2(a) shows a homodyning MT based on the Costas receiver . In this scheme, the RF signal is split in two and mixed with the oscillator output and a phase-delayed version of it. At each arm of the detector the RF signal is frequency shifted to baseband and to 2 × ω′RF and therefore each mixer is followed by a low-pass filter (LPF) that removes high-frequency components. The downconverted in-phase, sI(t) and quadrature, sQ(t), signals can be expressed as:
- Heterodyning MT: In order to simplify the MT frontend and reduce the number of components operating at RF, the RF signal is converted to an IF. As shown in Fig. 2(b), heterodyning MT is essentially one arm of the homodyning MT with the exception that LO is set to oscillate at ω′RF ± ωIF and that the mixer is followed by a BPF centered at IF instead of a LPF. The IF signal acquires the form of:
- SH-MT: In SH-MTs the RF signal is amplified, split in two, and mixed with itself at a mixer, which is followed by an BPF, as can be seen in Fig. 2(c). In this scheme, an electrical oscillator is not required since the RF carrier is used as reference. Hence, the mixer output is proportional to the square of the RF signal: 8].
- ED-MT: The ED-MT shown in Fig. 2(d) presents several advantages over the SH-MT: it does not require phase alignment, it requires less and cheaper components, and it has better sensitivity . An ED is nothing else than a diode operating as a half-wave rectifier followed by a LPF to suppress high-frequency components and reduce the ripple. Assuming that the ripple is negligible, the output of the ED, sED(t), can be mathematically calculated from the analytical signal of sDSB : Eq. (17) can be calculated applying Bedrosian’s theorem , which states that the Hilbert transform of a product of two signals whose spectra do not overlap is the product the low-pass signal and the Hilbert transform of the high-pass signal: Eqs. (10), (16), and (18), we get an expression for sED(t):
In the previous analytical study the insensitivity of the carrier assisted downconversion schemes, SH and ED, has been shown. In contrast, in homodyning and heterodyning receptions the phase noise in converted to amplitude noise, requiring narrow linewidth lasers and increasing the system cost. Therefore, for RF signals corrupted by significant phase noise, carrier assisted downconversion clearly outperforms that based on homodyning/heterodyning. It is important to note that this may not be the case for highly spectraly pure RF signals, where the mixing gain in homodyning/heterodyning based MTs improves its sensitivity.
3. Results and discussion
3.1. Experimental setup
Figure 3 shows the experimental setup that includes an OFDM-PON OLT, the proposed PON-to-RoF bridge, a simple BS, and a MT based on ED. A 2-Gbps OFDM signal was generated offline in Matlab and loaded into a Tektronix AWG7122B Arbitrary Waveform Generator (AWG) operating at a sampling rate of 6.25 GS/s. 128-subcarriers were used from which 100 were modulated using quadrature phase shift keying (QPSK) and 28 were left unmodulated to be used in the pilot assisted equalization. A 20% cyclic prefix was added. In order to mitigate low-frequency 2nd order nonlinear distortion, the OFDM signal was electrically upconverted to fIF = 1.75 GHz. The output of the AWG was amplified by a 10-GHz amplifier before being combined with a bias current using a broadband bias-T. This signal was used to drive the PON laser. Afterwards the signal was transmitted over a 25-km standard single mode fiber (SSMF) span. In the PON-to-RoF bridge, the signal was combined in a 3-dB optical coupler (OC) with an unmodulated laser, denoted in Fig. 3 as LO laser, whose polarization is controlled to match that of the signal optical field. The employed multiquantum-well (MQW) DFB lasers were supplied by Alcatel and showed a threshold current of 10 mA and a nominal linewidth of 1 MHz . The frequency difference between the PON laser and the LO laser was set to 10 GHz by setting the bias current of the first to 46.2 mA and of the second to 42.8 mA. The two DFB lasers presented a temperature sensitivity of 10 GHz/°C and their temperatures were controlled using independent temperature electrical controls with 0.01 °C precission. One of the OC outputs was connected to an Advantest Q8384 optical spectrum analyzer (OSA). The other output was amplified using an Alcatel SOA biased at 100 mA, attenuated by a variable optical attenuator (VOA), which accounted for the distribution splitting losses, and transmitted over a 1250-m SSMF span that emulated the fiber in the in-building distribution network. At the BS, the optoelectronic conversion was performed using a 12-GHz bandwidth PD. The PD was followed by a high-pass electrical filter (3-GHz cut-off frequency) to remove the generated low frequency components, after which a 10dB electrical attenuator emulated the wireless channel. Downconversion from RF to IF, was carried out employing an ED based on an 18-GHz-bandwidth detector from Herotek (DT1018P), whose output was amplified by a low noise amplifier before being digitalized by a Tektronic DPO72004B real time oscilloscope and demodulated in Matlab.
In Fig. 4, the power spectral densities at different stages of the system are shown. Figure 4(a) represents the optical spectrum at the output of the OC where the reference and the modulated can be seen. Due to the wavelength resolution of the OSA, 0.01 nm, the modulation sidebands cannot be distinguished from the carrier, however, the modulated tone presents a slightly broader spectrum. The lower power of the modulated signal is caused by the attenuation of the 25-km SSMF span. In Fig. 4(b), the normalized power spectral density of the photogenerated current is presented. The low frequency signal appears at ωIF, whereas the up-converted signal, composed of the carrier and upper and lower modulation sidebands, appear at ωRF around 10 GHz. It is important to note that, while at baseband the phase noise is completely cancelled, at 10 GHz some amount of phase noise appears, as can be seen in the RF carrier. After filtering the low-frequency components, the power spectral density shown in Fig. 4(c) is obtained. In comparison to Fig. 4(b), it is clear the suppression of the low-frequency components. Additionally, since the spectra were captured at different times, the center frequency of the generated RF signals was affected by the long term frequency drift (which will be analyzed in detail later). Finally, Fig. 4(d) shows the normalized power spectral density of the signal at the output of the LNA after the downconversion to ωIF. In this spectrum, the generated second order nonlinear components can be seen justifying the use of IF. In regards to signal quality, comparing the average signal to noise ratio (SNR) of the PON signal in Fig. 4(b) with that of the downconverted signal in Fig. 4(d), there is a penalty of around 4 dBs. It is important to note that the spectrum shown in Fig. 4(d) was obtained without the electrical attenuator that emulated the wireless channel and, therefore, the SNR penalty will increase as the path losses become greater.
3.2. Frequency stability and phase noise
In the previous subsection, we briefly talked about the phase noise and long term stability of the generated RF signal. In this subsection, we further analyze both impairments to show that, even when DFB lasers are employed, a phase-insensitive MT allows the recovering of the signal. The frequency stability of the generated signal was analyzed by switching off the modulation stage and performing the discrete time Fourier transform of the signal captured in the oscilloscope. The frequency peaks of 500 runs were computed (the time between runnings was around 8s) and represented in Fig. 5(a), which shows a frequency drift of ±150 MHz. The long-term frequency fluctuations of a stabilized semiconductor laser can be linearly modeled in terms of the temperature and current fluctuations, ΔT and ΔI, according to:Eq. (21). The maximum deviation of the generated RF signal is given by:
On the other hand, the RF phase noise has been measured using the curve shown in Fig. 5(b), which was obtained by averaging the spectra of 500 independent runs. Lorentzian curve fitting reveals that the combined linewidth of the generated signal is 2.1 MHz, which is a reasonable value considering that the linewidth of each laser was in the order of 1 MHz. The common metric for phase noise in microwave electrical oscillators is the single sideband (SSB) phase noise at a particular frequency offset, typically 1 MHz . The same metric is used in several optical techniques generating RF signals with high spectral purity, such as optical sideband injection locking (OSBIL) . In Fig. 5(b), it can be seen that the SSB phase noise at a frequency offset of 1 MHz is around −3 dBc/Hz, which is much higher than the −100 dBc/Hz reported in . In Subsection 3.3, however, we report that the phase insensitivity of ED-based MTs can be used to successfully downconvert and demodulate RF signals with a 2.1-MHz linewidth.
3.3. OFDM transmission performance
The modulation performance was assessed in terms of the error vector magnitude (EVM), which was estimated from the constellation diagram and is related to the bit error rate (BER) according to Eq. (13) in . Figure 6 shows the obtained EVM when the VOA was swept to change the distribution losses, Ldist. As a result of that, also the received optical power, Prec(t) changed according to Prec[dBm] = Pin[dBm] − Ldist[dB], where Pin is the power of the modulated tone before the attenuator. The distribution losses can alternatively expressed as the split size, which is related to Ldist through split size = 10Ldist. Results are given for four different cases: with and without optical amplification in the bridge and with and without fiber after the bridge. In Fig. 6(a), it can be appreciated that the inclusion of the 1250-m SSMF does not cause a significant power penalty in the detected signal at a BER of 10−3 (EVM = 0.32), which is adopted as forward error correction (FEC) limit . Using the popular Reed-Solomom (255,239) FEC, a BER of 10−3 is reduced down to 10−6 . However, comparing the EVM curves with and without amplification, amplification induces a power penalty of around 2.5 dB. This signal degradation is attributed to the combined effect of the SOA nonlinearities and the noise induced by the amplified spontaneous emission. Therefore, from the point of view of the sensitivity, the use of amplification seems to be harmful. However, when the EVM curves are plotted against the distribution losses, which account for both the splitting and fiber losses, Fig. 6(b), it is clear that when amplification is used, the same converted PON signal can be distributed to 64 BSs while when no amplification is used, only 4 BSs can be served.
In this paper, a novel low-cost photonic PON-to-RoF bridge has been presented. The proposed system is transparent to modulation format and frequency grid, and can convert PON signals over a broad wavelength range to the complete microwave/mm-wave band. Therefore, it is compatible with future WDM-PONs. We have experimentally demonstrated that phase-insensitive MTs enables the use of non-narrow linewidth lasers in the PON-to-RoF bridge, allowing the use of integrable DFB lasers at the PON-to-RoF bridge. If optical amplification is employed, the bridge is capable to deliver a 2-Gbps OFDM signal to up to 64 BSs at a distance of 1250 m.
References and links
1. T. Koonen, “Trends in optical access and in-building networks,” in Proceedings of European Conference on Optical Communcations (2008), paper We.2.A.1.
2. C. H. Lee, W. Sorin, and B. Y. Kim, “Fiber to the home using a PON infrastructure,” J. Lightwave Technol. 24(12), 4568–4583 (2006). [CrossRef]
3. H. Kosek, Y. He, X. Gu, and X. N. Fernando, “All optical demultiplexing of closely spaced multimedia radio signals,” J. Lightwave Technol. 25(6), 1401–1409 (2007). [CrossRef]
4. W. Choi and J. C. Andrews, “Downlink performance and capacity of distributed antenna systems in a multicell environment,” IEEE Trans. Commun. 6(1), 69–73 (2007).
5. Z. Pi and F. Khan, “An introduction to millimeter-wave mobile broadband systems,” IEEE Commun. Mag. 49(6), 101–107 (2011). [CrossRef]
6. T. J. Hall, R. Maldonado-Basilio, S. Abdul-Majid, J. Seregelyi, R. Li, I. Antolín-Pérez, H. Nikkhah, F. Lucarz, J. L. de Bougrenet de La Tocnaye, B. Fracasso, P. Pajusco, C. Karnfelt, D. Bourreau, M. Ney, R. Guemri, Y. Josse, and H. Liu, “Radio-over-Fibre access for sustainable Digital Cities,” Ann. Telecommun. 68(1–2), 3–21 (2013). [CrossRef]
7. G. Chang, Z. Jia, H. C. Chien, A. Chowdhury, Y. T. Hsueh, and J. Yu, “Convergence of broadband optical and wireless access networks,” Proc. SPIE 7234, 723402 (2009).
8. C.-S. Choi, Y. Shoji, and H. Ogawa, “Millimeter-wave fiber-fed wireless access systems based on dense wavelength-division-multiplexing networks,” IEEE Trans. Microwave Theory Technol. 56(1), 232–241 (2008). [CrossRef]
9. C. T. Lin, W. R. Peng, P. C. Peng, J. Chen, C. F. Peng, B. S. Chiou, and S. Chi, “Simultaneous generation of baseband and radio signals using only one single-electrode Mach-Zehnder modulator with enhanced linearity,” IEEE Photonics Technol. Lett. 18(23), 2481–2483 (2006). [CrossRef]
10. G. K. Chang, “Convergence of optical and wireless access networks,” Tutorial at OMC workshop on optical/wireless integration for enhanced broadband access and transmission, Optical Fiber Conference (2008).
11. J. Yu, Z. Jia, G. Chang, and X. J. Xin, “Broadband convergence of 60-GHz RoF and WDM-PON systems with a single modulator for bidirectional access networks,” in Proceedings of European Conference on Optical Communications (2008), paper 6.5.2.
12. L. Chen, J. Yu, S. Wen, J. Lu, Z. Dong, M. Huang, and G. K. Chang, “A novel scheme for seamless integration of ROF with centralized lightwave OFDM-WDM-PON system system,” J. Lightw. Technol. 27(14), 2786–2791 (2009). [CrossRef]
13. Z. Chunlei, G. Ling, and Z. Pengtu, “An overview of integration of RoF with PON,” in Proceedings of International Conference on Computer Applications and System Modeling (2010), Vol. 15, pp. 40–43.
14. Y. Shao, N. Chi, J. Fan, and W. Fang, “Generation of 16-QAM-OFDM signals using selected mapping method and its application in optical millimeter-wave access system,” IEEE Photonics Technol. Lett. 24(5), 1301–1303 (2012). [CrossRef]
15. I. Gonzalez-Insua, D. Plettemeier, and C. Schaffer, “Simple remote heterodyne radio-over-fiber system for gigabit per second wireless access,” J. Lightwave Technol. 28(16), 2289–2295 (2010). [CrossRef]
16. A. Islam, “Experimental demonstration of heterodyned radio-over-fiber system using unlocked light sources and RF homodyning at the receiver,” in Proceedings of Optoelectronics and Communication Conference (2010), pp. 714–715, paper 9A2-1.
17. I. Aldaya, G. Campuzano, C. Gosset, and G. Castañón, “Phase-insensitive RF envelope detection allows optical heterodyning of MHz-linewidth signals,” IEEE Photonics Technol. Lett. 25(22), 2193–2196 (2013). [CrossRef]
18. Z. Dong, X. Li, J. Yu, Z. Cao, and N. Chi, “8 × 9.95-Gb/s ultra-dense WDM-PON on a 12.5-GHz grid with digital pre-equalization,” IEEE Photonics Technol. Lett. 25(2), 194–197 (2013). [CrossRef]
19. M. Funabashi, H. Nasu, T. Mukaihara, T. Kimoto, T. Shinagawa, T. Kise, K. Takaki, T. Takagi, M. Oike, T. Nomura, and A. Kasukawa, “Recent advances in DFB lasers for ultradense WDM applications,” IEEE J. Quantum Electron. 10(2), 312–320 (2004). [CrossRef]
20. J. M. Tang, R. P. Giddings, X. Q. Jin, J. L. Wei, X. Zheng, E. Giacoumidis, E. Hugues-Salas, Y. Hong, C. Shu, J. Groenewald, and K. Muthusamy, “Real-time optical OFDM transceivers for PON applications,” in Proceedings of Optical Fiber Communication Conference (2011), paper OTuK3.
21. R. P. Braun, “Tutorial: fibre radio systems, applications and devices,” in Proceedings of European Conference on Optical Communication (1998), vol. 2, pp. 87–119.
22. U. Gliese, S. Norskov, and T. Nielsen, “Chromatic dispersion in fiber-optic microwave and millimeter-wave links,” IEEE Trans. Microwave Theory Technol. 44(10), 1716–1724 (1996). [CrossRef]
23. R. Daniels and R. Heath, “60 GHz wireless communications: emerging requirements and design recomendations,” IEEE Trans. Veh. Technol. pp. 41–50 (Sept. 2007). [CrossRef]
24. S. Haykin and M. Moher, Introduction to Analog & Digital Communications, 2nd ed. (John Wiley and Sons, 2007).
25. A. Poularikas, The Handbook of Formulas and Tables for Signal Processing (CRC Press LLC, 1999), chap. 15 The Hilbert Transform.
26. E. Bedrosian, “Memorandum RM-3439-PR: A product theorem for Hilbert transform,” Tech. Rep., United States Air Force (project RAND) (1962).
27. C. Palavicini, G. Campuzano, B. Thedrez, Y. Jaouen, and P. Gallion, “Analysis of optical-injected distributed feedback lasers using complex optical low-coherence reflectometry,” IEEE Photonics Technol. Lett. 15(12), 1683–1685 (2003). [CrossRef]
28. R. Wanner, R. Lachnert, G. R. Olbrich, and P. Russer, “A SiGe monolithically integrated 278 GHz push-push oscillator,” in Proceedings of IEEE International Microwave Simposium (2007), pp. 333–336.
29. R. P. Braun, G. Grosskopf, D. Rohde, and F. Schmidt, “Low-phase-noise millimeter-wave generation at 64GHz and data transmission using optical sideband injection locking,” IEEE Photonics Technol. Lett. 10(5), 728–730 (1998). [CrossRef]
30. R. Shafik, S. Rahman, and R. Islam, “On the Extended Relationships Among EVM, BER, and SNR as Performance Metrics,” in Proceedings of International Conference on Electrical and Computer Engineering (2006), pp. 408–411.
31. C. T. Lin, J. Chen, W. J. Jiang, L. Y. He, P. T. Shih, C. H. Ho, and S. Chi, “Ultra-high data-rate 60 GHz radio-over-fiber systems employing optical frequency multiplication and adaptive OFDM formats,” in Proceedings of Optical Fiber Conference (2011), paper OThJ6.
32. T. Walker, “Optical Transport Network (OTN) Tutorial,” Tech. Rep., International Telecommunication Union (2004).