1. Introduction

With the exponentially increasing of customer needs for broadband services, passive optical network (PON) is considered to be the most promising candidate that can economically provide high bandwidth to end-users [1]. Recently, there have been growing interests in the new type of optically amplified large-split long-reach PON (LR-PON) [2–6]. Integrated the access network with the metro network within the 100-km target range, LR-PON claims to considerably reduce the capital and operational expenditures by increasing the coverage of the central office and consolidating the O/E/O conversion interfaces inside the existing networks. Besides, LR-PONs can simplify the network hierarchies and thus reduce the network latency, which is very important for the real-time broadband services.

To realize next generation LR-PON systems, several multiplexing schemes are viable candidates to support more ONUs with higher data rate, such as time-division multiplexing (TDM), orthogonal frequency-division multiplexing (OFDM), and wavelength-division multiplexing (WDM). In particular, OFDM LR-PON, which boasts to offer high spectral efficiency and flexible bandwidth allocation, has attracted a lot of attention recently [7]. Both TDM and OFDM LR-PONs can easily support numerous ONUs by using an optical splitter with large splitting ratio and can broadcast the aggregated data to all ONUs on a single wavelength, thus have the benefit of simple wavelength management [6]. However, the aggregated data rate on a single wavelength will be very high (40 Gbps or higher), and each ONU has to process very high-speed aggregated data in order to receive and transmit a small portion of data (e.g. <1/128) [6]. In [7], a 108-Gbps OFDM PON over single wavelength is demonstrated employing two Mach-Zehnder modulators, polarization multiplexing and two receivers. For cost-sensitive ONUs, the proposed scheme will be too complicated and expensive to be a practical solution in the near future. An alternative solution is WDM LR-PON, where each ONU can be assigned a wavelength, thus ONUs doesn’t have to process the aggregated data and the speed requirement of each transceiver can be considerably lower. Besides, instead of using a large-splitting-ratio optical splitter, WDM LR-PON can use a wavelength multiplexer/demultiplexer (waveguide grating router, WGR), and the loss of WGR is much lower and insensitive to wavelength number. Unfortunately, even though the speed requirement of each transceiver is lower in a WDM system, numerous color transmitters are costly and undesired. To preserve the colorless upstream architecture, an RSOA (reflective semiconductor optical amplifier) is employed for upstream service [8,9]. Currently, employing RSOAs are still facing some technical challenges such as limited bandwidth, ASE noise, Rayleigh backscattering needed to be overcome. Besides, the systems fail to allocate bandwidth flexibly among ONUs of different wavelengths [4]. In particular, since a LR-PON is needed to be coexisted with the current PONs, such as 10-Gbps PON (XG-PON) and Gigabit PON (G-PON). As shown in Fig. 1 [10], excluding guard bands, the available enhancement bands provide very limited bandwidth at C-band for future LR-PONs. For the reasons stated above, a more practical solution will be a hybrid WDM/OFDM LR-PON, which optimizes the trade-off between the number of wavelength channels and the cost of transceivers. For instance, if a cost-effective transceiver can provide 32-Gbps capacity, only four wavelength channels are required for each direction (downstream and upstream) to support a LR-PON with 128 ONUs and 1-Gbps data rate per ONU.

 

Fig. 1 Wavelength allocations of current G-PON/XG-PON and proposed LR-PON. (DS: downstream; US: upstream)

Download Full Size | PDF

To reach the target transmission distance of up to 100 km, nonetheless, will require better performance and more expensive components. To further reduce costs, intensity-modulation-direct-detection (IMDD) is expected. Additionally, it is desirable to generate high-speed signals by low-bandwidth cost-effective transceivers, such as commercially matured 10-GHz directly modulated DFB lasers (DMLs) or electro-absorption modulated lasers (EMLs) and 10-GHz PIN detectors, assisted by spectrally efficient modulation format, such as quadrature amplitude modulation (QAM). Nonetheless, one critical drawback of DMLs or EMLs is the generated optical signal will be double-side band (DSB) and frequency-chirped, which will result in detrimental dispersion- and chirp-related power fading. As a result, after 100-km transmission over single-mode fiber (SMF), the bandwidth of the transmission system is limited to a few GHz [11,12]. Thanks to the advance in digital-signal-processing (DSP) technology, both power levels and modulation levels of OFDM subcarriers can be adaptively allocated to efficiently utilize very limited transmission bandwidth at baseband. Accordingly, our previous work has demonstrated that using 128-QAM and 3-GHz bandwidth to achieve a 21-Gbps LR-PON over 100-km SMF [11]. Moreover, to increase available transmission bandwidth, it has been proposed that modifying the chirp parameter of DMLs or EMLs to be negative [13] or replacing the deployed fibers by the negative-dispersion fibers [14]. These schemes, however, are too complicated and expensive to be feasible. Coded OFDM, which is a visible scheme to overcome fading in wireless systems [15], has been used to coped with power fading in SMF transmission [16], but at the cost of decreasing bandwidth efficiency and increasing complexity. Nonetheless, unlike wireless system, the fiber channel response is relatively stable, and power fading is periodic and predictable. Consequently, secondary low-fading bands have been proposed to carry supplementary OFDM subcarriers to achieve full utilization of transceiver bandwidth [12]. Nevertheless, within the range of 0–100 km, the lowest supporting data rate of 7 Gbps in [12] implies that the number of 1-Gbps ONU is restricted to only 7. Hence, to support 128 ONUs, 19 wavelength channels are required and this will reduce the channel spacing and increase difficulties in wavelength management.

In this work, utilizing a 10-GHz IMDD transmission system, we propose a dynamic multi-band OFDM subcarrier allocation scheme to avoid power fading effects in LR-PONs of up to 100 km. According to the different frequency responses of different SMF lengths, the power level and modulation level of each OFDM subcarrier are dynamically modified to maximize the transmission capacity. As such, this work successfully achieves a superior performance of 34.78 Gbps over 100-km SMF by economically using a 10-GHz EAM and a 10-GHz PIN. Compared with the earlier work in [11], the experimental results confirmed at least 57% increased of data rate after 100-km SMF transmission. Moreover, considering the 10-GHz bandwidth of the EAM, the narrowest available bandwidth within 100-km SMF transmission system does not occur at 100 km, but at ~40 km. As a result, this work also investigates the transmission capacities over all possible cover radii of (LR-) PONs, i.e. 20–100 km, to guarantee at least 33-Gbps obtainable capacity. Besides, subcarrier-to-subcarrier intermixing interference (SSII), which has been theoretically revealed to be harmful after transmission in [17], is also experimentally measured and discussed in this work. Finally, this experiment demonstrates a cost-effective LR-PON of up to 100 km, which can provide 1-Gbps data rate to each of 32 ONUs. Moreover, the system could simply be upgraded to support 128 ONUs by using only four wavelengths for each direction (downstream and upstream). Thus, compared with the previous proposals of LR-PONs, our proposed 33-Gbps hybrid WDM/OFDM LR-PON optimizes the trade-off between the number of wavelength channels [2–6,8,9,12] and the cost of transceivers [7], based on cost-effective 10-GHz EML and PIN based IMDD scheme.

2. Criteria and transmission bandwidth of feasible IMDD LR-PONs

Based on the trade-off between wavelength channel number and transceiver cost and the requirements of next-generation PON (NG-PON) discussed in [18], some criteria have been contemplated for a feasible LR-PON. Though some of the proposed numbers are open to discussion and may change with the development of technologies, these are the important issues that need to be addressed prudently: (a) Sustainable symmetric data rate of each ONU is 1 Gbps for supporting future broadband multimedia services [18]; (b) Power splitting ratio (N) is 32 for a reasonable optical loss budget and the aggregated data rate on a single wavelength is $\ge 32$ Gbps; (c) Supporting ONU number is $\ge 128$for sharing the capital and operational expenditures; (d) Using WDM technology to stack four (W = 4) 32-Gbps PON, thus 128 ($N\times W=128$) ONUs can be achieved by four wavelengths for each direction, and (e) Cost effective transceiver design is achieved by a 10-GHz IMDD scheme. Notably, four wavelengths for each direction (downstream and upstream) are recommended due to very limited wavelength resources, as shown in Fig. 1. Thus, the proposed architecture of a LR-PON and the corresponding parameters of the above criteria are illustrated in Fig. 2 , and the proposed LR-PON can meet the requirements of NG-PON [18] by using a cost-effect and feasible architecture. Wavelength management of four channels is not technique challenging, but the most crucial part turns out to be achieving the expected data rate of >32 Gbps via a cost-effective transceiver. In particular, while the IMDD scheme is preferred, an optical OFDM signal is generated in the form of DSB. Namely, an electrical OFDM subcarrier will be translated into two conjugated optical subcarriers on both sides of an optical carrier. After transmission in fiber link, however, the residual dispersion destroys their conjugated property and results in a well-known power fading. Moreover, since an optical signal modulated by a DML or an EML is chirped, the additional phase modulation also weakens the conjugated property between subcarriers. Accordingly, both frequency chirp and fiber dispersion will affect the received power. Considering small-signal approximation, the received power, P _r, of an OFDM subcarrier [17,19,20]

 

Fig. 2 Proposed feasible architecture for a cost-effective LR-PON.

Download Full Size | PDF

where α is the chirp parameter, ${\beta }_2$ is the group velocity dispersion (GVD) parameter (${\beta }_2=-21.66$ ps²/km for SMF), L is the fiber length, f is the frequency of the electrical OFDM subcarrier, and P_s represents the received power without fading. Therefore, as the residual dispersion increases such that $|2{\pi }^2{\beta }_{2}L{f}^2-{\tan }^{-1}\alpha |$ approaches $\pi /2$, the received electrical subcarrier will suffer serious fading to fail signal detection. Equation (1) also indicates that positive chirp will make fading happen with lower positive residual dispersion (${\beta }_2<0$). In any case, frequency components that suffer from high power fading are not suitable to carry subcarriers.

Via a network analyzer, Fig. 3 displays the measured frequency responses of an EAM-based transmission system over 20–100-km SMF. Notably, as the SMF transmission distance is increased from 20 km to 100 km, the first null point of the frequency response, f _null, shifts left from 10.18 GHz to 4.8 GHz and the available bandwidth of the 1st frequency band (i.e. baseband) is decreased from 5.56 GHz to 3.5 GHz. Using traditional baseband modulation schemes, the available bandwidth will be seriously limited to the 1st f _null. However, allocating multi-band OFDM subcarriers to the first and secondary low-fading bands, the total available bandwidth is increased dramatically. While 10-GHz transceivers are technique-matured and cost-effective, the total available bandwidth of the multiple frequency bands within 10 GHz has not been thoroughly investigated. If 3-dB power fading is used as the criteria to determine which frequency bands could carry subcarriers, the frequency bands which suffer from < 3-dB power fading could be defined as passbands and the others are forbidden bands. From Eq. (1), for a system with positive residual dispersion, the m-th 3-dB power fading will occur at frequencies f_m:

 

Fig. 3 Frequency responses of 20–100-km SMF transmissions.

Download Full Size | PDF

where m is a positive integer and $\lfloor x\rfloor$ denotes the largest integer less than or equal to x. Therefore, the passbands are defined at frequencies of < f ₁, f ₂ ~ f ₃, f ₄ ~ f ₅, and so on, and the other bands are the forbidden bands. While the chirp parameter of the EAM in our experiment measured by the method in [20] is about 0.53 at the bias voltage of –1 V, the corresponding passbands and forbidden bands after SMF transmission are theoretically shown in Fig. 4(a) . Although the bandwidth of the 1st passband (i.e. f ₁) is reduced after longer SMF transmission, the frequency of the 2nd passband also decreases, and it can be used to carry subcarriers without the requirement of high-speed (>10 GHz) electronics. Considering 10-GHz transceivers, the total bandwidth of passbands within 10 GHz is exhibited in Fig. 4(b). The total bandwidth increases after the distance of ~40 km due to the joining of the 2nd passband, but it decreases after ~80-km transmission owing to the appearance of the 2nd forbidden band. Thus, fully utilizing all the available passbands and properly allocating OFDM subcarriers can overcome the difficulty of very limited bandwidth after transmission. The following section will present the multi-band EAM-based transmission by allocating OFDM subcarriers in different passbands based on the analysis results shown in Fig. 4. Since the total bandwidth is not monotonously decreasing as transmission distance is increasing, in additional to 100-km transmission, the capacities of 20-km, 40-km, 60-km, and 80-km transmissions are also investigated to verify the feasibility of the proposed scheme over PONs of various sizes.

 

Fig. 4 Simulation results of (a) the passbands and forbidden bands and (b) the 1st-passband bandwidth and the total bandwidth of passbands <10 GHz, with α of 0.53.

Download Full Size | PDF

3. Experimental setup and results

Figure 5 illustrates the experimental setup with the insets displaying the corresponding electrical spectra for an optical OFDM transmission system over up to 100-km SMF transmission. The experimental setups of 20-km, 40-km, 60-km, and 80-km SMF are similar, except that the transmission distances are different. A point-to-point transmission is adopted to emulate a long-reach OFDM-PON, since the loss of a remote node in PONs can be compensated by an optical amplifier. The baseband electrical OFDM signal is generated by an arbitrary waveform generator (AWG, Tektronix^® AWG7122) using Matlab^® programs. The signal processing of the OFDM transmitter consists of serial-to-parallel conversion, QAM symbol encoding, inverse fast Fourier transform (IFFT), cyclic prefix (CP) insertion, and digital-to-analog conversion (DAC). The sampling rate and DAC resolution of the AWG are 12 GS/s and 8 bits, respectively. In order to utilize the bandwidth of both passbands as shown in Figs. 3 and 4, the two-band electrical OFDM signal is composed of two streams of OFDM signals with the FFT size of 512 generated owing to the limited AWG bandwidth of only 7.5 GHz. Nevertheless, since the channel number of the AWG is only 2, the signal of the 2nd band is emulated by directly up-converting the signal from the channel 2 of the AWG, instead of independent I- and Q-channels.

 

Fig. 5 Experimental setup with spectrum illustration. (a) channel 1, (b) channel 2, (c) channel 2 after up-conversion, and (d) combination of channels 1 and 2.

Download Full Size | PDF

With the detailed parameters of the multi-band OFDM at different transmission distances given at Table 1 at the end of this section, and to avoid similar scenario being repeated, the following paragraphs only describe the parameters of the 40-km and 100-km transmissions, because the transmission conditions of these two cases are severer than the others. That is, as shown in Fig. 4 (b), the 40-km SMF transmission has only one passband and the available bandwidth is the minimum, while the 100-km SMF transmission has the longest distance in spite of two available passbands. Thus, these two cases could be regarded as the performance baselines of one-band and two-band OFDM signals over 20–100-km SMF transmissions.

Table 1. OFDM subcarrier allocations, data rates, power penalties, receiver sensitivities and power budgets of 20–100-km SMF transmissions

View Table

For the case of 100-km SMF, as shown in inset (a) of Fig. 5, the channel-1 OFDM signal is located in the 1st passband, which consists of 23.44-MSym/s 64-QAM and QPSK symbols encoded at the 2nd–144th and 145th–180th subcarriers, respectively. Hence, the whole 179 subcarriers of the channel-1 signal occupy 4.195-GHz bandwidth (46.8 MHz – 4.22 GHz), yielding a total data rate of 21.78 Gbps. The channel-2 OFDM signal has the same symbol rate, but the modulation formats are 32-QAM and QPSK encoded at the 2nd–44th and 45th–76th subcarriers, respectively, shown in inset (b) of Fig. 5. After being up-converted to 7.992 GHz (corresponding to the frequency of the 341st subcarrier), the up-converted OFDM signal occupies the 265th–339th and 343rd–417th subcarriers (6.21 GHz – 9.77 GHz) located in the 2nd passband. Therefore, the up-converted signal with 3.52-GHz bandwidth is used to emulate the signal of the 2nd band with the total data rate of 13 Gbps, shown in inset (c) of Fig. 5. As shown in inset (d) of Fig. 5, through a power coupler, both two bands are then combined to achieve a total data rate of 34.78 Gbps.

For the case of 40-km SMF, both the channel-1 and channel-2 OFDM signals can be only located in the 1st passband. The channle-1 OFDM signal is consisted of 23.44-MSym/s 128-QAM and 32-QAM symbols encoded at the 2nd–94th and 95th–135th subcarriers, respectively. Namely, these 134 subcarriers occupy 3.14-GHz bandwidth (46.8 MHz – 3.16 GHz) to yield a total data rate of 20.06 Gbps. With the same symbol rate, 64-QAM and 32-QAM formats are encoded at the 2nd–28th and 29th–51st subcarriers of the channel-2 signal, respectively, and then the signal is up-converted to 4.382 GHz, i.e. the frequency of the 187^st subcarrier. Thus, by up-converting the channel-2 signal, the 136th–185th and 189th–238th subcarriers are generated at 3.19 GHz –5.58 GHz to emulate the signal of the 2nd band with a total data rate of 12.98 Gbps. After combining the signals of the both bands via a power coupler, the 33.04-Gbps OFDM signal is obtained.

Furthermore, the combined OFDM signals are then sent to the EAM to generate the optical DSB OFDM signals. After 40 km or 100 km of SMF transmission and direct-detection, the received electrical signals are both captured by a digital oscilloscope (Tektronix^® DPO 71254) with a 50-GS/s sampling rate and a 3-dB bandwidth of 12.5 GHz. An off-line Matlab^® DSP program is used to demodulate the OFDM signals and the demodulation process includes synchronization, FFT, one-tap equalization, and QAM symbol decoding. Lastly, from the constellation, the signal-to-noise ratio (SNR) is measured and used to calculate the bit error rate (BER).

Figures 6(a) and (b) show the relative received powers of each subcarrier after 40-km and 100-km SMF, respectively. The received powers are calculated by the relative power gain of one-tap equalization, and they correspond with power fading. In Fig. 6(a), since there is only one passband within 10 GHz, the OFDM signal can only be located in the 1st passband. On the contrary, the OFDM signal after 100-km SMF is located in two passbands, and the subcarriers at ~8 GHz get power gain, because of the interaction between the frequency chirp and dispersion described in Eq. (1). Moreover, for comparison, the red curves of theoretical power fading according to Eq. (1) are also plotted in Fig. 6, and they agree well with the measured relative received powers. Unfortunately, even though OFDM subcarriers are allocated in different passbands to avoid the serious power fading, the frequency response of each passband is not flat. Therefore, comparing the measured SNRs before and after transmission depicted in Fig. 7 by blue circles and green diamonds, respectively, severe SNR degradation occurs at the edges of passbands. In addtion, the frequency response of the transmitter is also uneven, especially the response of the AWG. To overcome these issues, a pre-emphasis algorithm is employed, and both the modulation levels and power levels are adjusted to optimize transmission performance. By introducing the pre-emphasis, smoother SNRs shown by the red triangles in Fig. 7 are attained to result in improved performances. The corresponding constellations before and after 40-km and 100-km SMF transmissions are shown in the insets of Figs. 7(a) and (b), respectively.

 

Fig. 6 Relative received powers and SNR penalties of each subcarrier after (a) 40 km and (b) 100 km.

Download Full Size | PDF

 

Fig. 7 SNR of each subcarrier and constellations of (a) 40-km and (b) 100-km SMF transmissions.

Download Full Size | PDF

Furthermore, comparing the SNRs before and after transmission without pre-emphasis in Fig. 7, the relative SNR penalties are also plotted by green dots in Fig. 6. The relative SNR penalty after 40-km transmission is consistent to the relative received power shown in Fig. 6(a), and this implies the penalty is mainly contributed by power fading. In Fig. 6(b), however, the SNR penalties of the subcarriers at 6.2–8.2 GHz are much severer than the degradation caused by power fading, compared with the other subcarriers. Actually, these additional penalties are caused by the dispersion- and chirp-relevant SSII [17]. In order to clarify the influence of SSII, the electrical spectra with and without 100-km transmission are shown in Figs. 8(a) and (b) , respectively, and only the 1st passband is used to carry subcarriers to simplify the investigation of SSII. The blue curves are the measured electrical spectra, and the simulation results of green curves are also drawn for comparison. Obviously, there is unexpected power increase outside the signal band. According to the SSII theory and the small-signal approximation given in [17], a chirped OFDM signal contains N _s subcarriers with the same power of P _s and the subcarrier spacing of $\Delta f$ at baseband, the power of the second-order SSII at the frequency of $n\Delta f$can be approximated as

 

Fig. 8 Experimental measured and simulated electrical spectra of (a) 1st band of optical B-to-B, (b) 1st band after 100-km SMF transmission and (c) 1st and 2nd bands after 100-km SMF transmission.

Download Full Size | PDF

where n is a positive integer less than 2N, ${\theta }_D$ equals to $2{\pi }^2{\beta }_{2}L\Delta f^2$ and P _c denotes the power of the optical carrier. Following Eq. (3), the theoretical SSII curve is plotted by the red curve in Fig. 8(b) to prove those additional frequency components are indeed caused by the SSII. While the in-band SSII (<4.2 GHz) would have small effects on the performance of subcarriers of the 1st passband, as shown in Fig. 8(b), the stronger out-of-band SSII (4.2–8.4 GHz) could be harmful to higher-frequency bands. Figure 8 (c) shows the electrical spectrum of the OFDM signal composed of two bands after 100-km transmission. Since the subcarriers at 6.2–8.2 GHz is overlapped with the out-of-band SSII induced by the 1st band, they suffer from severer SNR penalties, as shown in both Figs. 6(b) and 7(b). Thus, the total bandwidth illustrated in Fig. 4(b) is not the only concern in a cost-effective IMDD transmission system, but the dispersion- and chirp-relevant SSII must also be considered.

Furthermore, the BER performances of the EAM-based OFDM signals are plotted in Fig. 9 . Over 40-km and 100-km SMF transmissions, the receiver sensitivities are −6.5 dBm and −9.5 dBm, respectively, and the transmission penalties at the FEC threshold (BER of $3.8\times {10}^{-3}$) are less than 2.8 dB and 2.5 dB, respectively. The optical power budget is crucial for designing the transmission systems of LR-PONs, and 100-km LR-PON needs the largest optical power budget among 20–100-km LR-PONs. The optical power following the EDFA as shown in Fig. 5 is about 16.5 dBm, thus the corresponding optical power budget of 100-km LR-PON is 26 dB, which can support the optical power loss of the following 50-km SMF and the optical splitter with 32 splitting ratio. Finally, Table 1 collates all the experimental results of 20-km, 40-km, 60-km, 80-km and 100-km SMF transmissions, including the multi-band OFDM subcarrier allocations, data rates, received power penalties, receiver sensitivities and power budgets. Both the cases of 20-km and 40-km transmissions adopt only one passband, while the others have two passbands. The minimum data rate of 33.04 Gbps is obtained after 40-km SMF, and the maximum of 40 Gbps is demonstrated after 20-km SMF. Moreover, the received power penalties and the receiver sensitivities of the 40-km and 100-km systems are worse than the others, since the available bandwidth is the minimum after 40-km SMF and the 100-km system suffers from the severest SSII.

 

Fig. 9 BERs of OFDM signals before and after 40-km and 100-km SMF transmissions.

Download Full Size | PDF

4. Conclusion

To summarize, this work has experimentally demonstrated optical DSB-OFDM transmission over 20–100-km SMF links using a cost-effective 10-GHz EAM and a 10-GHz PIN. Although the frequency responses of the 20–100-km EAM-based transmission systems are seriously impaired by power fading, this work efficiently and dynamically allocates different modulation levels and power levels of OFDM subcarriers in different bands to cope with uneven frequency responses and to avoid serious power fading. The experimental results show the superior performance of at least 33 Gbps over 20–100-km transmissions and the record data rate of 34.78 Gbps after 100-km SMF based on a 10-GHz EAM-based IMDD scheme. Compared with our previous work [11], at least 57% data rate enhancement is achieved. Thus, by using the proposed multi-band OFDM subcarrier allocation scheme, the cost-effective EMLs with positive chirp parameter become available, and power fading is on longer the major bottleneck of LR-PONs. Besides, the investigation and measurement of SSII also reveal that the bandwidth limited by power fading is not the only concern; SSII must be also taken into consideration as designing the transmission system of LR-PONs. In conclusion, this experiment demonstrates a cost-effective LR-PON, which can provide more than 1-Gbps data rate to each of 32 ONUs. While 128 ONUs are required, the system could simply be upgraded by using only four wavelengths for both downstream and upstream, respectively.