1. Introduction

Due to its relatively low complexity and high stability at optical network units (ONUs), on-off keying (OOK) modulation format has been widely employed in passive optical networks (PONs) for decades. However, with the progress and popularities in recent Internet services and cloud network computing, such a simple modulation configuration reveals a great insufficiency in handling the soaring demands of data rate. In the system feasibility consideration, increasing the data rate of OOK format is difficult and inefficient because of its poor spectral efficiency and electronics bandwidth bottlenecks. Moreover, the extension of optical network distribution seriously restricts the data rate of OOK format due to the chromatic dispersion inherent in transmission fiber. Therefore, it’s an imperative mission to upgrade the legacy PONs with higher data capacity in the transmission throughput and longer distribution reach of both the downstream and upstream signals while still maintaining a low complexity in ONUs.

Orthogonal frequency division multiplexing (OFDM) technique has been widely applied in wired and wireless communications for decades. The orthogonal nature among data subcarriers and the employment of high-bit-rate vector signals of OFDM format guarantee a very high spectral efficiency, which effectively relieves the bandwidth constraint of the electronic devices and makes it a very promising solution in PONs. Moreover, the long symbol period also promotes OFDM as an effective solution to mitigate the inter-symbol interference (ISI) induced by chromatic dispersions (CD) and polarization mode dispersions (PMD) in optical fiber, and consequently extends the network reaches of PONs.

To date, OFDM applied in optical fiber communications can be mainly categorized as coherent optical OFDM (CO-OFDM) [1, 2] and direct-detection optical OFDM (DDO-OFDM) [3, 4]. Compared with DDO-OFDM, CO-OFDM technique possesses less communication bandwidth demands and better receiving sensitivities. In addition, the application of CO-OFDM in PONs also demonstrates a high splitting ratio [5]. Unfortunately, these fascinating features are only available when a high-cost receiver installation is adopted, which commonly comprises a local oscillating laser, an optical/electrical phase-locked loop (PLL), and a balanced photo-detecting set. In contrast, DDO-OFDM technique usually has no requirement of local oscillation laser and the related PLL circuit at the receiver side. Therefore it exhibits a relatively low complexity and results in a low cost expense of ONUs, which is a paramount consideration from system and service providers. Furthermore, based on the similarity in receiver implementation, the conventional OOK-based ONUs can be easily upgraded for DDO-OFDM receiving with some supplementary DSP circuits. Although DDO-OFDM technique can be adopted to meet the requirement of PONs easily, the square-law detection nature of the photo-detector will introduce signal-signal beating interference (SSBI) after O/E conversion [6], which is inevitable in DDO-OFDM scheme and is deleterious to signal quality. A guard band is conventionally allocated between the optical carrier and OFDM signal to keep the OFDM signal away from this SSBI noise, which limits the corresponding spectral efficiency. In CO-OFDM, a common approach to increase data capacity is to apply a super channel configuration that modulates multiple OFDM signal bands on a single optical carrier and then demodulate them one by one at the receiver, so that the bandwidth of the employed DSP circuits can be greatly saved due to parallel process [7]. If we apply such a super channel configuration in a conventional DDO-OFDM scheme, the beating interferences between signal bands will become more severe and complicated. To promote the data capacity in downstream transmission with minimum modification in traditional ONU architecture, we have proposed a simple receiving system for multiband DDO-OFDM data format [8, 9]. With a simple optical single-sideband (SSB) filter, we can always sift the desired OFDM signal band without SSBI contaminations. Thus the data capacity can be enlarged and the spectral efficiency can be enhanced simultaneously.

Besides, to implement a low-complexity and low-cost ONU, a major trend in PONs is to centralize the light sources at central office (CO), which means the downstream signals should be recycled for carrying upstream signals at the ONUs to save the local light sources. For the PON systems with CO-OFDM downstream, the local optical carrier is mandatory at ONUs. Although this local source can also be applied for upstream signals, it cannot meet the source-free ONU requirement. Since the optical carrier in DDO-OFDM is only for demodulating the desired OFDM signal at the ONUs, the downstream DDO-OFDM format has great potentials to achieve such a carrier-reuse function at ONUs by an optical carrier extraction setup.

Several bidirectional DDO-OFDM long reach PON systems by applying intensity modulation techniques have been presented with good system performance [10, 11]. Although intensity modulation usually possesses a better cost effectiveness and lower complexity, the quality of the modulated signals is always limited by some inherent properties. For examples: for better linearity, the modulation index in MZM or EAM is pretty limited, which results in a relatively high carrier-to-signal power ratio (CSPR) and thus too much optical power is wasted on optical carrier. Another problem of intensity modulation is that it requires real value signal modulation which wastes half of the precious modulation bandwidth of the employed modulation components. To further enhance the transmission capacity and receiving sensitivity in downstream, we apply field modulation technique to the generation of downstream DDO-OFDM signals, thus >100 Gb/s scale transmission data rate can be achieved in this proposed system. However, since the OFDM signal is encoded in electric field, the square-law detection of the photodetector induces SSBI noise which can be relieved by the proposed optical SSB filtering mechanism. In this paper, we propose and experimentally demonstrate a light source centralized bidirectional PON system in the both short-reach and long-reach optical distribution network (ODN) scenarios. The downstream signal is composed of a double-sided 6-band DDO-OFDM signal modulated on only one optical carrier. Based on the adaptive modulation configuration in signal allocation [12], the aggregated data capacity can achieve 150 Gb/s in single polarization. Under the forward error correction (FEC) code limit, the splitting ratios in the proposed PON systems can be as high as 180 and 1024 for short-reach and long-reach scenarios, respectively. Moreover, a carrier-reused mechanism is applied for carrying upstream signal at each ONU. To make a comparison, a 12.5 Gb/s non-return to zero (NRZ) quadrature phase-shift keying (QPSK) modulation format and a conventional 10.66-Gb/s NRZ-OOK signal are employed in the upstream signals. The corresponding bit-error rates (BERs) measurements are conducted for evaluating the system performances and the experimental results prove the robustness of the proposed source centralized bidirectional PON system.

2. Principles

Figure 1 schematically depicts the conceptual setup of the proposed bidirectional PON system with the spectral arrangements of the downstream and upstream signals, the proposed simple receiving system setup, and the carrier recycle and re-modulation scheme at each ONU. Each function block is described as follows.

 

Fig. 1 Conceptual diagram of the proposed bidirectional multiband DDO-OFDM PON system. (a) the spectral arrangement of the downstream and upstream signals; (b) the concept of optical SSB filter in each ONU; (c-d) simulated and experimental spectral distributions of the proposed simple receiving mechanism.

Download Full Size | PDF

2.1 A simple multiband DDO-OFDM receiving mechanism with an optical SSB filter

Enhanced by high level modulation format, OFDM potentially can carry a high volume of transmission data. However, in signal generation and demodulation, the related digital signal processing (DSP) complexity scales up with the data rate and the effective bandwidth [13], thus the scalability in physical implementation is pretty limited by the electronic bottlenecks. To effectively relieve the demand on high-end DSP circuits, while still maintain a high data rate in transmission throughput, we employ a multiband configuration for the downstream DDO-OFDM signal in the proposed PON system. As shown in Fig. 1, the downstream transmitter in the central office (CO) generates a super-channel OFDM signal with 2N bands carried on both sides of the employed optical carrier. The spectral arrangement of the proposed downstream multiband DDO-OFDM signal is illustrated as Fig. 1(a). At each ONU, we apply an optical single-side band (SSB) filter to sift out the spectral range from the optical carrier to the desired OFDM signal band before photo-detection. Such filtering keeps the desired signal band at the outmost position in the received spectrum, as shown in Fig. 1(b). If we insert a guard band with an equivalent bandwidth of one OFDM band between the optical carrier and the most inner band, after the square-law detection in a photo-detector, the induced SSBI noise will distribute from DC to the frequency next to the desired signal band. Thus, the desired band can be free from the deleterious SSBI noise. Figures 1(c) and 1(d) display the resulting electrical spectra of the proposed receiving mechanism by numerical simulation and experimental measurements respectively. Besides, by employing the proposed optical SSB filter, both the upper and lower side-bands of the optical carrier can carry distinct OFDM information without incurring any RF fading problem. Consequently, the maximum receiver’s bandwidth required at ONUs is only one half of the occupied optical spectrum.

2.2 Adaptive modulation in signal bands to enhance spectral efficiency

As more signal bands are added to the multiband configuration, due to non-zero linewidth of the employed light source, those signal bands located at higher RF frequencies suffer more chromatic dispersion induced phase deviation between the optical carrier and signal bands, which limits either the transmission distance or the acceptable QAM size in a multiband DDO-OFDM system [14]. Since less phase deviation is expected for those signal bands that are closer to the optical carrier, meaning more coherent between signal bands and optical carrier, we can regard these bands as “strong” channels in communication system. In contrast, for the more distant signal bands, they can be regarded as “weak” channels due to less coherence. To guarantee a least acceptable performance, the employed QAM level of the conventional optical OFDM is typically limited by the weakest channel if a uniform format is applied on all the signal bands. The utilization of such uniform format severely wastes the precious spectral efficiencies of the other stronger channels. Thus, according to the channel condition of each signal band, we apply a simple adaptive modulation format on each individual OFDM band to increase the data capacity and spectral efficiency of the downstream signals [12]. Therefore a higher data density is applied to those signal bands located on stronger channels, and vice versa. With such an adaptive modulation scenario, the PON system can receive more benefits in transmission capacity, performance robustness, and dynamic network flexibility with less hardware resources.

Without employing the proposed optical SSB filtering and adaptive modulation mechanism, the guard band for accommodating SSBI noise always shares half of the communication bandwidth in conventional DDO-OFDM system which results in a spectral efficiency R of only m/2, where m = log₂M indicates bits per symbol of the applied M-ary QAM. However, in the proposed system, a guard band with only one signal band is required and higher QAM level can be applied on the strong channels. Thus, the overall spectral efficiency of the proposed signal band arrangement can be theoretically represented as Eq. (1),

2.3 The carrier reused upstream signal and the related coherent receiving scheme

To reduce the cost of the ONUs, we employ a carrier-reuse architecture to save the light sources in the ONUs. Since we preserve a guard band on the both sides of the optical carrier for accommodating the deleterious SSBI noise in the proposed downstream multiband DDO-OFDM, the optical carrier can be easily extracted from the downstream signal for carrying upstream signal with high optical signal-to-noise ratio (OSNR) and little contamination from downstream signal. Therefore, by applying a simple optical carrier extraction mechanism to sift out the optical carrier for re-modulation, the light sources can be centralized at CO and the ONU can still be well functioned with less facilities. Then, the upstream signal can be re-modulated in either OOK or QPSK format as a baseband signal with little interference from the downstream multiband DDO-OFDM signals. The corresponding signal arrangement of the downstream and upstream signals can be illustrated as in Fig. 1(a).

3. Experimental setups

Figure 2 is the experimental setup for the proposed bidirectional PON system. In this system, the proposed adaptive modulation downstream O-OFDM signal is generated at CO. For the ODN, we employed both the short-reach PON and long-reach PON, with 25 km and 100 km of standard single mode fiber (SSMF), respectively. An additional in-line EDFA is applied for compensating for the transmission loss in the long-reach case. We employ full-duplex in the bidirectional transmission system, as shown in Fig. 2, to avoid Rayleigh backscattering (RB) because the related RB reduction mechanism is beyond the scope of this paper [15]. At the ONU, a multiband DDO-OFDM receiver based on the proposed optical SSB filtering is applied for demodulating the downstream signal. For the upstream signal, after optical carrier extraction, we applied two modulation formats: 10.66 Gb/s OOK and 12.5 Gb/s QPSK signals for both the short-reach and long-reach ODNs. To receive the upstream signal, a PIN detector and a balanced receiver are applied for OOK and coherent QPSK signals at the central office, respectively. The detailed architecture of each section in the experiment is detailed described as follows.

 

Fig. 2 Experimental setup of the proposed system. The upstream is demonstrated in OOK format and QPSK format respectively. (a) the implementation of optical frequency comb generation; (b) the short-reach and long-reach of optical distribution network; (c) the generated optical comb; (d) the 6-band OFDM signal; (e) the generated downstream signal; (f) the overlap of the desired signal bands; (g) the extracted optical carrier; (h) the re-modulated upstream QPSK signal. (Resolution bandwidth of all spectra is 0.01 nm.)

Download Full Size | PDF

3.1 The transmitter for downstream DDO-OFDM signal in central office (CO)

A 6-band DDO-OFDM signal is generated at the central office for the experimental demonstration. The only employed light source is a tunable external cavity laser (ECL) operated at 1552.52 nm with 100-kHz linewidth. As shown in Fig. 2, this continuous wave (CW) laser is first evenly split into three paths by an optical splitter. The upper two paths are arranged for the generation of DDO-OFDM downstream signal composed of the multiband O-OFDM signal generation path and the optical carrier insertion path. We leave the light source in the lowest path in the CO for QPSK upstream demodulation.

In the multiband O-OFDM signal generation path, the first section is a frequency comb generator, achieved by two-stage optical up-conversion, as depicted in Fig. 2(a). Two sinusoidal clock sources with their phase locked are applied to modulate the two Mach-Zehnder modulators (MZMs) individually. The first MZM (MZM1), modulated by a 15.75-GHz sinusoidal wave, is operated at optical carrier suppression (OCS) mode of the MZM. A 25/50 optical interleaver (IL) is embedded between the two MZMs to achieve a better optical carrier suppression ratio (>40 dB) due to the limited extinction ratio (20 dB) of MZM1. The second MZM (MZM2) is linearly modulated with a 6.25-GHz sinusoidal wave to generate 6 up-converted RF tones to carry the downstream OFDM signal bands, whose spectrum is as shown in Fig. 2(c). The second section is built for the modulation of O-OFDM signal, which comprises an arbitrary waveform generator (AWG) and an in-phase/quadrature-phase modulator (IQM) for optically uploading the OFDM signals. The baseband OFDM signal is offline encoded with MATLAB^TM and then digital-to-analog (D/A) converted by the AWG at a sampling rate of 12 GSa/s. The generated electrical OFDM signal is applied on the 6 optical tones by field modulating the employed IQM for better modulation linearity and thereby a higher signal quality. The generated O-OFDM signal bands are then combined with the phase-synchronized optical carrier in the optical carrier insertion path by a 3-dB optical coupler to establish the multiband O-OFDM downstream signal. Figures 2(d) and 2(e) displays the corresponding optical spectra before and after the carrier insertion. Using this carrier insertion mechanism, the optical carrier-to-signal power ratio (CSPR) can be easily adjusted. In our experimental demonstration, the CSPR is set at 8 dB to obtain better performance uniformity among the 6 signal bands [3]. After power boosting by an EDFA, this multiband O-OFDM signal is sent to the SSMF for downstream transmission at a launch power of 1.7 dBm to prevent from serious fiber nonlinearities.

The inner, middle and outer bands of the 6 O-OFDM bands are correspondingly allocated at 9.5-, 15.75- and 22-GHz away from the optical carrier. With the adaptive modulation configuration, we apply the OFDM signal with a QAM size of 64, 32 and 16 for the inner, middle and outer bands, respectively. Since we have only one set of AWG, we modulate all the 6 bands with the same QAM size simultaneously but evaluate the desired signal band at the corresponding QAM size one by one. As a result, when detecting the inner bands, the modulation QAM size is set at 64, while detecting the middle bands and outer bands, the QAM sizes are respectively set at 32 and 16 for all the channels in the experimental demonstration. In our OFDM configuration, 230 subcarriers out of 512 FFT size carry the corresponding QAM signals, which results in a baseband OFDM signal occupying 5-GHz bandwidth after inserting a 1/16 cyclic prefix (CP) for the chromatic dispersion compensation. The brief DSP procedure is illustrated in the upper left of Fig. 2. The D/A converted OFDM signal is composed of continuous frames, where 20 training symbols and 180 data symbols are arranged in each frame. With this data structure, the signal band carries 20-, 25-, and 30-Gb/s for 16-QAM, 32-QAM, and 64-QAM, respectively. Thus, the ensemble data rate of the 6-band DDO-OFDM signal is 150 Gb/s occupying 49-GHz bandwidth. The effective data rate is about 125.55 Gb/s after removing the training symbols, CP, and 7% forward error correction (FEC) overhead.

3.2 The receiving for downstream DDO-OFDM signal in optical network unit (ONU)

After transmission, the downstream signal is passively split to all the ONUs. For simplicity, a variable optical attenuator (VOA) is applied to manipulate the splitting ratios [16]. A successive 3-dB optical coupler evenly separates the received signal for downstream demodulation and for upstream re-modulation by recycling the optical carrier from the downstream signal. As mentioned in section 2, an optical SSB filter is employed before the photo-receiver to retrieve the desired signal band and prevent it from SSBI contamination. The applied optical SSB filter is passive, which is manually wavelength and bandwidth tunable. The best fitting filter order is found to be about 1.2 in Gaussian order. Based on the performance of the received signals, the optimized central wavelength settings (offset from optical carrier frequency) are 6 GHz for inner bands, 9.13 GHz for middle bands, and 12.25 GHz for outer bands, while the 3-dB passband bandwidth settings are 12 GHz, 18.25 GHz, and 24.5 GHz, respectively. The solid line in Fig. 3 depicts the optical spectrum of the double-sided 6-band OFDM signal, while the dashed lines represent the filter shapes that are employed to select the three upper side bands, respectively. To effectively suppress the un-desired signal bands, the settings of central wavelengths and 3-dB passband bandwidths is quite critical. With the matured integrated optical circuit manufacturing techniques, we believe that these filter specifications are achievable. Furthermore, if a more ideal optical SSB filter with sharper roll-off edge is available, the signal band arrangement can be more compact. An overlapped spectrum after correspondingly filtering the three bands of the upper side, along with the optical carrier, is displayed in Fig. 2(f). Since the detection scheme is direct detection, the O/E down-conversion is achieved by only a single-end photo-receiver, and the received OFDM signal is then analog-to-digitally (A/D) converted by a real-time oscilloscope at a sampling rate of 80 GSa/s. The signal demodulation and decoding is then conducted with MATLAB^TM offline. The corresponding DSP procedure is provided in Fig. 2. As indicated, before CP removal, we apply a digital down-conversion to demodulate the I and Q signals. After the FFT, QAM signals are retrieved, and the one-tap equalization is adopted to neutralize the channel response. Besides, to compensate for the common phase error (CPE), we also introduce a zero-overhead phase compensator to take care of this issue without any loss of data capacity [3]. The final DDO-OFDM system performance is evaluated by taking the received signals’ error-vector magnitude (EVM) values, defined in IEEE 802.11a-1999^TM standard, and estimating the corresponding bit error rates (BERs) from the measured EVMs [17] in our experiment.

 

Fig. 3 The spectral responses of the applied optical filters and the transmitted multiband signal.

Download Full Size | PDF

3.3 The upstream signal: OOK vs. QPSK

For upstream transmission, we send two data formats: 10.66-Gb/s OOK signal and 12.5-Gb/s QPSK signal from the ONU as a comparison. The corresponding receivers are then deployed at the CO accordingly. For re-modulating the upstream signal, we employ a fiber Bragg grating (FBG) centered at 1552.52 nm with 0.08-nm pass-band bandwidth and an optical circulator to extract the optical carrier from the downstream signal. For OOK, a 2²³-1 pseudo-random bit sequence (PRBS) is generated by a pulse-pattern generator (PPG) at 10.66-Gb/s to modulate the single-drive MZM. As for the QPSK, two 6.25-Gb/s PRBSs with the length of 2¹⁵-1 are generated by two independent PPGs to modulate the I/Q modulator (IQM2) in the ONU. The extracted optical carrier and the corresponding 12.5 Gb/s QPSK modulated upstream signal are displayed in Fig. 2(g) and Fig. 2(h), respectively. After power boosting, the upstream signal is fed into another 100-km SSMF to avoid the deleterious RB from the downstream signals. In the CO, the corresponding receivers are employed, as depicted in Fig. 2. A commercial 3R receiver is applied for detecting the OOK signal. The BER measurements for OOK upstream signal are performed by an error detector in real time. On the other hand, the received upstream QPSK signal is demodulated with a self-homodyne coherent receiving system. Since the upstream carrier is recycled from the downstream one, the local reference can be provided by the original laser directly. Thus, the local reference and the received QPSK signals are sent into a 90° optical hybrid simultaneously and the successive balanced photo-detector will demodulate the received QPSK signal from optical to electrical domain. After the A/D conversion with a real-time oscilloscope, the required DSPs, such as constant modulus algorithm (CMA) equalization mechanism [18], phase synchronization, as well as the signal de-mapping, are completed with MATLAB^TM offline. The BER is evaluated by error counting. Note, due to the simple architecture of PON system, neither optical nor electrical dispersion compensation is applied for both the OOK and QPSK upstreams for a fair comparison.

4. Experimental results

To establish a benchmark of the proposed bidirectional PON system, the BER measurements of both the back-to-back and transmission performance are conducted in the downstream and upstream scenarios. Moreover, we apply the 7% FEC threshold at BER = 3.8x10⁻³ [19] as the performance benchmark for all the signals.

4.1 Downstream transmission in short-reach (25 km) scenario

In the short-reach scenario, the transmission link in the ODN is 25-km long without in-line amplification. Figure 4 exhibits the BER curves of the 6 bands. As shown in Fig. 4, by using polynomial fitting curves, the averaged received power sensitivity at the specified FEC threshold of all the 6 signal bands is estimated to be about −27.5 dBm in back-to-back, while the threshold sensitivity is about −27.3 dBm after 25-km transmission. The sensitivities of all the signal bands spreads within 1 dB and about 0.2 dB penalty after 25-km transmission is observed. In Fig. 4, we noticed that a highly uniform performance among all the signal bands is achieved by employing the adaptive modulation format after fiber transmission. Since there exists a nonzero laser linewidth of about 100-kHz in the light source, a phase deviation between the optical carrier and OFDM signals band will gradually lose their coherence as the increases of either RF frequency or transmission distance due to chromatic dispersion. Therefore, the arrangement of 16-QAM format at the most distant channel, namely the weakest channel in this PON system, requires a higher per bit power to resist such impairment. In contrast, for the closest channel, also defined as the strongest channel, less loss of coherence between optical carrier and OFDM band is experienced. Thus 64-QAM format is employed, which results in a 1.5 times data rate of the outmost band and the total transmission capacity can be enhanced with the same channel bandwidth. An intermediate signal format, 32-QAM, is then applied in the middle channel. With this adaptive modulation format, the performance difference at the same FEC threshold is almost negligible among the 6 bands. For the overall performance, we only observed about 0.2 dB penalty after transmission. Since the launch power at the CO output is set to be 1.7 dBm to keep the signal immune from fiber nonlinearities, the 0.2-dB transmission penalty is mainly contributed by sampling instability due to offline process. An evaluation of power splitting ratio is displayed in Fig. 5. Under 1.7 dBm optical launch power at CO, we obtained about −4.5 dBm received optical power after 25 km transmission and a 0.8 dB insertion loss from VOA. Thus the potential splitting ratio of this short-reach PON can achieve about 1:180 at an average receiving sensitivity of −27.2 dBm for FEC threshold. For a short-reach PON system with 128 ONUs, we can have about 1.6 dB operational power margin.

 

Fig. 4 The BER curves of short-reach scenario.

Download Full Size | PDF

 

Fig. 5 The splitting ratio of short-reach scenario.

Download Full Size | PDF

4.2 Downstream transmission in long-reach (100 km) scenario

In the long-reach scenario, the transmission link extends to 100-km. An in-line EDFA is then installed before the variable optical attenuator (VOA) to compensate for the power loss in transmission. Shown in Fig. 6, the received power sensitivity at the specified FEC threshold is about −27.5 dBm in back-to-back condition, while that in 100-km transmission is about −27 dBm, after applying polynomial fitting curves. The performance of all the transmitted OFDM signal bands are almost the same due to two major reasons: one is the aforementioned non-uniform modulation format arrangement based on the adaptive modulation concept, and the other one is the applied digital CPE compensator in the signal demodulation. Since after 100-km fiber transmission, the signal accumulates four times dispersion with respect to the short-reach scenario, the common phase deviation is therefore enlarged. Such an effective CPE compensator without any overhead employs all tentative decisions of the received OFDM signal to estimate the corresponding CPE. The final results are revealed in Fig. 6, which indicates an almost identical performance to the short-reach scenario. The observed 0.5-dB transmission penalty, which is slightly larger than that of the short-reach scenario, is mainly contributed by the slight OSNR degradation due to the deployed in-line EDFA.

 

Fig. 6 The BER curves of long-reach scenario.

Download Full Size | PDF

The splitting ratio evaluation is also conducted, as depicted in Fig. 7. The optical launch power at CO is still set at about 1.7 dBm. To compensate for the power loss after 100-km fiber transmission, an in-line amplifier with 28 dB gain is employed which results in a maximum received power of about 3.25 dBm (i.e. only one ONU), where an additional 5-dB power loss is attributed to the employment of optical filter with 0.8-nm passband bandwidth followed by the in-line EDFA to reject the excess ASE noise. The maximum splitting ratio in this long-reach PON is found to be about 1:1024, as shown in Fig. 7. However, in such circumstance, there’s almost no operational power margin left.

 

Fig. 7 The splitting ratio of long-reach scenario.

Download Full Size | PDF

4.3 Upstream transmission scenarios

In the upstream transmission, we evaluate the performance difference between OOK format and QPSK format side-by-side for a vivid comparison. Figure 8 depicts the BER curves as a function of the received optical power for both cases. At first, we evaluate the performance of the 10.66-Gb/s OOK format for back-to-back, short-reach and long-reach scenarios. As we can expect, the BER curves of the long-reach scenario exhibits an obvious power penalty, about 1.5 dB at BER = 10⁻³, due to the accumulated chromatic dispersion, as shown in the BER curves and the corresponding eye diagrams in Fig. 8. As for the QPSK case, since QPSK bears a smaller effective bandwidth than it of the OOK signal at the same data rate, it has much stronger tolerance to chromatic dispersion. Based on this concern, we employ a 12.5-Gb/s QPSK format instead. With the proposed optical recycle mechanism at ONU and self-homodyne coherent receiving system at CO for the upstream QPSK signal, we found that the receiving sensitivities of all of the scenarios in ODN locate at about −45 dBm at the specified FEC threshold, as illustrated in Fig. 8. The penalty after long-reach QPSK transmission is about 0.5 dB. Compared it with the OOK, the power sensitivity for QPSK is much lower, by about 4 dB, in the long reach scenario, thus QPSK can enjoy much more system margin.

 

Fig. 8 The BER curves of OOK and QPSK.

Download Full Size | PDF

5. Conclusion

We have proposed and successfully demonstrated a double-sided multiband DDO-OFDM PON system based on a bidirectional configuration. In the downstream transmission of both short-reach and long-reach scenarios, the proposed simple receiving system via an optical SSB filter in each ONU guarantees the immunity of SSBI noise. An adaptive modulation configuration with 64-QAM, 32-QAM, and 16-QAM is also proven to improve the spectral efficiency while still maintains a uniform performance. The receiving power sensitivities at the specified FEC threshold are −27.5 dBm and −27 dBm before and after the long-reach transmission respectively, and the power splitting ratio achieves 1:1024. To overcome the accumulated chromatic dispersion and to achieve better receiving sensitivity, we employed a 12.5-Gb/s QPSK format in the upstream transmission. By applying the self-homodyne coherent receiving scheme, a received power sensitivity of about −45 dBm is obtained which is much lower than the required sensitivity of OOK.