1. Introduction

Recently, the bandwidth demand for the access network has witnessed a sharp increase driven by various services like business IP traffic, super HD video, mobile traffic backhaul and social networking, etc [1]. Passive optical network (PON), as an optimal technique for the access network, has also evolved to keep pace with the need. Owing to the development of passive splitters and burst mode transceivers, two typical examples of time division multiplexing-passive optical network (TDM-PON), Gigabit PON (GPON) based on ITU-T G.984 and Gigabit Ethernet PON (GEPON) based on IEEE.802.3ah are widely being deployed worldwide [2, 3]. TDM-PON has the advantage of low installation and maintenance cost of infrastructure, while the inherent power splitting architecture leads to insufficient power budget that severely constrains the number of subscribers and the transmission distance. Moreover, the access rate enjoyed by every optical network unit (ONU) is also restricted because of the shared bandwidth allocated to different time slots. Given this, 10G EPON and next generation PON (XGPON) have been standardized to increase the access rate and support full service [4]. With the further evolution of access network and the increasing bandwidth demand, the future access network beyond 10G TDM/TDMA should be able to converge heterogeneous services and support more than 1000 users with more than 1Gb/s access rate in a system distance of 60~100km [5, 6]. Multiple candidates have been proposed to satisfy the requirement. Wavelength division multiplexing (WDM) technology has been introduced [7]. In WDM-PON, each ONU occupies one dedicated wavelength thus it does not suffer splitting loss. But WDM-PON features a higher cost especially with the increasing number of ONUs thus a hybrid TDM/WDM-PON has been put forward [8]. Besides, to extend the distance and increase the number of subscribers, ultra-dense WDM-PON and coherent PON have been investigated [9, 10]. However, the former is sensitive to the quality of laser source and wavelength stability of the multiplexing and de-multiplexing devices. The latter scheme suffers from high cost and complicated implementation of coherent detection in ONUs. Orthogonal frequency division multiple access (OFDMA) technology has also attracted a lot of attractions since the orthogonality of subcarriers allows higher spectral efficiency and overall high speed access rate [11]. However, rooted in broadcast mechanism for OFDMA downstream, every ONU has to receive all the data while only extracts its own subcarriers thus the efficiency of devices bandwidth utilization is poor. To date, with the convergence of optical access network and metro-area network (MAN) to consolidate versatile data services with improved power efficiency, it is anticipated to combine current technologies with disruptive evolutions to support significantly increased data capacity for a large number of subscribers with wider area coverage and longer feeding distance [12].

The space division multiplexing (SDM) based optical communication has been proposed to be a favorable solution to accomplish the capacity crunch threat of the optical fiber transmission system. Few mode fibers (FMFs) [13] and multicore fibers (MCFs) [14] are intensively studied as the physical media for SDM communication systems. With the advanced signal processing techniques and sophisticated spatial-domain multiplexer/de-multiplexer (Mux/De-Mux) components, it has been proved feasible to utilize the spatial dimensions to enhance the capacity orders of magnitude together with coherent detections in long haul transmission systems [15, 16]. The SDM technology is also possible to penetrate into access network criteria from long distance applications just like the TDM/WDM technologies. Being able to unearth the spatial dimension resource, exploration in both capacity and subscription number are expected. Although the FMF based access network examples have been reported very lately [17, 18], the differential modal dispersion and modal interference may hinder its deployment in the access network region and MCF is actually a better choice owing to its well-controlled inter-core crosstalk and almost identical transmission quality compared with standard single mode fibers (SSMF). Zhu et al demonstrated a 7-core based optical access network using traditional TDM-PON technologies [19]. However, the access data rate and the fiber link distance are quite limited (2.5 Gb/s and 11.3 km). To keep pace with the requirements imposed by NG-PON2 and beyond, the MCF-based SDM scheme needs to incorporate with the standardized WDM technology and affordable advanced modulation formats to fully unveil its potential in the optical access network while no such results have been reported, to the best of the authors’ knowledge. As a universal platform for wired/wireless data services, the optical access network plays even more important role in the 4G/5G mobile data transmission [20] and it is also interesting to envision the application of MCF in the fiber/wireless converged networks.

In this paper, we move forward to propose a novel wavelength-spatial division multiplexing (WSDM) optical access network architecture with MCFs and advanced modulation formats. The MCF we developed and fabricated has seven cores in a hexagonal array and its geometrical and optical parameters are described in details in the following section. Being essential for any SDM experiments, fan-in/fan-out spatial Mux/De-Mux devices are in-house developed using chemical etching process and fiber bundle manufacturing technique. After that, we demonstrated a proof of concept experiment with 58.7km MCF and a pair of fan-in/fan-out devices. To increase the access rate in the cost-sensitive access network, we adopted the optical orthogonal frequency division multiplexing quadrature phase shifted keying (O-OFDM-QPSK) modulation format to intensity-modulate the optical wavelengths and the optical signal is directly detected in ONU. 10 wavelengths with 25 GHz channel spacing from an optical comb generator were employed and each wavelength was modulated with 5Gb/s (4.11Gb/s net rate) signal. The six outer cores of MCF were used as parallel channels between the optical line terminal (OLT) and ONUs. 10-wavelengths were then amplified to boost the power and were injected into 6 cores simultaneously after power splitter and a fan-in component. An aggregated 300Gb/s (229.34Gb/s usable) downstream (DS) data-rate has thus been achieved. For one ONU, 5Gb/s non-return-to-zero (NRZ) on-off-keying (OOK) signal is generated for the upstream (US) transmission in the same core and negligible performance degradation is obtained for DS signal. As an emulation of high speed mobile backhaul transmission, we generated 20Gb/s per wavelength QPSK signal that is transmitted in the inner core of MCF. The US mobile backhaul signal was coherently detected in the OLT. Both DS and US optical signal exhibit acceptable performance with sufficient power budget.

2. Proposed WSDM access network architecture

The proposed DS/US WSDM optical access network architecture is illustrated in Fig. 1. In our proposed architecture, the most prominent feature is that a physically isolated fiber channel (the inner core of a typical 7-core MCF for example) is allocated to the wireless data transmission such as the mobile backhaul transmission considering the mobile internet demand is booming. The outer cores of MCF are employed as the parallel channels for DS/US transmission. In the OLT block, m wavelengths are utilized as the laser source. For each wavelength in one subset OLT, it is power split by$N+1$ in which N representing the number of outer cores of MCF for DS/US wired signal transmission. $1/(N+1)$ of the signal power is left as the local oscillator for coherent detection of mobile backhaul signals in the OLT. In this way, this configuration can support $N\times m$subscribers only employing m laser sources that can lower the expense compared with the same situation in WDM-PON. To further enhance the capacity with affordable cost and complexity, downstream signal on each wavelength is suggested to intensity-modulated with optical OFDM scheme, which is spectral efficient and bandwidth flexible. After the modulation, the $N\times m$ branches from ${\lambda }_1$ to ${\lambda }_m$ are multiplexed respectively by N m-wavelength Mux devices like array waveguide gratings (AWGs). Afterwards, N sets of multi-wavelength signals are amplified by N erbium-doped fiber amplifiers (EDFAs). After getting through the circulators, the N sets signals are injected into N outer cores of MCF taking advantage of the fan-in device. Subsequently, the DS signals are transmitted in the MCF, and output to N independent single mode fibers by the fan-out device. The details of fan-in/fan-out device will be illustrated in the next section. Signals from each core containing wavelength from ${\lambda }_1$ to ${\lambda }_m$ are demultiplexed respectively and each ONU enjoys one dedicated wavelength. $N\times m$ ONUs can be supported by this configuration. With the abundant bandwidth provided by dedicated wavelength, each ONU will be able to support an additional TDM-PON with the low-loss fan-out and wavelength DeMux devices. Therefore our proposed WSDM optical access network has the potential to deliver multi-giga-bit services to a substantial number of subscribers. For US transmission, typical colorless ONU technology with wavelength reuse from DS transmission can be deployed as many prior arts presented [21]. As an alternative, we can use a tunable laser and an optical intensity modulator in the ONU side as an upstream transmitter when the link loss is large, as illustrated in Fig. 1 and subsequent experiments. With a tunable laser, the wavelengths used in US and DS can be offset to eliminate Rayleigh backscattering noise.

 

Fig. 1 Proposed WDM/SDM optical access network architecture.

Download Full Size | PDF

3. Development of 7-core multicore fibers and fan-in/fan-out device

MCF emerged as a possible solution to fulfill the capacity requirement of optical fiber transmission system few years ago. There exist 7 cores and 19 cores MCFs in a hexagonal array and 12 cores MCF in a ring structure [22, 23]. To ease the fabrication of fan-in/fan-out device, we applied the stack and draw process to fabricate 7-core homogeneous MCF whose properties of cores are almost the same. Before the fabrication, we used the commercial finite element analysis software-COMSOL Multiphysics to simulate the optical properties of optical fiber such as effective area, macro bending loss and so on, and to calculate the electric field distribution of the model. Then we integrate the data in MATLAB to figure out the mode coupling coefficient (MCC) between adjacent cores. The MCF crosstalk can be calculated with the aid of MCC [24]. Compared with the MCF we fabricated before [25–27], a trench-assisted index profile is optimized on the basis of G.657.B3 fiber under the guidance of simulations. The refractive index profile designed is shown in Fig. 2(a). The cladding diameter is 150μm and the core pitch is set as 42μm, respectively. The crosstalk between adjacent cores is suppressed to be as low as −45dB/100km. The effective area is about 77$\mu m^2$. The attenuation of inner core is measured as 0.24dB/km at 1550nm, while that of outer cores is about 0.32dB/km. Such difference is caused by excess loss due to a little small outer cladding thickness [28]. The cross section view of the MCF used for our experiments is shown in Fig. 2(b).

 

Fig. 2 (a) Refractive index profile of one MCF core; (b) cross section view of fabricated MCF (c) cross section view of fabricated fiber bundles.

Download Full Size | PDF

The low-loss and reliable connectivity between the MCF and the SSMF is of the most significance for any MCF based applications. There exist several methods like free-space coupling, tapered MCF connector (TMC), etched fiber bundles, and so on [29, 19, 30]. We chose the chemical etching method to fabricate fiber bundle that is used to be the fan-in/fan-out devices. Generally, 7 bare SSMFs are chemical etched until the cladding diameter matches with the MCF core pitch. The etched SSMFs are then inserted into a standard high precision ceramic ferrule (126μm inner diameter) and the resulted fiber bundle is arranged in hexagonal lattice just like the MCF core arrangement. The cross section of fiber bundles we fabricated is shown in Fig. 2(c). In [30], after etching, a long transition region exists and the cladding diameter of SSMF decreases all the way to the end thus a cone shape appears from the side view. Therefore it is difficult to control the end diameter and special holder must be designed instead of the standard telecom grade ceramic ferrule. Alternatively, we developed proprietary technology of chemical etching to reduce the length of transition region sharply and to keep a uniform etched diameter long enough for fiber bundle fabrication. The side view of our etched SMFs is shown in Fig. 3(a). It can be seen that after a short tapering region, uniform etched fiber diameters can be maintained precisely to be 42μm with sufficient longitudinal dimensions for the ease of fiber bundle arrangement inside the ceramic ferrule. The MCF is also chemical etched to 125μm to be able inserted into another ceramic ferrule with identical size and precision tolerance. After end polishing, both ceramic ferrules are aligned in the 6-dimension free-space alignment platform. The laser source and power meter are used to monitor the core-to-core alignment performance. At last both ferrules with glue are fixed in the glass tube with the exposure of ultra-violet light. The photo of our fabricated fan-in/fan-out device is shown in Fig. 3(b). The crosstalk between adjacent cores is measured as about −50dB. The insertion loss is show in Table 1. The inaccuracy of the MCF core pitch and the uncertainty of fiber bundle position determine the insertion loss largely. The improved hands-on experience and rigorous process control during every step will help to further reduce the insertion loss.

 

Fig. 3 (a) Side view of etched fiber bundles. (b) Picture of fan-in/fan-out device.

Download Full Size | PDF

Table 1. Insertion loss of fan-in/fan-out device

View Table

4. Experimental demonstration of WDM/SDM access network

To verify the feasibility of our proposed WSDM access network architecture, we conducted a proof of concept experiment and the experimental setup is depicted in Fig. 4. For DS transmission, ten wavelengths with 25GHz channel spacing are selected by a waveshaper (Finisar WaveShaper 4000s) from an optical frequency comb generator (OFCG) which is seeded by an external cavity laser (ECL) centered at 1550.12nm. Then the ten continuous waves (CWs) are intensity modulated with 5Gb/s baseband OFDM-QPSK signal. The transmitted signal is generated by MATLAB program originated from 2¹⁵-1 pseudorandom binary sequence (PRBS), and then mapped into QPSK modulation formats. For real-valued OFDM signal generation, Hermitian symmetry must be ensured and the first subcarrier is abandoned for eliminating noises near DC component. We adopt 128 points for the IFFT/FFT process. Thus the effective number of subcarriers in our system is 63. The length of cyclic prefix (CP) and frame are 13 and 139, respectively. The training sequence includes 11 OFDM symbols, in which one is used for frame synchronization and others are used for channel estimation. The time-domain OFDM signal is D/A converted by an arbitrary waveform generator (AWG 7122C, Tektronix, operated at sampling rate of 5GS/s) to drive the optical intensity modulator (MX-LN-20, Photline Technologies, biased at the quadrature point) after RF amplification. Boosted by an EDFA, the optical OFDM signals are power split by a 1:8 power coupler and simultaneously injected into six outer cores of the MCF through the fan-in device, and the optical spectra of amplified optical OFDM signals is shown in Fig. 5(a). After 58.7km MCF transmission, the signals are output into six single mode fibers through fan-out device. At the receiver side of every single mode fiber, after pre-amplification and filtering, 10 wavelengths are de-multiplexed and one wavelength is selected and directly detected by a photodetector (PD) with 2.4GHz bandwidth and then sampled by a 20GS/s digital sampled oscilloscope (DSO, Tektronix CSA7404B). The optical spectra of the signal under test is illustrated in Fig. 5(b). Demodulation and bit error ratio (BER) counting are implemented offline. During the offline DSP, down-sampling of the received signal, frame and frequency synchronization are implemented before CP removing and FFT operation. Afterwards, de-mapping is followed by channel estimation. In OB2B configuration, the fiber is replaced by a variable optical attenuator (VOA) with equal loss, so the comparison of the transmission performance is under fair condition.

 

Fig. 4 The experimental setup schematic diagram (OC: optical coupler, PC: polarization controller, WSS: wavelength selective switch, AWG: arbitrary waveform generator, IM: intensity modulator, VOA: variable optical attenuator, EDFA: erbium doped fiber amplifier, ECL: external cavity laser, DCM: dispersion compensation module).

Download Full Size | PDF

 

Fig. 5 optical spectrum of optical OFDM signals: (a) for ten wavelengths (b) for one channel after de-multiplexing.

Download Full Size | PDF

The BER performance of QPSK based OFDM DS signal centered at 1550.12nm from 6 outer cores at various received optical power after MCF transmission and in OB2B setup is shown in Fig. 6, with constellation diagrams of received power at −14dBm, −17dBm and −21dBm from core 1 as the inset. The detected optical signal at other wavelengths in different cores present similar performance thus we just plot one wavelength channel to save the space. No significant transmission performance deterioration is observed between OB2B and 58.7km MCF transmission for all six spatial channels, because the adequate cyclic prefix (CP) is used to conquer the impact of chromatic dispersion (CD) and polarization mode dispersion (PMD) of MCF. The BER can be kept under 7% Forward Error Correction (FEC) limit at BER = 3.8 × 10⁻³ at the received optical power as low as −16dBm. Therefore an aggregated 300Gb/s DS capacity (229.34Gb/s usable) has been realized with 10 wavelengths and 6 cores, a combination of spectral and spatial dimensions. A BER floor has been observed when the received optical power excess −15dBm for all six spatial channels both in OB2B or fiber transmission. This is mainly due to the relatively poor optical signal-to-noise ratio (OSNR, about 25dB) of the multi-wavelength optical frequency comb generator and also the inherently high peak to average power ratio (PAPR) of the OFDM signal, which will drive the PD working in a saturation mode leading to signal distortions. Moreover, the inter-symbol interference (ISI) caused by the low bandwidth PD we used in our experiment also contribute to the BER floor. By using a multi-wavelength source with better OSNR and bandwidth sufficient PD, together with PAPR reduction methods [31], BER performance can be improved.

 

Fig. 6 The BER curve for DS transmission (inset figures: constellation diagrams of received DS signals from core 1 with optical power at −14dBm, −17dBm and −21dBm).

Download Full Size | PDF

For the US transmission from the ONU side, 5Gb/s OOK signal is modulated onto a tunable laser and then launched into core 1 of the MCF through a circulator. The US signal is dropped by another circulator in the OLT side before amplification and filtering. The real-time oscilloscope captured the detected signal and the BER counting is implemented offline. The accumulated chromatic dispersion in the fiber link is compensated in the OLT side by a tunable dispersion compensation module (TDCMXPU-0096, TERAXION) whose compensation range is from −800ps/nm to 800ps/nm. The wavelength used in US is centered at 1556.55nm, which is different from that of DS in the same core, so that the penalty induced by Rayleigh backscattering effect is negligible. The DS transmission results are shown in Fig. 7(a). We noted that in the bidirectional case, no significant performance degradation in DS transmission is observed thanks to the independent wavelengths used in the two directions, despite that the overall power budget for DS is reduced by about 1.5dB which results from the two circulators used in bidirectional configuration. The US transmission performance is shown in Fig. 7(b) and eye diagram of received US signal at different received optical power is also depicted in Fig. 7(c). The reason to use OOK modulation format in US transmission is to reduce the transmitter complexity in the ONU side.

 

Fig. 7 (a) The BER curve for DS transmission in core 1 with and without US transmission; (b) The BER curve for US transmission in core 1; (c) The eye diagrams at received optical power of −8dBm and −15dBm.

Download Full Size | PDF

As proposed in the architecture shown in Fig. 1, the inner core of the MCF is reserved for large capacity wireless data service and we thus performed a US transmission of 20Gb/s QPSK IQ modulated signal from ONU to OLT as an emulation of the high speed mobile backhaul application. A CW laser from ECL is modulated by an IQ modulator with 20Gb/s QPSK signal generated by BER tester (BERT).The signal reusing the same wavelength is transmitted in the inner core of MCF such that the interference of Rayleigh backscattering is eliminated. After amplification and filtering, the signal with −1dBm power is coherently detected (Tektronix OM4006D) at the OLT. After coherent detection, the output electrical signal is digitalized by a real-time oscilloscope (DSA 72504D) and then offline digital signal processing is implemented using the traditional DSP flow without polarization demultiplexing [32]. The mobile backhaul transmission result is shown in Fig. 8, where the constellation diagram and the eye diagram at received optical power −17dBm are inserted as inset. The received power satisfying the BER requirement can be as low as −19dBm, showing that more than 18 dB power budget can be guaranteed. The coherent receiver is placed in the OLT side thus its cost and power consumption can be shared by many ONUs or mobile base stations.

 

Fig. 8 The BER curve for mobile backhaul transmission (inset figures: constellations and eye diagrams at received optical power of −17dBm).

Download Full Size | PDF

5. Conclusion

We have proposed and experimentally demonstrated a duplex WSDM optical access network utilizing our in-house developed 7-core MCFs and fan-in/fan-out devices. The proof of concept experiment proves the capability of the MCF based access network in terms of long reach transmission (58.7km), large capacity (300Gb/s aggregation DS data rate) and massive count of users (60 ONUs in current experiments). It is apparently that by using the affordable directly detected O-OFDM signal transmission together with multiple spatial channels in one fiber, optical access data rate could be significantly enhanced without adding much complexity in the ONUs. Since the optical signals transmitted in multiple cores are spatially De-Muxed by fan-out devices, sufficient power level can be maintained thus it can be further power split to support standard TDM-PONs to further increase the user number and the feeding area. Therefore, the additional spatial dimension we introduced in this architecture is prominent for future evolution of flexible and manageable optical access networks with the help of adaptive modulation formats and efficient signal processing.

With the development of MCF, more spatial channels are expected to be feasible in one pipeline thus it emerges as an ideal platform for future data-driven multi-service featured communications. As demonstrated by our experiments, 20Gb/s per wavelength US transmission in a physical isolated core has been successfully established and it can be used to support the mobile backhaul like wireless applications. We also demonstrated a tunable laser based US transmission to enable bidirectional transmission in a single core. Another interesting application scenario is the intra data-center communication [33]. High-density fiber cable connections between servers and racks produce congestion problem. The MCF integrated with various passive/active devices will become a viable solution by providing much higher data density per cross-section of the fiber.