1. Introduction

Wavelength division multiplexed (WDM) passive optical network (PON) technology has been considered to be a powerful means to realize an access network that is capable of handling the ever-increasing demands of data bandwidth, security, and scalability by local subscribers [1]. In the implementation of practical WDM PON networks the most critical issue is how to realize low cost multiwavelength light sources and wavelength routers.

Among various multiwavelength light sources the ASE spectrum-sliced incoherent light source is very attractive since stable multiple wavelength channels can be easily obtained by simply slicing an ASE spectrum through use of a multiwavelength filter. However, the fact that the network performance such as data rate, power budget, and deployment distance is severely limited by both low optical power and excess intensity noise of the spectrum-sliced sources [2, 3], still prohibits the ASE spectrum-sliced light source based PON technology from being deployed in real access networks.

One simple approach to tackle the low power issue of the ASE spectrum sliced source is to optically amplify the weak signals after a spectrum slicing filter or in the middle of a transmission link [4, 5]. The use of erbium-doped fiber amplifiers (EDFAs) and linear optical amplifiers (LOAs) have been proposed so far [5, 6].

In this paper, we propose a high performance, fully bidirectional, ASE spectrum-sliced light source based PON architecture that utilizes distributed Raman amplification and pump recycling techniques to sort out the issues of both low light source power and overall system simplicity. More specifically, we make use of distributed Raman amplification over a transmission fiber to obtain lossless signal transmission for the relaxation of the tight optical power margin. At the same time, we recycle residual Raman pump power to pump an EDF-based upstream broadband ASE source located at a remote node (RN). Note that significant Raman pump power is unused and wasted after Raman amplification in an optical fiber unless the optical fiber is long enough to attenuate all of the pump power [7].

Using our proposed architecture a range of benefits is expected to be achieved. First, subscriber units including a RN and optical network units (ONUs) can be constructed without active light sources since an upstream broadband ASE source is generated by recycling residual Raman pump power. Second, Fully-centralized light source control is possible since light sources at ONUs are remotely controlled at a central office [8]. Third, the network performance limitation caused by low optical power of ASE spectrum-sliced sources is significantly reduced owing to lossless/amplified signal transmission by distributed Raman amplification. Fourth, fully bidirectional signal transmission can be achieved since distributed Raman amplification is easily implemented without using isolators. Fifth, low-cost and wavelength-insensitive (colorless) ONUs without temperature control can be readily realized by use of Raman pump-recycled erbium ASE and electroabsorption modulators (EAMs).

Although Raman amplification has been considered to be less favorite for such cost-sensitive applications as access networks, our proposed scheme could overcome the cost sensitive overall performance issue owing to the active light source free subscriber end configuration that is based on Raman pump recycling. Note that EDFA-or LOA-based conventional PON schemes require separate, active upstream light sources at subscriber ends, which result in the increase of overall system cost.

Our architecture is implemented by use of polarization-insensitive EAMs, which enable us to avoid both signal power loss and signal-to-noise ratio degradation caused by additional polarization filtering process [9]. The proposed system performance is evaluated at a data rate of 622 Mbit/s over both 25 km and 50 km single-mode fiber (SMF) links.

2. Proposed WDM-PON architecture

 

Fig. 1. Our proposed bidirectional WDM-PON architecture

Download Full Size | PDF

Figure 1(a) shows our proposed bidirectional WDM-PON architecture. An erbium fiber ASE is employed as a downstream light source at a central office and is spectrally sliced by using a conventional AWG to generate N downstream signals. The sliced downstream beams are modulated by use of EAMs. The N modulated signals are then multiplexed with another AWG and transmitted over a transmission fiber. Note that the multiplexing AWG is a cyclic AWG that exhibits a periodic wavelength passband and is also used for upstream signal demultiplexing [5]. The wavelengths of the downstream and upstream signals are allocated to have a wavelength separation corresponding to the cyclic AWG wavelength period. The conventional AWGs after the ASE sources also play a role as band-limiting filters to fit the ASE output spectra into the allocated wavelength bands.

One of the key points in our proposed scheme is to use distributed Raman amplification over a transmission fiber to obtain lossless, bidirectional signal transmission. Raman pump laser diodes (LDs) are located at the central office for the purpose of fully-centralized control of all active light sources within the proposed system. Using the fact that non-negligible pump power remains unused after distributed Raman amplification in case of sufficient fiber loss compensation over a short reach fiber [7], residual Raman pump power is recycled as a pump beam for an EDF based upstream ASE source after the transmission fiber. The residual Raman pump beam is separated from the downstream signals using a pump/signal WDM at a RN and coupled into a length of EDF via a circulator. The generated ASE light is spectrally sliced through a conventional AWG and distributed to ONUs for upstream data transmission. On the other hand, the downstream signals are demultiplexed at the RN using a cyclic AWG and distributed to each subscriber. Note that the RN does not require any independent electrical power supply since the RN ASE generator is controlled by the Raman pump source located at the central office. The EDF located at the RN can be considered just as a passive medium for conversion of 14XX residual Raman pump power into a broadband ASE source for upstream signal generation.

Each of the ONUs consists of a EAM, a photoreceiver, and a circulator. The distributed downstream signals from the cyclic AWG share a common path between the RN and the ONUs with the modulated upstream signals owing to periodic wavelength response of the cyclic AWG. The multiplexed upstream signals at the RN are transmitted over the transmission fiber and demultiplexed through the cyclic AWG at the central office.

3. Experimental demonstration

3.1. 25 km upstream transmission

 

Fig. 2. Experimental setup for the demonstration of upstream data transmission using the proposed architecture.

Download Full Size | PDF

In order to prove the concept of the proposed architecture we constructed an experimental setup as shown in Fig. 2. Using this setup we performed only upstream transmission experiments of 5 WDM channels. The forward Raman pumping scheme, in which pump laser diodes (LDs) were located at the central office was adopted for the purpose of fully-centralized light source control. The Raman pump consisted of two laser diodes operating at 1455 and 1465 nm, respectively and the total pump power was ~350 mW. After Raman amplification over a 25 km SMF the residual pump power was measured to be ~99 mW. The residual Raman pump power was coupled into a 20 m long EDF with a peak absorption of 6 dB/m at 1530 nm via a 1460/1550 nm WDM coupler. We could not perform bidirectional transmission experiments with both downstream and upstream signals simultaneously in this demonstration since cyclic AWGs were not available in our lab at the moment. We believe that the pump-to-downstream signal crosstalk issue in the forward Raman pump scheme would be negligible due to low relative intensity noise less than -120 dB/Hz of our Raman pump used in this experiment [10].

Figure 3(a) shows the measured optical spectrum of the ASE generated from the 20 m EDF by recycling the residual Raman pump. The total ASE power was ~8 dBm. It is clearly evident from the graph and the power measurement that the residual Raman pump generated high quality of an ASE spectrum with a power level sufficient for its use as an upstream light source. The generated ASE was spectrally sliced to generate 5 upstream WDM signals at 1550, 1550.8, 1551.6, 1552.4, and 1553.2 nm through an AWG and the measured optical power of each signal was ~-7 dBm. Each upstream signal was modulated at 622 Mbit/s by use of a polarization-insensitive EAM and wavelength multiplexed by another AWG for upstream transmission. The data rate was limited to 622 Mbit/s due to excess intensity noise of spectrum-sliced source. The channel bandwidth of the AWG used in this experiment was 0.4 nm, which is not suitable for error-free 1.25 Gbit/s incoherent signal transmission [11, 12]. However, the data rate limitation can be sorted out by increasing the ratio of the optical to electrical bandwidth through use of AWGs with a channel bandwidth larger than 0.5 nm.

The signal power after the EAM was measured to be ~-13 dBm. After transmission over a 25 km long, distributed Raman amplified SMF, the upstream signals were fed on a 622 Mbit/s upstream receiver through an AWG. As an upstream signal receiver we employed a 10 Gbit/s p-i-n FET receiver attached with a ~470 MHz electrical filter.

Figure 3(b) shows the measured optical spectra of the upstream signals for both cases, i.e. before 25 km transmission and after 25 km transmission. Owing to distributed Raman amplification the transmission fiber loss was totally compensated and lossless transmission was thus achieved. The on-off Raman gain was ~6 dB, which is just enough for compensation of the 25 km SMF propagation loss (~5 dB). At such a small on-off Raman gain level Rayleigh scattering-induced multipath interference would not be a problem. The spectrally measured OSNRs of the signals after 25 km transmission were larger than 27 dB, which were high enough for error-free data detection at a data rate of 622 Mbit/s.

In order to quantify the proposed system performance we carried out the eye diagram and BER measurements. The results are summarized in Fig. 4. Error-free and penalty-free transmission was readily achieved for the upstream transmitted signals. No noticeable WDM channel crosstalk-induced penalty was observed due to the high SNR values. The small error floors observed in the graph are believed to be associated with nontrivial spontaneous-spontaneous beat noise that is caused by insufficient ratio of the optical to electrical bandwidth of a detected incoherent light signal [11, 12].

 

Fig. 3. Measured optical spectra of (a) an ASE generated from a 20 m EDF by recycling the residual Raman pump (~99 mW) and (b) 5 upstream signals.

Download Full Size | PDF

 

Fig. 4. Measured BERs and eye diagrams for 25 km upstream transmission.

Download Full Size | PDF

3.2. 50 km upstream transmission

Another potential benefit in our proposed architecture is that we can easily increase the signal transmission length to obtain wide area coverage without adding complicated system configurations owing to relaxed signal power margin. The increase of area coverage due to longer signal transmission distance leads to decrease of the number of central offices, which allows for significant reduction of both capital equipment costs and network operating expenditures for service providers.

In order to demonstrate the benefit we performed an additional experiment of 50 km upstream transmission of the 5 WDM channels. The Raman pump power had to be increased to ~650 mW to provide Raman gain high enough for compensation of signal propagation loss over a 50 km SMF. After Raman amplification the residual pump power was measured to be ~30 mW. The total ASE power from the EDF pumped by the residual Raman pump was 3 dBm. The sliced ASE channels through an AWG were modulated at 622 Mbit/s and wavelength multiplexed by use of another AWG. The signal power after the EAM was measured to be ~-19 dBm while it was ~-13 dBm before the EAM. The modulated single channel power of ~-19 dBm might be too low to be transmitted over a 50 km SMF without signal amplification and be detected by a low cost p-i-n FET receiver when considering the multiplexing AWG loss at the RN (~4 dB), the fiber propagation loss (~10 dB), and the demultiplexing AWG insertion loss (~4 dB) at the central office.

Figure 5(a) shows the measured optical spectra of the upstream signals for both cases, i.e. before and after 50 km transmission. Owing to distributed Raman amplification the transmission fiber loss was totally compensated and even a ~5 dB net signal gain was achieved. The spectrally measured OSNRs of the signals after 50 km transmission were ~18 dB, which were high enough for error-free data detection at a data rate of 622 Mbit/s. The measured BER curves are shown in Fig. 5(b). Error-free transmission was obtained with no penalty for all of the five channels.

 

Fig. 5. (a). Measured optical spectra of 5 upstream signals. (b) Measured BERs. These measurements were performed for the case of 50 km upstream transmission.

Download Full Size | PDF

4. Conclusion

We have proposed and experimentally demonstrated a high performance, distributed Raman amplification-based PON architecture that incorporates polarization-insensitive EAMs and ASE spectrum-sliced light sources. Using the architecture, we readily achieved not only the relaxation of the tight signal power margin through distributed Raman amplification over a transmission link, but also the realization of active light source-free subscriber ends.

The required Raman pump power should be increased for accommodation of more users or/and longer coverage distance. This would be a possible drawback of our architecture. There must be an optimum configuration of our proposed architecture in terms of Raman pump power, coverage distance, and the number of users. Our next step is to find the optimum configuration to make the proposed WDM PON architecture more practical and cost-effective.

As a matter of fact, at a data rate of 622 Mbit/s the use of directly-modulated FP-LDs would be cheaper than our proposed scheme due to relatively high cost of EAMs [13]. Note that the experimental demonstration in this paper was restricted to 622 Mbit/s simply due to the limited channel bandwidth of the AWGs available in our laboratory. Surely, our architecture should be able to work at 1.25 Gbit/s or beyond when AWGs with a channel bandwidth larger than 0.5 nm are adopted to mitigate the spontaneous-spontaneous beat noise. We believe that our proposed architecture should be a cost-effective solution for future high capacity access networks, comparable to or better than the FP-LD based one at 1.25 Gbit/s or beyond.