1. Introduction

As demand for greater bandwidth to the home continues to grow, research interest in new optical access network architectures to satisfy the bandwidth demands also continues to evolve. The bandwidth available to the end-user has generally increased with successive TDM-PON architectures starting with broadband PON (BPON) with downstream transmission at 622 Mb/s, Ethernet PON (EPON) with 1.25 Gb/s downstream, and now Gigabit PON (GPON) with 2.5 Gb/s downstream signals. The need for even higher speed access connections will push PON networks to transmit at 10 Gb/s on the downstream path, and there has already been significant research in this area [1, 2].

In addition to higher bit rates, another significant trend in PON research has been movement toward longer reach PONs. As PON systems grow in length, the boundaries between metropolitan and access networks will begin to blur. A key motivation for this vision of combined metro/access networks is the potential consolidation of central offices, thereby saving considerable cost in operating expenses [2]. Long reach PONs were preceded by high split ratio TDM-PONs [3] which were typically all passive for the short transmission distances. Early references to long-reach PONs or "super-PONs" ranged over a variety of topologies [4, 5], but most have required some form of active amplification in the fiber plant to overcome the larger losses associated with longer fiber transmission distances. Still, these “super-PONs” were able to achieve high split ratios at moderate bit rates. Recently, high split ratio and long reach were combined with the demonstration of a 1024-split access network over 110 km [6], but this architecture employed a bidirectional amplification stage in the fiber plant and the use of FEC detection. Long-reach WDM-PON system demonstrations have also generally employed remote node amplification to attain reach lengths on the order of 100 km [7, 8]. Recently, we proposed a purely passive long reach 10 Gb/s TDM-PON access system based on a low attenuation fiber and the duobinary modulation format [9]. Here, we expand upon, and extend, the results in [9] demonstrating the feasibility of high split ratio PON systems with a purely passive fiber plant over a 100 km distance. The system is advantaged by use of an ultra-low-loss G.652-compliant fiber with up to 3 dB lower loss over the 100 km system length than standard single-mode fiber, thus enabling a split over twice as many subscribers. We show that the duobinary format has the necessary characteristics of high tolerance to dispersion, SBS, and SPM for this application, allowing the high launch powers necessary while eliminating the need for dispersion compensation. We also find that a chirp-managed laser [10], an alternative transmitter technology with similar spectral characteristics and transmission performance to duobinary, can be used successfully for this application as the downstream signal. Finally, we demonstrate an evolution of the basic long-reach TDM-PON architecture with a WDM/TDM-PON system that serves more subscribers by splitting each WDM wavelength over up to 128 users, expanding on results shown in [11].

2. Characterization and TDM-PON system experiments and results

We begin by characterizing 10 Gb/s duobinary signals in terms of the properties that make this format especially well-suited to this application of a long-reach PON system. An important nonlinear impairment is stimulated Brillouin scattering (SBS), which is generally an issue for high-launch power systems. SBS can be a limiting factor which may be addressed for NRZ systems with high-SBS threshold fiber [12] and/or frequency dithering, both of which serve to increase the SBS threshold. However, duobinary is a carrier-less format that exhibits very high tolerance to SBS to the extent that it can be ignored for the launch power levels necessary in a long PON span considered here. Measurement results of the reflected power from Rayleigh scattering and SBS in a typical set-up with a circulator (not shown) for both duobinary and NRZ are shown in Fig. 1(a). Duobinary shows no discernible SBS component in the reflected power up to a launch power of 18 dBm for a 100 km span of standard single-mode fiber. Our measurements have shown no significant difference in the SBS threshold behavior of the ultra-low-loss fiber and standard single-mode fiber.

 

Fig. 1. (a) Reflected power for duobinary and undithered NRZ signals from an SBS measurement set-up with 100 km of standard single-mode fiber. (b) Required OSNR to achieve a BER value of 10^-3 with NRZ and duobinary signals over 100 km span of ultra-low-loss fiber. The wavelength for all data was 1550 nm, modulated at 10.3125 Gb/s.

Download Full Size | PDF

Another potentially limiting impairment for a long-reach PON requiring high launch power is self-phase modulation (SPM). We compared the SPM thresholds for duobinary and NRZ signals for a 100 km fiber span by making required OSNR measurements as a function of launch power while keeping the received power fixed. In these experiments ASE noise was added to the signal under test after propagating through the fiber and the OSNR required to achieve a fixed BER of 1×10^-3 was determined. The signal and noise were filtered with a 0.25 nm optical filter before detection. Experimental results are given in Fig 1(b) for 10 Gb/s NRZ (undithered and dithered) and duobinary signals over the 100 km span of ultra-low-loss G.652-compliant fiber.

A first point to note from the data of Fig. 1(b) is the 4-5 dB advantage enjoyed by duobinary over NRZ in the linear regime with low channel launch powers. This is due to the enhanced dispersion tolerance of duobinary [13, 14], a significant factor for a 100 km span and key advantage of duobinary for this application. Next, we observe for both formats that the required OSNR decreases as the input power increases as the eye opens due to the interaction of SPM and dispersion. Eventually, the fiber nonlinearities (SBS and SPM) begin to severely degrade the signal quality and result in an abrupt increase in the OSNR requirement. For duobinary, the limiting impairment is SPM, as we observed in Fig. 1a that SBS is suppressed by this format for this range of launch powers. For undithered NRZ, the dominant nonlinear impairment is SBS which limited the optimal launch power to about 13 dBm. When frequency dithering is applied to the laser, the SBS threshold for NRZ increases substantially such that the limiting nonlinearity becomes SPM. However, we note that the duobinary signal still possesses an approximately 2 dB greater SPM threshold over the 100 km span than dithered NRZ. This characteristic, along with its inherently greater dispersion tolerance, makes duobinary an attractive format for a long reach PON network. We performed similar experiments with the standard G.652-compliant fiber and found no significant differences in the nonlinear behavior of the two fibers.

The general experimental setup used for the TDM-PON downstream transmission experiments with duobinary and NRZ modulation formats is illustrated in Fig. 2. A distributed feedback (DFB) laser at 1550.12 nm was externally modulated at 10.3125 Gb/s (10 Gb/s Ethernet rate) with either an NRZ or duobinary signal using an Mach- Zehnder modulator (MZM). The pseudorandom bit sequence (PRBS) used was of length 2³¹-1. The modulated signal was then amplified and launched into a 100 km span of fiber. Two different fiber types were studied in these experiments; standard single-mode G.652-compliant fiber and an ultra-low-loss G.652-compliant single mode fiber (Corning^® SMF-28^® ULL fiber). The ultra-low-loss fiber had average attenuation of 0.17 dB/km while that of the standard fiber was 0.20 dB/km, so the total 100 km span loss including connectors for the two fiber types were 17 dB and 20 dB, respectively. The channel launch power into the span was controlled with a variable optical attenuator (VOA). Following propagation through the fiber, the signal was attenuated with another VOA adjusted to produce the equivalent loss of a 1:32, 1:64, or 1:128 passive optical splitter, with 17, 20, and 23 dB of loss, respectively. These splitter loss values represent a 2 dB increase over the nominal value in each case. The downstream signal was detected with an avalanche photodiode (APD) detector, which output electrical signals to a separate clock recovery unit and a bit error ratio tester (BERT) for measurement. The nominal sensitivity of the APD detector at BER = 1×10^-12 was -26.5 dBm. The set-up was slightly altered for the previously described SPM characterization of the modulation formats by noise-loading the signal after transmission through the span via a 3 dB coupler and measuring the required optical signal-to-noise ratio (OSNR) for a given BER value of 1×10^-3.

 

Fig. 2. Downstream TDM-PON experimental setup.

Download Full Size | PDF

Single channel TDM-PON downstream transmission experiments were carried out with both duobinary and NRZ 10.3125 Gb/s signals over the ultra-low-loss fiber and standard single-mode fiber with various split loss values to compare performance. Some of the experimental results obtained in terms of the measured BER as a function of launch power are shown for 1:128 split ratios in Fig. 3(a) and Fig. 3(b). These results demonstrate that duobinary and the ultra-low-loss fiber represent a significantly advantaged modulation format and optical fiber combination that showed the maximum split ratio and lowest launch power with error-free performance in our experiments. For example, in Fig. 3(a) we compare the performance of duobinary vs. NRZ for the 100 km span of ultra-low-loss fiber with a 1:128 split ratio. Duobinary shows error-free transmission while NRZ is limited by SPM, dispersion, and SBS to a minimum BER of about 1×10^-9 at 15 dBm launch power. Fig. 3(b) presents data for duobinary systems with 1:128 split ratios, comparing the performance over the two 100 km fiber spans. Error-free transmission is achievable over the ultra-low-loss fiber span, but the standard fiber span with 3 dB higher loss is not error-free because of SPM limitations at the higher launch power required. While not shown, we also observed that error-free NRZ downstream transmission over the ultra-low-loss fiber span with a split ratio up to 1:64 is feasible, but is possible over the span of higher attenuation standard fiber only for a 1:32 split or smaller.

 

Fig. 3. Downstream signal transmission results for (a) duobinary and NRZ with 1:128 split over ultra-low-loss fiber, (b) duobinary with 1:128 split over both fibers.

Download Full Size | PDF

A chirp-managed laser (CML) driven at 10.3125 Gb/s was also briefly evaluated as a potential downstream transmitter as an alternative to duobinary implemented with a MZM. This transmitter comprised of a directly-modulated laser with optical filter has been previously shown to have comparable characteristics to duobinary in terms of enhanced dispersion tolerance and nonlinear tolerance [10, 15]. The results for transmission of a 10.3125 Gb/s CML at 1553 nm over the 100 km ultra-low-loss fiber span with a split ratio of 1:128 (23 dB split loss) are shown in Fig. 4(a) along with the duobinary data. The CML also produced error-free transmission for this system for launch powers > 14 dBm with only a small (<1 dB) penalty relative to the duobinary transmitter, and thus represents a feasible alternative to the duobinary transmitter for this application.

 

Fig. 4. (a) Downstream signal transmission results for CML and duobinary transmitters over 100 km ultra-low-loss fiber TDM-PON with 1:128 split ratio. (b) Upstream transmission results at 1.25 Gb/s.

Download Full Size | PDF

Finally, upstream signal transmission experiments were also performed using a 1.25 Gb/s directly modulated distributed feedback (DFB) laser at 1553 nm over the ultra-low-loss fiber, assuming separate upstream and downstream fibers. The laser output power was 6 dBm and the upstream receiver that would be in the central office location was comprised of an EDFA pre-amplifier with 5.5 dB noise figure and a 0.25 nm bandwidth optical filter in front of a PIN photodetector. A wider filter could also be used in the receiver. The measured gain of the pre-amplifier at the low input power (-34 dBm) corresponding to a 1:128 split was about 34-35 dB. The total system loss was varied with the VOA on the customer side of the fiber span and the BER was measured. The upstream results are shown in Fig. 4(b). The upstream signal was error-free (< 1×10^-12) for a total PON loss of about 46 dB or less. Given that a PON over the ultra-low-loss fiber with a 1:128 split has a total loss of 40 dB, there is margin available to allow the use of a lower power upstream laser or to increase the upstream bit rate.

3. WDM/TDM-PON experiments and results

A WDM/TDM-PON system may be an alternative to the TDM-PON just discussed, allowing more subscribers to be served on the same fiber plant. In such a system, each individual wavelength in a WDM architecture is split among N users just as in the simple TDM-PON system. This allows the total number of subscribers to be increased by the number of wavelengths. The 8-channel experimental system configuration considered here and used for downstream transmission is illustrated in Fig. 5. Seven of eight WDM channels spaced by either 100 GHz or 200 GHz (minus the 5th channel) were modulated at 11.1 Gb/s with a Mach-Zehnder modulator (MZM1) in the duobinary format, driven with a PRBS of length 2³¹-1. The fifth channel in the wavelength plan at 1552.52 nm was modulated separately at the same bit rate and pattern length by a different modulator MZM2 and then combined with the other 7 channels, which were first de-correlated from each other and the test channel by a 10 km length of standard single-mode fiber. The bit rate of 11.1 Gb/s represents 10.3125 Gb/s Ethernet plus 7% FEC overhead.

 

Fig. 5. Experimental set-up used for WDM/TDM-PON downstream transmission.

Download Full Size | PDF

After combination, the 8 channels were amplified in an EDFA with an output power of 30 dBm. A VOA controlled the total launch power into the 100 km span of the fiber under test. The same two fiber spans were tested: 100 km of the ultra-low-loss fiber, and 100 km of standard single-mode fiber. After transmission through either fiber span, channel #5 was selected for testing with a grating filter that simulated the demultiplexer shown in Fig. 5. The filter/demultiplexer had a loss of 5 dB. A VOA was adjusted to provide the appropriate amount of attenuation corresponding to various split ratios. As before, the signal was detected with an APD detector, amplified, and passed to a bit error rate tester (BERT) for measurement along with the recovered clock signal.

BER measurements were made over each 100 km fiber span, with split ratios for each wavelength up to 1:128, and with 100 GHz and 200 GHz channel spacing. The results for 1:64 and 1:128 wavelength split ratios are shown in Fig. 6 as Q values for channel #5 as a function of the launch power per channel.

 

Fig. 6. Q values of downstream channel #5 as a function of launch power/channel. (a) 1:64 split per wavelength, (b) 1:128 split per wavelength.

Download Full Size | PDF

The optimal launch power per channel in all configurations is limited by nonlinear impairments, namely cross-phase modulation (XPM) and self-phase modulation (SPM). Consider first the 1:64 wavelength split ratio data in Fig. 6(a). The maximum Q at optimal launch power is up to about 2 dB higher with the ultra-low-loss fiber than with the standard fiber for a given channel spacing of either 100 or 200 GHz, producing significantly better performance for the same number of subscribers. Both channel spacing values produce optimal performance well above the standard FEC threshold with the ultra-low-loss fiber, while only the 200 GHz spacing results exceed this threshold for the standard fiber. Furthermore, we note that the 100 GHz system with ultra-low-loss fiber performs slightly better than the 200 GHz system over the standard fiber for this split ratio and requires approximately 2 dB lower launch power at maximum Q. This confirms that the ultra-low-loss fiber based WDM/TDM-PON system can serve twice as many subscribers as a standard fiber system, either by allowing a 2x increase in split ratio or by allowing twice as many wavelengths to be transmitted in the same optical bandwidth.

The results obtained for a 1:128 split ratio per wavelength are shown in Fig. 6(b). The ultra-low-loss fiber systems with either channel spacing can use either standard or enhanced FEC and deliver error-free operation, while the standard fiber system requires the more powerful enhanced FEC even with 200 GHz channel spacing. Transmission with 100 GHz spacing was not performed over the standard fiber span because it could not meet even the enhanced FEC threshold. We also observe that the differences in optimal performance between the two fiber spans are even greater with the 1:128 wavelength split ratio. For 200 GHz channel spacing, the maximum Q value obtained with the ultra-low-loss fiber was more than 3 dB greater than with the standard fiber. Similarly, even the 100 GHz spacing system over the lower attenuation fiber performed significantly better than the 200 GHz system over the standard fiber by about 1.7 dB. So in comparison, the ultra-low-loss fiber system can serve twice as many subscribers in the same optical bandwidth with significantly better overall transmission performance.

Upstream transmission was also performed over the 100 km ultra-low-loss fiber span with a 6 dBm directly modulated laser (DML) at 1552.52 nm modulated at 1.25 Gb/s. and 2.5 Gb/s. The receiver was the pre-amplified PIN photodetector described earlier. Results in terms of BER as a function of optical power into the photodetector are shown in Fig. 7(a). Error-free transmission was achieved for the 1.25 Gb/s upstream signal with up to a 1:128 split ratio, and for the 2.5 Gb/s upstream signal up to a 1:64 split ratio. These configurations correspond to upstream TDM bit rates of about 9.7 Mb/s and 39 Mb/s. The OSNR values of the received signals were 16.8 dB for a 1:64 split ratio and 13.8 dB for a 1:128 split.

Finally, we note that a slightly different WDM/TDM-PON architecture may be desirable in some circumstances in which we seek to avoid cross-channel nonlinear impairments and the use of FEC. In this case, the system in Fig. 5 may be changed by reversing the order of the wavelength demultiplexer and the 100 km fiber spans. That is, all wavelengths are amplified together and then immediately demultiplexed in the central office, and then transmitted as single channels over separate fiber spans. This is essentially the same as a pure TDM-PON system, but using a common amplifier for many individual systems, each of which would use a different downstream wavelength. Transmission results showing the received downstream BER at 1552.52 nm as a function of total EDFA output power are given in Fig. 7(b) for an 8-channel system with 1:128 TDM split ratio, demonstrating that an EDFA output power ≥ 29 dBm yields error-free downstream transmission without the use of FEC.

 

Fig. 7. (a) WDM/TDM-PON system upstream results for a DML modulated at 1.25 Gb/s and 2.5 Gb/s. (b) Downstream transmission BER vs. EDFA output power results for 1552.52 nm channel for alternative WDM/TDM-PON system architecture.

Download Full Size | PDF

4. Summary and conclusions

We have demonstrated purely passive long reach (100 km) 10 Gb/s TDM-PON and WDM/TDM-PON systems enabled by the use of ultra-low-loss optical fiber with attenuation 0.17 dB/km and downstream duobinary signal transmission. No in-field optical amplification is required. In both cases, a TDM split ratio up to 1:128 is supported, providing a downstream bandwidth per subscriber of almost 80 Mb/s. Downstream transmission using chirp-managed laser technology was also shown to be feasible for the application, demonstrated in the TDM-PON system. Upstream bit rates demonstrated were 1.25 Gb/s and 2.5 Gb/s using a directly modulated laser at the customer site. For the TDM-PON configuration, no FEC is required. For the WDM/TDM-PON configuration, standard FEC may be used for downstream signals with a channel spacing of either 100 GHz or 200 GHz. For the alternative WDM/TDM-PON configuration proposed, no FEC is required as each wavelength is transmitted over a separate fiber, eliminating cross-channel nonlinear effects, but using a high-power amplifier to amplify all channels together before transmission.