By Topic

• Abstract

SECTION I

## INTRODUCTION

Next Generation Passive Optical Networks (NG-PONs) focus on two major objectives. One objective is to increase the delivered bit rate to each subscriber. The second objective is to extend the reach to each customer in order to reduce deployment costs [1]– [2][3]. By employing wavelength division multiplexing (WDM) technology in future PONs, the guaranteed bit rate can be increased for each user at the cost of higher optical losses introduced by additional optical multiplexers/demultiplexers (mux/demux). Additionally, this technology offers point-to-point connectivity with dedicated bandwidth from central offices (COs) to optical network units (ONUs), higher security, and better quality of service [2].

Time division multiplexing (TDM) can be incorporated together with WDM technology, which is referred as hybrid WDM/TDM technique, to reduce the cost and share the increased bandwidth among subscribers. The cost is shared among all subscribers using TDM splitters for each optical channel. Fig. 1 depicts a hybrid WDM/TDM architecture used in later sections of this paper. This basic hybrid network architecture consists of an optical line terminal (OLT), a feeder line, a bidirectional optical amplifier, remote node (RN), access lines including TDM trees, and ONUs. Different configurations can be used at receiver/transmitter of ONUs such as reflective semiconductor optical amplifiers (RSOA) depending on the employed technology. With hybrid WDM/TDM, significant optical losses introduced by splitters, mux/demux, and circulator which reduce the optical power in the network. The optical amplification can be inserted in hybrid WDM/TDM NG-PONs at RN (as in Fig. 1) [4]. Among amplification techniques, SOAs are good candidates for amplification, as they offer moderate to high optical gain, they are inexpensive, integrable with photonic devices, and capable of intensity modulation with high data rates [5]. As the energy-efficiency is becoming an issue for NG-PONs, a proper amplification technique is required for this reason as well. Different optically pumped amplification approaches are used such as erbium doped fiber amplifiers (EDFA) [6] and Raman amplifiers [7]. However, electrically pumped amplifiers are highly desirable as a cost-effective solution.

Fig. 1. Hybrid WDM/TDM PON architecture.

This paper is based on our previous work [8], [9] on efficient SOA-based amplification techniques in NG-PONs. We demonstrate experimentally significant improvement of this technique in terms of reach, and the number of subscribers. The proposed technique is emulated in upstream/downstream (US/DS) scenarios for multi-gigabit NG-PONs.

The proposed extension scheme is not only limited to the hybrid WDM/TDM architecture in Fig. 1, but also different configurations such as hybrid TWDM NG-PON2 [10] can utilize this technique. However, the focus of the paper is to show the significance of the proposed amplification configuration considering the coexistence of the current and the future PON technologies. The paper is organized as follows: In the next section, the network configuration is described in detail for US/DS. Section 3 presents US/DS experimental results with respect to differential (quadrature) phase shift keying (D(Q)PSK) signals. Finally, Section 4 concludes the paper.

SECTION II

## ARCHITECTURE OF THE CONSIDERED NG-PON

Fig. 2 shows the network architecture with the proposed amplification technique. US and DS are considered separately throughout the paper which run with continuous wave lasers. The TDM technology can be easily introduced in the architecture without any drawback on the amplification concept. The architecture can support different FTTx (Fiber to the x) applications where x can be home (H), curb (C), building (B), and cell (C). The extension technique proposed in this paper, offers cost-efficient solution, that is a small modification in the remote node (RN) or local exchange, thereby eliminating the need of expensive components, and in this way, the cost among is shared among the larger number of subscribers.

Fig. 2. Architecture of the emulated PON: a) downstream (DS), b) upstream (US) setup configuration.

As shown in Fig. 2, the NG-PON extender is composed of a delay line interferometer (DLI), two circulators, and a high-gain SOA. The setup is designed for D(Q)PSK signals as they have higher tolerance against nonlinearities of the system compared to on-off keying (OOK) signals [11]. The signals are first converted from DPSK to OOK using DLI.

The input signal to the NG-PON extender enters the DLI for phase to intensity conversion. The DLI free spectral range (FSR) matches the channel bit rate and it enables the phase-to-intensity conversion for any number of channels at the same time. The output signals from the DLI are logically inverted. These two signals co-propagate and enter at the same time into the SOA. The high level signal enters from one side and the low level signal consisting of noise enters from the other side which lets the SOA saturate, and deplete the gain medium. The SOA should be bidirectional with same gain in both directions. As the SOA is in the saturation regime, the spaces experience higher gain than marks. As a result, the SOA works as a limiting amplifier and lowers fluctuations of the high level signals. We name our configuration the “saturated collision amplifier” (SCA), as it needs to operate in saturation region and signal “collision” occurs due to the counter-propagation. In this way, the extinction ratio of the signal is increased. Since DPSK signals have a constant optical power in each bit slot, the input power to the SOA is stable. As a consequence, the SCA configuration has only minimal bit-pattern effects [12].

### 2.1 Downstream Configuration

The experimental setup is for 12 × 10 Gbit/s return-to-zero (RZ)-DPSK Dense WDM/TDM-PON as shown in Fig. 2(a). The central office is emulated by a Pulse Pattern Generator (PPG) with Clock (Clk) at 10 GHz, the NRZ to RZ-DPSK converter, twelve distributed feedback (DFB) lasers operating at the following wavelengths: 1537.39 nm, 1538.9 nm, 1540.52 nm, 1542.93 nm, 1544.93 nm, 1550.11 nm, 1550.89 nm, 1551.72 nm, 1552.52 nm, 1553.32 nm, 1554.134 nm, 1556.55 nm; a dual drive electrical amplifier followed by a Mach–Zehnder modulator (MZM), two arrayed waveguide gratings (AWG), and an EDFA. $2^{31} - 1$ pseudo random binary sequence (PRBS) is generated by the PPG. With PCs we denoted the polarization controllers. The MZM is loaded from both sides to operate at push-pull in order to increase the extinction ratio. The wavelengths are demultiplexed and multiplexed again to de-synchronize the data patterns on each wavelength. To compensate the mux and demux losses, we use an EDFA after the mux. The wavelengths are at the ITU grid with 100 GHz spacing [13]. There is a variable optical attenuator shown as VOA1 in the feeder line to emulate the losses caused by the fiber from OLT to RN. The NG-PON extender (see Fig. 2(a)) is placed in the RN. VOA2 emulates the losses on the access lines. Finally, ONU is emulated by an avalanche photo diode (APD) and a bit error ratio tester (BERT). The output power of each channel is set to be 10 dBm.

### 2.2 Upstream Configuration

The US experimental setup uses cost effective chirped managed lasers as ONUs [14]. As displayed in Fig. 2(b), the US transmitter consists of a 4 × 10-Gbit/s DPSK directly modulated CML scheme. The driving electrical signal is encoded in inverse return-to-zero (IRZ) format before being sent to the CML. By direct modulation, a corresponding frequency shift is generated due to the adiabatic chirp of the laser. As the optical phase is a time integral of the instantaneous frequency, an IRZ pulse generates a phase shift of $\Delta\varphi = 2\pi\int^{T}_{0}\Delta fdt$, where $\Delta f$ is the optical frequency deviation and T is the pulse duration. As a consequence, in order to obtain a phase shift of $\pi$ with a 50%-duty-cycle IRZ signal, a maximum frequency shift of about 10 GHz is required. The wavelengths are at 1535.82 nm, 1536.61 nm, 1537.40 nm, and 1538.19 nm, respectively.

SECTION III

## EXPERIMENTAL INVESTIGATIONS

The measurements are performed by having the optical power budget in mind. The feeder budget is defined as the optical power loss from CO to the RN. The access budget is the loss between RN and ONUs, their sum yields the total optical power budget. In case of DS, the VOA1, which is the attenuator of the feeder line, is incremented in steps of 2 dB and a corresponding BER curve is measured with respect to receiver input power that is controlled by VOA2. For US case, the measurement is performed in the opposite way. The results are presented individually in the following subsections.

### 3.1 Downstream

Fig. 3 represents the back-to-back measurement results of the receiver sensitivity versus $\log_{10}$ BER at the total bit rate of 10 Gbit/s for each channel with no NG-PON extender. We assume here forward error correction (FEC) according to the standard [15], so that our target BER is $10^{-3}$. As an example, at 1556.55 nm the receiver sensitivity is −31 dBm as shown Fig. 3. The optical power budget is calculated as the transmitter output power (0.7 dBm) minus the receiver sensitivity −31 dBm, that is 31.7 dB. The results in Fig. 3 vary at lower BERs owing to the ASE noise caused by pre-amplification before the feeder line. There is always a different gain when ASE-noise dominates with respect to each wavelength.

Fig. 3. BER versus receiver sensitivity of DS.

The total optical power budget is 31.7 dB, this budget serves up to 64 customers (18 dB loss), 20 km feeder fiber (4 dB loss), 10 km access fiber (2 dB), and 7 dB extra loss is predicted for circulators and demux of DS. The guaranteed bit rate for each customer is 15.625 Mbit/s and the peak rate is 10 Gbit/s. The total number of users covered without SCA amplification is $64 \times 12 = 768$ (64 users on each channel). In order to expand the access budget we place the power budget extender in the RN as in Fig. 2. The bit rate should match the SOA recovery time. The SOA used in the experiment has recovery time of 16 psec. The SOA should have enough time to recover the gain for the next bit. The SCA has very high potential to perform amplification at much higher bit rate than 10 Gbit/s because it can mitigate the influence of SOA gain dynamic effect as the intensity in the SOA is always kept constant.

By incrementing VOA1 in steps of 2 dB, the receiver sensitivity for each channel is measured. As a result we acquire a so-called isoBER curve displayed in Fig. 4. The bias current of SOA is 250 mA, it is optimized for the targeted BER. The input signal power to the SOA from either side is set to be the same. A variable optical delay line is used to adjust the delay between two arms of the DLI and, thus, synchronize the signal outputs. The isoBER curve explains how much loss can be compensated by using the NG-PON extender. In Fig. 4, the total optical power budget (red dot on the curve) is selected to be 53.4 dB, that is 20 dB in feeder budget and 33.4 dB in access budget. We can always sacrifice the feeder budget as we are interested to extend our access budget. 33.4 dB access budget can serve up to 256 customers (one output of NG-PON extender), and 7.6 dB extra loss compensation which can be predicted for circulators, and access line fibers. In total we can support 512 customers. The feeder budget is limited by the saturation power of the SOA and the access budget is by ASE-noise and receiver sensitivity. We already showed [9] that our SCA configuration can enhance the access budget by far in comparison to a conventional SOA-based configuration.

Fig. 4. isoBER curve of power budget extender at 1556.55 nm.

We also evaluate our scheme by inserting fibers in feeder and access line. The transmission fibers used in the experiment is single-mode fiber (SMF). We considered 25.26 km feeder fiber and 10.56 km access fiber, Fig. 5 illustrates isoBER curves for back-to-back configuration and with fiber based system. The performance degradation is mainly due to the fiber loss at the access line. However, the optical power budget at the given example is 47.5 dB including 35.82 km fiber, which is still compatible with XG-PON1 class N1 [15]. The access budget would still allow 28 dB losses excluding fiber loss (2 dB). Consequently, $12 \times 2 \times 256 = 6144$ can still be served plus 4 dB extra loss compensation.

Fig. 5. isoBER curves of DS for two cases.

Three regions are marked in Fig. 4. Each region represents a different type of limitation. The operating area is limited by SOA gain, ASE-noise, and receiver sensitivity.

### 3.2 Upstream

Fig. 6(a) shows the bit error ratio measurements without the NG-PON extender. Back-to-back transmission is considered during the whole US measurements. If we assume FEC limit at BER $10^{-3}$, the receiver sensitivity is −32 dBm. If the transmitter outputs 7 dBm at 1538 nm, the total optical power budget is 39 dB. Now, the extender is used in the setup and the measurements are repeated. All the neighboring channels emit relatively at high power levels leading to nonlinearities in the SOA. The bias current is 300 mA for the SOA. We can see from Fig. 6(b) that the extender can boost up the power budget to 60 dB (30 dB access budget and 30 dB feeder budget). This access power budget allows for 10 km fiber (2 dB), 1:256 splitting ratio (24 dB) and 4dB Mux. Considering output of both circulators and four channels, 2,048 clients can be supported. In order to evaluate the system performance in terms of nonlinearities caused by the amplification, we compared a conventional configuration (only a SOA not SCA) as a NG-PON extender and our SCA setup. Fig. 6(c) depicts the BER versus receiver input power when only the conventional configuration is integrated in the RN (before the DLI). Large power penalty is observed when the neighboring channels have high input power levels. However, SCA can mitigate the nonlinear effects (mainly XPM) in the transmission. In Fig. 6(d) the upper curve (triangles) shows BER when conventional SOA scheme used and the lower one (circles) corresponds to SCA. A significant improvement is obvious when using SCA as a NG-PON extender. It reveals that, the extender is independent of power of the neighboring channels, it does not deteriorate the signals on the channels with lower input power levels, as it is the case in a DWDM transmission. Let it be mentioned that the conventional SOA configuration increases the power budget also but only to some extend if the neighboring channels have low input power levels at the input of the SOA. Thus, the SCA scheme that is applied here outperforms the conventional amplification configuration.

Fig. 6. a) back-to-back receiver sensitivity, b) IsoBER curve for US transmission, c)single SOA configuration with different power on neighboring channels, d) SCA and SOA comparison.

### 3.3 Single and Multi-Wavelength DWDM Performance of NG-PON Extender

In addition to single-channel operation, we also investigated the multi-channel (WDM) performance of the proposed NG-PON extender in DS direction. Measurements are carried out for 1, 2, 4, 8, and 12 channels, respectively. Back-to-back BER measurement are the results of the measurements in Fig. 7. The number followed by $\lambda$ denotes the number of wavelength channels and the corresponding isoBER curves belong to 1556.55nm. In other words, we turn off all channels except 1556.55 nm for the $1\lambda$ curve, and then turn on the neighboring channels to measure the rest of the isoBERs for 1556.56 nm. Fig. 7 indicates that the access budget reduces as we increase the number of channels from 1 to 12. The access budget varies at least 5.2 dB as we go from 1 $\lambda$ to 12 $\lambda$ s. The wavelength gain competition and ASE noise contribution of the SOA is the main reason for this behavior. Accordingly, increasing the number of wavelengths does not necessarily increase the number of clients but the data rate per client can be increased by reducing the splitting ratio. However, the setup gives the opportunity of converting DPSK to OOK with only DLI for multiple channels, which is a cost effective solution in terms of the cost at RNs.

Fig. 7. isoBER curves SCA DPSK for different number of channels, all curves belong to 1556.55 nm.

### 3.4 DQPSK Modulated Signals Utilizing the NG-PON Extender

The DQPSK configuration can be used in either US/DS scenario depending on NG-PON technology requirements. Here the DS scenario is demonstrated. Fig. 8 shows the DQPSK measurement setup at 1550 nm. It consists of two PPGs $(2^{7} - 1)$ for in-phase (I) and quadrature (Q), respectively, at 10 Gbit/s each. Different data streams are uploaded on each PPG. A single drive I/Q modulator is used to generate DQPSK signal which is followed by a dual drive amplifier. The VOA1 emulates the losses of feeder trunk. The RN is replaced by DQPSK optical demodulator and the amplification scheme is the same as the previous section except that we have four outputs from the RN, $I_{1}$, $I_{2}$, $Q_{1}$, and $Q_{2}$ as well as two SOAs. The ONU is in analogy to Fig. 2. In the following measurements we only consider $I_{1}$, as all the others have the same performance. We first show our back-to-back results to see how much access losses can be tolerated in the absence of the NG-PON extender. Obviously, from Fig. 9, for a transmitter output power of 3 dBm the total optical power budget at BER of $10^{-3}$ (assuming FEC) is 29.8 dB (dashed line in the Fig. 9). 64 (18 dB loss) users can be covered if we assume 10 km access fiber, 10 km feeder budget and the rest for mux/demux of the RN. However, in this case we have twice the number of outputs than DPSK case, thus the number of users is calculated to be $4 \times 64 = 256$. We benefit from higher bit-rate, higher bandwidth efficiency (because of the narrower spectrum of DQPSK), and more covered customers. Fig. 10 shows isoBER curves for two different configurations. The line with circle markers displays the SCA scheme which has remarkably 4 dB difference in comparison to ordinary SOA amplification scheme in RN (triangle line). 4 dB gain in access budget doubles the number of users. We consider the maximum achievable access budget on the isoBER curves. Therefore, the loss in feeder truck can be considered lower in order to achieve more access budget.

Fig. 8. DQPSK measurement setup for one channel at 20 Gbit/s.
Fig. 9. DQPSK back-to-back results with no SCA in the RN for $I_{1}$.
Fig. 10. back-to-back isoBER curves, DQPSK SCA versus SOA for $I_{1}$.

According to SCA isoBER curve, the total optical power budget is 40 dB (6 dB feeder budget and 34 dB access budget). 30 km feeder fiber, 256 users, 10 km access fiber, and 8 dB extra losses for mux/demux in US/DS directions. When considering four outputs of the RN, we can serve $4 \times 256 = 1024$ users on one channel, thanks to the SCA configuration.

SECTION IV

## CONCLUSION

Significant improvements in terms of power budget (i.e. reach) of a previously introduced NG-PON extender scheme [9] have been presented. An optical power budget of 47.5 dB has been demonstrated experimentally to support an optical access line of 36 km of SMF and to serve 6,144 subscribers in a downstream scenario. This is the highest served number of subscribers to the authors' best knowledge with this extension configuration. The demonstrated scheme is compatible with DPSK and DQPSK modulated signals. Thereby, flexible modulations formats can be used for the data transmission for increasing the total optical power budget as well as the bit rate. A cost-effective US transmitter architecture has been presented which can equally be used in central offices or for DS transmission. The achieved US optical power budget have been shown to be 60 dB for 40 Gbit/s directly-modulated DPSK transmission, 2,048 subscribers can be served in this case. The US/DS can be combined together for bidirectional transmission according to NG-PONs technology and cost requirements. The nonlinearities are negligible, dispersion caused penalty should be included in the margin. One can reduce the number of customers per wavelength to have more margin, at the same time data rate per user increases. The modulation flexibility of the extender has been adapted to DQPSK signal for a single channel DS scenario that can be modified to multi-channel transmission as well. $4 \times 256 = 1024$ users on one channel can be supported with more bandwidth efficiency in this case. It has been shown that the SCA scheme allows us to decrease the nonlinear effects caused by neighboring channels in the link. The proposed extender can be considered as a cost-effective, flexible, and easily-deployable scheme for NG-PON.

## Footnotes

No Data Available

## References

No Data Available

## Cited By

No Data Available

None

## Multimedia

No Data Available
This paper appears in:
No Data Available
Issue Date:
No Data Available
On page(s):
No Data Available
ISSN:
None
INSPEC Accession Number:
None
Digital Object Identifier:
None
Date of Current Version:
No Data Available
Date of Original Publication:
No Data Available