By Topic

IEEE Quick Preview
  • Abstract
The proposed stacked WDM-OFDM-PON system architecture, (a) the OFDM signals?? time waveform, (b) the frequency domain waveform of OFDM signals, (c) the spectra of the upstream channels.



With the popularity of broadband services of the terminal users, such as Internet video, online gaming, high-quality Internet protocol TV (IPTV), etc, the demands of bandwidth in access networks is rapidly increasing, which drives the development of next generation passive optical network (NG-PON) [1]– [2][3]. To address this issue, great effort has been expended on researching the high-speed, cost-effective, flexibility bandwidth allocation and “future-proof” NG-PON system. Among these techniques, WDM-based stacked PON systems have attracted a great deal of research and development interest [4]– [5] [6] [7] [8] [9][10], due to their capable of providing cost-effective way for increasing the overall bit rate and transmission reach of networks. Furthermore, the orthogonal frequency division multiplexing (OFDM) is widely considered as one of the strongest candidate for stacked WDM-based PON system, owing to its unique advantages of superior tolerance to chromatic dispersion impairments, dynamic provision of multi-granularity bandwidth allocation both in time and frequency domains, and fully exploiting the rapid advances in modern digital signal processing technology [8]– [9] [10] [11] [12] [13][14].

In fact, to cost-effectively implement stacked WDM-OFDM-PON system while maintaining good system performance, the challenges as following should be considered [7]– [8] [9][10], [14]: (a) high speed low-cost colorless transmitters deploying in optical network unit (ONU) for uplink, (b) tunable optical filter (TOF) in ONU used for selecting wavelength, (c) high optical power budget for supporting the increased power loss caused by large split ratios, (d) long reach between the optical line terminal (OLT) and ONU for benefits in network operation. In recent years, several proposals [7], [8], [10], [14]– [15] [16][17] have been demonstrated to achieve one or several factors of mentioned above. In colorless upstream transmitter's schemes such as those proposed in [15]– [16][17] and [18], [19], the reflective semiconductor optical amplifier (RSOA) or Fabry–Perot laser diode (FP-LD) based on the wavelength reuse and self-seeding was used as tunable transmitters, whereby the reflective approaches lack the transmission capability. As for high power budget and long reach transmission, in [20], the authors have presented deploying the erbium-doped fiber amplifier (EDFA) as relay amplifier, which extended the system reach but changed the passive feature of PON system. In [21], [22], the authors have adopted a coherent receiver in OLT to achieve high sensitivity therefore improving power budget. Nonetheless, the coherent technique of these schemes used for stacked WDM-based PON, multi-coherent transceivers are required, which increases the system complexity and cost.

In our previous work [9], a low-cost, symmetric 40-Gb/s stacked WDM-OFDM-PON architecture based on directly modulated DFB with completely compatible optical distribution network (ODN) has been demonstrated and achieved 30.4 dB power budget for supporting 1 : 256 split ratios with 25 km distance. A specially designed electronic controlled liquid crystal TOF was deployed in ONU for both downstream wavelength selection and upstream signals chirp management, and an EDFA was used as pre-amplifier to improve downstream sensitivity. Note, however, that there are views that the deploying of EDFA in each ONU is unacceptable for subscribers due to its high cost. Moreover, the properties of TOF are very important, such as the order of filter and the steep induced the adjacent-channel crosstalk, which decides chirp management effect and downstream performance. Therefore, the properties of TOF should be investigated so as to improve the whole system performance.

In this paper, we further optimize and experimentally demonstrate this stacked WDM-OFDM-PON system to further extend transmission distance and improve receiver sensitivity thus achieving high power budget to support more terminal users. By using a specially designed linearity avalanche photo diode (APD) to replace the receiver consisting of EDFA and PIN in ONU of our proposed architecture, the system cost is reduced while increasing the downstream receiver sensitivity. Meanwhile, the common APD is also used to replace the PIN of our previous scheme in OLT, which improves system power budget. Moreover, the detailed investigations by experiment and simulation are undertaken, for the first time, to explore the impact of properties of TOF on selecting downstream wavelength and managing upstream signals' chirp in terms of receiver sensitivities. Results show that, the non-steep profiles of TOF significantly affect the adjacent-channel crosstalk for downlink and chirp management for uplink, therefore reducing achievable receiver sensitivity. Experimental results also indicate that system performance is almost the same as single wavelength system when the adjacent-channel power ratio is more than 15 dB.



Fig. 1 depicts the proposed stacked WDM-OFDM-PON architecture, which comprises OLT, a single mode fiber (SMF) link without any relay amplifier, and ONUs. At the OLT, four transmitter and receiver (TX/RX) units are multiplexed by an arrayed waveguide grating (AWG) to achieve 40-Gb/s aggregate rates. In each TX/RX, a continuous wave light is injected into the Mach–Zehnder modulator (MZM), which is driven by the electronic OFDM signals. Here, OFDM signals are generated offline following the procedure presented in [23], which mainly include data mapping, inverse fast Fourier transform (IFFT), cyclic prefix inserting, OFDM symbol serial-parallel and digital-to-analog conversion. Through AWG multiplexing, the signals are then amplified by a bi-directional optical amplifier (OA). This OA deployed in the OLT is used to compensate the high optical loss of bi-directional signals. After the SMF, the downstream signals are split by a 1 × N splitter (N denotes the numbers of ONU) at remote note (RN), and then routed to each ONU over distribution fibers.

Figure 1
Fig. 1. Proposed stacked WDM-OFDM-PON system architecture. (a) OFDM signals’ time waveform. (b) Frequency domain waveform of OFDM signals. (c) Spectra of the upstream channels.

At the ONU side, downstream signals pass through a TOF which is simultaneously used for downstream wavelength selection and upstream signal chirp management. This takes full advantage of the optical device of network thus reducing system cost. In addition, the specially designed linear APD, as shown in the inset (i) of Fig. 1, is used to detect the downstream OFDM signals. Note that, different from the common APD used in digital communication system, not limiting electronic amplifier but linear electronic amplifier is integrated in this APD, which increases the linear range of APD thus improving OFDM signals receiver sensitivity. Besides, compared to our previous scheme [9], only an APD is deployed in each ONU thus reducing the ONU complexity and users’ cost. As for uplink, the thermally tuned DFB laser as transmitter is driven by the 10-Gb/s data to generate upstream signal. This DFB has thermally tunable wavelength range of 3.0 nm with stable output power, so each ONU is able to tune any channel with 0.8 nm. After the TOF, upstream signal pass through the power coupling and fiber transmission, and is pre-amplified by OA and demultiplexed by AWG, and then injected into the common APD for detecting.



Experimental Setup

Following the configuration of Fig. 1, an experiment is set up for investigating the performance of this stacked WDM-OFDM-PON system. As for downlink, four ITU-T standard wavelengths working at 1543.73 nm, 1544.53 nm, 1545.33 nm, and 1546.13 nm are used as downstream optical source and injected into the MZM. The 10-Gb/s single-drive MZM (JDSU OC-92) is biased at quadrature point for linear electronic-to-optical conversion and driven by the OFDM signal. The OFDM data is generated offline and continuously output by 20-GS/s sampling rate arbitrary waveform generator (Tektronix AWG7122C), with 16-QAM symbol mapping, 512 point IFFT with Hermitian symmetry operation and 128 data-bearing subcarriers. In addition, the cyclic prefix of 16 samples is added to alleviate the inter-symbol interference incurred by chromatic dispersion. The corresponding time and frequency domains of OFDM signal output from AWG7122C is shown as the inset (a) and (b) of Fig. 1, respectively. Through the 0.4-nm AWG multiplexing, the downstream OFDM signals are boosted 10 dBm per wavelength by a bi-directional EDFA, and then are distributed to each ONU. A variable optical attenuator employed in the RN is used to emulate the loss of the optical splitter. Passing through a 100-GHz bandwidth TOF [25] channel selection, downstream signals are detected by an APD [as shown in inset (i) of Fig. 1] and then sampled by the 20GS/s real-time scope (Agilent DSA 91304A).

For the uplink, directly modulated Distributed feedback (DFB) laser is used as upstream transmitter. This DFB has the thermoelectric cooler (TEC) module to realize the wavelength continuous adjustment. And in this experiment, the DFB is biased at 75 mA, and it has the thermally tuned range from 15 °C to 65 °C with 9 dBm stable output power, which is corresponding to 3.0 nm wavelength range. Owing to these properties of DFB, the DFB laser can be easily tuned to any ITU-T wavelengths with 0.8 nm channel spacing by proper adjusting the parameter (reference resistor) of TEC. The corresponding spectra are shown in (c) of Fig. 1. In addition, the 10-Gb/s Pseudo Random Binary Sequence (PRBS) data with the word length of Formula$2^{15} - 1$ is used to drive this DFB to generate upstream signals, and the 100-GHz bandwidth TOF for selecting downstream wavelength is also employed to mitigate upstream signal chirp. Passing through splitter and SMF, the processed upstream signals are fed into the OLT. Prior to reception at OLT, the signals are pre-amplified by EDFA with an optical gain of 20 dB. The amplified signals are fed to AWG and then detected by a common APD used in digital communication for upstream bit-error-ratio (BER) performance evaluation.

Experimental and Simulation Results

First, we investigate the impact of the properties of TOF on downstream system performance, and the corresponding results are shown in Fig. 2. Here, we defined the adjacent-channel power ratio (ACPR) to depict different edge profile TOF. The Fig. 2(a) depicts the schematic diagram of ACPR. It refers to the signal power (through TOF) ratio of main channel (useful signal) to adjacent channels, which indicates the crosstalk of adjacent channels’ signals due to non-steep edge of TOF. Fig. 2(b) shows the spectrum of signals with 5 dB ACPR. In this experiment, the TOF is tuned to make the downstream wavelength locate at different position of TOF to emulate different ACPR cases. The BER performances with different ACPR for 100-km SMF transmission are measured, as shown in Fig. 2(c). Single wavelength and four wavelengths with different ACPR cases are all demonstrated in this Fig. 2(c). It is noted that, since no obvious BER differences among four channels for fixed ACPR case, we only select the sensitivity under the worst case to analyze. Results show that, with an increase in the ACPR, BER performance can achieve significant improvement. This is because that the low ACPR would induce high adjacent-channel crosstalk components. Moreover, compared with single wavelength case, the system with ACPR induced by TOF profile higher than 15 dB achieves negligible power penalty at a forward error correction (FEC) limit of BER@4e-3 [24].

Figure 2
Fig. 2. (a) Schematic of ACPR definition. (b) Spectrum of filter profile and the signals with 100-GHz channel spacing after the TOF. (c) BER performance versus different ACPR.

At the same time, the filter properties are also vital for processing upstream signals’ chirp. Since there are no various profile filters in our lab, the effects of TOF are analyzed by commercial simulation software OptiSystem 7.0. To make simulation results approach to experimental ones, all parameter values of system devices (including DML, APD and so on) are adjusted to achieve the best fit with experimental results. The Gaussian type TOF with 100 GHz 3 dB bandwidth is used in our simulation. It is used to filter out the chirp-induced spectrum broadening low-frequency components, therefore enhancing the extinction ratio (ER) of signals. The BER performances of signal with various orders TOF are shown in Fig. 3(a). With the orders of TOF increasing, BER performances become better. Result also shows that, as for low order TOF case, the error floors are observed at high BER performance. These are attributed to the steep filter edge of high-order TOF that filter out the majority dispersion components induced by chirp. The steep properties of different order filter are shown in inset of Fig. 3(a). As expected, similar performance improvements achieved by the high order filter can also be found with Chebyshev and Butterworth TOF. Moreover, as for different order TOF, the tolerances of DFB wavelength shift are also evaluated, which is presented in Fig. 3(b). Here, the power penalty concept is defined as receiver sensitivities difference of BER@ 4e-3 between wavelength shift position and the optimal wavelength-offset filter configuration. The zero frequency point is referring to optimal wavelength-offset location for chirp management. From Fig. 3(b), we can obtain that, for a fixed power penalty, a reduction in filter order can increase the tolerance of laser wavelength shift. This originates from the fact that the low-order filter gives less steep filter edge as shown in the inset of Fig. 3(a), leading to a large frequency detuning range for chirp management.

Figure 3
Fig. 3. Simulation results (a) upstream BER curve with different order Gaussian TOF, (b) power penalty of BER@ 4e-3 with frequency detuning for different order Gaussian TOF.

Based on the discussion and analysis above, to achieve better performance both for upstream and downstream, we deploy our designed electronic controlling liquid crystal TOF [25] with the similar properties of 12-order Gaussian TOF in our system. The spectrum of this TOF is demonstrated in Fig. 4(a), which measured by an optical spectrum analyzer with a resolution bandwidth of 0.16 pm. It is clearly that our specially designed TOF has the steeper filter edge so that it can achieve larger ACPR (the ACPR 42 dB). These properties make our TOF can be used in system for simultaneously processing chirp and selecting downstream signal. The upstream signals spectrum before and after TOF with the optimal wavelength offset from carrier wavelength, as well selected downstream wavelength is presented in Fig. 4(b). As shown in this figure, downstream wavelengths can achieve 0.4 nm offset from the upstream ones so that the Rayleigh backward scattering induced crosstalk can be mitigated. It is also clearly observed that, the chirp caused spectra broadening is filtered out, thus improving the ER of upstream signal up to be 7.4 dB. In comparison with the case without filter, the eye diagrams are clearly open after the filtering processing, which is presented in Fig. 4(c).

Figure 4
Fig. 4. (a) The profile of our TOF (b) downstream OFDM spectrum after the TOF, and upstream OOK signals spectrum before and after filtering, (c) and (d) is the eye diagrams without and with TOF, respectively.

As mentioned in [9], power budget of our proposed system is limited by downstream transmission. In order to improve system power budget, we also investigate the downstream launched optical power effect on the receiver sensitivity, and the corresponding results are demonstrated in Fig. 5(a), where 100 km SMF is considered. It is shown that, with launched power increasing, the sensitivity degrades and the error floors occur, resulting from the high power-induced fiber nonlinearity in this system. Whereas, the low launched power will produces small power budget. As the analysis above, to achieve the high power budget while keeping high receiver sensitivity, the downstream launched power is set at 10 dBm in our system.

Figure 5
Fig. 5. (a) Increased downstream launched power effect on system sensitivity, (b) BER and constellation diagrams of downstream signals, (c) BER curves of upstream signals.

The BER performance for downstream and upstream link is shown in Fig. 5(b) and (c). Both figures show the single wavelength back-to-back (BtB) case, as well measurements through different distance fibers with four wavelength channel. Here, we only select the sensitivity under the worst case among four channels to analyze. As for downlink, the receiver sensitivities with BtB, 25 km, 50 km, and 100 km SMF transmission are −29.2 dBm, −28.6 dBm, −27.8 dBm, and −27.4 dBm, respectively at a FEC BER limit of 4e-3 [24], as shown in Fig. 5(b). In comparison with the single wavelength BtB transmission, there is almost no obvious performance degradation when four transmission distance cases are observed in the downstream direction, Therefore, the chromatic dispersion induced penalty is almost negligible in downlink. Similarly, from the BER curves of Fig. 5(c), we can easily obtain that, the sensitivities of BtB, 25 km, 50 km, and 100 km SMF transmission are −38.4 dBm, −37.6 dBm, −36.2 dBm, and −35.4 dBm at FEC threshold of BER@4e-3. The negligible sensitivity differences among different distances are also observed in this figure, which are attributed to the filtering effect of suppressing chirp and the residue chirp caused dispersion effects along with the fiber transmission. Compared to the BtB without TOF case, the sensitivity @4e-3 is improved by 5 dB for 100 km SMF, which results in significant power budget improvement.

Finally, according to the signals’ sensitivity and launched power, we evaluate power budget of this stacked WDM-OFDM-PON system, as demonstrated in Table 1. As for downstream link, the output power is 10 dBm per wavelength, and the signal sensitivity is −28.6 dBm@4e-3 and −27.4@4e-3 for 25 km and 100 km SMF in this paper, respectively. With 3-dB insertion losses of our designed TOF, the power budget is 35.6 dB and 34.4 dB, respectively for 25-km and 100-km SMF. In uplink, with the 9-dBm output power, and 6-dB insertion loss of TOF due to detuning filter of the chirp signals, we can achieve that the power budget for uplink is 40.4 dB and 38.4 dB, respectively for 25 km and 100 km SMF. Therefore, the power budget of the whole system is 35.6 dB and a 1 : 512 split ratio splitter can be installed in the RN for 25-km case, 34.4 dB for 100-km fiber. Furthermore, from this table, it is clearly seen that, by replacing PIN by a common APD in the OLT and pre-amplified receiver by our specially designed linear APD, system power budget can be further improved.

Table 1
TABLE 1 Power Budget Comparison Between [9] and This Paper in the 40-Gb/s Stacked WDM-OFDM-PON System


We present a low-cost, high legacy compliant, high capacity stacked WDM-OFDM-PON system based on direct detection OFDM signals for downstream and direct modulated DFB laser with OOK signal for upstream. The impacts of the edge steep and orders of TOF in this proposed system on both downstream and upstream performance in term of BER are investigated in detail by experiment and simulation. Results show that, the profiles of TOF significantly affect the adjacent-channel crosstalk for downstream signals and chirp management for upstream signals. Based on the verified analysis, we deploy our designed TOF with steep filter edge in ONU to simultaneously select the downstream signal and process the dispersion induced by upstream signal chirp. By this experiment setup, we demonstrate the error-free transmission performance of this symmetric 40-Gb/s WDM-OFDM-PON system with 100-km distance. Experimental results show that, 35.6 dB power budget over 25-km SMF transmission is achieved in our system, which can support more than 500 users. As for 100 km fiber transmission, 34.4 dB power budget can be obtained. Owing to its simplicity and good performance, the proposed architecture may be valuable for practical implementation in future next generation PON system.


This work was supported in part by the National Natural Science Foundation of China under Grant 61271216, Grant 61090393, Grant 61221001, and Grant 60972032; by the National 973 Program of China under Grant 2010CB328205, Grant 2010CB328204, and Grant 2012CB315602; and by the National “863” High-Tech Program of China. Corresponding author: S. Xiao (e-mail:


No Data Available


No Photo Available

Meihua Bi

No Bio Available
No Photo Available

Shilin Xiao

No Bio Available
No Photo Available

Hao He

No Bio Available
No Photo Available

Jun Li

No Bio Available
No Photo Available

Ling Liu

No Bio Available
No Photo Available

Weisheng Hu

No Bio Available

Cited By

No Data Available





No Data Available
This paper appears in:
No Data Available
Issue Date:
No Data Available
On page(s):
No Data Available
INSPEC Accession Number:
Digital Object Identifier:
Date of Current Version:
No Data Available
Date of Original Publication:
No Data Available

Text Size