Browse

• Abstract

SECTION I

## INTRODUCTION

With the increment of personal communication demand, high-speed passive optical network (PON) techniques have attracted more and more attention; among which, orthogonal frequency division multiplexing (OFDM) PON is considered to be a promising candidate for the next-generation high-speed PON for its high spectral efficiency (SE), chromatic dispersion (CD) and polarization mode dispersion (PMD) tolerance and advantages in bandwidth allocation and control [1], [2], [3], [4]. Recently, a lot of high-speed OFDM-PON systems have been demonstrated, the downstream capacity of which has far exceed 100 Gb/s and even reached several terabits per second [5], [6], [7], [8]. As the capacity of the OFDM-PON continues increasing, how to effectively and flexibly allocate the bandwidth among the huge number of ONUs also becomes a hot topic. A wonderful frequency- and time-domain partitioning for OFDMA system to provide multiservices is proposed in [2], where the different subcarriers in different time slot are configured for different applications. However, OFDM-PON system is highly electrical device dependent, and the bandwidth and speed of the current electrical devices, such as the modulation bandwidth of the modulator, the samplerate and bandwidth of the ADC/DAC, and the speed of the DSP, are not powerful enough for high-speed OFDM-PON systems. To solve this problem, a novel subwavelength bandwidth access scheme is proposed in [9], [10], [11], where the total datastream is divided into several smaller bands, and every band is generated and received, respectively. This provides a promising solution to realize high-speed optical OFDM system with relatively low-speed electrical devices. However, in subwavelength bandwidth access scheme, the minimum unit of signal receiving and processing in the ONU is the OFDM band, which is not very convenient to provide personalized or flexible bandwidth allocation since the bandwidth of one band is usually still too wide for a single ONU, and may also make the ONUs system complex and expensive according to the research of real-time OFDM systems in [12] and [13].

In this paper, a novel subband access scheme for OFDM-PON is proposed, where the granularity of the bandwidth allocation can break through the limitation of the band and be as small as the bandwidth of a single OFDM subcarrier. In the optical line terminal (OLT), the total downstream signal is still divided into several bands and generated, respectively, to release the device requirement, but in the ONU, there will be no OFDM-band concept, and the ONUs are able to access any subcarriers from the whole downstream OFDM signal, both inside a OFDM band and across adjacent OFDM bands. What is more, focusing only on the desired subcarriers allows the ONUs to use low-speed devices to realize high system capacity, which is very helpful to reduce the system cost in access network. Based on the proposed scheme, a 1-Tb/s WDM-OFDM-PON system is experimentally demonstrated to prove the feasibility of the proposed scheme, and the transmission performance is also evaluated.

SECTION II

## PRINCIPLE AND ALGORITHM

The schematic of the proposed scheme is described in Fig. 1. In the OLT, the baseband binary data is first modulated in quadrature amplitude modulation (QAM) or other format to increase the SE, and then, to release the requirement of the electrical devices, the total data is divided into some parts and every part is processed individually. Before OFDM modulation, in order to realize the subband access, some training symbols (TSs) are added to compensate for the transmission penalty and help the ONUs to extract their desired subcarriers, which will be explained below in detail. Next, the signal is multiplexed in the OFDM method, and the generated OFDM bands are upconverted to different optical carriers to form the broadband downstream OFDM signal. There are several available technologies to generate multiple wavelength, such as optical frequency comb (OFC), multiwavelength laser, etc. In the proposed scheme, since there will be no band limitation in the ONUs and the ONUs are able to “selectively” receive their desired subcarriers, the structure of the optical distribution network (ODN) is very simple, where only single mode fiber (SMF) and power splitter are equipped. The schematic spectrum of the broadband downstream OFDM signal is shown in Fig. 1(a), where different colors represent different frequencies and it is also clear the whole downstream signal are composed of several OFDM bands separated by the small frequency intervals. To better explain the principle of the proposed scheme, it is assumed that the subcarriers allocated to ONU $k$ are colored green and the bandwidth of which is smaller than that of a complete OFDM band. Compared with direct detection, coherent detection can help to access any part of the signal without aliasing, so in ONU $k$, a tunable laser is used as the local oscillator (LO) to downconvert the desired subcarriers to the baseband through coherent detection. Under the permission of that the desired subcarriers are involved, in the coherent receiver, narrow bandwidth photodetector can be used to reduce the system cost and filter the undesired subcarriers out at the same time. The schematic spectrum of the signal after coherent detection is shown in Fig. 1(b), where the desired subcarriers, which are colored in green, are moved to the baseband and some subcarriers of other ONUs are also involved. Then, the signal is sampled by ADC and processed in the DSP, where the desired subcarriers are recognized and extracted with the help of the TS. The schematic spectrum of the desired subcarriers is also shown in Fig. 1(c), and then, the data carried on them are recovered. In Fig. 1(c), the subcarriers allocated to ONU $k$ are not adjacent, showing the bandwidth allocation flexibility of the proposed scheme, and, since, after the bandwidth-limited PD, the bandwidth of the signal is much narrower than that of a whole OFDM band, the bandwidth and speed requirement of the ADC and DSP will be significantly reduced.

Fig. 1. Schematic of the proposed scheme. (a) Schematic spectrum of the downstream high-speed OFDM signal. (b) Schematic spectrum of the received signal after bandwidth limited detection. (c) Schematic spectrum of the specific subcarriers of one ONU after digital processing.

The frame structure applied to realize the subband access scheme and flexible subcarrier-level bandwidth allocation is shown in Fig. 2, where the horizontal axis represents frequency and the vertical axis represents time, different colors represents different OFDM bands, and every small square represents the data carried on one subcarrier in one time slot, including TS1, TS2, and payload.

Fig. 2. Frame structure of the proposed scheme.

The procedure of the signal receiving and processing in the ONU is shown in Fig. 3. The received signal is first downconverted to the baseband by coherent detection, and then, the start point of the OFDM datastream are detected by frame synchronization, and the spectrum of the received signal is also inserted in Fig. 3(a).

Fig. 3. Procedure of the signal receiving and processing in the ONU.

To make it easier to understand, it is assumed the baseband modulation format is quadrature phase shift keying (QPSK). In TS1, the data carried on the even-index subcarriers are nonzero (QPSK data), while the data carried on the odd-index subcarriers are set zero. According to the time–frequency property of the discrete Fourier transform (DFT), after inverse DFT, in time domain, the two halves of TS1 will be completely the same [14]. Since half of the subcarriers of TS1 are zero, to maintain the power of the datastream constant, the data of TS1 are multiplied by a factor $\sqrt{2}$. TS1 of different bands can be generated individually as long as maintaining the specific spectrum structure of the whole downstream OFDM signal.

As Fig. 1 shows, the laser diode in the OLT and the LO in the ONUs are not synchronized, so the frequency shift should be corrected. It is equivalent that there is an intermediate frequency (IF) carrier in the downconverted signal. According to the principle of the OFDM, the orthogonality of the OFDM subcarriers can be maintained only if the frequency domain sampling point is just at the center of every subcarrier. The frequency of the equivalent IF carrier can be completely random, which may destroy the orthogonality of the subcarriers. In the ONUs, although only several subcarriers are received, in frequency domain, the data carried on the even-/odd-index subcarriers of TS1 are nonzero, while the data carried on the odd-/even-index subcarriers are zero. So, in the received subcarriers, the two halves of TS1 (excluding the cyclic prefixes) are completely the same in time domain when it is transmitted, but there should be a phase shift between the two halves of TS1 because of the equivalent IF carrier. It is assumed the number of the samples of TS1 is $2L$ (excluding the cyclic prefixes), and $d_{i}$ is the $i$th sample point. In theory, the phase shift should be the same between any pair of $d_{i}$ and $d_{i + L}$, to remove the impact of noise, the average value of the phase shift is used to correct the frequency shift TeX Source \eqalignno{\Delta \varphi = &\, \left[\sum_{i = 1}^{L} angle (d_{i + L}/d_{i})\right]\Bigg/L&\hbox{(1)}\cr Data = &\, Data_{r}\times e^{-j \cdot {\Delta \varphi \times R \over L} \cdot t}.&\hbox{(2)}}

In (2), $Data_{r}$ is the received data, $Data$ is the compensated data, and $R$ is the samplerate. It should be noted that the phase shift calculated by (1) will be in the range of $0 \sim \pi$ because of the “$angle$” operation, while the realistic phase shift between $d_{i}$ and $d_{i + L}$ may exceed $\pi$ if the frequency difference between the received signal and the LO is large, and the phase shift should be more precisely expressed as $\Delta \varphi + k\pi$, where $k$ is an integer. If the $\Delta\varphi$ in (2) is replaced by $\Delta\varphi + k\pi$, (2) can be rewritten as TeX Source $$Data = Data_{r}\times e^{-j \cdot {\Delta \varphi \times R \over L} \cdot t}\times e^{- j \cdot 2\pi \cdot k \cdot {R \over 2L} \cdot t}.\eqno{\hbox{(3)}}$$

Compared with (2), in (3), there is an additional term $e^{- j \cdot 2\pi \cdot k \cdot (R/2L) \cdot t}$, where $k \cdot (R/2L)$ is just the bandwidth of $k$ subcarriers. In other words, after the compensation defined by (2), the positions of frequency domain sampling will be just at the center of every subcarrier as Fig. 3(b) shows, but the relative position of the received subcarriers is still unknown because there is an unknown number $k$, and that is why TS2 is needed. Unlike TS1, in the proposed scheme, on the subcarriers of TS2, there are the pseudorandom binary sequences (PRBS), and the PRBS data on TS2 of the whole downstream OFDM signal is known in every ONU, so TS2 provides an identity for every subcarrier of the whole downstream OFDM signal.

In the proposed scheme, only a part of the subcarriers are received in the ONU, after the compensation defined in (2), the signal carried on the received subcarriers is already able to be recovered. So, the TS2 of the received signal is only a part of the TS2 of the whole downstream OFDM signal. It is known that PRBS has the maximum correlation with itself, so self-correlation operation can be applied to recognize the desired subcarriers from the received signal.

It is assumed that the total number of the subcarriers of the downstream OFDM signal is $N$, which is also the length of the TS2 of the whole downstream OFDM signal, the number of the received subcarriers in the ONU is $n$, and the TS2 of the whole downstream OFDM signal and the TS2 of the received signal are noted as $TS_{t}$ and $TS_{r}$, respectively. The value of the index $k$, which maximums TeX Source $$\sum_{p = 1}^{n} TS_{r} (p) \cdot TS_{t}^{\ast} (p + k) \quad (0 \leq k \leq N - n)\eqno{\hbox{(4)}}$$ will indicate the exact position of the received subcarriers during the whole OFDM signal. Fig. 3(c) shows the spectrum of the received signal after subcarrier recognition. Then, the indexes of the received subcarriers can be calculated; the ONU is able to extract the data carried on the subcarriers allocated to them, the spectrum of which is shown in Fig. 3(d). At last, the data carried on the desired subcarriers are demodulated, and the original data are recovered. According to the explanation of the proposed algorithm, it can be seen that the proposed algorithm has similar computational complexity compared with the Schmidl algorithm.

In Fig. 2, $\hbox{ONU}_{1} \sim \hbox{ONU}_{3}$ illustrates three typical bandwidth allocation strategies with the proposed scheme: the subcarriers allocated to $\hbox{ONU}_{1}$ are adjacent with each other and just inside an OFDM band; the subcarriers allocated to $\hbox{ONU}_{2}$ are adjacent but belong to different OFDM bands; and the subcarriers allocated to $\hbox{ONU}_{3}$ are not adjacent and belong to different OFDM bands. It is very clear that the granularity of the bandwidth allocation is as small as the bandwidth of just one subcarrier and the subcarriers allocated to one ONU do not have to be adjacent or in one OFDM band; they can be at any position as long as they are involved in the received signal after coherent downconversion.

SECTION III

## EXPERIMENT SETUP

To prove the feasibility of the proposed scheme, a 1-Tb/s WDM-OFDM-PON experimental system is also established, and the setup of the system is depicted in Fig. 4. There are mainly three parts in the system: OFC generation, OFDM signal generation and modulation, and signal detection and processing. To make it clear, in Fig. 4, the blue lines represent for optical signals and the red ones for electrical signals.

Fig. 4. Setup of the experimental system. (a) The schematic spectrum of the OFC; (b) The schematic spectrum of the downstream signal; (c) The schematic spectrum of the received signal in one ONU.

### 3.1. OFC Generation

To increase the system flexibility, recirculating frequency shift (RFS) structure [15] is applied in the proposed scheme to generate OFC, where the number of the tones and the free-space range (FSR) of the comb can be adjusted very conveniently. As shown in Fig. 4, the complex modulator is biased at the signal-side-frequency-shift status; every time the signal passes the modulator, it will be upconverted/downconverted, and after several rounds, there will be OFC generated. The modulator is driven by a 12-GHz microwave source, which also defines the FSR of the generated OFC. A Finisar Waveshaper is used in the loop as a bandpass filter (BPF) to control the number of the tones to 50, and considering the number of the tones and the maximum input optical power of the modulator, the output power of the laser diode is set to about 0 dBm, whose linewidth is about 10 kHz. In the loop, there is also an erbium-doped fiber amplifier (EDFA) to compensate the power loss and a polarization controller (PC) to adjust the signal polarization status before the modulator to achieve maximum modulation efficiency. The schematic spectrum of the generated OFC is shown in Fig. 4(a).

### 3.2. OFDM Signal Generation and Modulation

In the experiment, the 10-Gb/s original $2^{15} - 1$ PRBS data are first modulated in QPSK format to increase the SE. The number of the subcarriers in one OFDM band is 1000, so the bandwidth of a single subcarrier is only 5 MHz, which is also the minimum granularity of the bandwidth allocation. Cyclic prefixes are added to combat with the CD originated from the fiber transmission. To realize broadband transmission with relatively low-speed devices and simple system structure, a 3-GHz IF carrier is used to upconvert the baseband OFDM signal and turn the complex baseband OFDM signal to real. All the above are processed offline, and then, Tektronix arbitrary waveform generator (AWG) 7122B, whose samplerate is 24 GS/s, is employed as the DAC to generate the electrical OFDM signal. An intensity modulator is used to modulate the IF OFDM signal to the OFC, and the schematic spectrum of the generated downstream OFDM signal is shown in Fig. 4(b). Since the FSR of the OFC is 12 GHz, the frequency of the IF carrier is 3 GHz, and the bandwidth of the baseband OFDM signal is 5 GHz, there will be a 1-GHz guard band between adjacent OFDM bands as Fig. 4(b) shows to avoid interchannel crosstalk and improve the transmission performance.

### 3.3. Signal Detection

As Fig. 4 shows, the downstream OFDM signal is delivered to the ONUs after SMF transmission. A tunable laser diode, whose linewidth is about 100 kHz and wavelength covers the whole downstream signal range, is employed as the LO, whose wavelength is adjusted in the range of the desired subcarriers. Coherent detection is used to downconvert the desired subcarriers to baseband without beat noise, and the schematic spectrum of the received signal is shown in Fig. 4(c). Then, the real and imaginary parts of the received signal are sampled by ADC, respectively. At last, the algorithm introduced in Section 2 is used to synchronize the received signal and extract the desired subcarriers. In the experiment, Tektronix digital phosphor oscilloscope (DPO) 72004B, whose samplerate is 50 GS/s, is used as the A/D converter. It is assumed the bandwidth for each ONU is several gigahertz, so the sampled signal is filtered by a 2-GHz digital filter and downsampled to 10 GS/s to simulate the characteristics of the commercial devices.

In PON applications, it is generally thought not cost effective to use coherent detection, which will make the system complex and expensive. But with the development of photonic integrated circuit (PIC), these days, even active devices, such as modulator and photo detectors, can be integrated on silicon, let alone the passive devices, such as hybrid and couplers. It is worth expecting that the coherent structure will be widely used with the technique development.

SECTION IV

## EXPERIMENT RESULTS

With the proposed scheme, a 1-Tb/s WDM-OFDM-PON experimental system is successfully established. The spectrum of the 50-tone OFC is shown in Fig. 5(a), where the original pump laser is on the right side. It is clear that, as the frequency goes far from the base frequency, the tone-to-noise ratio (TNR) also decreases, which is caused by the accumulated noise (mainly phase noise) during the loop transmission. Fortunately, the TNR of the 50th tone is still higher than 20 dB, and the power fluctuation of the 50 tones is smaller than 3 dB, which indicates the quality of the OFC is good enough for modulation. The spectrum of the modulated downstream WDM-OFDM signal is shown in Fig. 5(b); however, it is not the same as the ideal spectrum in Fig. 4(b), which is limited by the characteristic of the intensity modulator and the resolution of the optical spectra analyzer.

Fig. 5. Spectra of the signal and BER curve of the downstream signal. (a) Spectrum of the 50-tone OFC; (b) Spectrum of the downstream signal; (c) Spectrum of the received signal in one ONU; (d) Constellation and BER curve of the received signal in one ONU.

In the ONU, after coherent detection and A/D conversion, the signal is filtered and downsampled in digital domain, and the spectrum of the signal is shown in Fig. 5(c). To prove that, with the proposed scheme, the subcarrier allocated to an ONU can be in different OFDM bands; the frequency of the LO is tuned to be in the center of two adjacent OFDM bands, so there is a 1 GHz guard band at the center of the spectrum of Fig. 5(c). The bandwidth of the received signal in the ONU is only 2 GHz, which is much smaller than the bandwidth of the 0.6-THz downstream signal and even smaller than the bandwidth of s single OFDM band, and is already under the processing ability of the commercial devices. In the experiment, the bandwidth allocated for the ONU under test is 1 GHz, so the ONU extracts 200 desired subcarriers from the received signal. In order to evaluate the transmission performance, the BER of the received signal is also calculated. The BER V.S. received power curve is shown in Fig. 5(d), where a typical constellation of the demodulated QPSK signal is also inserted.

SECTION V

## CONCLUSION

In this paper, a 1-Tb/s OFDM-PON system with subband access scheme has been proposed and experimentally demonstrated. The proposed scheme breaks through the limitation of the OFDM band and allows the ONUs to receive any subcarriers from any positions of the whole downstream OFDM signal. What is more, in the proposed scheme, the minimum granularity of bandwidth allocation is equal to the bandwidth of a single subcarrier, which can be as small as several thousand Hertz if needed. It should be noted that, in the experiment, to simplify the structure of the experimental system, the data of the OFDM bands completely the same, although the transmission performance is given at the end of the paper, which may be different from the real system according to the analysis in [16]. But the results of the paper are still meaningful, which shows the feasibility of the proposed scheme and also shows that the proposed scheme provides a promising solution for the realistic high-speed OFDM-PON deployment with cost-effective commercial devices.

## Footnotes

This work was supported in part by the NSFC under Contracts 61132004, 60932004, 60736002, 60807026, and 61090391, by the National Program on Key Basic Research Project (973) under Contract 2012CB315703, by the Program for New Century Excellent Talents in University (NCET-10-0520), by the State Key Laboratory of Advanced Optical Communication Systems and Networks, China (2008SH03), and by the Open Fund of Key Laboratory of Optical Communication and Lightwave Technologies (Beijing University of Posts and Telecommunications). Corresponding author: M. Chen (e-mail: chenmh@tsinghua.edu.cn).

The authors are with the Tsinghua National Laboratory for Information Science and Technology (TNList), Tsinghua University, Beijing 100084, China and also with the Department of Electronic Engineering, Tsinghua University, Beijing 100084, China.

## 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