By Topic

• Abstract

SECTION I

## Introduction

While 100-Gb/s Ethernet is currently under extensive research and development for next-generation Ethernet transport systems [1], Terabit/s Ethernet has already been mentioned as a future direction for transport system evolution [2]. Digital coherent detection [3], [4], [5], [6], [7] is considered to be a promising technique for future high-speed transmission because of its high receiver sensitivity and capability to compensate for transmission impairments such as chromatic dispersion (CD) and polarization-mode dispersion (PMD), which critically impact the performance of high-speed transmission. A key component needed for digital coherent detection is the electronic analog-to-digital converter (ADC). The sampling speed of the ADC used in recent research demonstrations is 50 Gsamples/s [3], [4], and it is expected to be limited to well below 100 Gsamples/s for the foreseeable future. This causes an electronic bottleneck for realizing single-channel Terabit/s transmission. To solve this electronic bottleneck issue, optical multiplexing techniques seem to be necessary. With optical time-division multiplexing (OTDM), demodulation of one out of the 32 tributaries of a single-carrier 640-Gb/s OTDM signal with polarization-division multiplexed (PDM) quadrature phase-shift keying (QPSK) was recently demonstrated by using a manual polarization demultiplexer followed by a pulsed optical local oscillator (OLO) operating at a repetition rate of 10 GHz [8]. More recently, this demodulation scheme has been extended to receive one of the 32 tributaries of a 1.28-Tb/s OTDM signal with PDM 16-QAM [9] and one of the 128 tributaries of a 5.1-Tb/s OTDM signal [10].

Another approach to realize optical multiplexing is to use a multi-carrier signal consisting of multiple subwavelength channels [11], [12], [13], [14], [15]. Recently, a novel multi-carrier coherent receiver where the coherent beating (or demodulation) between all the modulated carriers and their corresponding OLOs is conducted in a single shared polarization-diversity optical hybrid has been proposed [16], aiming to simplify the receiver front end, as compared with conventional wavelength-division multiplexing (WDM). The carrier separation is performed afterward by an integrated arrayed waveguide grating (AWG) array to ease the timing alignment among the carriers, demonstrating the first simultaneous demodulation of all the signal tributaries of a Terabit/s signal [16]. In this paper, we present in more detail the design of the multi-carrier coherent receiver, and the experimental results obtained with this receiver for complete demodulation of a 1.12-Tb/s multi-carrier signal consisting of ten 112-Gb/s PDM-QPSK carriers (subchannels) spaced at 50 GHz. We also describe how the cyclic feature of the AWG array can be used to receive carriers that are not adjacently spaced. Finally, we extend the multi-carrier coherent receiver design to accommodate closely spaced carriers, e.g., under the orthogonal frequency-division multiplexing (OFDM) condition [11], [12], [13], [14], [15], for high-spectral-efficiency (SE) Terabit/s transmission.

SECTION II

## Principle and Design

Fig. 1(a) shows the schematic of a conventionally configured receiver for a 1.12-Tb/s multi-carrier signal with 10 × 112-Gb/s subchannels $({\rm S}_{1}, {\rm S}_{2}, \ldots, {\rm S}_{10})$, where all the subchannels are first separated by a wavelength demultiplexer (DMUX) and then individually detected by 10 digital coherent receivers consisting of 10 polarization-diversity optical hybrids, each of which is input with a subchannel signal $({\rm S}_{\rm n})$ and a corresponding OLO reference $({\rm R}_{\rm n})$. The inphase (I) and quadrature (Q) components of two orthogonal polarization states of the subchannels are then sampled by 40 ADCs followed by a digital signal processing (DSP) unit. A key drawback of this approach is the need for 10 polarization-diversity hybrids, which not only causes large size and cost but also difficulty in time matching all the subchannels. Free-space-optics-based hybrids have the advantages of being low loss, athermal, and wavelength independent over a broad wavelength range, but their relatively bulky size further undermines the practicality of this conventional approach. In our proposed receiver setup, as shown in Fig. 1(b), only one polarization-diversity hybrid is used with one input port being connected to the entire multi-carrier signal and the other input port being connected to a multi-carrier OLO source. This shared hybrid design only requires one phase control to stabilize the 90° I/Q phase offset, which is advantageous as compared with the conventional design where multiple phase controls are needed for multiple independent hybrids. The multi-carrier OLO source can be constructed by ten independent lasers followed by a wavelength multiplexer or by one laser followed by a multi-carrier generator. The multi-carrier generator can be based on cascaded modulators [14] or based on recirculating frequency shifting [11], [15]. The output ports of the hybrid are connected to four DMUXs, which were realized by integrating four 1 × 10 AWGs on a planar lightwave circuit (PLC), as shown in the inset of Fig. 2(b).

Fig. 1. Schematics of (a) a conventional receiver setup and (b) the proposed receiver setup for a 1.12-Tb/s multi-carrier signal with 10 × 112-Gb/s subchannels. The inset in (b) shows the layout of the integrated 4 × 40 AWG array. PD: Photodetector. OLO: Optical local oscillator.
Fig. 2. Measured passbands of (a) the 10 output ports of each of the four AWGs and (b) one of the 40 AWG-array outputs over the C-band. The center frequency of each passband is aligned to that of a particular subchannel (sc).

The size of the AWG array is 5.7 cm × 4 cm, and the path lengths for its 40 outputs were matched within ±0.01 mm. Furthermore, each of the AWGs was designed to be cyclic every 500 GHz so that this receiver front end can receive any such multi-carrier signal over a broad wavelength range. The input and output ports of the 4 × 40 AWG array were fiber-pigtailed and connectorized. Fig. 2(a) shows the measured fiber-to-fiber transmittance at each of the 40 output ports of the AWG array, and Fig. 2(b) shows the passbands of one of the 40 outputs over the C-band, illustrating the cyclic feature of this device. Over the C-band, the average loss at the passband centers in Fig. 2(b) is $\sim$3 dB, and the loss variation is within ±0.2 dB. It is important to align the passbands of the four AWGs. We achieved accurate alignment of the passband center frequencies of the four AWGs to within ±2 GHz, as indicated in Fig. 2(a), by designing each AWG to have 10 vernier inputs and selecting the vernier input that gives the best alignment after fabrication. The passbands are of first-order Gaussian type with a 3-dB bandwidth of 32 ± 2 GHz. We note that this coherent receiver front end is well suited for further hybrid or monolithic integration with photodiodes.

SECTION III

## Experimental Setup

Fig. 3 shows the experimental setup of the generation of a 1.12-Tb/s multi-carrier signal and its detection with the compact receiver front end. Ten external-cavity lasers with 100-kHz linewidth were used as the light sources. The 1.12-Tb/s multi-carrier signal contained ten 50-GHz-spaced 112-Gb/s PDM-QPSK subchannels with the even and odd subchannels being modulated by two separate I/Q modulators before being combined by a 100/50-GHz interleaver. Polarization multiplexing was achieved by first splitting the signal into two paths by an optical coupler (OC) and then recombining them in a PBS, using polarization controllers (PCs) and a decorrelation delay of 431 symbols. The drive signals were pseudo-random bit sequence (PRBS) of length $2^{15} - 1$. The signal was then attenuated by a variable optical attenuator (VOA) followed by an erbium-doped fiber amplifier (EDFA) to obtain different OSNR before entering the signal port of a free-space-optics-based 2 × 4 polarization-diversity hybrid. A copy of the 10 laser sources was amplified by a booster EDFA with 25-dBm output power and decorrelated from the transmitter lasers by 20 km of SSMF fiber before entering the OLO port of the hybrid. The OLO power to signal power ratio was set to about 15 dB. After coherent mixing, the four outputs of the hybrid were connected to the integrated 4 × 40 AWG array. At the 40 outputs of the AWG array, simultaneous demodulation of all the signal tributaries of the 1.12-Tb/s signal is completed. The BER performance of the entire 1.12-Tb/s signal was measured one subchannel at a time by connecting the four outputs associated with each subchannel to four 30-GHz photodetectors, which were followed by four 50-GS/s ADCs in a Tektronix real-time sampling scope. Sampled waveforms of length $2 \times 10^{6}$ each were stored and processed offline with typical digital coherent detection processes [6], [7], which included resampling, blind equalization for polarization demultiplexing, frequency and phase estimation, and data recovery.

Fig. 3. Experimental setup of the generation and detection of the 1.12-Tb/s multi-carrier signal. Insets are the measured optical spectra of the signal and the multiple carriers serving as the OLOs. MUX: Multiplexer. INT: Interleaver. PBS: Polarization-beam splitter.
SECTION IV

## Experimental Results

Fig. 4 shows the measured BER performance of each of the 10 subchannels of the 1.12-Tb/s signal as a function of OSNR. Here, the definition of OSNR follows the convention of the signal power per 112-Gb/s subchannel divided by the noise power in a 0.1-nm bandwidth for ease of comparison. All ten subchannels perform similarly, and at $\hbox{BER} = 10^{-3}$, the mean required OSNR is 17 dB/subchannel, which is 0.5 dB higher than that for a single 112-Gb/s signal measured without the AWG array, indicating small penalty due to the combined use of the shared hybrid and the AWG array. The required OSNR for the entire 1.12-Tb/s channel at $\hbox{BER} = 10^{-3}$ is 27 dB. The signal performance could be further improved by designing the AWG passband to be higher order Gaussian.

Fig. 4. Measured BER performance of each of the 10 subchannels of the 1.12-Tb/s multi-carrier signal as a function of OSNR per subchannel. The single-channel 112-Gb/s performance is plotted for comparison.

Fig. 5 shows the measured BERs of the ten subchannels at $\hbox{OSNR} = 16.8\ \hbox{dB}$, showing similar performance across the subchannels and a mean BER of $1.4 \times 10^{-3}$. We also verified similar BER performance when the signal was tuned inside the C-band. Typical constellations of the recovered x- and y-polarization components of a subchannel at 35-dB OSNR are shown as the insets, showing error-free operation at high OSNR.

Fig. 5. Measured BERs of the 10 subchannels of the 1.12-Tb/s channel as a function of subchannel index at a mean OSNR of 26.8 dB. Insets show typical recovered constellations at $\hbox{OSNR} = 35\ \hbox{dB}$.
SECTION V

Due to the cyclic feature of the AWG array and the broadband nature of the polarization-diversity optical hybrid used, the multi-carrier coherent receiver is capable of receiving multiple modulated carriers that are not adjacent to each other. Fig. 6 shows two spectral arrangements, where the fifth subchannel is shifted from its original location at 192 THz by 500 GHz and 1.5 THz, while other subchannels remain at their original locations. Note that the frequency shifts are multiples of the cyclic period of the AWG array, i.e., 500 GHz. Fig. 7 shows the BER performances of the fifth subchannel after these frequency shifts, which are similar to that prior to the frequency shift. This ability to receive nonadjacent carriers may be utilized to support interesting applications such as wavelength contention avoidance and selective distribution of subchannels. Note, however, that the use of nonadjacent carriers may require a more sophisticated cancellation of the propagation time difference and, therefore, more buffering in the receiver.

Fig. 6. Measured optical spectra of a 1.12-Tb/s multi-carrier signal whose fifth subchannel is shifted from its original location by (a) 500 GHz and (b) 1.5 THz.
Fig. 7. Measured BER performances of the fifth subchannel at three different locations in the C-band. The OSNR quoted here refers to the subchannel OSNR.
SECTION VI

## Discussions

To support high-SE transmission, the frequency spacing between adjacent carriers in a multi-carrier signal is desired to be as small as possible. Coherent optical OFDM (CO-OFDM) [11], [12], [13], [14] allows the modulated carriers to be closely spaced at the modulation symbol rate (or baud rate) without incurring coherent crosstalk penalty. In this section, we discuss how the receiver concept described above could be extended to the case where the subchannels are closely spaced in the OFDM condition. Fig. 8 shows the schematic of a multi-carrier CO-OFDM transmitter with M frequency-locked carriers generated by a multi-carrier generator [14], [15]. The frequency-locked carriers are then separated by a DMUX before being individually modulated with PDM-OFDM by an I/Q modulator array consisting of 2M I/Q modulators and an M polarization-beam combiner (PBC). Due to the orthogonality among the carriers, no guard band is needed between adjacent carriers [11], [12]. The modulated carriers are then combined by a M:1 coupler. The optical spectra right after the laser, the multi-carrier generator, and the coupler are also illustrated in Fig. 8 as insets.

Fig. 8. Schematic of a multi-carrier CO-OFDM transmitter with frequency-locked carriers. Optical spectra at locations (a)–(c) are also illustrated. DMUX: Wavelength demultiplexer. PBC: Polarization-beam combiner.

Since no guard band is used between adjacent carriers, it is needed for the AWG array used in the proposed multi-carrier coherent receiver to have sharp passband so that for each AWG output port, the corresponding modulated carrier is efficiently passed while the adjacent OLOs are sufficiently rejected. This can be achieved by making the AWG to have a high-order Gaussian-like passband. For example, with a second-order Gaussian-type passband, the rejection at the center of the neighboring modulated carriers is 48 dB when the passband loss at the edge of the center modulated carrier is 3 dB.

Fig. 9 shows the schematic of a multi-carrier coherent receiver for receiving the multi-carrier CO-OFDM signal. The inset illustrates the spectrum of the OLO sources and the high-order Gaussian passbands of the AWG arrays imbedded in the multi-carrier coherent receiver. The OLOs can be generated by a multi-carrier generator, as shown in Fig. 9, to potentially save size, power, and cost, as compared with the conventional approach, where multiple independent lasers are used as OLOs. Also, because of the use of the shared optical hybrid, no wavelength demultiplexing is needed to separate the multiple OLO carriers from the multi-carrier generator. This provides additional potential reduction in complexity. Note that the number of OLOs at the receiver does not have to be equal to the number of modulated carriers at the transmitter. Banded detection where multiple carriers are simultaneously received per digital sampling can be used to reduce the number of OLOs and to improve the sampling efficiency [17]. Note also that to provide common mode rejection, balanced detection (rather than the single-ended detection shown in Fig. 1) is needed following the coherent demodulation. To achieve this, a 2 × 8 polarization-diversity hybrid followed by an 8 × 8 M AWG array would be needed for the complete demodulation of a M-carrier coherent optical signal.

Fig. 9. Schematic of a multi-carrier coherent receiver with AWG array for receiving the multi-carrier CO-OFDM signal. Inset illustrates the optical spectrum at the output of the multi-carrier generator and the high-order Gaussian-type passband of the AWG arrays imbedded in the multi-carrier coherent receiver.
SECTION VII

## Conclusion

We have demonstrated a compact multi-carrier coherent receiver front end consisting of an integrated 4 × 40 cyclic AWG array following a shared polarization-diversity hybrid that is capable of complete demodulation of a 1.12-Tb/s multi-carrier signal containing 10 × 112-Gb/s PDM-QPSK subchannels over the C-band. The cyclic feature of the AWG array can be utilized to receive modulated carriers that are not adjacently spaced for applications such as wavelength contention avoidance. Applying this multi-carrier receiver for closely spaced multi-carrier CO-OFDM signals has also been discussed. It is expected that photonic integration will play an important role in future Terabit/s transmission by allowing for high-performance coherent detection with reduced complexity and potentially much improved cost effectiveness.

### Acknowledgment

The authors would like to express their appreciation to A. H. Gnauck for fruitful discussions and the loan of the polarization-diversity receiver. They are also grateful to A. Adamiecki for the loan of the PRBS generator. The authors wish to thank A. R. Chraplyvy for his support and encouragement.

## Footnotes

Corresponding author: X. Liu (e-mail: xiang.liu@alcatel-lucent.com).

## 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