TerabitPerSecond Optical SuperChannel Receiver Models for Partial Demultiplexing of an OFDM Spectrum
 Author: Reza Ahmed Galib, Rhee JuneKoo Kevin
 Publish: Journal of the Optical Society of Korea Volume 19, Issue4, p334~339, 25 Aug 2015

ABSTRACT
Terabitpersecond (Tb/s) transmission capacity for the next generation of longhaul communication networks can be achieved using multicarrier optical superchannel technology. In an elastic orthogonal frequency division multiplexing (OFDM) superchannel transmission system, demultiplexing a portion of an entire spectrum in the form of a subband with minimum power is critically required. A major obstacle to achieving this goal is the analogtodigital converter (ADC), which is powerhungry and extremely expensive. Without a proper ADC that can work with low power, it is unrealistic to design a 100G coherent receiver suitable for a commercially deployable optical network. Discrete Fourier transform (DFT) is often seen as a primary technique for understanding partial demultiplexing, which can be attained either optically or electronically. If fairly comparable performance can be achieved with an alloptical DFT circuit, then a solution independent of data rate and modulation format can be obtained. In this paper, we investigate two distinct OFDM superchannel receiver models, based on electronic and alloptical DFTtechnologies, for partial carrier demultiplexing in a multiTb/s transmission system. The performance comparison of the receivers is discussed in terms of biterrorrate (BER) performance.

KEYWORD
Orthogonal frequency division multiplexing (OFDM) , Superchannel , Partial demultiplexing , Coherent modulation , Fast Fourier transform (FFT) processor

I. INTRODUCTION
To meet the forecasted growth rate of internet traffic, on the order of Tb/s, enhancing the capacity of a singlecarrier transmission system is critically challenging, due to the large penalty from fiber impairments and also a lack of highspeed commercial electronics, which are most likely to remain unavailable in the foreseeable future. Even though the capacity of a transmission system can be increased by incorporating higherorder modulations, such as PM16QAM or PM64QAM [1], it becomes more susceptible to noise, and its reach also decreases as the modulation order increases. In this regard, optical superchannel technology can be viewed as a strong candidate to satisfy tangible Tb/s transmission, where coherently generated, dense multiple carriers are demultiplexed and decoded by orthogonal frequency multiplexing (OFDM) [2], or by Nyquist wavelength division multiplexing [3]. In a superchannel transmission system, the required capacity is usually attained by combining a large number of optical carriers, where each carrier maintains a certain phase relation to the others and is modulated at a lower bit rate, so as to be terminated at a receiver for a service interface [4]. The first successful longhaul superchannel transmission of 1.15 Tb/s over a distance of 10,000 km was reported in [5], while a recordbreaking 26 Tb/s superchannel transmission experiment was reported in [2]. In this paper, the superchannel is formed using OFDM technology because of its potentially high spectral efficiency, finegranularity, and high tolerance of transmission impairments like optical nonlinearity [6], chromatic dispersion [7], and polarization mode dispersion [8].
One of the crucial requirements for successfully deploying optical superchannel technology is the flexibility in carrier assignments, where an arbitrary number of multiple adjacent carriers are grouped to form subbands that can be demultiplexed and demodulated independently from other subbands. This enables the superchannel to provide flexible services of various requirements elastically, to handle differing capacity, modulation format, spectral efficiency, number of carriers, bandwidth, etc., for each subband. A typical superchannel receiver can demultiplex and demodulate a small number of carriers in a group (
i.e. a subband) out of a large number of superchannel carriers by using electronic numerical OFDM and Nyquist filtering after optical heterodyning coherent detection, or by means of alloptical OFDM. A key technique to demultiplex the OFDM carriers is the DFT, which can be realized either electronically or purely optically [9]. These two DFT approaches can form two distinct OFDM superchannel receiver models as partial carrier demultiplexers for subband demultiplexing. A major burden on the electronic DFTbased partial carrier demultiplexer is the powerhungry ADC. If the DFT can be performed on a continuous signal using an optical DFT circuit, then the power and bandwidth consumptions of the system can be reduced significantly. In this paper, we investigate these two options for Tb/s optical superchannel OFDM receivers that can terminate a subband.II. PRINCIPLE OF PARTIAL DEMULTIPLEXING IN AN OFDM SUPERCHANNEL TRANSMISSION SYSTEM
In an OFDM superchannel environment, the spectrum can be extremely wide, since hundreds of OFDM carriers are multiplexed over a single wavelength. In this work, we focus on how to demultiplex a portion of the entire OFDM spectrum. For a better understanding, we intuitively divide the entire spectrum into multiple OFDM subbands and decode each subband with partial demultiplexing, as shown in Fig. 1. There are no restrictions on the number of subbands in the spectrum, nor on the number of carriers in each subband. Due to the orthogonality property of the OFDM [10], the carriers in each subband can have a spacing of Δ
f , corresponding to the reciprocal of the symbol rate. There is no guard band between subbands; in other words, the subbands are orthogonal to each other.In this paper, by considering alloptical and electronic DFT technologies, we investigate two OFDM superchannel receiver models for partial carrier demultiplexing with minimum power and bandwidth consumptions.
III. SUPERCHANNEL SYSTEM MODEL
In this section, we discuss our system model for the Tb/s optical OFDM superchannel transmission system, which includes transmitter and receiver designs. We first describe the transmitter setup, and then the receiver setup for partial carrier demultiplexing.
3.1. OFDM Superchannel Transmitter Model
In the transmitter, a set of frequencylocked carriers is assumed to be generated from a modelocked laser (MLL) synchronized at its fundamental frequency [11], a comb generator employing a concatenated, overdriven optical phase modulator [2], or a recirculating frequency shifter, as shown in Fig. 2. Similar to a DWDM approach, we deploy an arrayed waveguide grating (AWG) device to split the frequency comb into its nonoverlapping Fourier components with a channel spacing of Δ
f = 1/T . Subsequently, the frequencylocked carriers are individually modulated and combined together to form a Tb/s optical OFDM superchannel. This transmitter design offers several definitive advantages over the conventional approach: (1) no electronic IFFT (inverse fast Fourier transformation) block is required; (2) no datarate limitations are imposed by electronics, including the digitaltoanalog converter (DAC); and (3) the carriers can be modulated individually.In this work, a total number
N = 64 carriers with an exact frequency spacing of Δf = 25 GHz is modeled to realize an OFDM superchannel transmission system, in which each carrier is individually modulated using coherent modulation, according to a pseudorandom binary sequence (PRBS) with a length of 2^{15}1. The modulated signals are then combined to form a 3.2 Tb/s OFDM superchannel.3.2. OFDM Superchannel Receiver Model
In this subsection, we present two optical superchannel receiver models as partial carrier demultiplxers. These can be termed as (1) an electronic FFTbased COOFDM, and (2) an alloptical demultiplexing and coherent demodulation OFDM superchannel receiver.
3.2.1. Electronic FFTbased COOFDM Superchannel Receiver
A schematic diagram of the coherent opticalOFDM (COOFDM) superchannel receiver is illustrated in Fig. 3. In this design, the FFT is performed electronically. The major components of the receiver include: (1) a homodyne coherent receiver; (2) two ADCs; (3) an electronic FFT circuit; and (4) a phase estimator, to eliminate the phase noise generated from an LO. Although the architecture of the homodyne receiver is quite complex, it is a preferred technology, because of its minimal electricalbandwidth requirement and 3dB higher sensitivity, compared to heterodyne detection [12].
In the first place, the transmitted coherent modulated optical signals are extracted, using a secondorder superGaussian filter with a bandwidth marginally larger than
K × Δf GHz, whereK denotes the number of carriers. Then the extracted timedomain optical OFDM signals are fed to a coherent receiver, where each component of the signals is coupled with an LO in a 90° optical hybrid, followed by an antialiasing lowpass filter, before being detected by the two photodetectors to obtain inphase (I) and quadrature (Q) components of the signals. In this work, to limit the bandwidth consumption of the system, the LO is tuned at the center frequency of the subband to be demultiplexed. The I/Q components of the signals are sent to an ADC to obtain timedomain sampling (TDS). The digitized I/Q components of the signals are then combined prior to serialtoparallel conversion. With a higher sampling rate, the cost and power consumption of the ADCs increase rapidly, so it is important to limit the sampling rate of an ADC. Subsequently, the timedomain samples are fed to an electronic FFT circuit to demultiplexk carriers in a subband. Finally, each carrier is equalized using the ViterbiandViterbi method [13], to compensate for the phase and amplitude distortions caused by the optical and electrical paths.3.2.2. Alloptical Demultiplexing and Coherent Demodulation OFDM Superchannel Receiver
The AODFTbased alloptical demultiplexing and coherent demodulation OFDM superchannel receiver is illustrated in Fig. 4(a). The major components of the receiver are (1) an AODFT circuit for partial carrier demultiplexing, and (2) a coherent receiver for opticaltoelectrical (O/E) conversion and symbol demodulation in each demultiplexed carrier. Therefore, we first introduce the AODFT process, and then the overall operation of the receiver is discussed.
The AODFT circuit is a simple device that can perform both serialtoparallel conversion and DFT processing alloptically. In this paper, the DFT is realized by tuning the phase of MachZehnder delayandadd interferometers (MZDI), couplers, and optical gates. With this device the DFT can be performed on a continuous signal, so no ADCs are required. In addition, the circuit consumes little or no power, as it consists of passive optical components. An exemplary illustration of an 8point AODFT circuit can be found in Fig. 4(b). For an 8point AODFT, we cascade 3 MZDIs for each carrier, with delay and phase adjustments in the upper and lower arms respectively. At the end of the AODFT circuit, an optical gate, such as an electroabsorption modulator (EAM), is placed at each output arm of the last cascading stage, to sample at the center of a symbol period. The transfer function of cascaded couplers with a delay line in the lower arm of a MZDI is expressed in Eq. (1) [11].
In Eq. (1),
τ andϕ _{m} denote the delay and phase shift in the lower arm of the interferometer, respectively, which are defined asτ =T / 2^{n} andϕ _{nm} = 2π (m / N) π . Here,n represents the stages of anN point DFT, andm <N /2.In this receiver design, the transmitted signals modulated by QPSK are demultiplexed first using a MZDIbased
k point AODFT circuit, as discussed above. The received signals at each of thek demultiplexed carriers are coupled with a reference LO in a 90° optical hybrid, followed by balanced photodetectors for opticaltoelectrical (O/E) conversion. Finally the symbols received at each carrier are equalized using the ViterbiandViterbi [13] method, to mitigate the amplitude and phase distortions.IV. RESULTS
In this section, we investigate the performance of the superchannel receivers in elastic demultiplexing at a transmission rate of 64 × 50 Gb/s over an amplified fiber, with an amplified spontaneous emission (ASE) noise model. The simulated transmitter model follows the design presented in Section 3.1. Because of the carrier spacing of 25 GHz, the symbol duration is assumed to be
T = 40 ps for orthogonality. A total number ofN = 64 OFDM carriers centered on 193.1 THz are coupled to form a 3.2 Tb/s OFDM superchannel. The generated OFDM signals are transmitted over a dispersioncompensated fiber; therefore, no OFDM overhead, such as a cyclic prefix, is required. For balanced coherent detection we use an LO laser, which is assumed to be noiseless and therefore works as a noiseless optical amplifier. In the simulation, the major performance parameter isbiterror rate , which is measured by comparing the received bit sequence to the transmitted bits and counting the number of errors, with respect to the optical signaltonoise ratio (OSNR). Typically the OSNR is defined as the ratio of the transmitted optical signal power to the ASE noise power with a reference bandwidth of 0.1 nm (approximately 12.5 GHz for 1550nm transmission). The total number of transmitted OFDM symbols in a single run is 512. Every BER point in the plots is obtained by iteratively running the simulation many times.Figure 5 compares the BER performances of the aforementioned receiver designs as a function of OSNR for a backtoback transmission situation. The data in the figure correspond to the BER results in demultiplexing of subbands with
k = 4, 8, and 16 carriers as 100, 200 and 400 GHz widths, respectively. In the case of the electronic FFTbased COOFDM superchannel, after balanced detection the signal passes through an antialiasing lowpass filter. Now we collectk = 4, 8, and 16 equidistant samples (no oversampling) at sampling frequencies off_{s} = 100, 200, and 400 GS/s respectively in a symbol period ofT_{s} = 40 ps, and then performedk point FFT to demultiplex thek carriers. Note that the red and blue curves show respectively the BER results of the electronic FFTbased COOFDM and AODFTbased coherent modulated OFDM superchannel receivers. In Fig. 5, the carriers are labeled as –N_{SB} / 2,  (N_{SB} / 21), ⋯ 2, 1, 1, 2, ⋯, (N_{SB} / 21),N_{SB} / 2, whereN_{SB} denotes the number of carriers in a subband and ±N_{SB} / 2 corresponds to the edgemost carriers in a subband.Figure 5(a) shows BER versus OSNR data for
N_{SB} =k = 4 demultiplexing carriers of a 100 GHz subband. For the electronic FFTbased COOFDM receiver, the BER of the center carrier of the subband (CR: 1) is ~10^{5} at 18.6 dB. We notice that the edgemost carriers (CR: 2) completely failed in decoding, due to aliasing effects between carriers located at +50 GHz and –50 GHz. We also experience a noticeable performance degradation of the carriers close to the edge of the decoded subband. In the case of an AODFTbased receiver, a 4point DFT is achieved by cascading two MZDIs followed by optical gates to demultiplexk = 4 carriers, as depicted in Fig. 4(b). Despite an equal gain in performance atevery carrier, we observe that the performance ofany carrier is not satisfactory under any OSNR values.Figures 5(b) and 5(c) present similar data plots for
N_{SB} =k = 8 andN_{SB} =k = 16 demultiplexing carriers as 200 GHz and 400 GHz subbands, respectively. For the electronic FFTbased COOFDM receiver, the measured BER is around 10^{6} at 18.6 dB for the carriers located around the center of the demultiplexed subband. As seen in Fig. 5(a), we notice a significant degradation in BER performance of the edgecarriers. For the AODFTbased coherent demodulation OFDM superchannel receiver, we cascade three and four MZDIs to demultiplexN_{SB} = 8 andN_{SB} = 16 carriers, respectively. The performances of all carriers are almost equal (around 10^{6} at 18.6 dB) for bothN_{SB} = 8, andN_{SB} = 16 cases. Aliasing is not observed in any case.One possible way to mitigate the effect of aliasing on the electronic FFTbased COOFDM superchannel receiver is oversampling. In this paper, two oversampling scenarios are considered: Case 1, 50% oversampling, and Case 2, 100% oversampling. In Case 1, 50% oversampling is achieved by sampling a signal with
k = 3N_{SB} /2 equidistant samples in a symbol period ofT . In Case 2, the number of equidistant samples in a symbol period is increased tok = 2N_{SB} . However, the number of demultiplexed carriers remainsN_{SB} for the both cases. Figure 6 shows BER versus OSNR performance forN_{SB} = 4, 8, and 16 demultiplexed carriers. In this figure the black and red curves respectively show the performance with 50% and 100% oversampling. From Fig. 6(a) we can see that the performance with 50% oversampling is overshadowed by the performance with 100% oversampling, in the case ofN_{SB} = 4 demultiplexing carriers. It is important to note that the impact of aliasing is nearly removed. Similarly, a marginal discrepancy in BER performance is observed in comparing the two oversampling cases in Figs. 6(b) and 6(c). Although oversampling can improve the performance of the receiver in subband decoding, it accordingly increases the power consumption of the system. To limit power consumption, 50% oversampling can be a seen as a good choice for reasonable performance.Several important conclusions can be drawn from the above simulation results: (1) With AODFT, all carriers can realize nearly equal gain in performance and no carriers suffer from aliasing, because the optical waveform of the symbol is demultiplexed before being sampled. (2) A 4point MZDIbased AODFT circuit failed in demultiplexing, and at least three cascaded 8point MZDIs or
N_{SB} ≥ 8 are required to obtain reasonable performance. (3) The performance of an electronic FFTbased COOFDM receiver improves with oversampling,i.e .k >N_{SB} , at the price of increased consumption of power and bandwidth. (4) The AODFT process has several advantages over the electrical FFT. It consists of passive optical components and hence consumes little or no power. Unlike for electronics, there is no datarate limitation, and also it offers simpler implementation. Therefore, the AODFT circuit can be directly adapted to any flexible modulation rate.V. CONCLUSION
In this paper, we model a 64 × 50 Gb/s optical OFDM superchannel transmission system with 64 carriers, and demultiplex the superchannel carriers flexibly in the form of a subband, using two distinct receiver designs. The receivers for partialcarrier demultiplexing employ electronic and alloptical DFT technologies respectively to form (i) an electronic FFTbased superchannel receiver, and (ii) an AODFTbased superchannel receiver. The BERperformance results reveal that the electronic FFTbased superchannel receiver suffers from aliasing, which can be mitigated by oversampling. The simulation results also reveal some minor limitations of MZDIbased AODFT implementation that require more than the first few MZDI cascading stages. Nonetheless, the AODFTbased superchannel receiver consumes no or little power. It neither suffers from aliasing nor requires any ADC/DAC. Therefore the significance of an AODFTbased partial carrier demultiplexer in optical superchannel technology could be immense.

2. Hillerkuss D., Schmogrow R., Schellinger T., Jordan M., Winter M., Huber G., Vallaitis T., Bonk R., Kleinow P., Frey F., Roeger M., Koenig S., Ludwig A., Marculescu A., Li J., Hoh M., Dreschmann M., Meyer J., Ezra S. B., Narkiss N., Nebendahl B., Parmigiani F., Petropoulos P., Resan B., Oehler A., Weingarten K., Ellermeyer T., Lutz J., Moeller M., Huebner M., Becker J., Koos C., Freude W., Leuthold J. 2011 “26 Tbit s1 linerate superchannel transmission utilizing alloptical fast Fourier transform processing,” [Nature Photon.] Vol.5 P.364371

[FIG. 1.] Conceptual diagram of the splitting of an OFDM superchannel spectrum into many subband (SB) plans for partial carrier demultiplexing.

[FIG. 2.] Transmitter setup for optical OFDM superchannel: (1) laser source, (2) optical multicarrier generation, (3) transmitter spectrum.

[FIG. 3.] Receiver setup for the electronic FFTbased COOFDM superchannel: (1) The LO is tuned to the center of the subband to decode. (2) Demultiplexing a subband of k optical carriers out of N = 64 carriers by performing FFT electronically. LO: local oscillator, PD: photodiode, LPF: low pass filter, A/D: analogtodigital, S/P: serialtoparallel, P/S paralleltoserial.

[FIG. 4.] (a) Receiver setup for the AODFTbased coherent modulated OFDM superchannel receiver. LO: local oscillator, PD: photodiode, LPF: low pass filter, A/D: analogtodigital. (b) Schematic illustration of an eightcarrier opticalDFT example [11].

[]

[FIG. 5.] BER versus OSNR performance of the two receiver models, for different numbers of demultiplexing carriers. The red and blue curves show respectively the BERs of the electronic FFTbased COOFDM and AODFTbased coherent modulated OFDM superchannel receivers. (a) NSB = 4, (b) NSB = 8, (c) NSB = 16.

[FIG. 6.] BER versus OSNR performance of the electronic FFTbased OFDM receiver in the cases of 50% (black curve) and 100% (red curve) oversampling for NSB = 4, 8, and 16 demultiplexing carriers. (a) NSB = 4, (b) NSB = 8, (c) NSB = 16.