Transmission of 344 Gb/s 16-QAM Using a Simplified Coherent Receiver Based on Single-Ended Detection

Transcription

1 Transmission of 344 Gb/s 16-QAM Using a Simlified Coherent Receiver Based on Single-Ended Detection Volume 8, Number 3, June 2016 Thang M. Hoang Mohammed Y. S. Sowailem Mohamed Morsy-Osman Mathieu Chagnon David Patel Stéhane Paquet Carl Paquet Ian Woods Odile Liboiron-Ladouceur David Plant DOI: /JPHOT Ó 2016 IEEE

2 Transmission of 344 Gb/s 16-QAM Using a Simlified Coherent Receiver Based on Single-Ended Detection Thang M. Hoang, 1 Mohammed Y. S. Sowailem, 1 Mohamed Morsy-Osman, 1,2 Mathieu Chagnon, 1 David Patel, 1 Stéhane Paquet, 3,4 Carl Paquet, 3,4 Ian Woods, 3,4 Odile Liboiron-Ladouceur, 1 and David Plant 1 1 Deartment of Electrical and Comuter Engineering, McGill University, Montreal, QC H3A 0E9, Canada 2 Deartment of Electrical Engineering, Alexandria University, Alexandria 12544, Egyt 3 TeraXion, Inc., Einstein, QC 2716, Canada 4 Ciena Cororation, Einstein, QC 2716, Canada DOI: /JPHOT Ó 2016 IEEE. Translations and content mining are ermitted for academic research only. Personal use is also ermitted, but reublication/redistribution requires IEEE ermission. See htt:// for more information. Manuscrit received March 25, 2016; revised May 3, 2016; acceted May 27, Date of ublication June 1, 2016; date of current version June 7, Corresonding author: T. M. Hoang ( thang. hoang@mail.mcgill.ca). Abstract: We demonstrate a single-wavelength, 344-Gb/s, 43-Gb 16-quadrature amlitude modulation (QAM) olarization division multilexed signal transmission over 800 km oerating below the hard-decision forward error correction (FEC) BER threshold of 3: using an InP dual-olarization in-hase/quadrature modulator. At the receiver, a simlified single-ended coherent receiver with novel digital signal rocessing (DSP) for self-beating noise removal is emloyed. Comared with the conventional single-ended detection, this signal rocessing method doubles the roagation distance and increases the SNR by more than 1 db. In addition, this DSP method rovides 10 db of the received otical ower dynamic range while oerating below the hard-fec BER threshold of 3:8 10 3, comared with the no dynamic range for a single-ended receiver without the roosed DSP. Index Terms: Coherent receiver, in-hase/quadrature (I/Q) modulator, single-ended detection (SED). 1. Introduction The need for higher caacities in short reach and metro otical communication links has created demand for systems oerating at unrecedented bit rates. This goal can be achieved by using several aroaches, one of which is to use high baud rates with relatively lower modulation efficiency, as demonstrated in [1]. In [1], a suer-nyquist filter was used to reduce the signal bandwidth of a 110 Gbaud olarization-division-multilexed quadrature hase shift keying (PDM-QPSK) signal to fit 100 GHz grid. Another alternative is to use a high sectral efficiency modulation format such as high order quadrature amlitude modulation (M-QAM) [2] [4]. At such high baud rate exeriments, digital-signal-rocessing (DSP) at the transmitter and receiver is necessary to mitigate the imairments of the transceiver and otical link. Additional increases in seed can be obtained via multilexing either in wavelength, e.g., dual-carrier 32 Gbaud PDM-16-QAM [5], or in time [6]. For an equal bit rate, modulating a single carrier from a single laser rovides a solution that is cost

3 effective relative to a hardware multilexed alternative which involves the use of multile wavelengths or fiber lanes to achieve the same total desired caacity. However, obtaining higher data rates with lower transmitter comlexity, including ackage size, and ower dissiation is challenging. Indium hoshide(inp) Mach Zehnder modulators (MZM) have both high bandwidths and low oerating voltages, with ublished values of V rangingfrom1to2.5vandbandwidthsfrom12to36ghz[7] [9]. In addition, InP devices can be integrated with other electrical or otical comonents such as lasers, RF amlifiers and otical amlifiers [7], [9], [10]. All of these features enable the reduction of both ackage size and ower consumtion. Several single wavelength transmissions have been demonstrated using InP IQ modulators, with different reorted values of distances, data rates, and modulation formats. Among those exeriments, 32 Gbaud PM-QPSK and PM-16-QAM transmission over 8000 km, and 960 km, resectively, was demonstrated at the soft decision forward error correction (SD-FEC) threshold of [7]. In addition, in [11] transmission of 28 Gbaud 64-QAM modulated signal over 40 km at a BER of 10 2 is shown. The above systems all utilize balanced-detection (BD) based coherent receivers. To recover the transmitted signals, coherent receivers linearly ma the in-hase (I) and quadrature (Q) comonents of the received otical signal to the electrical domain by mixing the received otical field with that of a local oscillator (LO). This direct maing enables digital signal rocessing, which then allows the use of sectrally efficient modulation formats and also transmission imairment mitigation [12]. There are two main aroaches, with their resective tradeoffs, to imlement coherent receivers recetion: i) balanced detection (BD), which is commonly used in commercialized integrated coherent receivers (ICR); and ii) single-ended detection (SED) [13], [14], [17] [19]. BD ICRs have a high common-mode rejection ratio which reduces self-beating noise and make them suitable for high fidelity alications (e.g, long-haul transmission). Alternatively, SED ICRs can be realized with a simler front-end and are thus more comact and less exensive. They can be alied when a larger enalty for both Q-factor and dynamic range is accetable [13], [15], [16]. Since a comact transceiver form factor is a desirable feature for metro alication [20], SED ICR research is of immense interest to reduce receiver comlexity. In this aer, we successfully demonstrate 344 Gbs transmission (300 Gbs ayload) below the hard decision forward error correction (HD-FEC) threshold of 3: over 800 km of standard single mode fiber using an InP dual olarization IQ modulator (DP-IQM) and SED ICR imlementation. Secifically, the system uses a 2.5 V InP DP-IQM on a single carrier oerating at symbol rate of 43 Gbaud 16-QAM and a SED ICR with DSP incororating an additional function that removes unwanted self-beating noise. The roosed algorithm allows for an additional 300 km of roagation and imroves the SNR relative to conventional SED receiver by more than 1 db. In addition, this DSP method rovides 10 db of received otical ower dynamic range while oerating below the HD-FEC BER threshold of 3: Exerimental Setu and DSP Procedure 2.1. Exerimental Setu Fig. 1 resents the setu used for transmission exeriments using the InP-based DP-IQM with DP-16QAM modulation. We used the same exerimental setu as well as arts of DSP blocks to demonstrate the feasibility of utilizing DP-IQM for high data rate exeriment [21]. The DP-IQM used in this exeriment is ackaged in a 41 mm by 19 mm module and uses two InP chi on carriers, as described in [8]. Fig. 2 shows a schematic illustrating the DP-IQM layout. The extinction ratios at 1550 nm (in db) of the child MZMs are and for the X olarization modulator and and for the Y olarization modulator. Over the entire C-band, the worst case extinction ratio is 25 db and the largest insertion loss of the modulator is 10 db. The V for this design is 2.5 V and the 3-dB bandwidth is greater than 35 GHz for all MZM structures. An external cavity laser (ECL) at nm, with a linewidth of 100 khz, and 15.5 dbm outut ower is used as the laser source. The four RF signal streams driving the InP-based DP-IQM are generated using a digital-to-analog

4 Fig. 1. Exerimental setu. Fig. 2. InP DP-IQM (left) schematic and (right) module. converter (DAC). The AC-couled 8-bit DAC, running at 65.7 GSs, has four differential channels with a maximum outut voltage of 1.2 V. The differential oututs from the DAC channels are amlified by the RF linear driver to obtain four 5 V RF single-ended signals to drive the InP DP-IQM. Afterwards, the otical signal is amlified to 23 dbm with a booster amlifier. The booster is followed by a variable otical attenuator (VOA) to adjust the signal ower launched into the fiber to 1 dbm, which was emirically found to be the otimum launch otical ower. The signal is then fed into an otical re-circulating loo controlled by switches. Each san in the loo has 80 km of SMF- 28e+ fiber, followed by an inline erbium-doed fiber amlifier (EDFA) with a noise figure of 5 db. The second san is followed by a 2-nm bandwidth tunable filter centered at a wavelength of nm. The outut signal from the re-circulating loo is connected to a 0.8 nm filter and is then reamlified to 9 dbm. Then, the re-amlified signal is attenuated using a VOA to swee the received ower rior to the coherent receiver. A free-sace-otics-based 90 otical hybrid is used to mix the received signal with a continuous wave (CW) light from another ECL laser with similar secifications to the one used at the transmitter. The otical hybrid is followed by balanced detection to eliminate the direct detection terms for referencing erformance. To switch from the balanced detection to SED configuration, we disconnected the inut to negative bias of the BD hotodetector. After the coherent detection, the four signals are samled by a 33 GHz 3 db bandwidth real-time oscilloscoe (RTO) running at 80 GS/s for offline DSP and serving as an 8-bit analog-to-digital (ADC) DSP Procedure As the system emloys both DACs and ADCs, DSP can be alied at both the transmitter and the receiver. This section describes the different DSP schemes alied at both ends of the transmission system, from signal generation to signal detection, in order to best generate and recover the desired waveforms. Fig. 3 illustrates the DSP rocess used in the transceiver. The transmitter DSP (Tx-DSP) starts with generating two indeendent streams of 16-QAM symbols uniformly distributed over four levels for dual olarization IQ transmission. The maximum number of symbols allowed by the

5 Fig. 3. DSP stacks at the (a) transmitter and (b) receiver side. DAC were generated. Considering our oversamling factor (DAC samling rate/baud rate) and the granularity of the DAC s memory, this turned out to be symbols. Because the maximumnumberofsymbolsallowedwerechosentofillthedac s memory, the imact of the attern length on the erformance arameters such as BER and SNR was not investigated exerimentally. After u-samling the signal from one samle er symbol (ss) to two ss, a ulse shaing filter is alied using a root-raised-cosine (RRC) ulse with a roll-off factor of 0.4. Afterwards, the signals are re-samled to the oerating rate of the DAC at 65.7 GSs. This is followed by non-linear comensation for the inherent nonlinear MZM transfer function to kee the constellation oints equally saced after electrical-to-otical conversion. Next, the frequency resonse of the transmitter comonents (including i) the DAC, ii) the RF linear driver, and iii) the DP-IQM) is comensated using an FIR filter for digital re-emhasis. The FIR filter was exerimentally otimized at 43 Gbaud and had 61 effective tas. Finally, the RF re-emhasis is followed by an 8-bit quantization rocess. At the receiver, the DSP starts with IQ demodulation using the novel DSP for a SED based ICR. The details of the roosed DSP are described in the next section. After demodulation, the signal is resamled to two ss and then assed through tyical DSP blocks: i) frequency domain chromatic disersion (CD) comensation, ii) laser frequency offset comensation based on the Fast Fourier Transform of the signal at the 4th ower, iii) matched filtering, iv) timing recovery, v) synchronization, and vi) olarization de-multilexing and hase noise mitigation using training-symbol least mean square (TS-LMS) algorithm for initial convergence and decision-directed LMS (DD-LMS) for steady-state oeration. After olarization de-multilexing, the SNR calculation is done by measuring the noise ower with transmitted symbols and subtracting that noise ower from the total received signal ower. The BER is directly counted over bits Proosed DSP for IQ Demodulation Using Single-Ended Detection The derivation begins by modeling the hotocurrent of any two oututs of an otical hybrid as I 1 ¼ I S þ I LO þ 2 ffiffiffiffiffiffiffiffiffiffiffiffi I S I LO cos I 2 ¼ I S þ I LO þ 2 ffiffiffiffiffiffiffiffiffiffiffiffi I S I LO cosð Þ ¼ I S þ I LO þ 2 ffiffiffiffiffiffiffiffiffiffiffiffi I S I LO ½cos cos þ sin sinš (1) where I S and I LO are the hotocurrents from the signal (S) and the LO, resectively; is the hase difference between the signal and the LO field; and is the otical hase shift of either the signal or the LO introduced in the second branch by the otical hybrid. In conventional AC couled SED, the I LO and the DC comonent of I S are removed. However, the AC comonent of I S still remains and degrades the fidelity of retrieved signal. Mathematically, (1) is a system of two indeendent equations with two unknowns and I S. For the roblem at hand, we roose an algorithm, based on the hase-shifting interferometry concet, to remove the self-beating noise ¼ I S þ I LO [22] [25]. The self-beating noise can be removed and I/Q comonents can be found from the following stes (the derivation is shown in the Aendix):

6 Fig. 4. Performance at received signal ower of 2 dbm and LO ower of 15.5 dbm. (a) Measured BER versus transmission distance. (b) (Inset) Measured SNR versus transmission distance and their constellations at 640-km transmission. Ste 1 Calculating : where v ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðv ¼ 2 4u wþ 2u (2) u ¼ 2ð1 cosþ v ¼ 2ð1 cosþði 1 þ I 2 Þ 4I LO sin 2 w ¼ðI 1 I 2 cosþ 2 þ I2 2 sin2 þ 4ILO 2 sin2 : (3) Ste 2 Calculating I/Q from : For instance, in the case of a 2 2 SED with a 90 hybrid ð ¼ =2Þ I ISED ¼ I 1 ¼ 2 ffiffiffiffiffiffiffiffiffiffi I S I LO cos I QSED ¼ I 2 ¼ 2 ffiffiffiffiffiffiffiffiffiffi I S I LO sin : (4) The benefits of using the roosed algorithm with resect to conventional SED are twofold: The self-beating noise ¼ I S þ I LO is removed at the exense of additional DSP comlexity. The comlexity of the roosed DSP is only 13 N real oerations (eight real adders and five real multiliers), where N is the number of rocessed samles er symbol for a 90 hybrid (N ¼ 2 in this case). The comlexity of a reviously roosed algorithm using a second order aroximation is 16 N real oerations (eight real adders and eight real multiliers) [17]. The hase-shift between the two otical branches of the hybrid can be any value in the range 0 G G, thus relaxing the fabrication accuracy of the otical hybrid. 3. Results and Discussion Fig. 4 shows the BER and the SNR versus the transmitted distance for a 43 Gbaud 16-QAM signal using four detection schemes: BD, conventional SED, SED using the resented DSP and SED using second order aroximation DSP [17]. The LO ower was set to 15.5 dbm and the received signal ower was 2 dbm. In Fig. 4(a), SED with the roosed DSP (curve shown using blue squares) outerforms the conventional SED (curve shown using red circles) and slightly outerforms the receiver using second order aroximation (curve shown using orange triangles) [17]. This BER versus distance curve tends to converge towards the BD curve (shown as

7 Fig. 5. Measured BER versus received signal ower after 640-km transmission and an LO ower of 15.5 dbm. black diamonds) at longer distances. Although it is observed that a 344 Gbs signal can be transmitted over 400 km oerating below the HD-FEC threshold using conventional SED, using the roosed DSP, the reach doubles and extends to 800 km. In Fig. 4(b), we show the effectiveness of the roosed DSP in terms of SNR imrovement of the SED aroach. Relative to conventional SED, the SED with this roosed DSP gains 2 db of SNR in the back-to-back configuration and aroximately 1 db of SNR at 800 km. The erformance of SED using roosed DSP is comarable with that of SED using second order aroximation in terms of SNR. Moreover, it is exerimentally shown, for the first time, that the back-to-back SNR of both SED using the roosed DSP and second order aroximation is only 1 db below the SNR from balanced detection. Fig. 5 shows the BER versus received signal ower at the hybrid for a distance of 640 km and LO ower of 15.5 dbm. With the received signal ower varying from 3 to7dbm,thelowest achievable BER using the conventional SED is only 4: and is above the HD-FEC. However using our roosed DSP, the algorithm enables oeration below the HD-FEC as well as a dynamic range of 10 db, from 3 to 7 dbm of received signal otical ower. At a received signal ower of 2 dbm, the roosed algorithm imroves the BER of SED from 6: down to 1: The BER of the BD at this same signal ower is 1: Note that the erformance of SED (both conventional and our roosed one) degrades at high received signal owers since the contribution of the remaining self-beating noise becomes significant in that regime. While the roosed SED has narrower dynamic range comared to BD, its BER is only marginally higher. 4. Conclusion We exerimentally demonstrated single wavelength, dual olarization, 43-Gbaud, 16 QAM transmission (344 Gbs) over 800 km of SMF below HD-FEC threshold of 3: using an InP dual-olarization IQ modulator and a SED coherent receiver, which emloy a novel DSP algorithm to eliminate the distortion caused by self-beating noise in a SED ICR. This signal rocessing algorithm also allow to achieve 10 db of dynamic range at 640 km and obtained an imrovement of 1 db in SNR comared to a conventional SED ICR. The use of the novel DSP algorithm resented in this aer is highly favorable, if a SED ICR is used at the receiver in a single carrier high-seed signal transmission for comact transceiver imlementation along with a comact InP modulator for the transmitter.

8 Aendix This Aendix shows the derivation of the closed-form solution of I/Q detection using haseshifting interferometry concet. With reference to (1), we can re-organize the equations as ffiffiffiffi I 1 I S cos ¼ 2 ffiffiffiffiffiffi I LO ffiffiffiffi I 2 I 1 cos ð1 cosþ I S sin ¼ 2 ffiffiffiffiffiffi : (5) I LOsin After taking the square of both sides of (5) and then taking their sum, we arrive at ði 1 Þ 2 sin 2 þ ½I 2 I 1 cos ð1 cosþš 2 ¼ 4I LO I S sin 2 ¼ 4I LO ð I LO Þsin 2 or 2ð1 cosþ 2 2ð1 cosþði 1 þ I 2 Þþ4I LO sin 2 þði1 I 2 cosþ 2 þ I 2 1 þ I2 2 2I 1I 2 cos þ 4I 2 LO sin2 ¼ 0: (6) Equation (6) is a quadratic equation in. The solutions to the above quadratic equation can be easily found to be ¼ v ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðv 2 4u wþ (7) 2u where u ¼ 2ð1 cosþ v ¼ 2ð1 cosþði 1 þ I 2 Þ 4I LO sin 2 w ¼ I1 2 þ I2 2 2I 1I 2 cos þ 4ILO 2 sin2 ¼ðI 1 I 2 cosþ 2 þ I2 2 sin2 þ 4ILO 2 sin2 : (8) Thus can be found without an additional hoto detector (PD), rovided the current of the LO is known and the +/ sign can be decided. The current of the LO can be found during a calibration rocess. Uon noting that only one of the signs is the correct solution, we shall now determineit.from(1),wecalculate sin 2 I 1 þ I 2 þ 2I LO 1 cos ¼ 2 þ 2I sin 2 LO 1 cos þ 2 ffiffiffiffiffiffiffiffiffiffiffiffi I S I LO ½cosð ÞŠþ2 ffiffiffiffiffiffiffiffiffiffiffiffi I S I LO cos from which we can write sin 2 2 ¼ I 1 þ I 2 þ 2I LO 1 cos 2I sin 2 ffiffiffiffiffiffiffiffiffiffiffiffi LO 2 I S I LO ½cosð Þþcos Š 1 cos ¼ v 4I LOsin 2 ffiffiffiffiffiffiffiffiffiffiffiffi 4ð1 cosþ I S I LO ½cosð Þþcos Š u v 4 ffiffiffiffiffiffi ffiffiffiffiffiffi I LO I LOsin 2 ffiffiffiffi þð1 cosþ I S½cosð Þþcos Š ¼ u v 4 ffiffiffiffiffiffi ffiffiffiffiffiffi ffiffiffiffi I LO I LOð1 cosþð1 þ cosþþð1 cosþ I S½cosð Þþcos Š ¼ u v 4 ffiffiffiffiffiffi ffiffiffiffiffiffi ffiffiffiffi I LOð1 cosþ I LOð1 þ cosþþ I S½cosð Þþcos Š ¼ : (9) u

9 ffiffiffiffiffiffi ffiffiffiffi In ractical alications, since it is often the case that I LOð1 þ cosþ > I S½cosð Þþcos Š, 2 is, thus, always less than or equal to v=u in (9). Therefore, minus sign in (7) should be chosen. The equation can now be used in the first ste to remove self-beating noise described in Section 2.3. References [1] J. Zhang, J. Yu, Z. Jia, and H-C. Chien, 400 G transmission of suer-nyquist-filtered signal based on single-carrier 110-GBaud PDM QPSK with 100-GHz grid, J. Lightw. Technol., vol. 32, no. 19, , Oct [2] Z. Zhang et al., Coherent transceiver oerating at 61-Gbaud/s, Ot. Ex., vol. 23, no. 15, , Jul [3] R. Rios-Muller et al., Sectrally-efficient 400-Gb/s single carrier transort over 7200 km, J. Lightw. Technol., vol. 33, no. 7, , Ar [4] F. Buchali, A. Klekam, L. Schmalen, and T. Drenski, Imlementation of 64 QAM at GBaud using 1.5 samles er symbol DAC and demonstration of u to 300 km fiber transmission, in Proc. IEEE Ot. Fiber Commun. Conf., San Francisco, CA, USA, 2014, [5] T. J. Xia et al., Transmission of 400 G PM-16 QAM channels over long-haul distance with commercial all-distributed Raman amlification system and aged standard SMF in field, in Proc. IEEE Ot. Fiber Commun. Conf., San Francisco, CA, USA, 2014, [6] G. Raybon et al., Single-carrier 400 G interface and 10-channel WDM transmission over 4800 km using all-etdm 107-Gbaud PDM-QPSK, in Proc. IEEE Ot. Fiber Commun. Conf., Anaheim, CA, USA, 2013, [7] S. Chandrasekhar, X. Liu, P. Winzer, J. E. Simsarian, and R. A. Griffin, Comact all-inp laser-vector-modulator for generation and transmission of 100-Gb/s PDM-QPSK and 200-Gb/s PDM-16 QAM, J. Lightw. Technol., vol. 32, no. 4, , Feb [8] G. Wang and I. Woods, Low V, high bandwidth, small form factor InP modulator, in Proc. IEEE Avionics, Fiber- Ot. Photon. Technol. Conf., Atlanta, GA, USA, 2014, [9] T. Tatsumi et al., A comact low-ower 224-Gb/s DP-16 QAM modulator module with InP modulator and linear driver ICs, in Proc. IEEE Ot. Fiber Commun. Conf., San Francisco, CA, USA, 2014,. Tu3H.5, [10] M. Smit et al., An introduction to InP generic integration technology, Semicond. Sci. Technol., vol. 29, no. 8,. 1 42, Jun [11] N. Kikuchi, R. Hirai, and Y. Wakayama, High-seed otical 64 QAM signal generation using InP-based semiconductor IQ modulator, in Proc. IEEE Ot. Fiber Commun. Conf., San Francisco, CA, USA, 2014, [12] S. J. Savory, Digital filters for coherent otical receivers, Ot. Ex., vol. 16, no. 2, , Jan [13] A. Carena, V. Curri, P. Poggiolini, and F. Forghieri, Dynamic range of single-ended detection receivers for 100 GE coherent PM-QPSK, IEEE Photon. Technol. Lett., vol. 20, no. 15, , Aug [14] Y. Painchaud, M. Poulin, M. Morin, and M. Tetu, Performance of balanced detection in a coherent receiver, Ot. Ex., vol. 17, no. 5, , Mar [15] Y. Yadin, M. Orenstein, and M. Shtaif, Balanced versus single-ended detection of DPSK: Degraded advantage due to fiber nonlinearities, IEEE Photon. Technol. Lett., vol. 19, no. 3, , Feb [16] Y. Feng, H. Wen, H. Zhang, and X. Zheng, 40-Gb/s PolMux-QPSK transmission using low-voltage modulation and single-ended digital coherent detection, Chin. Ot. Lett., vol. 8, no. 10, , [17] X. Zhou, J. Yu, and D. Qian, A novel DSP algorithm for imroving the erformance of digital coherent receiver using single-ended hoto detection, in Proc. IEEE ECOC, Brussels, Belgium, Se. 2008, [18] Y.-K. Huang et al., Filterless recetion of Gb/s WDM channels using single-ended hotodiodes and digital interference reduction, in Proc. IEEE ECOC, Amsterdam, The Netherlands, Se. 2012, [19] C. Xie et al., Colorless coherent receiver using 3 3 couler hybrids and single-ended detection, Ot. Ex., vol. 20, no. 2, , [20] D. Rafique, Fiber nonlinearity comensation: commercial alications and comlexity analysis, J. Lightw. Technol., vol. 34, no. 2, , Jan [21] M. Sowailem et al., 400 G single carrier 500 km transmission with an InP dual olarization modulator, IEEE Photon. Technol. Lett., vol. 28, no. 11, , Jun [22] Z. Malacara and M. Servin, Interferogram Analysis for Otical Testing. 2nd ed. CRC: Boca Raton, FL, USA, [23] T. M. Hoang, M. Morsy-Osman, M. Chagnon, Q. Zhuge, D. Patel, and D. Plant, Phase diversity method for otical coherent receiver, in Proc. IEEE CLEO, San Jose, CA, USA, 2015, [24] T. M. Hoang et al., Phase-diversity method using hase-shifting interference algorithms for digital coherent receivers, Ot. Commun., vol. 356, , [25] X. F. Meng et al., Two-ste hase-shifting interferometry and its alication in image encrytion, Ot. Lett., vol. 31, no. 10, , 2006.