IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO. 2, MARCH/APRIL 2020 3500108

Phase-Coded Microwave Signal Generation Based ona Segmented Silicon Mach–Zehnder Modulator

Gangqiang Zhou , Linjie Zhou , Liangjun Lu, Yuyao Guo, and Jianping Chen

Abstract—We demonstrate novel phase-coded microwave signalgeneration based on a two-segment single-drive push–pull siliconMach–Zehnder modulator. The coding signal and the microwavefrequency are each applied to one of the separate segments. Thephase-coded microwave signal is generated by properly settingthe working point of the modulator. The precise π phase shift isdetermined by the sign of the coding signal. Both theoretical anal-ysis and experimental demonstration are performed to verify theconcept. The modulator successfully generates a 10-GHz (20-GHz)phase-coded microwave signal with a 5-Gb/s (10-Gb/s) coding datarate and the measured pulse compression ratio is 133.0 (156.8).

Index Terms—Microwave photonics, silicon modulator, siliconphotonics, phase-coded microwave signal.

I. INTRODUCTION

IN MODERN radar systems, in order to obtain a high resolu-tion and a large detection range simultaneously, phase-codedmicrowave signals have been widely used [1]. Traditionally,the phase-coded microwave signal is generated in the electricaldomain. However, due to the bandwidth limitation of the electri-cal devices, the operation frequency is usually low. High-speedmicrowave signal generation by photonic approaches possessesthe merits of a larger bandwidth and more flexible tunabil-ity compared to electrical ones. One method for phase-codedmicrowave signal generation is to first produce two opticalsidebands by RF modulation and then perform phase codingon one sideband [2]–[6]. After optical-to-electrical conversion,a phase-coded microwave signal is generated. An optical filter[2]–[4] or a polarization maintaining fiber [5], [6] is necessaryto separate the two sidebands. Another method is based onthe optical carrier phase-shifting technique using a modulatorusually with two branches, such as a dual-parallel Mach-Zehndermodulator (MZM) [7], a dual-polarization modulator [8], adual-polarization dual-parallel MZM [9]. In these structures,

Manuscript received April 3, 2019; revised June 24, 2019; accepted June 27,2019. Date of publication July 1, 2019; date of current version July 16, 2019.This work was supported in part by the National Natural Science Foundation ofChina under Grants 61705129 and 61535006, and in part by the Shanghai Mu-nicipal Science and Technology Major Project under Grant 2017SHZDZX03.(Corresponding author: Linjie Zhou.)

The authors are with the State Key Laboratory of Advanced Optical Commu-nication Systems and Networks, the Shanghai Institute for Advanced Communi-cation and Data Science, Shanghai Key Lab of Navigation and Location Services,and the Department of Electronic Engineering, Shanghai Jiao Tong University,Shanghai 200240, China (e-mail: zhougq_9@sjtu.edu.cn; ljzhou@sjtu.edu.cn;luliangjun@sjtu.edu.cn; guoyuyao1992@sjtu.edu.cn; jpchen62@sjtu.edu.cn).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/JSTQE.2019.2926198

one branch is modulated by the coding signal to change thephase of the optical carrier, and the other branch is driven bythe microwave signal with the optical carrier suppressed togenerate the two sidebands. The phase-coded microwave signalis generated upon combining the two branches. Besides theabove two common methods, phase-coded microwave signalscan also be generated from a dual-drive MZM by properly settingthe working point and the coding signal [10], [11]. All-opticalgeneration of binary phase-coded microwave signal based oncross-polarization modulation in a highly nonlinear fiber hasalso been demonstrated [12].

With the recent rapid development of integrated microwavephotonics (MWP), MWP systems can be reduced in footprintand complexity considerably [13]. Indium phosphide (InP) [14],silicon nitride (SiN) [15], and silicon-on-insulator (SOI) [16] arethe three key material platforms for monolithic MWP integra-tion. The InP platform enables the integration of various passiveand active photonic components. However, it is not compatiblewith the complementary metal-oxide-semiconductor (CMOS)process and lacks the potential for low-cost and mass production.On the SiN platform, it is hard to realize high-speed modulationunless hybrid integration with other materials is employed. TheSOI platform features CMOS compatibility and high-densityphotonic integration as the waveguide size and bending radiusare much smaller than those on the InP and SiN platforms.

Electro-optic modulators on the SOI platform have beenintensively investigated in recent years [17]–[22]. They havebeen employed in multiple integrated MWP signal generationand processing applications. For example, an integrated mi-crowave photonic filter (MPF) has been implemented by usinga high-speed phase modulator (PM) followed by a thermallytunable high-Q micro-disk resonator (MDR) and a high-speedphotodiode (PD) [23]. An integrated optical frequency combhas been realized by using a micrometer-scale ring resonatormodulator (RRM) [24]. A frequency-doubled microwave signalhas been generated by employing a coupling-modulated ring res-onator [25]. As for phase-coded microwave signal generation, asilicon micro-ring modulator together with an external intensitymodulator has been used [26]. In this method, to realize theprecise π phase shift, the amplitude of the coding signal shouldbe set properly.

In this paper, we demonstrate phase-coded microwave signalgeneration using a two-segment single-drive push-pull siliconMZM. The segmented silicon modulators have previously beendeveloped for pulse-amplitude-modulation (PAM) [27], [28]. Atwo-segment MZM has been used to realize 128 Gb/s PAM-4

1077-260X © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

https://orcid.org/0000-0003-2791-0453https://orcid.org/0000-0002-2792-2959https://orcid.org/0000-0002-1186-7633mailto:zhougq_9@sjtu.edu.cnmailto:ljzhou@sjtu.edu.cnmailto:luliangjun@sjtu.edu.cnmailto:guoyuyao1992@sjtu.edu.cnmailto:jpchen62@sjtu.edu.cn

3500108 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO. 2, MARCH/APRIL 2020

modulation [27]. A three-segment MZM has been used for114 Gb/s PAM-8 modulation [28]. Here, we show that such seg-mented modulators can also be used for phase-coded microwavesignal generation. The longer segment is driven by a codingsignal and the shorter segmented is driven by a microwavesignal. The principle is similar to the phase-coded microwavesignal generation based on a dual-drive modulator. However,compared to the dual-drive configuration, the single-drive onehas a smaller capacitance as the PN junctions in the two armsare connected in series. This greatly improves the modulationbandwidth [29]. To the best of our knowledge, this is the firsttime to realize phase-coded microwave signals using only oneintegrated modulator on the SOI platform. Besides, the precise πphase shift is independent of the amplitude of the coding signal.Both theoretical analysis and experimental demonstration areperformed to verify the concept. Microwave signals at 10 GHzand 20 GHz frequencies have been generated with phase-codeddata rates of 5 Gb/s and 10 Gb/s, respectively.

The rest of this paper is divided into the following sections.Section II presents the device structure of the segmented modula-tor. In Section III, theoretical analysis is performed for the phase-coded microwave signal generation. Section IV describes thecharacteristics of the segmented modulator. The experimentalresults are reported in Section V. Finally, we draw conclusionsin Section VI.

II. DEVICE STRUCTURE

Fig. 1(a) shows the structure of the segmented siliconMZM with two signal-drive push-pull traveling-wave electrodes(TWE). The modulator consists of two multimode interferom-eters (MMIs) as the input optical splitter and the output opticalcombiner. This segmented modulator was originally designedfor PAM-4 modulation driven by two binary electrical signalswith an identical peak-to-peak voltage. Therefore, the two seg-ments have lengths of 3 mm and 1.5 mm. Each segment isintegrated with a pair of horizontal PN junctions connected inseries. The modulator is based on an imbalanced MZI with anarm length difference of 90 μm. Thus, the operation point of themodulator can be changed by varying the input light wavelength.To facilitate RF packaging, we add 1.5-mm-long auxiliary TWEat the end of the short segment and put the GSGSG electrodepads along the chip edge. The waveguide is routed away at theend of the output MMI and the auxiliary TWE has no modulationeffect on the device.

The inset of Fig. 1(a) shows the cross-section of the mod-ulation arms. The designs of the PN junction and TWE areidentical for the two segments. The waveguide width is 500 nmand the height is 220 nm with a 90-nm-thick slab. The horizontalPN junction is formed inside the waveguide and is offset fromthe center by 190 nm towards the n-doped region. The p-typedoping concentration is ∼4 × 1017 cm−3 and the n-type dopingconcentration is ∼8 × 1017 cm−3. The concentrations of theheavily n++- and p++-doping regions are ∼1 × 1020 cm−3 toform good ohmic contacts. They are separated by 500 nm awayfrom the edge of the rib waveguide to make a compromisebetween the free-carrier absorption loss and the modulation

Fig. 1. (a) Schematic structure of the segmented Mach-Zehnder modulator.The inset shows the cross-section of the modulation arms. (b) Microscope imageof the fabricated device. (c) Scanning capacitance microscopy image of onemodulation arm. The blue color represents the n-type doping region and thebright green color represents the p-type doping region.

bandwidth. The TWE is made of two layers of aluminum metals.The metal line is optimized to have low microwave attenuationand 50 Ω characteristic impedance, and meanwhile make themicrowave effective refractive index match to the optical groupindex. The widths of metal-1 and metal-2 are 17 μm and 45 μm,respectively. The gap separations between the signal and groundmetal traces for metal-1 and metal-2 layers are 35μm and 50μm,respectively. A direct current (DC) bias line is connected to then++-doping region in the middle of the two arms of the MZI sothat the PN junctions can be reverse-biased.

The device was fabricated using CMOS-compatible processesin IME (now AMF), Singapore. The top silicon layer and theburied silicon oxide layer thicknesses are 220 nm and 2 μm,respectively. 248-nm deep ultra-violet (DUV) photolithographywas used to pattern the silicon waveguides. Fig. 1(b) shows themicroscope image of the fabricated device. The two-segmentmodulator has a footprint of 0.7 mm × 3.6 mm. Fig. 1(c)shows the scanning capacitance microscopy (SCM) image of onemodulation arm. A horizontal PN junction is clearly discernedinside the waveguide.

III. THEORETICAL ANALYSIS

A binary coding signal c(t) is applied on the longer segmentand a microwave signal with an amplitude of Vrf is applied

ZHOU et al.: PHASE-CODED MICROWAVE SIGNAL GENERATION BASED ON A SEGMENTED SILICON MZM 3500108

on the shorter segment. Assuming the electric field of the inputoptical wave is 1, then the output electric field can be written as

Eout =1

2ej[2πfct−(ϕ1+ϕ2)]

[ej(−θ1+θ2) + ej(θ1−θ2+Δθ)

](1)

where fc is the frequency of the optical carrier, ϕ1 = Vd1/Vπ1and ϕ2 = Vd2/Vπ2 are the phase shifts when the DC bias volt-ages Vd1 and Vd2 are applied on the two segments, respectively,θ1 and θ2 are the phase shifts caused by the coding signal and themicrowave signal, respectively, and Δθ is the phase differencebetween the two arms. The phase shifts are given by

θ1 = πc(t)/(2Vπ1) (2)

θ2 = πVrf cos (2πfrf t)/(2Vπ2) (3)

where frf is the frequency of the microwave signal, Vπ1 and Vπ2are the half-wave voltages of the longer and shorter segments,respectively. The generated photocurrent in the PD is

iout ∝ Eout(t)Eout(t)∗

∝ 12{1 + cos [πVrf cos (2πfrf t)/Vπ2− πc(t)/Vπ1 −Δθ]} (4)

Using Bessel expansion, we could get the photocurrent of thePD as

iout ∝ 12

⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎪⎪⎩

[J0 (θ3) + 2

∞∑m=1

(−1)mJ2m (θ3)

× cos (4mπfrf t)] cos (θ4)

−2∞∑

m=0(−1)mJ2m+1 (θ3)

× cos [(4m+ 2)πfrf t] sin (θ4) + 1

⎫⎪⎪⎪⎪⎪⎪⎪⎪⎬⎪⎪⎪⎪⎪⎪⎪⎪⎭

(5)

where Jn is the nth-order of the first Bessel function,θ3 = πVrf/Vπ2 , and θ4 = πc(t)/Vπ1 +Δθ. Neglecting the DCand the higher-order terms, we could get the output of the PD as

iout ∝ J1 (πVrf/Vπ2) cos (2πfrf t) sin [−πc(t)/Vπ1 −Δθ](6)

When Δθ = π, we have

iout ∝ J1 (πVrf/Vπ2) cos (2πfrf t) sin [πc(t)/Vπ1] (7)From (7), we observe that the amplitude of the generated

microwave signal is modulated by a sine function. When c(t)> 0, the phase of the generated microwave is “0”. When c(t) <0, the phase of the generated microwave is “π”. Therefore, thephase of the generated microwave signal changes by “π” andthe magnitude remains the same at the two levels of the codingsignal. Although the amplitude of the generated microwavesignal depends on the amplitude of the coding signal, theπ phaseshift of the generated microwave signal is independent. To get astrong phase-coded microwave signal, the peak-to-peak voltage(Vpp) of the coding signal should set close to Vπ1 . We arbitrarilychoose to apply the coding signal on the longer segment and themicrowave signal on the shorter one. The reverse should alsowork. Fig. 2 shows the simulated phase-coded microwave signaltogether with the driving microwave and coding signals.

Fig. 2. (a) Waveform of the 20 GHz microwave signal. (b) Waveform of the5 Gb/s square wave coding signal. (c) Generated 20 GHz phase-coded microwavesignal.

Fig. 3. (a, b) Transmission spectrum of modulator when the voltage is appliedon (a) the longer segment and (b) the shorter segment. (c, d) Extracted phaseshift versus bias voltage for (c) the longer segment and (d) the shorter segment.

IV. CHARACTERIZATION OF THE MODULATOR

Figs. 3(a) and 3(b) show the measured transmission spectra ofthe segmented MZM when the reverse bias voltage is applied onone arm of the longer segment and of the shorter segment, respec-tively. The fiber-to-fiber insertion loss of the segmented MZM isabout 16.5 dB at zero bias. Given the fiber-to-chip coupling lossof 3 dB/facet, the on-chip insertion loss of the modulator is about10.5 dB at zero bias. Figs. 3(c) and 3(d) show the extracted phaseshift as a function of the bias voltage. For the longer segment,the π phase shift is achieved at the −7 V bias voltage and themodulation efficiency is 1.53 V·cm at −1 V bias. For the shortersegment, Vπ is larger than 10 V and the modulation efficiencyis 1.65 V·cm at −1 V bias. The modulation efficiency (VπL) iscalculated based on the expression VπL = πVbiasL/Δϕ, where

3500108 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO. 2, MARCH/APRIL 2020

Fig. 4. (a)–(d) Small-signal electro-electro (EE) responses of the TWE in(a, c) the longer segment and (b, d) the shorter segment with the auxiliary TWE.Both transmission (EE-S21) and reflection (EE-S11) are measured. (e) and(f) Small-signal electro-optic (EO) responses of (e) the longer segment and(f) the shorter segment.

Vbias is the applied reverse bias voltage, Δϕ is the phase shiftat this voltage and L is the length of the phase shifter [30]. Weattribute the slight difference in modulation efficiency to thefabrication error-induced non-uniformity.

Next, we measured the electro-electro (EE) S-parameters ofthe TWE at various DC bias voltages using a 67-GHz vectornetwork analyzer (VNA, Keysight, N5247A). The RF signalfrom the VNA was applied to TWE through a 40-GHz ground-signal (GS) probe and the transmitted RF signal was fed backto the VNA through another GS probe. Figs. 4(a) and 4(b)show the EE-S21 curves of the first TWE (longer segment)and the second TWE (shorter segment concatenated with theauxiliary one). Both the EE-S21 curves were normalized to thestarting frequency of 100 MHz. For the first TWE, the 6.4-dBbandwidth is 23.9 GHz at−1 V bias and it increases to 29.7 GHzat −3 V bias. For the second TWE, the 6.4-dB bandwidth is24.8 GHz at −1 V bias and increases to 30.2 GHz at −3 Vbias. Because the second TWE incorporates two equal sections,the short segment in the modulator essentially has a much largerbandwidth than the measured value. Figs. 4(c) and 4(d) show theEE-S11 responses of the two TWEs. The RF reflection is below−10 dB from 100 MHz to 40 GHz, indicating reasonably goodimpedance match of the modulator TWE with the RF cable. TheEO-S21 response of the segmented modulator was measuredusing a lightwave component analyzer (LCA, Keysight,N4373C). Fig. 4(e) shows the EO-S21 responses of the longer

Fig. 5. Experimental setup for generating the phase-coding microwave sig-nal. PPG: pulse pattern generator; AMP: electrical amplifier; PC: polarizationcontroller; EDFA: erbium-doped fiber amplifier; BPF: bandpass filter; PD:photodiode; OSC: oscilloscope.

segments. The 3-dB EO bandwidth is 17.1 GHz at −1 V biasand increases to 24.7 at −3 V bias. Fig. 4(f) shows the EO-S21responses of the shorter segment. The 3-dB EO bandwidth is32.9 GHz at −1 V bias, much larger than that of the longersegment [27]. The 3-dB EO bandwidth is below the 6.4-dB EEbandwidth probably due to the mismatch between the microwaveeffective refractive index and the optical group index. The char-acteristic impedance of the TWE is not exactly equal to 50 Ω,which also slightly decreases the 3-dB EO bandwidth.

V. EXPERIMENTAL RESULTS

Fig. 5 depicts the experimental setup for the phase-coded mi-crowave signal generation. The pseudo-random binary sequence(PRBS) signal was generated from a pulse pattern generator(PPG, Keysight, N4960A and N4951B). The Vpp of the PRBSsignal was amplified to 7 V and applied to the longer segment.The microwave signal was generated from a microwave source(Anapico, APSIN20G). The Vpp of the microwave signal wasalso amplified to 7 V and applied to the shorter segment. TheDC bias voltage was applied through a DC probe to ensure themodulator to work in the carrier-depletion condition. The DCvoltages on the longer and shorter segments were both 2 V. Lightat the 1549.6 nm wavelength from a tunable laser was set tothe transverse-electric (TE) polarization before coupling to themodulator, which is the minimum optical transmission (carriersuppression) point. The modulated signal was amplified by anerbium-doped optical amplifier (EDFA) and filtered before itwas detected by a 50-GHz photodiode (PD, U2t, XPDV2120R).The output of the PD was received by a real-time oscillo-scope (Lecroy, LabMaster 10-36Zi-A) with a sampling rateof 80 GS/s.

We set the microwave frequency at 10 GHz and the codingsignal at a data rate of 5 Gb/s with a pseudo-random binarysequence (PRBS) pattern with a length of 27 − 1. Fig. 6(a)shows the measured waveform and Fig. 6(b) shows the zoom-inof the signal. The recovered phase change from the measuredsignal was obtained through the Hilbert transform as shown inFig. 6(c). It can be seen that the phase changes between “0” or“π”.

In order to verify the pulse compression ability of phasecoding, we first calculated an autocorrelation trace based on

ZHOU et al.: PHASE-CODED MICROWAVE SIGNAL GENERATION BASED ON A SEGMENTED SILICON MZM 3500108

Fig. 6. (a) Measured waveform of a 10-GHz microwave signal phase-codedby a 27 − 1 PRBS signal with a coding data rate of 5 Gb/s. (b) Zoom-in ofthe generated phase-coded signal. (c) Recovered phases from the phase-codedsignal.

the simulated signal and calculated it based on the experimen-tally recorded signal. Fig. 7(a) shows the autocorrelation of thesimulated signal. The insets are the zoom-in of the main lobe ofthe autocorrelation traces. The peak-to-side lobe ratio (PSR) is10.75 dB. The pulse compression ratio (PCR) is 127. Fig. 7(b)shows the autocorrelation of the measured signal. The PSR is10.17 dB and the full-width at half-maximum (FWHM) of themain lobe is about 0.191 ns. The RF signal length is 25.4 ns, sothe PCR is 133.0.

The microwave frequency and the coding data rate can beeasily changed by resetting the RF source and PPG. Figs. 8(a)and 8(b) show the overall and zoom-in waveforms of a 20-GHzmicrowave signal phase-coded at a data rate of 10 Gb/s. Fig. 8(c)presents the recovered phase change. We also calculated theautocorrelation of the simulated and measured signals at thisfrequency. Fig. 9(a) shows the autocorrelation of the simulatedsignal. The peak-to-side lobe ratio (PSR) is 10.73 dB. The pulsecompression ratio (PCR) is 127. Fig. 9(b) shows the autocor-relation of the measured signal. The PSR is 8.42 dB and theFWHM of the main lobe is about 0.081 ns. The RF signal lengthis 12.7 ns, so the PCR is 156.8.

Figs. 10(a) and 10(b) show the comparison of the autocor-relation main lobe between the simulated and the measuredresults. The FWHM measured from the experiment is smaller

Fig. 7. Autocorrelation traces of (a) the simulated and (b) the measured10-GHz phase-coded microwave signals at a coding data rate of 5 Gb/s. Theinsets show the zoom-in of the main lobe.

Fig. 8. (a) Measured waveform of a 20-GHz microwave signal phase-codedby 27 − 1 PRBS signals with a coding data rate of 10 Gb/s. (b) Zoom-in ofthe generated phase-coded signals. (c) Recovered phases from the phase-codedsignal.

than that from the simulation. Thus, it results in a higher pulsecompression ratio. Figs. 10(e) and 10(f) present the electricalspectra of the measured signals. There are extra frequencycomponents around 5 GHz when the coding data rate is 5 Gb/s.When the coding data rate is 10 Gb/s, the 10 GHz frequencycomponent is also present. It is possibly caused by the amplitude

3500108 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO. 2, MARCH/APRIL 2020

Fig. 9. Autocorrelation traces of (a) the simulated and (b) the measured20-GHz phase-coded microwave signals at a coding data rate of 10 Gb/s. Theinsets show the zoom-in of the main lobe.

Fig. 10. (a, b) Measured (black line) and simulated (red line) main lobes.(c, d) Filtered (black line) and simulated (red line) main lobes. (e, f) Powerdensity spectrum of the measured signal. (g, h) Power density spectrum ofthe measured signal after filtering. The left column shows the results of themicrowave frequency at 10 GHz with a coding data rate of 5 Gb/s and the rightcolumn shows the results of the microwave frequency at 20 GHz with a codingdata rate of 10 Gb/s.

Fig. 11. Experimental setup for measuring the SFDRSHD of the shortersegment.

chirp when the binary phase-shift keying (BPSK) modulation isrealized in the longer segment. Besides, due to the nonlinearityof the modulator, high-order harmonic frequency componentsof the applied microwave signal may generate. From Figs. 10(e)and 10(f), we can observe that the high-order harmonic fre-quency components are very small. The measured spurious-freedynamic range (SFDR) for the second-order harmonic distortion(SHD) of the shorter segments is SFDRSHD = 90.32 dB·Hz1/2(See Appendix). These frequency components could be elim-inated by an electrical filter or an antenna. We removed thesefrequency components by using a Butterworth filter. The powerdensity spectra after filtering are shown in Figs. 10(g) and 10(h).Then we calculated the autocorrelation of the measured signalafter filtering. The retrieved FWHM of the main lobe is about0.219 ns and 0.107 ns at the coding rates of 5 Gb/s and 10 Gb/s,respectively. The corresponding PCR is 116.0 and 118.7, closeto the simulation results. Figs. 10(c) and 10(d) compare theautocorrelation main lobe between the filtered experimentalresults and the simulation results. They agree very well witheach other. Thus, we can infer that those unwanted frequencycomponents lead to higher PCR in the unfiltered signal.

VI. CONCLUSION

We have demonstrated a novel integrated photonic approachto generate phase-coded microwave signals based on a two-segment single-drive push-pull silicon MZM. The phase codingsignal is applied to the longer segment and the microwavesignal is applied to the shorter segment. As a proof-of-principledemonstration, we successfully generate two phase-coded mi-crowave signals at 10 GHz with a coding rate of 5 Gb/s and at20 GHz with a coding rate of 10 Gb/s. The PCR of the measuredphase-coded microwave signal is 133.0 and 156.8 at 10 GHzand 20 GHz, respectively. The successful generation of a phase-coded microwave signal in an optical beam marks a significantstep forward in full integration of microwave photonic signalgenerators on a single chip and it can find potential applicationsin radar systems.

APPENDIX

Fig. 11 shows the experimental setup for measuring theSFDRSHD of the shorter segment. The microwave signal gen-erated from a microwave source (Rohde&Schwarz, SMB100A)was applied to the shorter segment. The frequency of the mi-crowave signal was 1 GHz. The input laser wavelength was set

ZHOU et al.: PHASE-CODED MICROWAVE SIGNAL GENERATION BASED ON A SEGMENTED SILICON MZM 3500108

Fig. 12. Measured SFDRSHD of the shorter segment at the quadrature opera-tion point. The dots represent the measured data and the straight lines representlinear fits.

at the quadrature point. The modulated light was amplified andfiltered before entering a 50 GHz PD. The DC voltage was set at2 V. The SFDRSHD was obtained by measuring the fundamentaland SHD components on an RF spectrum analyzer (Keysight,N9000B). Five data points were measured and linearly fitted forboth fundamental and SHD components, as shown in Fig. 12.The fundamental and the SHD has linear and quadratic responsesto the input RF power with a slope of 0.998 and 2.005, respec-tively. The noise floor of the RF spectrum analyzer is -163 dBm.The SFDRSHD is hence 90.32 dB·Hz1/2.

REFERENCES

[1] M. I. Skolnik, Introduction to Radar System, 3rd ed. New York, NY, USA:McGraw-Hill, 2001, pp. 349–354.

[2] Z. Li, W. Li, H. Chi, X. Zhang, and J. Yao, “Photonic generation of phase-coded microwave signal with large frequency tunability,” IEEE Photon.Technol. Lett., vol. 23, no. 11, pp. 712–714, Jun. 2011.

[3] H. Y. Jiang et al., “Photonic generation of phase-coded microwave signalswith tunable carrier frequency,” Opt. Lett., vol. 38, no. 8, pp. 1361–1363,Apr. 2013.

[4] W. Chen et al., “Photonic generation of binary and quaternary phase-coded microwave waveforms with frequency quadrupling,” IEEE Photon.J., vol. 8, no. 2, Apr. 2016, Art. no. 5500808.

[5] C. Wang and J. Yao, “Phase-coded millimeter-wave waveform generationusing a spatially discrete chirped fiber Bragg grating,” IEEE Photon.Technol. Lett., vol. 24, no. 17, pp. 1493–1495, Sep. 2012.

[6] Z. Li, M. Li, H. Chi, X. Zhang, and J. Yao, “Photonic generation ofphase-coded millimeter-wave signal with large frequency tunability usinga polarization-maintaining fiber bragg grating,” IEEE Microw. WirelessCompon. Lett., vol. 21, no. 12, pp. 694–696, Dec. 2011.

[7] W. Li, L. X. Wang, M. Li, H. Wang, and N. H. Zhu, “Photonic generationof binary phase-coded microwave signals with large frequency tunabilityusing a dual-parallel Mach–Zehnder modulator,” IEEE Photon. J., vol. 5,no. 4, pp. 5501507–5501507, Aug. 2013.

[8] F. Zhang, X. Ge, B. Gao, and S. Pan, “Phase-coded microwave signalgeneration based on a single electro-optical modulator and its applica-tion in accurate distance measurement,” Opt. Express, vol. 23, no. 17,pp. 21867–21874, Aug. 2015.

[9] S. Zhu et al., “Photonic generation of background-free binary phase-codedmicrowave pulses,” Opt. Lett., vol. 44, no. 1, pp. 94–97, Jan. 2019.

[10] Z. Tang, T. Zhang, F. Zhang, and S. Pan, “Photonic generation of aphase-coded microwave signal based on a single dual-drive Mach-Zehndermodulator,” Opt. Lett., vol. 38, no. 24, pp. 5365–5368, Dec. 2013.

[11] W. Li, W. T. Wang, W. H. Sun, L. X. Wang, and N. H. Zhu, “Photonicgeneration of arbitrarily phase-modulated microwave signals based on asingle DDMZM,” Opt. Express, vol. 22, no. 7, pp. 7446–7457, Apr. 2014.

[12] W. Li, W. T. Wang, W. H. Sun, and N. H. Zhu, “All-optical generation ofbinary phase-coded microwave signal based on cross-polarization modu-lation in a highly nonlinear fiber,” Opt. Lett., vol. 39, no. 6, pp. 1561–1564,Mar. 2014.

[13] D. Marpaung, J. Yao, and J. Capmany, “Integrated microwave photonics,”Nature Photon., vol. 13, no. 2, pp. 80–90, Feb. 2019.

[14] M. Smit et al., “An introduction to InP-based generic integration technol-ogy,” Semicond. Sci. Technol., vol. 29, no. 8, Jun. 2014, Art. no. 083001.

[15] D. J. Moss, R. Morandotti, A. L. Gaeta, and M. Lipson, “New CMOS-compatible platforms based on silicon nitride and Hydex for nonlinearoptics,” Nature Photon., vol. 7, pp. 597–607, Jul. 2013.

[16] D. Thomson et al., “Roadmap on silicon photonics,” J. Opt., vol. 18, no. 7,Jun. 2016, Art. no. 073003.

[17] G. T. Reed, G. Mashanovich, F. Y. Gardes, and D. J. Thomson, “Siliconoptical modulators,” Nature Photon., vol. 4, pp. 518–526, Jul. 2010.

[18] X. Li et al., “Highly efficient silicon Michelson interferometer modula-tors,” IEEE Photon. Technol. Lett., vol. 25, no. 5, pp. 407–409, Mar. 2013.

[19] D. Patel et al., “Design, analysis, and transmission system performanceof a 41 GHz silicon photonic modulator,” Opt. Express, vol. 23, no. 11,pp. 14263–14287, Jun. 2015.

[20] Y. Zhou, L. Zhou, F. Su, X. Li, and J. Chen, “Linearity measurement andpulse amplitude modulation in a silicon single-drive push–pull Mach–Zehnder modulator,” J. Lightw. Technol., vol. 34, no. 14, pp. 3323–3329,Jul. 2016.

[21] M. Li, L. Wang, X. Li, X. Xiao, and S. Yu, “Silicon intensity Mach–Zehnder modulator for single lane 100-Gb/s applications,” Photon. Res.,vol. 6, no. 2, pp. 109–116, Feb. 2018.

[22] J. Sun et al., “A 128 Gb/s PAM4 silicon microring modulator withintegrated thermo-optic resonance tuning,” J. Lightw. Technol., vol. 37,no. 1, pp. 110–115, Jan. 2019.

[23] W. Zhang and J. Yao, “On-chip silicon photonic integrated frequency-tunable bandpass microwave photonic filter,” Opt. Lett., vol. 43, no. 15,pp. 3622–3625, Aug. 2018.

[24] I. Demirtzioglou et al., “Frequency comb generation in a silicon ringresonator modulator,” Opt. Express, vol. 26, no. 2, pp. 790–796, Jan. 2018.

[25] Y. Zhong et al., “Microwave frequency upconversion employing acoupling-modulated ring resonator,” Photon. Res., vol. 5, no. 6, pp. 689–694, Dec. 2017.

[26] Y. Xie, L. Zhuang, and A. J. Lowery, “Silicon microring modulator-basedRF mixer for millimeter-wave phase-coded signal generation,” Opt. Lett.,vol. 42, no. 14, pp. 2742–2745, Jul. 2017.

[27] A. Samani et al., “Experimental parametric study of 128 Gb/s PAM-4transmission system using a multi-electrode silicon photonic Mach Zehn-der modulator,” Opt. Express, vol. 25, no. 12, pp. 13252–13262, Jun. 2017.

[28] A. D. Simard, B. Filion, D. Patel, D. Plant, and S. LaRochelle, “Segmentedsilicon MZM for PAM-8 transmissions at 114 Gb/s with binary signaling,”Opt. Express, vol. 24, no. 17, pp. 19467–19472, Aug. 2016.

[29] L. Chen, C. R. Doerr, P. Dong, and Y.-K. Chen, “Monolithic silicon chipwith 10 modulator channels at 25 Gbps and 100-GHz spacing,” Opt.Express, vol. 19, no. 26, pp. B946–B951, Dec. 2011.

[30] D. Patel et al., “High-speed compact silicon photonic Michelson inter-ferometric modulator,” Opt. Express, vol. 22, no. 22, pp. 26788–26802,Nov. 2014.

Gangqiang Zhou received the B.S. degree from the School of Physics andOptoelectronic Engineering, Xidian University, Xian, China, in 2016. He iscurrently working toward the Ph.D. degree with the Department of Electronic En-gineering, State Key Laboratory of Advanced Optical Communication Systemsand Networks, Shanghai Jiao Tong University, Shanghai, China. His researchinterests include the design of silicon modulators and integrated microwavephotonics.

Linjie Zhou received the B.S. degree in microelectronics from Peking Univer-sity, Beijing, China, in 2003, and the Ph.D. degree in electronic and computerengineering from the Hong Kong University of Science and Technology, HongKong, in 2007. From 2007 to 2009, he was a Postdoctoral Researcher with theUniversity of California, Davis, Davis, CA, USA. He is currently a Professorwith the State Key Laboratory of Advanced Optical Communication Systemsand Networks, Shanghai Jiao Tong University, Shanghai, China. His researchinterests include silicon photonics, photonics integration, and plasmonic wave-guide devices.

3500108 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO. 2, MARCH/APRIL 2020

Liangjun Lu received the B.S. degree from the Department of Optical Engi-neering, Zhejiang University, Hangzhou, China, in 2011, the Ph.D. degree inelectronic engineering from Shanghai Jiao Tong University, Shanghai, China.He is currently a tenure-track Assistant Professor with the State Key Laboratoryof Advanced Optical Communication Systems and Networks, Shanghai JiaoTong University, Shanghai, China. His research interests include silicon opticalswitches and silicon programmable optical processors.

Yuyao Guo received the B.S. degree from the School of Renewable Energy,North China Electric Power University, Beijing, China, in 2014, and the M.S.degree from the School of Mechanical Engineering, Beijing Institute of Tech-nology, Beijing, in 2017. He is currently working toward the Ph.D. degreewith the State Key Laboratory of Advanced Optical Communication Systemsand Networks, Department of Electronic Engineering, Shanghai Jiao TongUniversity, Shanghai, China. His research interests include photonics packagingand silicon hybrid laser.

Jianping Chen received the B.S. degree from Zhejiang University, Hangzhou,China, in 1983, and the M.S. and Ph.D. degrees from Shanghai Jiao TongUniversity, Shanghai, China, in 1986 and 1992, respectively. He is currently aProfessor with the Department of Electronic Engineering, State Key Laboratoryof Advanced Optical Communication Systems and Networks, Shanghai JiaoTong University, Shanghai, China. His main research interests include photonicdevices and signal processing, optical networking, and sensing optics. He is alsoa Principal Scientist of the 973 projects in China.

/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages false /GrayImageDownsampleType /Bicubic /GrayImageResolution 1200 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.00083 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages false /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages false /MonoImageDownsampleType /Bicubic /MonoImageResolution 1600 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.00063 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/CreateJDFFile false /Description >>> setdistillerparams> setpagedevice