Flat-top bandpass microwave photonic filter withtunable bandwidth and center frequency basedon a Fabry–Pérot semiconductor optical amplifierFAN JIANG,1 YUAN YU,1,2 TONG CAO,1 HAITAO TANG,1 JIANJI DONG,1 AND XINLIANG ZHANG1,3

1Wuhan National Laboratory for Optoelectronics & School of Optical and Electronic Information, Huazhong University of Science and Technology,Wuhan 430074, China2e-mail: yuan_yu@mail.hust.edu.cn3e-mail: xlzhang@mail.hust.edu.cn

Received 29 January 2016; revised 12 June 2016; accepted 16 June 2016; posted 17 June 2016 (Doc. ID 258526); published 14 July 2016

We propose a flat-top bandpass microwave photonic filter(MPF) with flexible tunability of the bandwidth and centerfrequency based on optical nonlinearities in a Fabry–Pérotsemiconductor optical amplifier (FP-SOA). Phase-invertedmodulation induced by cross-gain modulation (XGM) andoptical spectral broadening induced by self-phase modula-tion (SPM) are exploited to achieve flat-top and bandwidthtuning, respectively. Wideband and continuous tuning ofthe center frequency is achieved by altering the bias currentof the FP-SOA. Experimental results demonstrate a flat-topsingle-passband MPF with its center frequency tunablefrom 6.0 to 18.3 GHz by adjusting the bias current from54.05 to 107.85 mA. The 3-dB bandwidth of the passbandwhen centered at 10.0 GHz is shown to be variable from680 to 1.43 GHz, by increasing the injected optical powerfrom −1 to �5 dBm. During the bandwidth tuning, theamplitude ripple within the passband is maintained at lessthan �0.5 dB. Excellent main to secondary sidelobe ratioexceeding 45 dB is achieved when the MPF is centered at18.3 GHz. © 2016 Optical Society of America

OCIS codes: (060.5625) Radio frequency photonics; (250.5980)

Semiconductor optical amplifiers; (190.4360) Nonlinear optics,

devices.

http://dx.doi.org/10.1364/OL.41.003301

Advantageous features inherent to photonics such as low loss,high bandwidth, immunity to electromagnetic interference(EMI), tunability, and reconfigurability have aroused increasinginterest in the microwave photonic filter (MPF) as a versatilesubstitute for its electronic counterpart in both broadbandwireless access networks and modern radar systems [1,2]. Asthe bandpass MPF with flat-top can maintain the fidelity ofthe processed radiofrequency (RF) signal, approaches to achieveMPFs with flat passband have been proposed. Superimposingtwo RF bandpass responses [3,4] has been proved to be a con-venient option, while the tunability is relatively limited.Directly transforming the flatness of the optical filter to the

obtained RF filter is verified as feasible as well [5,6], with effortsdevoted to finely engineering the fiber Bragg gratings (FBGs).Elaborately programming the tap coefficients of a multi-tapfinite impulse response (FIR) MPF can also construct thedesired flat-top passband [7–11]. However, in this case the sys-tem would get increasingly complex with increased taps.Optical nonlinearity is expected to be a good candidate to real-ize agile control over the obtained MPF passband [1].Stimulated Brillouin scattering (SBS) in fiber has been pro-posed to implement an ultra-flat rectangular MPF by digitallycontrolling the multiple frequency components in the pump[12]. Since a long span of fiber is used as the gain medium,making such a system compact is not easy.

In this Letter, we propose and demonstrate a flat-top band-pass MPF with flexible tunability of the bandwidth and thecenter frequency based on nonlinear effects in a Fabry–Pérotsemiconductor optical amplifier (FP-SOA). Linear amplifica-tion by the FP-SOA has been used to realize the phasemodulation to intensity modulation (PM-IM) conversion inour previously proposed single-passband MPF scheme [13].Furthermore, the resonant gain of an SOA assisted with distrib-uted feedback (DFB) grating has been used to selectively amplifythe optical signal [14] and further employed to generate anMPF[15]. In these three Letters [13–15], only selective linear ampli-fication process was exploited. However, in this proposedscheme, besides the mapping technique by selectively amplify-ing the optical sidebands, the self-phase modulation (SPM) ofFP-SOA is used to control MPF bandwidth while the cross-gainmodulation (XGM) is used to realize flat-top passband. Here bydriving the FP-SOA into saturation with strong enough incidentoptical power, the XGM effect is intentionally aroused [16].Then an inverted copy of the input modulated signal is directlyapplied onto the output comb spectrum of the FP-SOA. Afterphotodetection an extra amplitude-scaled and phase-invertedRF response [17] is thus generated in addition to the RF pass-band generated by the inputmodulated signal. A flat-topMPF istherefore obtained after the subtraction between the two anti-phase RF passbands. Meanwhile, optical spectral broadeninginduced by the SPM effect [18] is employed to adjust the

Letter Vol. 41, No. 14 / July 15 2016 / Optics Letters 3301

0146-9592/16/143301-04 Journal © 2016 Optical Society of America

bandwidth of the obtained RF response by controlling theinjected optical power. The tuning of the center frequency isrealized by altering the bias current. In the experiment, aflat-top single-passband MPF with its center frequency tunablefrom 6.0 to 18.3 GHz is demonstrated. The 3-dB bandwidth ofthe MPF when centered at 10.0 GHz is variable from 680 to1.43 GHz by increasing the optical power injected into theFP-SOA from −1 to�5 dBm. Excellentmain to secondary side-lobe ratio (MSSR) exceeding 45 dB is obtained when the centerfrequency of the MPF is tuned to half of the free spectral range(FSR) of the FP-SOA, for instance, 18.3 GHz.

The operation principle is schematically illustrated in Fig. 1.Figure 1(a) displays the formation principle of the bandpassMPF based on phase modulation and an FP-SOA. The inputphase-modulated signal and the characteristic comb gain spec-trum of the FP-SOA [19] are represented by the green-arrowedlines and the magenta comb line, respectively. The frequencyseparation of the gain spectral peaks equals the FSR of theFP-SOA, which is determined by cavity length. When locatednear one spectral peak of the FP-SOA, the phase-modulatedsignal will experience the PM-IM process, and a bandpassRF filter can be achieved [13]. The center frequency of the filteris equal to the frequency interval between the carrier andthe nearby spectral peak. Therefore, the center frequency ofthe MPF can be tuned by shifting the gain spectrum of theFP-SOA or by altering the carrier wavelength.

Figure 1(b) shows the bandwidth adjustment of the MPF byexploiting the SPM effect in the FP-SOA. The bandwidth ofthe MPF is determined by the linewidth of the laser source andthe spectral width of the nearby gain spectral peak of the FP-SOA. If a continuous wave (CW) laser with narrow linewidth isused as the light source, the bandwidth of the MPF is almostdetermined by the nearby gain spectral peak of the FP-SOA.However, when the optical power injected into the FP-SOAis strong enough, carrier consumption-induced nonlinearitieswill become significant, and the spectrum of the input signalwill be broadened due to the SPM effect. Stronger SPM can begenerated by increased input signal power and leads to larger

spectral broadening. The broadened optical spectrum (repre-sented by the green curve in Fig. 1(b)) will induce MPF band-width to increase (represented by the red curve). Therefore, thebandwidth of the MPF can be controlled simply by adjustingthe input signal power.

The XGM effect is used to flatten the microwave passband,as shown in Fig. 1(c). When the input intensity modulatedpower is high enough to drive the SOA into saturation, thegain of the SOA will be modulated by the input signal [16].Therefore when the FP-SOA is saturated by the PM-IM con-verted signal, the obtained output comb spectrum of the FP-SOA will be intensity modulated as well. After photodetection,the counter-phase-modulated comb spectrum of the FP-SOAwill generate an extra RF response (represented by the blue-dashed curve in Fig. 1(c)), which is weaker in amplitudeand inverted in phase compared to the RF response generatedby the input signal (represented by the black-dashed curve).After subtraction, the top part of the RF response inducedby the PM-IM conversion will be rightly offset by theXGM-induced response. Thus, the passband of the obtainedMPF can be flattened as indicated by the red solid curve.

The effectiveness of the proposed MPF is experimentallyverified with the setup illustrated in Fig. 2. A CW light emittedfrom a laser source (LS, Koheras BasiK E15) is launched into aphase modulator (PM) and phase-modulated by the RF signalemitted by a vector network analyzer (VNA). A polarizationcontroller (PC1) is used to adjust the polarization state of thesignal injected into the PM. After power adjusted by a variableoptical attenuator (VOA), the modulated signal is injected intothe FP-SOA through an optical circulator (CIR). PC2 is placedbefore the FP-SOA for polarization adjustment. The phase-modulated signal is then processed by the FP-SOA and launchedinto a photodetector (PD) with large bandwidth via the CIR.The frequency response of the MPF is measured by theVNA (AnritsuMS4647B) with a typical intermediate frequencybandwidth (IFBW) of 1 kHz being set. The fundament func-tional element within the setup is the FP-SOA, a customizedproduct from INPHENIX. For both the input and output sides,the facet reflectivity is 30%, and the maximum output power isabout 10 dBm. With a multiquantum-well active region and acavity length of 1200 μm, the FP-SOA shows an FSR of36.6 GHz. To ensure stable performance of the FP-SOA, biasand temperature controllers with high stability are preferred.

The output optical spectrum from the FP-SOA is character-ized via an optical spectrum analyzer (Yokogawa AQ6370C)with the resolution bandwidth (RBW) set at 0.02 nm.Figure 3 demonstrates the optical spectra at the output ofthe FP-SOA and the corresponding MPF responses when thelaser wavelength is closer to one gain spectral peak of theFP-SOA and exactly in the middle of two gain spectral peaks,

(a) PM-IM Conversion by Unbalanced Gain

(b) Spectral Broadening by SPM Effect

(c) Anti-phase Modulation by XGM Effect

RF Frequency

Tunable Center

RF Frequency

Tunable Bandwidth

RF Frequency

Flattened Top

Phase-Modulated Input Signal

Φ=π

Φ=0

ωRFFSR

Intensity ModulatedFP-SOA Output Comb Spectrum

Spectrum Broadened

FP-SOA Gain Spectrum

ωRF

ωRFωRF

Fig. 1. Schematic illustrations of (a) the forming of the MPF,(b) the broadening of the bandwidth, and (c) the flattening of the pass-band top.

LS

CIRPC1

PC2

VNA

VOAPM

FP-SOA

PD1

2

3

Fig. 2. Experimental setup of the proposed microwave photonicfilter.

3302 Vol. 41, No. 14 / July 15 2016 / Optics Letters Letter

respectively. In the former case, two RF passbands will be gen-erated within the 40-GHz operation bandwidth of the PD. Thedominant passband is contributed by the carrier signal and thenearest spectral peak, and the secondary passband is contrib-uted by the carrier signal and the next nearest spectral peak.In the latter case, a single RF passband centered at FSR/2 willbe obtained, as now the RF passband formed by the carrier andthe left nearest spectral peak is overlapped with that formed bythe carrier signal and the right nearest spectral peak. In this case,improved MSSR can be expected from the subtraction effectbetween the two overlapped passbands [13]. During the experi-ment, the temperature of the FP-SOA is set at 25°C, and thewavelength of the optical carrier is fixed at 1549.73 nm. Theseparation between the optical carrier and the gain spectralpeak, for instance, the center frequency of the MPF, is adjustedby altering the bias current of the FP-SOA, as the gain spec-trum of the FP-SOAwill be red-shifted when the bias current isincreased [13]. The power of the input phase modulated signalis adjusted by the VOA to drive the FP-SOA into differentstates.

In Fig. 3(a), the optical spectra output from the FP-SOA isshown and can be used to analyze the nonlinear effects in theFP-SOA. For Fig. 3(a), the bias current of the FP-SOA is69.86 mA, and the nearest FP-SOA comb spectral peak tothe optical carrier is spaced 0.08 nm (i.e., 10.0 GHz) awayon the right. To obtain the demonstrated optical spectra,the input optical carrier is phase-modulated by a 10.0-GHzRF signal prior to injecting into the FP-SOA. These spectralpeaks with a spacing of one FSR are the output comb spectralpeaks of the FP-SOA. When the input optical power is−20 dBm, the input signal only experiences linear amplifica-tion by the FP-SOA (seen from the black-dashed curve).

However, when the input power is increased to �5 dBm,the FP-SOA is driven into gain saturation state. As can be seenfrom the spectrum in the magenta solid curve, most of thecomb peaks are obviously suppressed due to the increased car-rier depletion induced by the increased input optical power.Now the strongest comb spectral peak is the one that locatesin alignment with the right modulation sideband of the inputsignal. Resulting from the XGM effect, these comb spectralpeaks are intensity modulated as indicated by their modulationsidebands. Figure 3(b) demonstrates the corresponding MPFresponses. As analyzed in Fig. 1(c), a flat-top bandpass MPFcentered at 10.0 GHz is obtained with the participation ofthe inversely modulated comb spectral peaks (in the red solidcurve). According to the inset, only �0.5 dB amplitude rippleis within the 3-dB bandwidth of 1.43 GHz. TheMPF shows anMSSR of 20 dB. The secondary sidelobe centered at 26.6 GHzis generated by the carrier signal and the next nearest combpeak on the left. As predicted in Fig. 1, by increasing the opticalinput power and stimulating proper nonlinear effects in theFP-SOA, the passband top of the MPF is flattened and thebandwidth is broadened compared to the case of linear ampli-fication (the blue-dashed curve).

For Fig. 3(c), the bias current of the FP-SOA is 107.52 mA.Now the injected optical carrier locates exactly in the middle oftwo adjacent comb spectral peaks and is phase-modulated byan 18.3-GHz RF signal. When the input optical power isincreased to �5 dBm, the modulation sidebands of the combspectral peaks induced by the XGM effect can be clearly seenfrom spectrum in the magenta solid curve. The correspondingMPF response is shown in Fig. 3(d). Since the passband formedby the carrier and the left nearest spectral peak is now over-lapped with the passband formed by the carrier signal andthe right-nearest spectral peak, a single passband centered at18.3 GHz is obtained with an improved MSSR exceeding45 dB. According to the inset, the amplitude ripple is about�1.5 dB within the 3-dB bandwidth of 1.28 GHz. The flat-ness of the passband is deteriorated since now the MPF pass-band is concurrently determined by two adjacent comb spectralpeaks. When the carrier signal is slightly deviated from thecenter, a flat microwave passband centered at FSR/2 can alsobe obtained [4].

Figure 4 demonstrates the bandwidth tunability based onthe SPM-induced optical spectral broadening. Keeping the biascurrent of the FP-SOA the same as that in Fig. 3(b), we alter theinput signal power from −1 to �5 dBm with a step size of1 dBm. The corresponding MPF responses are shown inFig. 4(a). It can be observed that when the optical power isincreased from −1 to �5 dBm, the 3-dB bandwidth of the

Fig. 3. FP-SOA output optical spectra and corresponding MPFresponses: (a) optical spectra when the laser wavelength is located0.08 nm away from the nearest FP-SOA comb spectral peak, RBWis 0.02 nm; (b) the normalized amplitude responses of the MPF whenit is centered at 10.0 GHz, inset shows detail view of the passband top;(c) optical spectra when the laser wavelength is aligned in the middle oftwo adjacent FP-SOA comb spectral peaks, RBW is 0.02 nm; (d) thenormalized amplitude responses of the MPF when it is centered at18.3 GHz, inset shows detail view of the passband top.

Fig. 4. Bandwidth tuning by the input optical power: (a) the MPFresponses when centered at 10.0 GHz; (b) the MPF responses whencentered at 18.3 GHz.

Letter Vol. 41, No. 14 / July 15 2016 / Optics Letters 3303

flat-top MPF is linearly increased from 680 to 1.43 GHz,correspondingly. An averaged bandwidth tuning coefficient of126 MHz/dBm is obtained. Then we repeat the power tuningprocess when the MPF is centered at 18.3 GHz and plot themeasured results in Fig. 4(b). Now the 3-dB bandwidth ofthe flat-topMPF is increased from 620 to 1.28 GHz, which cor-responds to a slightly different bandwidth tuning coefficient of113MHz/dBm. Considering that the input power variation alsoresults in the spectral drift of the FP-SOA and thus the centerfrequency drift of the MPF, proper adjustment of the bias cur-rent is necessary to compensate for the power-induced driftingduring bandwidth tuning. Note that the input optical powershould be neither too high nor too low. Too-low input opticalpower cannot arouse strong enough XGM to flatten the top oftheMPF, while too-high input optical power will deteriorate thestability of theMPFbecause of the possible injection locking andhigh-order modulation sideband. In the experiment, the inputoptical power is controlled between −1 and �5 dBm so thatwithin the whole frequency tuning range the MPF can haveflat-top, tunable bandwidth and stable frequency response.

Continuous tunability of the center frequency for the pro-posed MPF is shown in Fig. 5(a). The center frequency of theMPF is adjusted at 7.0, 7.5, 8.0, 8.5, and 9.0 GHz whenthe bias current is set at 59.60, 61.69, 64.73, 67.38, and69.16 mA, respectively. With the incident optical power keptat −1 dBm, a center frequency tuning coefficient of 200 MHz/mA is obtained. The stability of the center frequency is deter-mined by the stability of laser wavelength as well as the stabilityof the gain spectrum of the FP-SOA, because the relative wave-length drift between the carrier wavelength and the gain spec-trum of FP-SOA will result in the spectral drift of the MPF.Owing to the high stability of the employed laser sourceand the bias and temperature controllers for the FP-SOA,no obvious drifting of the frequency response is observed.During the tuning process, the 3-dB bandwidth of about1 GHz and the amplitude ripple of about �0.5 dB are main-tained. The wideband tunability of the proposed MPF is dem-onstrated in Fig. 5(b). By adjusting the bias current from 54.05to 107.85 mA, the center of the MPF is tuned from 6.0 to18.3 GHz. The lower limit of the center frequency of theMPF (i.e., the minimum frequency separation between the op-tical carrier and the nearest comb spectral peak of the FP-SOA)is observed to be 6.0 GHz in the experiment. When the opticalcarrier gets too close to the resonant comb peak of the FP-SOA,the FP-SOA easily gets injection locked to the optical carrierand no desired MPF will be achieved. The upper frequencylimit (i.e., 18.3 GHz) equals to half of the FSR of the FP-SOA.

In conclusion, we have proposed and demonstrated a flat-top MPF with flexible tunability of the bandwidth and thecenter frequency based on an FP-SOA. Phase-inverted XGMon the output comb spectrum of the FP-SOA is exploitedto flatten the passband top of the MPF. Convenient bandwidthtuning is realized by controlling the SPM-induced optical spec-tral broadening through the input optical power. This methodprovides a novel approach to realize bandwidth tuning by con-trolling the intensity of phase modulation, which can easily becontrolled in various nonlinear media. Moreover, by shiftingthe gain spectrum of the FP-SOA by adjusting the bias current,wideband and continuous tuning of the center frequency isrealized. To construct the proposed MPF with agile tuningof the passband, only a single CW laser source, a singlePM, a single FP-SOA, and a single PD are demanded. Witha simple structure a monolithically integrated version of thesystem can be implemented based on the InP/InGaAsPplatform.

Funding. NSFCMajor International Joint Research Project(61320106016); National Science Fund for DistinguishedYoung Scholars (61125501); Science Fund for CreativeResearch Groups (2014CFA004); National Natural ScienceFoundation of China (NSFC) (61501194); Natural ScienceFoundation of Hubei Province (2015CFB231).

REFERENCES

1. J. Capmany, B. Ortega, and D. Pastor, J. Lightwave Technol. 24, 201(2006).

2. J. Yao, IEEE Microwave Mag. 16(8), 46 (2015).3. L. Zhuang, C. G. H. Roeloffzen, M. Hoekman, K.-J. Boller, and A. J.

Lowery, Optica 2, 854 (2015).4. Y. Yu, J. Dong, B. Luo, F. Jiang, and X. Zhang, Opt. Commun. 286, 95

(2013).5. T. Chen, X. Yi, L. Li, and R. Minasian, Opt. Lett. 37, 4699 (2012).6. G. Liang, C. Xiangfei, and Y. Jianping, IEEE Microw. Wireless.

Compon. Lett. 23, 362 (2013).7. B. Vidal, J. L. Corral, and J. Martí, Opt. Express 14, 581 (2006).8. X. Yi and R. A. Minasian, J. Lightwave Technol. 26, 2578 (2008).9. D. Yitang and Y. Jianping, IEEE Trans. Microw. Theory Tech. 58,

3279 (2010).10. S. Minhyup, C. M. Long, W. Rui, S. DongSun, D. E. Leaird, and A. M.

Weiner, IEEE Photon. Technol. Lett. 23, 1618 (2011).11. X. Xue, X. Zheng, H. Zhang, and B. Zhou, Opt. Express 20, 26929

(2012).12. L. Yi, W. Wei, Y. Jaouen, and W. Hu, “Ideal rectangular microwave

photonic filter with high selectivity based on stimulated brillouin scat-tering,” in Optical Fiber Communication Conference, Los Angeles,California (2015), paper Tu3F.5

13. Y. Yu, J. Dong, E. Xu, X. Li, L. Zhou, F. Wang, and X. Zhang,J. Lightwave Technol. 29, 3542 (2011).

14. J. Xiong, R. Wang, T. Pu, L. Lu, T. Fang, Z. Wei, G. Sun, J.Zheng, and P. Xiang, IEEE Photon. Technol. Lett. 25, 1114(2013).

15. Y. Deng, M. Li, J. Tang, S. Sun, N. Huang, and N. Zhu, IEEE Photon.J. 7, 1 (2015).

16. T. Durhuus, B. Mikkelsen, C. Joergensen, S. L. Danielsen, and K. E.Stubkjaer, J. Lightwave Technol. 14, 942 (1996).

17. F. Coppinger, S. Yegnanarayanan, P. D. Trinh, and B. Jalali, IEEETrans. Microw. Theory Tech. 45, 1473 (1997).

18. G. P. Agrawal and N. A. Olsson, IEEE J. Quantum Electron. 25, 2297(1989).

19. Y. Yamamoto, IEEE J. Quantum Electron. 16, 1047 (1980).

Fig. 5. Tuning of the filter center frequency by adjusting the biascurrent of the FP-SOA: (a) continuous tuning of the center frequencyfrom 7.0 to 9.0 GHz; (b) wideband tuning of the center frequencyfrom 6.0 to 18.3 GHz.

3304 Vol. 41, No. 14 / July 15 2016 / Optics Letters Letter