We report the design, fabrication and characterization of an integrated frequency discriminator on InP technology for microwave photonic phase modulated links. The optical chip is, to the best of our knowledge, the first reported in an active platform and the first to include the optical detectors. The discriminator, designed as a linear filter in intensity, features preliminary SFDR values the range between 67 and 79 dB.Hz2/3 for signal frequencies in the range of 5-9 GHz limited, in principle, by the high value of the optical losses arising from the use of several free space coupling devices in our experimental setup. As discussed, these losses can be readily reduced by the use of integrated spot-size converters improving the SFDR by 17.3 dB (84-96 dB.Hz2/3). Further increase up to a range of (104-116 dB.Hz2/3) is possible by reducing the system noise eliminating the EDFA employed in the setup and using a commercially available laser source providing higher output power and lower relative intensity noise. Other paths for improvement requiring a filter redesign to be linear in the optical field are also discussed.
©2013 Optical Society of America
Microwave photonics, a broad and multidisciplinary research discipline dealing with the generation, transmission and processing of microwave signals in the optical domain, has been the subject of considerable research eﬀort during the past few decades . In particular, and thanks to the outstanding properties of optical fibers as a transmission medium, analog optical links have been actively explored over the years as a promising technology to transmit and distribute RF/microwave signals in a wide variety of scenarios, including CATV systems, antenna remoting and RADAR applications. However, the high end-to-end RF losses and device non-linearities, which arise from both low modulation eﬃciencies of typically available commercial devices and intrinsic non-linear behavior of modulators and photodetectors, persist as the main limitations in the performance of traditional intensity-modulated direct detection (IM-DD) microwave photonic links (MWPLs). Thus, much research has been focused towards improving figures of merit of MWPLs at the system level using diﬀerent techniques .
As an alternative route for improvement, phase modulated direct detection (PM-DD) RoF links have been proposed in order to achieve lower distortion and higher linearity as compared to non-linearized IM-DD links for the transport of high-speed wireless signals. In PM-DD MWPL a phase-modulated signal is converted to intensity modulation (PM-IM conversion) using an optical discriminator, thereby allowing a simple direct detection scheme. The interest in such a scheme stems from two reasons; first, a phase modulator (for example a conventional lithium niobate PM) can provide high linearity  and its operation does not require biasing. The second reason is that there is an additional degree of freedom in tailoring the characteristic of the optical filter discriminator to enhance the RoF link performance. Thus, in such an approach, the photonic discriminator is designed for increasing the link linearity and/or suppressing the noise in the RoF link.
Experimental demonstrations using diﬀerent discrete component filter architectures and technologies have been reported, including Mach-Zehnder interferometers (MZI) [3–6] and Fiber Bragg Gratings (FBGs) . However, the MZI solution suffers from large nonlinearities [3,4], while the FBG based configurations require circulators and lack of programmability. In both cases, the resulting architectures are bulky and thus prevent the possible implementation of a compact discriminator.
To overcome many of the limitations outlined above, integrated optics solutions have also been proposed. Ring loaded Mach Zehnder structures were first proposed by Xie et al , while Wyrwas and Wu [9,10] theoretically proposed the use of finite impulse response (FIR) photonic integrated circuits (PICs) in order to improve the limitations on the performance of MZI discriminators due to nonlinearities. The roadmap towards experimental demonstration has started following different technology platforms. In  researchers from the University of Twente have reported the design, fabrication and characterization of a photonic chip frequency discriminator using the Si3N4 based TriPleXTM waveguide technology manufactured by CMOS compatible fabrication equipment. The photonic discriminator consisted of five optical ring resonators (ORRs), fully programmable using thermo-optical tuning. The chip was employed in a balanced photonic link simultaneously featuring minimum second and third order SDFR values at one bias point. Despite the fact that the obtained values for the dynamic range, 90 dB.Hz2/3 are still far from the 113 dB.Hz2/3 predicted for the specific discriminator design, there is a clear understanding that the path leading to improvement passes through reducing the waveguide propagation loss to below 0.2 dB/cm and using spot-size converters to reduce the coupling loss to below 0.5 dB/facet. Work on Silica on Silicon discriminators has been recently published [12,13] which reports the successful operation of both FIR and infinite impulse response (IIR) discriminators. In particular, a cascaded MZI FIR lattice filter and a ring assisted MZI (RAMZI) IIR filter. For both types of discriminators, a > 6 dB improvement in the link’s third order output intercept point (OIP3) over an IM-DD MZM link has been demonstrated.
InP based integrated optics technologies have not been yet explored for the implementation of FM-DD discriminators, to the best of our knowledge. The interest behind this option resides on one hand in that it allows the easy integration of photodiodes (and thus the implementation of a complete FM-DD receiver) and secondly that it allows optical amplification that can be employed to combat losses. In this paper, we report the design, fabrication and experimental characterization of the first photonic integrated circuit working as a full FM-DD receiver using a readily available InP generic technology platform .The chip is based on a conveniently cascaded structure of ring-assisted MZI IIR lattice filters, along with high-performance balanced photodiodes integrated on the same chip. This work is organized as follows. In section 2 a general description of the PIC concept is provided, along with details on the design procedure, chip layout and manufacturing process. In section 3, experimental results are reported on both the optical filters and system level performance in terms of RF gain and linearity. A discussion on the potential paths towards the improvement of the discriminator performance metrics is also provided. Finally, a Summary and Conclusions are exposed in section 4.
2. PIC description
A conceptual diagram of the chip structure can be seen in the upper part of Fig. 1 , while a picture of the manufactured PIC is shown in the lower part. Firstly, the light is coupled into one of the chip facets and processed by an input filter. The purpose of this filter is to serve as a band splitter, dividing the spectrum of the FM optical signal into two separate bands. Each of these two bands are subsequently fed into two different filters, termed Filter 1 and Filter 2, but this time nominally designed to provide a linear ramp in optical intensity [7,11]. Once filtered, both bands are finally detected in a balanced, high-bandwidth integrated photodiode.
All the aforementioned filters were implemented using the same architecture, whose general structure is shown in Fig. 2 . As it can be seen, multiple ring resonators are coupled to both branchs of a Mach-Zehnder Interferometer (MZI), resulting in an IIR-equivalent optical filter. Given a certain transfer function designed using well-known digital filter synthesis techniques, a recursive algorithm finds the coupling coefficients of the directional couplers and the phases of both the ring resonators and MZI arms such that the optical structure exhibits the same periodic response . This is, the digital filter response is mapped into the optical domain, with a frequency period that depends on the free spectral range (FSR) of the ring resonators, and a filter order that is equal to the total number of rings. An interesting advantage of using filtering structures based on MZIs resides in the fact that there are two possible inputs/outputs per filter. Furthermore, in an ideal situation with no losses, the sum of optical powers at both outputs must be equal to the power present at the input. This is, an output port has an amplitude response that is complementary to the other output port. Thanks to this special feature, the unused input/output ports of the MZIs can be employed for testing purposes, allowing for individual characterization of each individual filter. Thus, these additional test waveguides were routed to the facets of our chip, in order to simplify experimental characterization (see the lower part of Fig. 1).
In our case, and due to space constraints, all three filters are second order. That is, they have one ring resonator coupled to each arm of the MZI, which was found to be enough for implementing a linear frequency discriminator. The input filter was designed using MatlabTMas a second-order Elliptic filter. It features a pass band normalized cut-off frequency of 0.33, and 0.25 dB attenuation, and a stop-band normalized frequency of 0.66 with 18 dB of minimum attenuation. The coefficients for the linear frequency discriminator using this structure were extracted from  and previously reported in . Table 1 shows the synthesized values of each coupling constant and optical phase, after applying the aforementioned recursive algorithm. Note that, although the linear optical filters present the same complementary response (and thus the same coupling coefficients), their optical phases must be different in order to ensure that they are properly aligned with the right pass-band of the Input Filter. Their FSR must also behalf of it, so the linear ramp is only applied to either the upper or lower band of the transmitted spectrum.
Thus, the perimeter of the ring resonators in Filter 1 and Filter 2 is twice that of the rings in the Input Filter, except for the small deviations in length used to introduce the appropriate phase offsets. Figure 3(a) -3(d) shows the ideal simulated responses of the optical filters.
The chip was manufactured under the EuroPIC EU FP7 project , using state-of-the-art InP generic integration technology. Deeply etched 2 μm wide rib waveguides, with an InGaAsP core over an InP substrate, allowed for low bending radius (150 μm). The total chip size is 6x6 mm2. Due to the high sensitivity of the filter response to the relative optical phases of the ring resonators and the MZI arms, multiple thermo-optic heaters were placed wherever needed, in order to allow both for post-fabrication fine tuning as well as for the tuning of the filter responses (see low part of Fig. 1). Two horizontal rows of metal pads were allocated on either sides of the chip for easier access with multi-contact DC probes. The different coupling constants obtained after the filter synthesis technique were implemented by means of 12 μm wide tapered MMIs in a so-called butterfly configuration , since the high confinement of the deeply-etched waveguides lead to unacceptable lengths using the more straightforward approach based on parallel waveguide couplers. Due to the high density of metal interconnects, air-bridge waveguide-metal crossings were employed, but an appropriate optical routing was done so as to minimize extra optical losses due to possible metal absorption. As for the balanced photodetectors, the specifications for each one are : bandwidth 35 GHz, responsivity 0.54 A/W, and large signal saturation current of 10 mA. The balanced configuration has a common mode rejection ratio specification of > 25 dB.
3. Experimental characterization and discussion
3.1 Optical discriminator filters
The experimental set-up used for characterization and adjustment of the optical filters composing the discriminator is shown in Fig. 4 .
A high-power broadband ASE source is coupled to each one of the three input waveguides, and the corresponding transfer function of each filter (Input Filter, Filter 1 and Filter 2) is recorded by measuring the optical spectrum at the output of the chip with a high-resolution Optical Spectrum Analyzer. In order to properly characterize the on-chip optical lattice filters, the input polarization of the light must be accurately controlled. This stems from the fact that the difference in effective index between TE and TM modes propagating within the waveguides of the structure will lead to a totally different response, since it heavily depends on the relative optical phases of the MZI arms and the ring resonators, as it has been previously mentioned. In order to do so, a free space optical set-up was arranged, where the light to be injected into the chip from the optical fiber is firstly collimated and passed through a linear polarizer, before being focused by a proper microscope objective into one of the chip facets. Multi-contact DC probes, connected to an array of current source modules, allowed for current injection into the thermo-optic heaters present on the chip.
The measured individual filter responses are shown in Fig. 5 , along with ideal design values. These curves were obtained after an iterative manual adjustment of the injected currents, where the transfer functions of the devices were firstly compensated and then aligned in wavelength. A good agreement with design targets is found. The curves are normalized to their respective insertion losses, which were calculated after normalization with a straight waveguide within the same chip. The estimated insertion losses for the Input Filter are 6.5 dB, while Filter 1 and Filter 2 show 6 and 6.2 dB, respectively. Filter1 ramp bandwidth is 19.34 GHz and ramp slope is 5.17x10−11 Hz−1, while Filter 2 ramp bandwidth is 16.69 GHz and ramp slope is 5.99x10−11 Hz−1. The total response can be seen in Fig. 5(d). The frequency span of the composed linear ramp is about 36.2 GHz, and the total insertion losses are about 13 dB in each branch.
3.2 Overall discriminator system level performance
Figure 6 shows the experimental set-up used for the characterization of the FM-DD system performance.
The output from a tunable laser (ANDO 1520-1620 nm tunability range, 0.7 dBm output power, RIN = −145dB/Hz) was fed via a polarization controller to an electro-optic phase modulator (JDSU, volt, modulation bandwidth 10 GHz, 3.5 dB insertion losses) modulated by two RF signals at f1 and f2 = f1 + 1 MHz which were supplied by an RF vector analyzer (AGILENT E8267C). We selected a modulation frequency f1 = 9 GHz for measurements where the optical carrier was placed in the middle point of the discriminator ramp curve (1549,1 nm) and modulation frequenciesf1 = 5 GHz and f1 = 7 GHz for measurements where the optical carrier was placed in the middle point of the discriminator positive and negative ramp sections (1549.184 nm and 1549.020 nm respectively). The modulated signal was fed to the discriminator chip via a free space polarizer and the output signal was extracted from the chip via an RF probe and passed through a 30 dB Gain RF amplifier before being fed to a spectrum analyzer. The operation point of the chip was controlled via multi-contact DC probes fed by current sources.
We first measured the third order input interception points (IIP3s) for the three selected modulation frequencies. The upper part of Fig. 7 shows the spectral diagram for each modulation frequency. Diagram (a) corresponds to f1 = 9 GHz and in this case, the wavelength of the optical carrier was chosen to match that of the discriminator central frequency, that is λ0 = 1549.11 nm, Diagram (b) corresponds to f1 = 7 GHz and for this case, as well as for that illustrated by diagram (c), which corresponds to f1 = 5 GHz, we carried out measurements with the optical carrier placed in the middle of the negative part of the discriminator slope λc- = 1549.02 nm and with the optical carrier placed in the middle of the positive part of the discriminator slope λc+ = 1549.18 nm. In each case the input RF power (i.e that of the sidebands) was increased until the measured output powers of the fundamental atf1and the third order intermodulation product at 2f2-f1were equal. The lower part of Fig. 7 shows the measured results together with the IIP3 level corresponding to an ideal intensity modulated (Mach-Zehnder modulator, MZM) MWPL. Measured results for the IIP3s outperform those predicted by simulation from the designed filter. For instance, 34 and 33 dBm as compared to 22, 2 and 23,5 dBm for f1 = 5 GHz andf1 = 7 GHz respectively and, , 26 and 29 dBm as compared to 22,7 and 23,3 dBm for f1 = 5 GHz andf1 = 7 GHz respectively and, , and 28 dBm as compared to 16,5 dBm for f1 = 9GHzand, . The reason behind this improvement is connected to the fact that the actual measured spectral characteristics of the filters as shown in Figs. 5(b) and 5(c) show a parabolic feature and thus are more linear as related to the optical field than as related to the optical intensity. This feature is expected to improve the value of the IIP3 as pointed out in .
As it can be observed, there is an improvement in the value of the IIP3 for all the cases as compared to the standard reference case of the intensity modulation link with volt used in other works [11,12]. Note that for f1 = 5GHz and f1 = 7 GHz the results are different depending on whether the optical carrier is placed at or at . In both cases the improvement is higher if the carrier is placed at (11 dB vs. 3 dB for f1 = 5 GHz and 10.8 dB vs. 6.2 dB for f1 = 7 GHz). This non-symmetric behavior is due to the fact the undesired band suppression is achieved more efficiently in the region below as it can be observed in Fig. 5(d). When the carrier is placed at the central frequency of the discriminatorand f1 = 9 GHz each sideband is affected by a different slope and this results in a degradation of the IIP3. All in all, the reported IIP3 values (26-34 dBm) are in the range of those reported for Si3N4 (25-36 dBm)  and below those recently reported for Silica on Silicon PLC discriminators  (38 dBm).
We then measured the output RF power versus input RF power characteristic for some of the cases shown in Figs. 5(a)-5(c). The results are shown in Fig. 8 together with the overall measured noise baseline, which was −130dB/Hz.
The measured values of the SFDR are in the range between 67 and 79 dB.Hz2/3(those predicted by simulation for a linear intensity profile are within a range of 66.2 and 71.3 dB.Hz2/3) which in principle are below those obtained for a Si3N4 IIR discriminator (90 dB.Hz2/3)  (no SFDR are reported for the Silica on Silicon PLC discriminator in ) in other contributions. However, as it will be discussed in the next subsection there is a considerable room for improvement for the InP discriminator.
The reported results in terms of SFDR for our discriminator where limited by the high value of the optical losses in the MWP link. The value of SFDR depends on the IIP3, the RF gain GRF and the noise density N as:19],
where , describe the spectral action of the discriminator over the fundamental Ω1 and the third order intermodulation 2Ω1-Ω2 tones respectively , is the detector responsivity, T the link losses from the EDFA output to the receiver input, and PEDFA the output power from the EDFA. In addition, Zin, Zout represent the input impedance of the modulator and the output receiver load impedance respectively.
Observation of Eq. (1) reveals that in order to increase the value of the SFDR (without changing the value of N), one has either to increase the IIP3, or the GRF Reducing the value of Vπ increases the value of GRF according to Eq. (2), but reduces the value of IIP3 in the same proportion according to Eq. (3) so no net improvement is obtained in the SFDR. A straightforward option is to increase the value of GRF by reducing the link losses (i.e. increase the value of T in Eq. (2)). In particular an overall loss figure of 24.9 dB was measured from the EDFA output to the receiver input. Here 4.4 dB correspond to losses from the EDFA output to the free space polarizer output, 7 dB to the chip input facet coupling and 13.5 dB to internal chip losses. Losses can be reduced replacing the free-space input coupling to the chip by a spot-size converter. An estimate of this reduction would be, according to reported results , in the range of 10 dB which, in turn, would increase the value of GRF by 20 dB and consequently the SFDR by 13.3 dB (80-92 dB.Hz2/3).
A second possibility to increase the SFDR without changing the filter design is to reduce the value of the noise density. In our case it is dominated by the EDFA amplified spontaneous noise relative noise (RIN) which is −130 dB.Hz, while the laser RIN is −145 dB.Hz. This can be readily achieved by replacing the low output power laser (0.7dBm) and the EDFA by a high output power laser (20 dBm). In this case the value of N would be reduced by 15 dB and therefore the SFDR improved by 10 dB (90-102 dB.Hz2/3). Further improvement could be achieved by employing a laser with a reduced RIN value. Devices featuring values of RIN = −160 dB.Hz are commercially available which could improve the SFDR by an extra 10 dB (100-112 dB.Hz2/3).
Finally, our filter was designed to be linear in intensity while it has been demonstrated  that a considerable improvement in the value of IIP3 can be achieved by a filter design featuring a linear behavior in the optical field. An improved performance is thus expected by a proper filter redesign.
4 Summary and conclusions
We have reported the design, fabrication and characterization of an integrated frequency discriminator on InP technology for microwave photonic phase modulated links. The optical chip is, to the best of our knowledge, the first reported in an active platform and the first to include the optical detectors. The discriminator, designed as a linear filter in intensity, features preliminary SFDR values the range between 67 and 79 dB.Hz2/3 for signal frequencies in the range of 5-9 GHz limited, in principle, by the high value of the optical losses arising from the use of several free space coupling devices in our experimental setup. As discussed, these losses can be readily reduced by the use of integrated spot-size converters improving the SFDR by 14.3 dB (80-92 dB.Hz2/3). Further increase up to a range of (100-112 dB.Hz2/3) is possible by reducing the system noise eliminating the EDFA employed in the setup and using a commercially available laser source providing higher output power and lower relative intensity noise. Other paths for improvement requiring a filter redesign to be linear in the optical field have also been discussed.
The authors wish to acknowledge the financial support given by the Research Excellency Award Program GVA PROMETEO 2008/092, Spanish MICINN Project TEC2010-21337, acronym ATOMIC; project FEDER UPVOV10-3E-492 and project FEDER UPVOV08-3E-008. J.S. Fandiño acknowledges financial support through grant FPU-2010 (ref: AP2010-1595). J. D. Doménech acknowledges both Beca de Formacion de Personal Investigador (FPI) research grant (ref: BES-2009-018381) and funding from Intel Doctoral Student Honor Programme. The authors are also grateful to the EuroPIC consortium (FP7 funded programme NMP 228839-2) for access to the design manual and design toolkit of the generic InP fab.
References and links
1. J. Capmany and D. Novak, “Microwave photonics combines two worlds,” Nat. Photonics 1(6), 319–330 (2007). [CrossRef]
2. T. Darcie and J. Zhang, “High-performance microwave-photonic links,” in Proc. IEEE Radio and Wireless Symposium, 125 –128 (Jan. 2008).
3. M. LaGasse and S. Thaniyavaru, “Bias-free high-dynamic-range phase-modulated fiber-optic link,” IEEE Photon. Technol. Lett. 9(5), 681–683 (1997). [CrossRef]
4. V. Urick, F. Bucholtz, P. Devgan, J. McKinney, and K. Williams, “Phase modulation with interferometric detection as an alternative to intensity modulation with direct detection for analog-photonic links,” IEEE Trans. Microw. Theory Tech. 55(9), 1978–1985 (2007). [CrossRef]
5. J. McKinney, K. Colladay, and K. Williams, “Linearization of phase-modulated analog optical links employing interferometric demodulation,” J. Lightwave Technol. 27(9), 1212–1220 (2009). [CrossRef]
6. B. Haas, V. Urick, J. McKinney, and T. Murphy, “Dual-wavelength linearization of optically phase- modulated analog microwave signals,” J. Lightwave Technol. 26(15), 2748–2753 (2008). [CrossRef]
7. T. Darcie, J. Zhang, P. P. Driessen, and J. J. Eun, “Class-B microwave-photonic link using optical frequency modulation and linear frequency discriminators,” J. Lightwave Technol. 25(1), 157–164 (2007).
8. X. Xie, J. Khurgin, J. Kang, and F. Choa, “Compact linearized optical FM discriminator,” IEEE Photon. Technol. Lett. 14(3), 384–386 (2002). [CrossRef]
9. J. Wyrwas and M. Wu, “Dynamic range of frequency modulated direct-detection analog fiber optic link,” J. Lightwave Technol. 27(24), 5552–5562 (2009). [CrossRef]
10. J. Wyrwas and M. Wu, “High dynamic range microwave photonic links using maximally linear FIR optical filters,” OFC/NFOEC, Los Angeles, CA, 2010, paper JWA43, (2010).
11. D. Marpaung, C. Roeloffzen, A. Leinse, and M. Hoekman, “A photonic chip based frequency discriminator for a high performance microwave photonic link,” Opt. Express 18(26), 27359–27370 (2010). [CrossRef] [PubMed]
12. J. M. Wyrwas, R. Peach, S. Meredith, C. Middleton, M. S. Rasras, K. Y. Tu, M. P. Earnshaw, F. Pardo, M. A. Cappuzzo, E. Y. Chen, L. T. Gomez, F. Klemens, R. Keller, C. Bolle, L. Zhang, L. Buhl, M. C. Wu, Y. K. Chen, and R. DeSalvo, “Linear phase-and-frequency-modulated photonic links using optical discriminators,” Opt. Express 20(24), 26292–26298 (2012). [CrossRef] [PubMed]
13. M. S. Rasras, E. Y. Chen, K.-Y. Tu, M. P. Earnshaw, F. Pardo, M. A. Cappuzzo, E. Chen, L. Gomez, F. Klemens, B. Keller, C. Bolle, L. Buhl, J. M. Wyrwas, M. C. Wu, R. Peach, S. Meredith, C. Middleton, and R. DeSalvo, “Reconfigurable linear optical FM discriminator,” IEEE Photon. Technol. Lett. 24(20), 1856–1859 (2012). [CrossRef]
14. “EuroPIC” project, Grant no. CP-TP 228839–2 under FP7 Work Programme 2008, Cooperation, Theme 4: Nanosciences, Nanotechnologies, Materials and New Production Technologies, http://europic.jeppix.eu/
15. C. K. Madsen and J. H. Zhao, Optical Filter Design and Analysis (John Wiley & Sons, Inc. 1999).
16. J. S. Fandiño, P. Muñoz, and J. Capmany, “Design of a Frequency Modulation – DirectDetection Microwave Photonic Link receivermonolithically integrated in InP generic technology,” in Proc. European Conference on Integrated Optics, April 16–19, Sitges (Spain), (2012).
17. P. A. Besse, E. Gini, M. Bachmann, and H. Melchior, “New 2x2 and 1x3 multimode interference couplers with free selection of power splitting ratios,” J. Lightwave Technol. 14(10), 2286–2293 (1996). [CrossRef]
18. A. Beling, H.-G. Bach, D. Schmidt, G. G. Mekonnen, M. Rohde, L. Molle, H. Ehlers, and A. Umbach, High-speed balanced photodetector module with 20 db broadband common-mode rejection ratio,” in Tech. Dig. Optical Fiber Commun. (OFC 2003), Atlanta, GA, Mar. 23–28, 1, pp. 339–340, paper WF4, (2003).
20. M. Hamacher, R. Kaiser, H. Heidrich, P. Albrecht, B. Borchert, K. Janiak, R. Loffler, S. Malchow, W. Rehbein, and H. Schroeter-Janssen, “Monolithic integration of lasers, photodiodes, waveguides and spot size converters on GaInAsP/InP for photonic IC applications,” in Proc. 2000 International Conference on Indium Phosphide and Related Materials, pp.21–24, (2000).