Open Access
Add to BibSonomy
Abstract
We have fabricated a compact and integrated 4-channel analog optical transceiver for radio over fiber application. In the fabricated module, the transmitter optical sub-assembly is composed of four directly modulated DFB laser chips integrated with an optical multiplexer based on an arrayed waveguide grating (AWG) using silica-based planar lightwave circuit (PLC) technology. The receiver optical sub-assembly consists of a PIN photodiode array integrated with an AWG-PLC-type optical de-multiplexer. For all the lanes, the 3 dB bandwidth exceeds 19.1 GHz and the measured spurious-free dynamic range (SFDR) is up to 90.5 dB⋅Hz2/3 when the input RF frequency is from 2 GHz to 14 GHz. Meanwhile, the electrical inter-channel crosstalk of the transceiver is less than −20 dB when the carry frequency is below 18.5 GHz. This module shows a great transmission performance in radio over fiber system. Under simultaneous 4-channel different 600 Mb/s 5-band 64QAM-OFDM RF signal operation, the measured error vector magnitude (EVM) performance below 8% is achieved after 15.5 km single-mode fiber propagation for all lanes. This scheme has potential applications in guiding high-dense, cost-effective and high-linearity analog optical transceiver design to realize high-capacity radio over fiber transmission systems.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
The fifth-generation (5G) mobile fronthaul system will require very-high capacity and ultra-low latency connection between the baseband units (BBUs) and remote radio heads (RRHs) [1,2]. To meet this requirement, there are some challenges such as the bandwidth constraints imposed by current common public radio interface (CPRI), and the increasing costs and latency due to the RF signal processing at RRH [3]. Radio over fiber (RoF) is considered as a promising candidate because it directly transports RF signals between BBU and RRH, and allows for centralization of signal processing and network management [4]. Obviously, the design of broadband, high-linearity and low-noise analog optical transceiver is the key in the RoF-based mobile fronthaul [5,6] and satellite communication [7]. Especially for large number of access, multiple-channel and high-performance optical transceiver is highly desired.
To date, some works have presented the design of multiple-channel optical transmitter or receiver [8–12]. In [10], monolithically integrated 4 × 25 Gb/s distributed Bragg reflector (DBR) laser array is reported to meet the demand of 100 Gb/s Ethernet transmission standard. After that, the first compact 100 Gb/s (4 × 25.8 Gb/s) Ethernet directly modulated transmitter optical sub-assembly (TOSA) with a 3 dB bandwidth of 18 GHz has been demonstrated [11,12]. However, all the 4-channel directly modulated optical transmitters mentioned above are designed for digital application rather than analog application. For RoF system, the modulation bandwidth, linearity, gain and noise figure of the directly modulated transmitter are the primary factors [13–16]. In addition, the spurious-free dynamic range (SFDR) is also a main figure of merit which is mainly limited by the linearity, gain and noise figure of the RoF systems. By considering these factors, many single-channel directly modulated modules have been investigated [17–19]. However, to the best of our knowledge, until recently no one have actually been able to design and fabricate a cost-efficient, multi-channel, and small-form factor fiber optical transceiver that combines in the same device for analog application.
In this paper, we have developed the first compact, 4-channel analog optical transceiver. By specially integrating and packaging, four individual directly modulated DFB lasers (DML) and an optical multiplexer based on an arrayed waveguide grating (AWG) using silica-based planar Lightwave circuit (PLC) technology are integrated in TOSA. Meanwhile, a PIN photodiode array and an AWG-PLC-type optical de-multiplexer integrated in the receiver optical sub-assembly (ROSA). At last, the 3 dB bandwidth exceeds 19.1 GHz and the measured SFDR is up to 90.5 dB⋅Hz2/3 when the input RF frequency is from 2 GHz to 14 GHz for all lanes. Moreover, by using the fabricated module, error-free transmission has been achieved for 600 Mb/s multi-band 64QAM-OFDM RF signal over 15.5 km single mode fiber (SMF) for each lane.
2. Module fabrication
Different from the widespread digital optical transceiver, the analog optical transceiver demands higher performance standards. Noise, intermodulation distortion, nonlinearities are all required to be validated in design. Therefore, many issues such as optical coupling, impendence matching, bonding wire, RF connector and transitions between different transmission lines are all needed to be considered. Putting the cost and convenience factor into consideration, in our scheme TOSA and ROSA are separated and connected to the analog driving circuit board by the flexible printed circuit (FPC) board, as shown in Fig. 1.
Figure 1(a) shows the structure of TOSA, and this TOSA is composed of FPC, four chips on carrier (CoC), four aspheric coupling lens and an AWG-PLC with tail fiber. All these components are packaged in a small metal box with size of 18mm × 20mm × 6mm. The FPC is fabricated using polyimide (dielectric constant ~3.4) type flexible copper-clad-laminate. Ground-signal-signal-ground coplanar microstrip transmission is used in FPC, and in this way, a laser driver (FL500) is connected to DML chip for bias current control. Moreover, there are four termination resistances which are close to DML chip in FPC to achieve impendence matching. The CoC is placed close to FPC, and thus the bonding wire that connects FPC and DML chip is shortened by less than 0.15mm long, contributing to the increase of modulation bandwidth. It should be noted that two bonding wires are used for each DML chip, and this operation can minimize parasitic inductance to mitigate the degradation of the high-frequency electrical signals. The bandwidth of TOSA can be further improved by resonance of bonding wire inductance and electrode capacitance [20,21]. After that, four aspheric coupling lens are used to effectively couple the light from DML chip to AWG-PLC. In our design, the numerical apertures (NA) in vertical and horizontal directions are 0.45 and 0.35 for the DML chips respectively, and the NA of the PLC is 0.17. The distances between CoC and aspheric Lens, aspheric Lens and PLC are 0.25 mm and 2.5 mm, respectively. Subsequently, an AWG-PLC integrated with SMF is used for optical multiplexer. The output optical power is about 2.5 dBm when DML is biased at 36 mA.
Figure 1(b) shows the structure of ROSA, and this ROSA is composed of FPC, two pieces copper heat sink, a four-channel linear trans-impedance amplifier (TIA), four PIN type photodetector (PD) array, an AWG-PLC with output facet 42 degrees polished and integrated with SMF fiber. Similarly, all these components are packaged in a small metal box with size of 18 mm × 22 mm × 6 mm. The aluminum nitride (AlN) with low thermal resistance is used as heat-sink in order to dissipate the heat from TIA and PD carrier. The thermal expansion coefficient and thermal conductivity of the used heat sink are 4.6 × 10−6/°C and 200 W/mK, respectively. The PD carrier where the PD array is mounted is attached to a predetermined position on the TIA. The used TIA (IN2864TA) with the 3-dB bandwidth of 25 GHz is operated at automatic gain control (AGC) mode. The 42° polished AWG-PLC can use total reflection to change the direction of the light field, which can effectively improve the coupling efficiency and make the coupling module compact. The measured responsivity of the ROSA is 0.43 A/W, 0.44 A/W, 0.43 A/W, 0.42 A/W for each lane respectively, and its linear dynamic response range are more than 27 dB.
To maximize the operating bandwidth and minimize the electrical crosstalk, it is essential to optimize the package of analog optical transceiver module. Two methods have been done in our work. The first one is to make the side of transmission lines metallization, and in this way, great grounding effect is achieved and the electromagnetic field of the microwave signal can be effectively constrained. The side of transmission lines metallization cannot only decrease the electrical crosstalk but also guarantee the signal integrity. The second one is to increase the width of the pitch between the RF signal lines to reduce the electrical crosstalk. Moreover, perforated operation should be done around the RF transmission lines. The termination resistances in FPC are set at 35 Ω because the measured resistance of the DML chip is about 10 Ω for all the lanes. By this means, a high-performance and compact 4 × 4 analog optical transceiver has been fabricated in our laboratory. Figure 2 shows the photograph of this optical transceiver.
3. Module performance
Figure 3(a) shows the light-current (L-I) characteristics of the TOSA. The threshold currents are around 4 mA for all the lines. Figure 3(b) displays the optical spectra of the TOSA when the bias current is set at 36 mA for all the lanes, and the out optical power are all above 2 dBm. Moreover, the central wavelength for all the lanes are 1271.38 nm, 1290.75 nm, 1311.57 nm and 1330.11 nm respectively, and the side mode suppression ratio (SMSR) for each lane is higher than 53.2 dB.
Fig. 3 (a) light-current (L-I) characteristics of the TOSA and (b) optical spectrum of the output optical signal when the bias current is set to 36 mA.
The small-signal modulation response S21 of the packaged TOSA and ROSA is measured by a vector network analyzer (VNA Anritsu MS464A) and shown in Fig. 4. The frequency response of the optical transceiver varies with the bias current from 17 mA to 41 mA, and the frequency of resonance peak gets higher when the bias current is larger. It could be clearly observed that the 3 dB bandwidth for all the lanes is up to 19 GHz when the bias current is 35 mA. Moreover, the results show that there is a little difference between the small-signal frequency responses of different lanes.
For analog application, the most straightforward limitation for the optical transceiver is the SFDR performance, and the SFDR performance is closely related to the linearity, gain and noise figure of the optical transceiver. In our test, a two-tone RF signal with the central frequencies of 3.99 GHz and 4.01 GHz is generated by a vector signal generator (R&S SMW200A) to modulate the Lane3 of the TOSA. A variable optical attenuator (Joinwit JW3303) is adopted to control the received optical power at the Lane3 of the ROSA. An electrical signal analyzer (ESA, R&S FSV13) is used to measure the power of the fundamental, IMD3 and noise floor. According to the L-I characteristics of the TOSA, the DC bias current is set at 36.5 mA in our test. By scanning the input RF power, the electrical power of IMD3 and fundamental components is monitored. The measured noise floor for four lanes is −142.1 dBm/Hz, −141.9 dBm/Hz, −142.5 dBm/Hz and −141.7 dBm/Hz, respectively. Figure 5(a) shows the measured output RF power with respect to input RF power and the calculated SFDR3 of Lane3 in the proposed transceiver. It could be observed that the SFDR3 can reach up to 95.3 dB·Hz2/3. The SFDR3 performance for all the four lanes at different central frequency from 2 GHz to 14 GHz are also tested and shown in Fig. 5(b). It could be clearly observed that the measured SFDR3 is slightly worse when the central frequency of the input RF signal is increased. However, the SFDR3 of 90.5 dB⋅Hz2/3 can be achieved even when the central frequency of the input RF signal is 14 GHz.
Fig. 5 (a) SFDR3 performance of Lane3 at 4 GHz and (b) SFDR3 for all the lanes at different RF frequency.
To test the crosstalk characteristic of the proposed transceiver, the same VNA is used, and the bias currents are applied to the DFB lasers of all lanes simultaneously. For Lane-X, the Y-X curve is first measured by connecting the output RF port of VNA to the transmitter of Lane-Y, and connecting the input RF port of VNA to the receiver of Lane-X. The X-X curve is then tested by connecting the output RF port of VNA to the transmitter of Lane-X, and connecting the input RF port of VNA to the receiver of Lane-X. At last, the crosstalk from Lane-Y to Lane-X is calculated by subtracting the Y-X curve to the X-X curve. As shown in Fig. 6, for Lane1, the crosstalk from Lane2 which is the adjacent channel, is only about −20 dB at a frequency of 22 GHz. For Lane2, the crosstalk from Lanes 1 and 3 which are adjacent channels, are about −20 dB at a frequency of 18.6 GHz. Similarly, for Lane3, the crosstalk from Lanes 2 and 4 which are adjacent channels, are about −20 dB at a frequency of 18.5 GHz. For Lane4, the crosstalk from Lane3 which is adjacent channel, is −20 dB at a frequency of 21 GHz.
The transmission performance of the proposed transceiver is also evaluated by using the experimental setup shown in Fig. 7. In this experiment, 5-band 64QAM-OFDM RF signal with the central frequency from 1 GHz to 14 GHz is generated by a vector signal generator (R&S SMW200A) for each lane. The occupied bandwidth for each band is 20 MHz, and the band gap is 10 MHz. Therefore, the total capacity of each lane is 20 MHz × 5 × log2(64) = 600 Mb/s. The generated RF signal is used to modulate the designed TOSA. After 15.5 km SMF propagation, the variable optical attenuator (VOA) is adopted to adjust the received optical power at the ROSA. At last, the 64QAM-OFDM RF signal is captured by a signal analyzer (R&S FSV13). After demodulation, the error vector magnitude (EVM) performance are measured. In our experiment, we have evaluated the transmission performance of this transceiver under discrete operation (DO) and under simultaneous 4-channel different 5-band 64QAM-OFDM RF signal operation (SO). For the transmission performance evaluation under SO, there are only two vector signal generators (R&S SMW200A and SMBV100A) in our laboratory, and the maximum operating carrier frequency of SMBV100A is only 3.2 GHz. Therefore, an electrical mixer and another signal generator (R&S SMF100A) is used to upconvert the generated intermediate frequency (IF) signal of SMBV100A to RF signal with the central frequency from 1 GHz to 14 GHz. In our test, the SMW200A is used to drive the lane to be measured, and the upconverted RF signal from SMBV100A is used to drive the other three lanes after a 1:3 electrical divider. It should be noted that the occupied bandwidth of the output signals of SMW200A and SMBV100A are same, and the carry frequency and the power of these four RF signals are set to the same value in each test.
Figures 8(a)-(d) show the measured EVM performance in terms of received optical power at the ROSA when the central frequency of the 64QAM-OFDM signal is 6 GHz for both optical back to back (OBTB) and 15.5 km SMF transmission. The transmission performance of this transceiver under DO and under SO are both tested. It can be clearly observed that these four Lanes have the similar receiver’s sensitivity at the EVM threshold (EVM of 8%) [22]. The receiver’s sensitivity is −21.9 dBm and −20.8 dBm for OBTB and 15.5 km SMF in the Lane3, respectively. 1.2 dB power penalty is negligible because the optical wavelength of the Lane3 is 1311.57 nm, which has zero dispersion. The Lane1 has the highest optical power penalty of 3.1 dB after OBTB and 15.5 km SMF transmission. This could be attributed to the fact that the optical wavelength of Lane1 is farthest to the zero-dispersion wavelength. Under SO, the induced power penalty from inter-channel crosstalk is 1.5 dB, 1.7 dB, 2.2 dB and 1.1 dB for Lane1, Lane2, Lane3 and Lane4 respectively.
Fig. 8 Measured EVM performance versus optical received power with/without 15.5 km SMF under DO and SO when the central frequency of RF signal is 6 GHz.
To further investigate the EVM performance of the proposed transceiver when the central frequency of the input RF signal above 10 GHz. The measured EVM performance in terms of received optical power at the ROSA when the central frequency of the 5-band 64QAM-OFDM signal is increase to 12 GHz as shown in Fig. 9. In the OBTB case, there are negligible performance difference between different lanes. After 15.5 km SMF propagation, Lane3 still has the best EVM performance due to its zero-dispersion wavelength of 1311.57 nm, and Lane1 has the worst EVM performance because it has the farthest wavelength to the zero-dispersion wavelength. Moreover, RF signal suffers from frequency-dependent fading caused by fiber dispersion, and higher RF frequency worse EVM performance. However, the EVM performance of Lane1 is 4.3% which is far less than the EVM threshold when the received optical power is −8 dBm. Similarly, under SO, the induced power penalty from inter-channel crosstalk is 1.6 dB, 3.8 dB, 3.7 dB and 2.9 dB for Lane1, Lane2, Lane3 and Lane4 respectively. Obviously, the induced power penalty is increased when carry frequency is higher, and this could be attributed to the improved crosstalk as shown in Fig. 6.
Fig. 9 Measured EVM performance versus optical received power with/without 15.5 km SMF under DO and SO when the central frequency of RF signal is 12 GHz.
We have also tested the EVM performance for each lane versus the central frequency of the input 5-band 64QAM-OFDM signal. By scanning the central frequency of the input RF signal from 1 GHz to 14 GHz, EVM performance are monitored by the signal analyzer. Figures 10(a)-10(d) illustrate the measured EVM performance for each lane in terms of different central frequency when the received optical power at the ROSA is −8 dBm. Similarly, the transmission performance of this transceiver under DO and under SO are both tested. It can be observed that the EVM performance will become worse as the central frequency increases, which is attributed to the frequency-dependent fading caused by fiber dispersion. Under SO, the measured EVM performance is degraded by a maximum value of 2.7% compared with DO when the carry frequency is 14 GHz, and this could be also attributed to the improved crosstalk as shown in Fig. 6. It can be easily observed that the EVM are below 8% and error-free transmission over a 15.5 km SMF has been achieved. These results indicate that the designed 4 × 4 optical transceiver is one of promising candidates for RoF mobile communication where the reach exceeds 15.5 km.
Fig. 10 Measured EVM performance for 64QAM-OFDM RF signal versus optical received power with/without 15.5 km SMF under DO and SO when the received optical power is −8 dBm.
Furthermore, to evaluate the stable performance of the proposed optical transceiver, both the EVM performance of 5-band 64QAM-OFDM RF signal and the corresponding environment temperature of the proposed module are measured when the central frequency of the input signal is 6 GHz under DO and OBTB. A hot air blower is used to vary the temperature of the transceiver indirectly. In the test, not only the temperature is varied but also mechanical vibration is induced by the wind. As shown in Fig. 11, when the temperature changes within 25 to 65°C, the measured EVM performance fluctuates within in 2.83%, and the measured EVM performance remains below EVM threshold (EVM of 8%). That means the proposed transceiver has a relatively stable performance.
4. Conclusion
We have fabricated a compact 4 × 4 directly modulated analog optical transceiver for the first time. For all the lanes, the SFDR3 are above 90.5 dB⋅Hz2/3 without any pre-distortion and post-compensation linearization technique. Moreover, the 3 dB bandwidth of the designed transceiver exceeds 19.1 GHz for each lane. Error-free transmission over a 15.5 km SMF with rate of 600 Mb/s for 64QAM-OFDM RF signal is demonstrated for each lanes when the central frequency is from 1 GHz to 14 GHz. The proposed technique allows the realization of multi-channel, broadband, high linearity and compact optical transceiver modules and shows a great performance in the analog optical communication systems such as radio over fiber and phased-array radar systems. Meanwhile, the technique may be a promising candidate for future 5G mobile communication network.
Funding
National “863” Program of China (2015AA016904); National Nature Science Foundation of China (NSFC) (61675083, 61505061); Fundamental Research Funds for the Central Universities HUST (2017KFKJXX010); Key project of R&D Program of Hubei Province (2017AAA046).
Acknowledgments
Thanks to Accelink Techn. Co., Ltd and Henan Shijia Photons Techn. Co., Ltd, who are essential in the development of DFB laser chip, PD array chip and AWG-PLC.
References
1. China Mobile Research Institute, “C-RAN: The road towards green RAN,” whitepaper, (2013).
2. N. Cvijetic, A. Tanaka, K. Kanonakis, and T. Wang, “SDN-controlled topology-reconfigurable optical mobile fronthaul architecture for bidirectional CoMP and low latency inter-cell D2D in the 5G mobile era,” Opt. Express 22(17), 20809–20815 (2014). [CrossRef] [PubMed]
3. Common Public Radio Interface (CPRI), Interface Specification, (2015).
4. X. Liu, H. Y. Zeng, N. Chand, and F. Effenberger, “Efficient mobile fronthaul via DSP-based channel aggregation,” J. Lightwave Technol. 34(6), 1556–1564 (2016). [CrossRef]
5. P. T. Dat, A. Kanno, N. Yamamoto, and T. Kawanishi, “5G transport networks: the need for new technologies and standards,” IEEE Commun. Mag. 54(9), 18–26 (2016). [CrossRef]
6. D. Zhu, J. Chen, and S. Pan, “Multi-octave linearized analog photonic link based on a polarization-multiplexing dual-parallel Mach-Zehnder modulator,” Opt. Express 24(10), 11009–11016 (2016). [CrossRef] [PubMed]
7. S. L. Pan, D. Zhu, S. F. Liu, K. Xu, Y. T. Dai, T. L. Wang, J. G. Liu, N. H. Zhu, Y. Xue, and N. J. Liu, “Satellite payloads pay off,” IEEE Microw. Mag. 16(8), 61–73 (2015). [CrossRef]
8. B. Pezeshki, “The advantages of hybrid optical integration, as demonstrated by a 4×25 Gb/s transceiver (TROSA),” in Optical Fiber Communication Conference, OSA Technical Digest (online) (Optical Society of America, 2017), paper M3B. 3.
9. W. M. Yao, M. K. Smit, and M. J. Wale, “Monolithic 300 Gb/s parallel transmitter in InP-based generic photonic integration technology,” IEEE J. Sel. Top. Quant. 24(1), 6100711 (2018). [CrossRef]
10. S. Matsuo, T. Kakitsuka, T. Segawa, R. Sato, Y. Shibata, R. Takahashi, H. Oohashi, and H. Yasaka, “4×25Gb/s frequency-modulated DBR laser array for 100-GbE 40-km reach application,” IEEE Photonics Technol. Lett. 20(17), 1494–1496 (2008). [CrossRef]
11. S. Kanazawa, W. Kobayashi, Y. Ueda, T. Fujisawa, K. Takahata, T. Ohno, T. Yoshimatsu, H. Ishii, and H. Sanjoh, “30-km error-free transmission of directly modulated DFB laser array transmitter optical sub-assembly for 100-Gbit/s application,” J. Lightwave Technol. 34(15), 3646–3652 (2016). [CrossRef]
12. S. Kanazawa, W. Kobayashi, Y. Ueda, T. Fujisawa, T. Ohno, T. Yoshimatsu, H. Ishii, and H. Sanjoh, “Low-crosstalk operation of directly modulated DFB laser array TOSA for 112-Gbit/s application,” Opt. Express 24(12), 13555–13562 (2016). [CrossRef] [PubMed]
13. K. V. Gasse, J. V. Kerrebrouck, A. Abbasi, J. Verbist, G. Torfs, B. Moeneclaey, G. Morthier, X. Yin, J. Bauwelinck, and G. Roelkens, “III-V-on-Silicon photonic transceivers for radio-over-fiber links,” J. Lightwave Technol.in press.
14. Y. T. Dai, Y. Cui, X. Liang, F. Yin, J. Q. Li, K. Xu, and J. T. Lin, “Performance improvement in analog photonics link incorporating digital post-compensation and low-noise electrical amplifier,” IEEE Photonics J. 6(4), 5500807 (2014).
15. A. Agarwal, T. Banwell, and T. K. Woodward, “Optically filtered microwave photonic links for RF signal processing applications,” J. Lightwave Technol. 29(6), 2394–2401 (2011). [CrossRef]
16. S. Li, X. Zheng, H. Zhang, and B. Zhou, “High linear radio-over-fiber system incorporating a single-drive dual-parallel Mach-Zehnder modulator,” IEEE Photonics Technol. Lett. 22(24), 1775–1777 (2010). [CrossRef]
17. Y. Ye, L. Deng, S. Chen, M. Cheng, M. Tang, S. Fu, M. Zhang, D. Zhang, B. Huang, and D. Liu, “Simultaneous suppression of even-order and third-order distortions in directly-modulated analog photonic links,” IEEE Photonics J. 9(3), 1–12 (2017). [CrossRef]
18. S. Chen, L. Deng, Y. Ye, X. Chen, M. Cheng, M. Tang, S. Fu, F. Luo, and D. Liu, “Experimental investigation on improved pre-distortion circuit for directly modulated radio over fiber system,” IEEE Photonics J. 9(5), 1–10 (2017).
19. Y. Ye, L. Deng, S. Chen, M. Cheng, M. Tang, S. Fu, M. Zhang, D. Zhang, B. Huang, and D. Liu, “A broadband and high linearity directly-modulated analog photonic link based on push-pull structure and digital signal post-compensation,” in Optical Fiber Communication Conference, OSA Technical Digest (online) (Optical Society of America, 2016), paper Th2A 0.22. [CrossRef]
20. L. Xie, J. W. Man, B. J. Wang, Y. Liu, X. Wang, H. Q. Yuan, L. J. Zhao, H. L. Zhu, N. H. Zhu, and W. Wang, “24-GHz directly modulated DFB laser modules for analog applications,” IEEE Photonics Technol. Lett. 24(5), 407–409 (2012). [CrossRef]
21. C. Chen, N. H. Zhu, S. J. Zhang, and Y. Liu, “Characterization of parasitics in TO-packaged high-speed laser modules,” IEEE Trans. Adv. Packag. 30(1), 97–103 (2007). [CrossRef]
22. X. Liu, F. Effenberger, N. Chand, L. Zhou, and H. F. Lin, “Demonstration of bandwidth-efficient mobile fronthaul enabling seamless aggregation of 36 E-UTRA-like wireless signals in a single 1.1-GHz wavelength channel,” in Optical Fiber Communication Conference, OSA Technical Digest (online) (Optical Society of America, 2015), paper M2J. 2. [CrossRef]
