Open Access
Add to BibSonomy
Abstract
We demonstrate 64 Gbps operation in a compact Si photonic crystal optical modulator that employs meander line electrodes and compensate for the phase mismatch between slow light and RF signals. Although low dispersion slow light increases the modulation efficiency, maintaining a sufficiently wide working spectrum, the phase mismatch becomes a limiting factor on the operation speed even when the phase shifter length is as short as 200 μm. Meander line electrodes broke this limit and enhanced the cutoff frequency by up to 31 and 38 GHz using 50 Ω and 20 Ω termination resistors, respectively. This allowed to use a group index of slow light higher than 20, and greatly improved the quality of the modulation characteristics at 25 and 32 Gbps. Clear open eye was observed even at 40–64 Gbps.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
In recent years, a large number of low-cost and high-speed optical transceivers have come into use as optical interconnects. In the current 100 Gbps Ethernet, 10–25 Gbps/lane are being used, whereas speeds higher than 40 Gbps/lane will be desired in the next era of 400 Gbps standards [1]. Therefore, one approach that is being considered is to use wavelength division multiplexing; however, the management of wavelengths becomes more severe. Although pulse amplitude modulation is an alternative candidate to reach higher speeds, it requires a larger signal amplitude, higher power consumption, and some sort of bit error correction. This means that phot-electronic circuits and systems become larger, more complicated, and costly. Therefore, simply increasing the baud rate is another promising option. For example, high-speed beyond 50 Gbps has been reported in Si modulators [2–5].
The miniaturization of devices not only contributes to low power consumption but also improves the degree of freedom in the layout and compactness of modules. High-speed small modulators that are currently being studied include resonators; plasmon waveguides; and those using Si photonics, electro-optic (EO) polymers, etc [6–8]. However, they have some disadvantages: a narrow working spectrum, a large optical loss, and/or low compatibility with complementary metal-oxide-insulator (CMOS) processes. We have studied Si photonic crystal waveguide (PCW) and lattice-shifted PCW (LSPCW) slow light modulators, which have a short phase shifter length, L = 200 μm, and a wide working spectrum of 15–20 nm [9–12], thus assuring a wide temperature tolerance; we have demonstrated the operation from 19°C to 124°C. We have previously reported an optimized p-n junction that exhibited 32 Gbps modulation with an extinction ratio higher than 3 dB at a drive voltage (Vpp) of 1.75 V. In that report, the on-chip passive loss was approximately 5 dB, and it demonstrated a well-balanced performance. However, it did not reach a bit rate over 40 Gbps; we only observed a barely open eye in the old study [9]. One reason for this is the phase mismatch between slow light and RF signals. When the frequency becomes higher or the group index, ng, of slow light becomes larger, this mismatch becomes more severe. In our previous study, we theoretically suggested that the phase mismatch could be compensated for by meander line electrodes and termination resistors that delay the RF signals, and predicted that the operation speed can be increased to 50 Gbps or more [11]. In this paper, we report the experimental demonstration of this prediction.
2. Device fabrication
Figure 1 shows the devices fabricated by CMOS process using KrF excimer laser exposure (248 nm) and phase shift masks (resolution <130 nm) on 200 mm silicon-on-insulator. The PCW was designed for operation in the C band (λ = 1530–1565 nm); the lattice constant a = 400 nm, and the hole diameter 2r = 190 or 205 nm for a Si thickness of 210 nm. The lattice shift in the LSPCW is usually used to flatten the group index ng spectrum, but in this study, we employed a PCW with no lattice shifts so that we can set an appropriate ng by changing the wavelength. We used a moderate ng of no higher than 40 to avoid a large induced loss particularly near the band edge. We also employed a simple p-n junction, for which the p-and n-type doping concentrations were set at NA = 1.05 × 1018 cm−3 and ND = 6.2 × 1017 cm−3, respectively. Highly doped regions for ohmic contact were arranged outside with a spacing of 4 μm to avoid optical absorption. The width of the signal (S) electrodes was 10 μm, and the distance from the ground (G) electrodes was 7 μm. The G electrodes were commonized to avoid the coupled slot-line modes of RF signals [11]. The width of the meander line electrodes was the same as the S electrodes’, and the length of the delay line part (Ld) was set at 422, 804, 1186, or 1568 μm for each device. Figure 2 shows the device where Ld = 1186 μm, and the one-way RF transmission loss added by this delay line part was estimated to be 0.75 dB from the S parameter measurement [11]. The termination resistors were integrated on the chip, and their values were varied between 20, 30, or 50 Ω for each device. The VπL for the DC bias were estimated to be ~0.6, ~0.4 and ~0.3 V∙cm for ng = 20, 30 and 40, respectively, all of which are much lower than usual Si rib-type modulators. The total on-chip loss was estimated to be 6–8 dB, which included a passive loss of 5–6 dB (a value measured for another device fabricated similarly) and additional loss under modulation of 1.0–2.6 dB; we tuned this value for each experiment, which are shown in each figure as the modulation loss (ML).
Fig. 1 Fabricated Si PCW Mach–Zehnder modulators. (a) Normal electrode device and dimensions of PCW. (b) Meander line electrode device.
Fig. 2 (a) Frequency response and (b) f3dB for different RL. The EO response was smoothed by using the moving average in the range of ~0.25 GHz.
3. Frequency response
We investigated the RL dependence of the frequency response using the Ld = 1186 μm device. The frequency responses for different values of RL are shown in Fig. 2(a). With the termination resistors, S11 showing the reflection of the RF signals was reduced as compared with the open termination, and the EO response was improved. Figure 2(b) shows the change in the cutoff frequency f3dB versus the RL characteristics. The measured f3dB for the open termination was 14 GHz, whereas it increased to higher than 30 GHz when RL = 20–50 Ω. The responses at RL = 20 and 30 Ω were better than that for RL = 50 Ω. In particular, f3dB reached 38 GHz when RL = 20 Ω. This is explained by considering the phase of the RF reflectivity at the termination, ΓL, which is given by
where Z0 is the characteristic impedance of the phase shifters. If RL − Z0 < 0, ΓL also takes a negative value, and this causes the phase inversion of the reflected RF signals, which suppresses the low-frequency response and results in a pre-emphasis of the high-frequency components. The frequency response, η(f), of the meander line electrode device is expressed as follows [11]: where Zg is the internal impedance of the signal generator, Vg is the drive voltage from the signal generator, γ is the complex propagation constant of the RF signals, for which α is the attenuation constant and βRF is the propagation constant, nd is the refractive index in the delay line, nRF is the junction, and Cpn is the capacitance of the p-n junction. The solid line in Fig. 2(b) shows the calculated f3dB when L = 200 μm, Z0 = 50 Ω, nd = 2, nRF = 4, ng = 20, α = 0, Rpn = 60 Ω, and Cpn = 50 fF. This fits well with the measurement results. Using RL = 50 Ω is a well-balanced choice when considering the modulation efficiency, whereas RL = 20 Ω is more advantageous for high-speed modulation although it decreases the low-frequency phase shift to 1 + ΓL = 1 + (20–50)/(20 + 50) = 0.57 times.Figure 3(a) shows the measured frequency responses for different ng at different wavelengths, where RL was fixed at 50 Ω. The frequency response improved as Ld became larger. Compared with that for ng = 20, which is a standard value employed in LSPCW modulators, the Ld dependence became larger for ng = 40. Figure 3(b) summarizes the Ld dependence of f3dB. When ng = 20 and Ld = 1186 μm, f3dB was evaluated to be 31 GHz, which is close to the theoretical value. Even for ng = 30 and 40, f3dB improved to ~30 GHz by lengthening the delay line part. This clearly shows the effect of the phase matching.
Fig. 3 (a) Frequency response and (b) f3dB for different Ld. The EO response was smoothed using the moving average in the range of ~0.25 GHz.
4. Modulation experiments
4.1 25- and 32-Gbps modulation
We operated the 50 Ω-terminated devices at 25 and 32 Gbps with relatively large values of ng of 30 and 40, which had not been employed so far due to the phase mismatch. We used a pulse pattern generator (PPG, Anritsu, MP1800A) and sampling oscilloscope (Keysight Tech., 86100C, 86109A) for the eye pattern observation, where the speed limit was 30 GHz. We set the initial phase difference between the two arms in the Mach-Zehnder interferometer at the quadrature point using the thermo-optic phase tuner. Figure 4(a) shows a 25 Gbps eye pattern at Vpp = 1 V, VDC = −0.5 V, and Fig. 4(b) shows a 32 Gbps eye pattern at Vpp = 2 V, VDC = −1 V. In both cases, a clear open eye was observed at such low voltages. The bit energy consumption in the pair of p-n junctions, W, is given by CpnVpp2/2 for the junction capacitance Cpn. Using a calculated value for Cpn of 50 fF [10], CpnVpp2/2 = 25 and 100 fJ/bit were evaluated when Vpp = 1 and 2 V operation, respectively. We also evaluated the consumption in the termination resistors as 0.8 and 2.5 pJ/bit, where the latter is dominant but is still moderately small.
Fig. 4 Eye patterns of meander line electrode devices. (a) 25 Gbps, ng = 40, Vpp = 1 V, VDC = −0.5 V. (b) 32 Gbps, ng = 30, Vpp = 2 V, VDC = −1.0 V.
4.2 40-Gbps modulation
Next, 40 Gbps modulation was carried out. Because the aforementioned PPG and sampling oscilloscope did not have sufficient bandwidths for this speed, we changed them to Alnair Labs, SeBERT-1040C and EYE-1000C, respectively; the latter has a temporal resolution of 1 ps. The results are shown in Fig. 5. Compared with the normal electrode device, the meander line electrode device showed less noise and a clearer eye opening. When ng = 40, the response of the normal electrode device did not respond sufficiently, whereas that of the meander line device was almost maintained.
Fig. 5 40 Gbps eye patterns. Vpp = 3 V, VDC = −3 V. (a) Normal electrode device. (b) Meander line electrode device.
4.3 56–64 Gbps modulation
We also tried 50–64 Gbps modulation with the 20 Ω-terminated device. We generated 50, 56, and 64 Gbps signals by using a 2:1 multiplexer (MUX, SHF 601 A) from 2-channel 25–32 Gbps signal outputs from Anritsu's PPG. Figure 6(a) shows the signal when Vpp = 5.2–5.3 V (we could not set a moderate low voltage after the electrical amplification because of the absence of appropriate attenuators), where the rise and fall times were slightly degraded after amplification by using RF amplifiers (SHF 810, spec: ≤50 GHz) in addition to an equalization filter (SHF Eq. (25) A-3 dB, Nyquist frequency ≈25 GHz). Figure 6(b) shows the eye pattern when the modulator was driven by this signal at the quadrature point. The clear open eye was obtained even for the noisy 64 Gbps signal. This means that the pre-emphasized response in the 20 Ω-terminated device improved the eye quality. These results still have room for optimization of the geometry and impedance of the electrodes as well as the drive signal quality. We expect that the total device size, including the footprint of the electrodes, is reduced to that of a normal electrode device through these improvements. In the experiment at 64 Gbps, the total bit energy consumption at the junctions and terminations was estimated to be 21 pJ/bit. This large value was caused by the high drive voltage as well as the low termination resistance. Different from the experiments in Fig. 4, we used ng = 20 for this experiment to avoid the phase mismatch particularly increased at this high speed. If we further optimize the meander line electrode so that ng = 40 is usable, the voltage will be reduced to half and the energy consumption will be reduced to ~5 pJ/bit.
Fig. 6 50–64 Gbps eye patterns. (a) RF signal generated by MUX. (b) Modulated signal by meander line electrode device.
Table 1 compares the performance between Si Mach-Zehnder modulators operating beyond 50 Gbps. Our device exhibited a comparable ER value even at a much shorter L of 200 μm, while Vpp = 5.2 V should be reduced to use CMOS drivers with low power consumption. A low Vpp will be obtained by the optimization of the meanderline electrode for higher ng.
Table 1. Comparison between Si Mach-Zehnder modulators.
5. Conclusion
In summary, we demonstrated the 64 Gbps operation of a compact Si photonic crystal modulator with 200-μm phase shifters, meander line electrodes, and termination resistors to compensate for the phase mismatch between slow light and RF signals. We observed f3dB = 31 GHz for ng = 20 and f3dB ≈30 GHz for ng = 40 with the 50 Ω termination. At 25–32 Gbps, the eye quality was much improved, even at low voltages. Furthermore, we observed f3dB = 38 GHz and a clear open eye at 50–64 Gbps with the pre-emphasized frequency response of the 20 Ω-terminated device. These results contribute to the development of CMOS-compatible modulators of compact size, low voltage, low power consumption, wide working spectrum, and a temperature tolerance toward 400 Gbps optical interconnects.
Funding
New Energy and Industrial Technology Development Organization (NEDO) (project #13004).
Acknowledgment
We thank Mr. T. Okitsu, SHF Japan Corp., and Mr. O. Takezawa, I-Wave Corp., for the use of their equipment.
References
1. “IEEE standard for ethernet - amendment 10: media access control parameters, physical layers, and management parameters for 200 Gb/s and 400 Gb/s operation,” IEEE Std 802.3bs-2017 (Amendment to IEEE 802.3–2015 as amended by IEEE’s 802.3bw-2015, 802.3by-2016, 802.3bq-2016, 802.3bp-2016, 802.3br-2016, 802.3bn-2016, 802.3bz-2016, 802.3bu-2016, 802.3bv-2017, and IEEE 802.3–2015/Cor1–2017) 1–372 (2017).
2. D. J. Thomson, F. Y. Gardes, J.-M. Fedeli, S. Zlatanovic, Y. Hu, B. P. P. Kuo, E. Myslivets, N. Alic, S. Radic, G. Z. Mashanovich, and G. T. Reed, “50-Gb/s silicon optical modulator,” IEEE Photonics Technol. Lett. 24(4), 234–236 (2012). [CrossRef]
3. M. Streshinsky, R. Ding, Y. Liu, A. Novack, Y. Yang, Y. Ma, X. Tu, E. K. S. Chee, A. E.-J. Lim, P. G.-Q. Lo, T. Baehr-Jones, and M. Hochberg, “Low power 50 Gb/s silicon traveling wave Mach-Zehnder modulator near 1300 nm,” Opt. Express 21(25), 30350–30357 (2013). [CrossRef] [PubMed]
4. H. Xu, X. Li, X. Xiao, P. Zhou, Z. Li, J. Yu, and Y. Yu, “High-speed silicon modulator with band equalization,” Opt. Lett. 39(16), 4839–4842 (2014). [CrossRef] [PubMed]
5. D. Patel, S. Ghosh, M. Chagnon, A. Samani, V. Veerasubramanian, M. Osman, and D. V. Plant, “Design, analysis, and transmission system performance of a 41 GHz silicon photonic modulator,” Opt. Express 23(11), 14263–14287 (2015). [CrossRef] [PubMed]
6. M. Pantouvaki, S. A. Srinivasan, Y. Ban, P. De Heyn, P. Verheyen, G. Lepage, H. Chen, J. De Coster, N. Golshani, S. Balakrishnan, P. Absil, and J. Van Campenhout, “Active components for 50 Gb/s NRZ-OOK optical interconnects in a silicon photonics platform,” J. Lightwave Technol. 35(4), 631–638 (2017). [CrossRef]
7. J. Sun, M. Sakib, J. Driscoll, R. Kumar, H. Jayatilleka, Y. Chetrit, and H. Rong, “A 128 Gb/s PAM4 silicon microring modulator,” in Opt. Fiber Commun. Conf. Postdeadline Papers (Optical Society of America, 2018), Th4A.7. [CrossRef]
8. C. Haffner, D. Chelladurai, Y. Fedoryshyn, A. Josten, B. Baeuerle, W. Heni, T. Watanabe, T. Cui, B. Cheng, S. Saha, D. L. Elder, L. R. Dalton, A. Boltasseva, V. M. Shalaev, N. Kinsey, and J. Leuthold, “Low-loss plasmon-assisted electro-optic modulator,” Nature 556(7702), 483–486 (2018). [CrossRef] [PubMed]
9. H. C. Nguyen, N. Yazawa, S. Hashimoto, S. Otsuka, and T. Baba, “Sub-100 μm photonic crystal Si optical modulators: spectral, athermal, and high-speed performance,” IEEE J. Sel. Top. Quantum Electron. 19(6), 127–137 (2013). [CrossRef]
10. Y. Terada, T. Tatebe, Y. Hinakura, and T. Baba, “Si photonic crystal slow-light modulators with periodic p–n junctions,” J. Lightwave Technol. 35(9), 1684–1692 (2017). [CrossRef]
11. Y. Hinakura, Y. Terada, H. Arai, and T. Baba, “Electro-optic phase matching in a Si photonic crystal slow light modulator using meander-line electrodes,” Opt. Express 26(9), 11538–11545 (2018). [CrossRef] [PubMed]
12. K. Hojo, Y. Terada, N. Yazawa, T. Watanabe, and T. Baba, “Compact QPSK and PAM modulators with Si photonic crystal slow-light phase shifters,” IEEE Photonics Technol. Lett. 28(13), 1438–1441 (2016). [CrossRef]
