Open Access
Add to BibSonomyAbstract
Individual channels of a five-channel microring silicon photonics transmitter are used for bit error ratio analysis and demonstrate error-free transmission at 10Gb/s. Two channels of the same transmitter are concurrently modulated using an 80GHz channel spacing comb laser and demonstrate open eye diagrams at 10Gb/s and 12.5Gb/s. Finally, concurrent modulation with tunable lasers is done to quantify optical power penalty for link bit error ratio versus channel spacing from +100GHz to −100GHz. When using a comb laser for concurrent modulation, no direct power penalty is observed for an 80GHz channel separation.
© 2015 Optical Society of America
1. Introduction
In order to meet the data traffic demand in data centers, dense wavelength division multiplexing (DWDM) architectures that use silicon microring resonator modulators (Si-RRM) have been proposed [1, 2]. These devices are desirable in such applications due to their small footprint, low insertion loss and inherent wavelength selectivity [3]. In DWDM links, channel count, spacing, and data rate are key factors in determining the overall link bandwidth. The inter-channel crosstalk of these devices has been investigated previously[4] to quantify the optical power penalty with decreasing channel spacing. The work presented here further extends on this study by using rings with higher cavity quality factors, larger drive voltages, and also studies the impact of comb lasers as optical sources versus only commercial benchtop tunable lasers. The optical sources used for DWDM have also been under investigation [5, 6]. In these works, a single cavity is designed to generate multiple optical tones at a desired frequency spacing and have been combined with modulators with high-speed data modulation [7]. Most recently, concurrent DWDM transmission with Si-RRMs was demonstrated at 10Gb/s with a 240GHz channel spacing [8]. The work presented in this paper demonstrates error-free (bit error ratio of less than 1 × 10−12) concurrent modulation at 80GHz channel spacing at 10 and 12.5Gb/s with a comb laser and quantifies the optical power penalty on channel bit error ratio (BER) as a function of channel spacing using a pair of tunable lasers.
2. Background & theory
BER is a unitless number defined as the ratio between incorrect and overall bits transmitted[9]. Due to the statistical nature of BER, an integration time is associated with the reported number. More accurately, a confidence limit of a reported BER value is defined by the integration time, T (in seconds) as [9]:
where f is the data rate, BER is the desired error ratio minimum value to be resolved, and CL is the confidence limit (between 0−1). For this work, an integration time of 4 minutes is used for a BER minimum of 1× 10−12 at a data rate of 10Gb/s. During concurrent modulation, one can expect an increase in BER from such effects as DC optical crosstalk, modulation sidebands [9], among others. Such deleterious effects as these can be overcome by an increase in received optical power in order to achieve error-free transmission. This increase in the optical power is defined as the optical power penalty (OPP).When designing ring-based DWDM links, the trade-off between available free spectral range (FSR) determined by the diameter, channel spacing, and individual channel modulation data rate must be carefully considered. Denser channel spacing, higher channel count, and faster data rates enable higher overall bandwidth. Denser channel spacing can also possibly enable narrower laser gain bandwidth which would enable higher optical powers [5] for optical comb generation. Carrier-depletion Si-RRMs have achieved up to 56Gb/s [10] and carrier-injection-based rings have demonstrated data rates of 20Gb/s and 25Gb/s [11, 12]. But these might not always be the most cost-effective choices given restrictions of crosstalk pertaining to increased receiver sensitivity, higher laser power, more strict thermal management, and sharper optical filtering. The need for higher optical sensitivity at the receiver is especially pronounced in the case of carrier-depletion Si-RRMs given their lower extinction ratio (ER) and higher insertion loss. Laser power consumption is also a significant part of the overall link power budget. As an example, with the use of 10µm diameter rings, 25 channels can be assigned into one FSR at a spacing of 80GHz, giving a bi-directional bandwidth of 500Gb/s for individual channel data rate of 10Gb/s. Increasing the data rate up to 12.5Gb/s for the same channel spacing increases the bandwidth to 625Gb/s. The aim of the work presented here is to quantify the associated OPP with this bandwidth increase.
3. Concurrent DWDM modulation
The experiments in this section show transmission from each channel individually and quantify system noise limits of error-free detection. Subsequently, concurrent, two-channel transmission with a comb laser demonstrates a possible DWDM implementation with a commercially available, low-noise, high output power, comb laser with a fixed 80GHz channel spacing. For the wavelength range of 1300−1320nm, a frequency separation of 80GHz corresponds to ~0.46nm. In the subsequent experiments, for the sake of brevity, we will refer to frequency values when discussing channel spacing and detuning.
3.1. Individual channel modulation
Figure 1(a) shows the optical transmission of the test structure which comprises a 350nm-wide bus waveguide to which five ring resonators (10µm diameter) are coupled with a through and drop gap of 250nm and 350nm, respectively, with a measured cavity Q factor of ~10,000. The radii of the rings were designed for an 80GHz channel spacing on a 250nm,2µm SOI wafer. However, due to process variation, the sequence of resonance wavelengths is switched for channels 5 and 7 as compared to the physical layout. Integrated silicon resistive heaters enable resonance tuning of the channel resonance wavelength/frequency similar to [13] and modulation is done by carrier injection.
Fig. 1 Figure showing a) output spectrum of the transmitter, b) experimental setup, c) individual channel eye diagrams, d) individual channel BER at 10Gb/s.
The experimental setup used for individual channel modulation is shown in Fig. 1(b). A commercial laser (Santec TSL-510) serves as the optical excitation source. The optical output is passed through a variable optical attenuator (VOA) and detected by a high-sensitivity photodetector (HSPD) (Discovery Semiconductor). The HSPD has two outputs which allow simultaneous measurement of the BER and electrical eye diagram via a BER module and a Digital Communications Analyzer (DCA). A 10Gb/s pre-emphasized electrical driving voltage with a 27 −1 PRBS word length is generated, similar to [8], with 2.5V,0.6V for VPP and VDC of DATA and 1.9V for VPP of/DATA, respectively. With these drive voltages, the rings achieve a 25GHz spectral shift in the Lorentzian lineshape. Each channel is modulated individually and the resulting eye diagrams are shown in Fig. 1(c) at −19dBm received optical power at the HSPD. Figure 1(d) shows each channel’s BER across a range of received powers input to the HSPD. As a point of reference, the BER of a commercial Lithium Niobate modulator (EOSpace) was also measured and plotted using the same optical setup described in Fig. 1(b). The marked difference between the BER performance of the Mach-Zehnder based reference modulator versus the ring-resonator based transceiver channels can be attributed to inter-symbol interference (ISI) due to the the bandwidth limitations of carrier-injection modulator.
3.2. Comb laser-driven concurrent modulation
Next, a comb laser (Innolume Gmbh) (described previously in [8]) is used to optically excite multiple channels simultaneously; Figure 2(a) shows the input spectrum into the photonic chip. The comb laser was biased such that one of the tones is aligned with channel 7 of the transmitter device. A representative RF and DC probe configuration is shown in the microscope image of Fig. 2(b). Channel 3 of the device is then thermally tuned to an optical tone +80GHz w.r.t to that of channel 7. Both of the channels are then modulated with two uncorrelated PRBS data sources. An optical bandpass filter with a 35GHz-wide square transmission bandwidth (slope of 1.5dB/GHz) is used to individually select and monitor the eye diagram of each channel.
Fig. 2 Figure showing a) input from comb laser to the transmitter, b) experimental setup with two RF probes and DC probe for thermal tuning, c) eye diagrams from concurrent modulation at 10Gb/s, d) eye diagrams from concurrent modulation at 12.5Gb/s.
Figure 2(c) shows output eye diagrams from concurrent 10Gb/s modulation of the two channels at an optically received power of −21dBm onto the HSPD. Channels 3 and 7 demonstrate a BER of 6.8 × 10−6 and 2.3 × 10−7, respectively. These values correspond to the curves shown in Fig. 1(d) for −21dBm input optical power when each channel is modulated individually. Due to non-optimal grating coupler design, ~20dB insertion loss from the photonic chip is measured. The reduced optical power at the HSPD is due to the lower maximum available power from the comb laser as compared to the commercial laser. From this data, no apparent degradation of BER is observed as a result of concurrent modulation of the channels 80GHz apart in frequency. To test the impact of optical sidebands that are generated due to modulation and increased spectral content due longer PRBS word lengths, the rings were concurrently modulated at 12.5Gb/s with a PRBS word length of 231 − 1 bits and optically excited by the comb laser. The eye diagrams are shown in Fig. 2(d). Channel 3 and 7 maintained their 80GHz channel spacing and show a BER of 9.9 × 10−4 and 1.4 × 10−3, respectively. The increase BER value contributes to ~2dB OPP, which can thus be attributed to the longer word length and higher data rate, which increases ISI and crosstalk.
4. Crosstalk analysis
The comb laser has been replaced by two tunable lasers that are used as the optical excitation source to quantify crosstalk mechanism such as DC optical leakage, generation of optical sidebands due to modulation, and ring resonator lineshape blue-shift, all as a function of decreasing channel spacing. In the following experiments, the channel frequency detuning is defined as the frequency of the ring modulator less the frequency of the optical tone used for MOD or the unmodulated optical tone. This definition is analogous to the frequency spacing between optical carrier tones used for data communication applications.
4.1. DC optical crosstalk
The experimental setup is shown in Fig. 3(a). Two commercial lasers serve as optical sources and a commercial lithium niobate Mach-Zehnder modulator (MOD) (EOSpace) is used as a reference modulator. The rest of the setup remains as described previously and shown in Fig.1(b). Due to the Lorentzian lineshape of the ring resonator, light from neighboring optical channels can leak into the victim ring resonator, hereto referred to as DC optical leakage. For ring resonator cavities with the aforementioned Q values, a static optical crosstalk of −17dB is expected for an optical signal 80GHz away from the cavity resonance frequency. To quantify BER degradation from DC optical leakage, Ch. 1 (RING) is modulated as mentioned previously, while a DC optical tone is tuned from +100GHz to −100GHz. This unmodulated tone is passed through the MOD to maintain consistent insertion loss. The resulting device BER and eye diagrams are shown in Fig. 3(c) and Fig. 3(d), respectively. There is no clear significant trend of BER degradation above measurement noise as a function of frequency detuning down to 35GHz between the modulated and unmodulated optical tones.
Fig. 3 a) experimental setup, b) schematic showing the DC optical tone (dashed line) relative to the modulated tone from the Si-RRM (solid) line. The Lorentzian shape for the on/off state of the ring is also shown in dashed/solid lines, c) output eye diagrams at various values of detuning d) BER vs. received power for various detuning cases.
4.2. External modulator and ring
Carrier injection causes a blue shift, or an increase, in the Si-RRM resonance frequency due to electro-optic interaction. To test BER degradation from optical sidebands due to modulation and the resonance blue-shift, the MOD shown in the setup in Fig. 3(a) is modulated at 10Gb/s with PRBS 231 − 1 and the RING is modulated as previously described. The resulting BER plots and eye diagrams are shown in Fig. 4. Condition 1, schematically shown in the inset of Fig. 4(e), is defined as when the MOD resonance is at a higher frequency w.r.t. to the ring resonance frequency. Condition 2 is thus defined as the opposite. In order to optimize the RING’s output eye, the laser is tuned to the higher frequency side of the Lorentzian lineshape; the on/off state of the RING lineshape is shown by the dashed/solid line. As can be seen from Fig. 4(a) and Fig. 4(b), there is no significant BER degradation as the frequency detuning between the MOD and RING is decreased down to 35GHz for condition 1. However, in condition 2, an OPP of ~3.5dB is observed for a BER of 1 × 10−9 for the MOD BER at a detuning of −35GHz. This value can be observed by extrapolation of the measured data shown in Fig. 4(c) as no errors were measured for higher power ranges for the given integration time. For the same detuning and BER value, we see ~1.5dB OPP for the RING and this is confirmed by the noisier eye diagram shown in Fig. 4(e). This degradation can be primarily attributed to the lineshape blue shift of the RING as a consequence of being tuned to the higher frequency side of the Lorentzian lineshape. Should the RING be optically tuned to the lower frequency side of the lineshape, this degradation would be observed for the +35Ghz detuning. For frequencies outside of 50GHz, there is no significant BER degradation, implying minor OPP due to generation of optical sidebands and lineshape blue-shift for a 10Gb/s data rate.
Fig. 4 Plots for MOD and Ring BER for a),b) condition 1 and c),d) condition 2 and e) output eye diagrams for individual channel modulation and various values of detuning. A schematic of the detuning conditions is also shown.
4.3. Concurrent ring modulators
The final experiment involves concurrent modulation of two ring channels, Ch. 3 and Ch. 7. In the optical setup, the MOD has been removed and both lasers are injected into the photonic chip at equal power levels using a fiber combiner. The resonance frequency of Ch. 3 is swept from higher frequency w.r.t. to Ch. 7 from +100GHz to +35GHz, as shown in condition 1 of Fig. 5(a). Condition 2 is defined as when Ch. 3 is swept away from Ch.7 to lower frequency values, going from −35GHz to −80GHz. This is analogous to the arrangement above where the MOD has been replaced by Ch.3 RING modulator. In these experiments, Ch. 7 is modulated with the tunable laser aligned to the lower frequency side of its lineshape, as shown in Fig.5(a), whereas Ch. 3 is modulated on the higher frequency side. Therefore, we can expect maximum BER degradation due to lineshape blue shift and optical sidebands in condition 1 as Ch. 3 is brought closer to Ch. 7 in frequency. This is primarily due to the blue-shift from Ch. 3 crossing into the optical transmission window of Ch. 7 at +35GHz detuning. Conversely, we can expect less degradation in condition 2 where BER degradation can be attributed to optical sidebands only as the two modulator lineshapes overlap less due to the modulation blue-shift. From the BER plots in Fig. 5(b) and Fig. 5(c), and the eye diagrams in Fig. 5(f), we can see an increase in BER as channel spacing is reduced between Ch. 3 and 7 from +100GHz to +35GHz. For condition 1, an OPP of ~1dB and ~3dB are observed for Ch. 7 and Ch. 3, respectively. This OPP can be attributed to the generation of optical sidebands due to modulation, as well as the Lorentzian blue-shift previously described. Furthermore, the difference in the two channels’ OPP suggests a range of penalties depending on each device and its operating conditions. As Ch. 3 crosses to lower frequencies and moves further away, from Fig. 5(d) and Fig. 5(e), we can observe the BER vary within less than 1dB, which is within measurement noise. In this same scenario, no significant BER degradation is observed for Ch. 7. The eye diagrams shown in Fig. 5(f) show the output eye diagrams for each channel modulated individually and for frequency detuning in the rate of +80GHz to −80GHz, which also show no clear degradation in condition 2.
Fig. 5 a) schematic showing frequency tuning to each ring and relative position of each channel’s resonance frequency for the two test conditions, b)-e) show various BER plots vs. received power for the two channels and test conditions, f) shows output eye diagrams for each channel modulated individually and at various frequency detuning values.
5. Conclusion
Each channel of a five-channel Si-RRM transmitter is individually modulated and demonstrates error-free transmission at 10Gb/s and PRBS 27 − 1. Two channels are then modulated concurrently at 80GHz channel spacing and at 10Gb/s with a comb laser. This comb laser has lower optical input power as compared to the tunable laser. This decrease causes higher BER for the channels. This is confirmed by the BER observed by individual channel modulation for the same value of received optical power. Therefore, no direct power penalty due to concurrent modulation is observed for this channel spacing. Finally, for concurrent modulation, at 12.5Gb/s and PRBS 231 −1 a power penalty of ~2dB is observed for the given channel spacing which can be attributed to noise from the comb laser, longer PRBS word length, ISI, and optical sidebands due to modulation.
In the next set of experiments, deleterious effects such as DC optical leakage, optical sideband generation, and Lorentzian lineshape blue-shift are investigated as a function of channel separation. It is found that optical sidebands contribute significantly to increased OPP for channels spacing of 35GHz at a 10Gb/s data rate as demonstrated by an external modulator and a single ring channel. This frequency spacing is not, however, a fundamental limit and depends on the optical filter used at the receiver, as well as the ring resonator cavity Q-factor (10,000 for the used devices). From concurrent ring modulation, we see 1−3dB power penalty at +50GHz when the two rings are biased as previously mentioned and at +35GHz, the eye is closed. From this data, we conclude that channel spacing as small as 60 − 50GHz can be used in DWDM applications with moderate power penalty for modulation at 10Gb/s.
The preceding experiments aim to quantify the design trade-offs between data rate and channel spacing. Future work remains to further analyze the OPP associated with denser channel spacing and higher modulation data rates. These numbers must be considered carefully depending on the application and the desired bandwidth as they directly impact overall link power consumption and cost. Cost-effective design of DWDM links will help successfully integrate silicon photonics in commodity markets such as data center and high-performance computers.
References and links
1. R. Beausoleil, J. Ahn, N. Binkert, A. Davis, D. Fattal, M. Fiorentino, N. Jouppi, M. McLaren, C. Santori, R. Schreiber, S. Spillane, D. Vantrease, and Q. Xu, “A Nanophotonic Interconnect for High-Performance Many-Core Computation,” Integrated Photonics and Nanophotonics Research and Applications (IPNRA) (Optical Society of America, 2008), paper ITuD2. [CrossRef]
2. A. Boletti, P. Boffi, P. Martelli, M. Ferrario, and M. Martinelli, “Performance analysis of communication links based on VCSEL and silicon photonics technology for high-capacity data-intensive scenario,” Opt. Express 23(2), 1806–1815 (2015). [CrossRef] [PubMed]
3. P. Dong, S. Liao, D. Feng, H. Liang, D. Zheng, R. Shafiiha, C.C. Kung, W. Qian, G. Li, X. Zheng, A. Krishnamoorthy, and M. Asghari, “Low Vpp, ultralow-energy, compact, high-speed silicon electro-optic modulator,” Opt. Express 17(25), 22484–22490 (2009). [CrossRef]
4. K. Padmaraju, X. Zhu, L. Chen, and M. Lipson, “Intermodulation crosstalk characteristics of WDM silicon microring modulators,” IEEE Photon. Technol. Lett. , 26(14), 1478–1481 (2014). [CrossRef]
5. G. Wojcik, D. Yin, A. Kovsh, A. Gubenko, E. Alexey, I. Krestnikov, L. Igor, S. Mikhrin, D. Livshits, D. Fattal, A. David, M. Fiorentino, and R. Beausoleil, ”A single comb laser source for short reach WDM interconnects,”Proc. SPIE 7230, 72300M (2009).
6. S. Joshi, C. Calo, N. Chimot, M. Radziunas, R. Arkhipov, S. Barbet, A. Accard, A. Ramdane, and F. Lelarge, “Quantum dash based single section mode locked lasers for photonic integrated circuits,” Opt. Express 22(9), 11254–11266 (2014). [CrossRef] [PubMed]
7. J. Muller, J. Hauck, B. Shen, S. Romero-Garcia, E. Islamova, S. S. Azadeh, S. Joshi, N. Chimot, A. Moscoso-Martir, F. Merget, F. Lelarge, and J. Witzens, “Silicon photonics WDM transmitter with single section semiconductor mode-locked laser,” Adv. Opt. Techn. ,4(2), 119–145 (2015).
8. C.-H. Chen, M.A. Seyedi, M. Fiorentino, D. Livshits, A. Gubenko, S. Mikhrin, V. Mikhrin, and R. Beausoleil, “A comb laser-driven DWDM silicon photonic transmitter based on microring modulators,” Opt. Express 23(16), 21541–21548 (2015). [CrossRef] [PubMed]
9. S. Haykin, An Introduction to Analog and Digital Communication (Wiley & Sons, 1989)
10. M. Pantouvaki, P. Verheyen, J. De Coster, G. Lepage, P. Absil, and J. Van Campenhout, ”56 Gb/s ring modulator on a 300mm silicon photonics platform,” European Conference on Optical Communication (ECOC) (IEEE, 2015) pp. 4–6.
11. S. Tanaka, A. Hayakawa, T. Akiyama, M. Imai, S. Sekiguchi, Y. Tanaka, and K. Morito, ”Wide-temperature-range, 25-Gbps error-free operation of integrated silicon photonic transmitter using wavelength-controlled micro ring modulator,”) European Conference on Optical Communication (ECOC) (IEEE, 2015) pp. 1–3.
12. R. Wu, C.-H. Chen, T.-C. Huang, K. Cheng, and R. Beausoleil, ”20 Gb/s carrier-injection silicon microcing modulator with SPICE-compatible dynamic model,” in 2015 International Conference on Photonics in Switching (PS) (IEEE, 2015) pp.304–306.
13. C. Li, R. Bai, A. Shafik, E.Z. Tabasy, G. Tang, C. Ma, C.H. Chen, Z. Peng, M. Fiorention, P. Chiang, and S. Palermo, ”A ring-resonator-based silicon photonics transceiver with bias-based wavelength stabilization and adaptive-power-sensitivity receiver,”) Solid-State Circuits Conference Digest of Technical Papers (ISSCC) (IEEE, 2013) pp. 124–125.
