First demonstration of long-haul transmission using silicon microring modulators

: We report error-free long-haul transmission of optical data modulated using a silicon microring resonator electro-optic modulator with modulation rates up to 12.5 Gb/s. Using bit-error-rate and power penalty characterizations, we evaluate the performance of this device with varying modulation rates, and perform a comparative analysis using a commercial electro-optic modulator. We then experimentally measure the signal integrity degradation of the high-speed optical data with increasing propagation distances, induced chromatic dispersions, and bandwidth-distance products, showing error-free transmission for propagation distances up to 80 km. These results confirm the functional ubiquity of this silicon modulator, establishing the potential role of silicon photonic interconnects for chip-scale high-performance computing systems and memory access networks, optically-interconnected data centers, as well as high-performance telecommunication networks spanning large distances.

We demonstrate, for the first time to our knowledge, error-free operation of a silicon microring resonator electro-optic modulator for modulation rates up to 12.5 Gb/s, and draw a comparison of this device with a commercial lithium niobate (LiNbO 3 ) Mach-Zehnder electro-optic modulator. To date, all proposed silicon photonic network-on-chip (NoC) architectures have employed silicon microring resonator modulators [1,[3][4][5][6][7]. The measurements demonstrated in this work may enable more accurate physical layer modeling of these photonic devices for optically-interconnected high-performance computing systems [36,37].
In this work, to the best of our knowledge, we also present the first experimental demonstration of an error-free long-haul (up to 80-km) transmission using a high-speed (up to 12.5-Gb/s) and compact (12-µm diameter) silicon microring modulator, demonstrating a bandwidth-distance product up to 1000 Gb-km/s. We experimentally measure less than a 1-dB power penalty associated with the bandwidth-distance product of 600 Gb-km/s, comparable to commercial telecommunications-grade modulators. We also perform bit-error-rate (BER) and power penalty characterizations for varying propagation distances, induced chromatic dispersions, and bandwidth-distance products.

Silicon microring resonator electro-optic modulator
The silicon microring resonator electro-optic modulator in this work comprises a microring resonator configured as a PIN carrier injection device [14]. The microring resonator has a 12µm diameter, optical quality factor of about 10,000, and is coupled to a straight waveguide with input and output ports (Fig. 1d). The waveguides are 450-nm wide and 260-nm tall. The waveguide of the microring has a 50-nm slab that is doped to form the PIN diode structure, with nickel silicide for the electrical contacts. The silicon modulator was fabricated using electron-beam lithography and reactive-ion etching [38]. The non-return-to-zero (NRZ) onoff-keyed (OOK) modulation signal is electro-optically encoded onto a wavelength channel when an incoming continuous wave (CW) source passes in and out of a shifting resonance of the microring resonator. When the resonance of the microring resonator is tuned onto the wavelength of the incoming signal, the light is coupled into the microring resonator, corresponding to a logical "0" bit. When the resonance of the microring resonator is tuned away from the wavelength of the incoming signal, the signal bypasses the microring resonator and leaves on the output port, corresponding to a logical "1" bit. The resonance shift is induced by the plasma-dispersion effect from injecting and extracting electrical carriers through the PIN diode, which is integrated into the microring resonator. To achieve high modulation rates that are typically limited by carrier lifetimes, the modulator is driven using a pre-emphasis method [1,14].

Experimental setup
The experimental setup for characterizing the modulation rate dependence of the silicon microring modulator (Fig. 1a) involves a tunable laser (TL) source generating a CW 1565-nm lightwave that is coupled on chip using a tapered fiber. The lightwave is then modulated on chip using the microring resonator, which is driven by a pulse pattern generator (PPG) generating a 2 7 -1 pseudo-random bit sequence (PRBS), followed by a pre-emphasis circuit. Off chip, the signal passes through an erbium-doped fiber amplifier (EDFA), a tunable grating filter (λ), and a variable optical attenuator (VOA). The signal is received by a high-speed PIN photodiode and transimpedance amplifier (PIN-TIA) receiver followed by a limiting amplifier (LA), and is evaluated using a BER tester (BERT). Both the PPG and the BERT are synchronized to the same clock. Using a power tap, a communications signal analyzer (CSA) is used to examine the temporal response of the signal before the receiver. Polarization controllers (PCs) are used throughout the setup. We actively modulate the microring resonator with a 2.5-V PP electrical signal with about a 1-V voltage bias.
The experimental setup for comparing the performance of the silicon modulator with a commercial LiNbO 3 Mach-Zehnder electro-optic modulator is similar to the aforementioned experimental setup, except the modulation occurs off chip using the LiNbO 3 modulator, before the optical signal is coupled on chip (Fig. 1b). Once on chip, this modulated signal passes by the silicon microring resonator off resonance.
In order to inspect the signal integrity degradation induced by chromatic dispersion effects, the lightwave is again modulated on chip using the microring resonator. Off chip, the signal passes through an EDFA and a tunable grating filter (λ) before passing through varying lengths of standard single-mode optical fiber (SSMF), set to 1-, 2-, 5-, 10-, 15-, 40-, 60-, and 80-km lengths, as well as the 0-km back-to-back case bypassing the optical fiber. After leaving the optical fiber, the signal travels through a VOA, is received by a high-speed PIN-TIA receiver followed by a LA, and is evaluated using a BERT. Again, both the PPG and the BERT are synchronized to the same clock, which is now set to either 10 or 12.5 GHz. Using a power tap, a CSA is used to examine the temporal response of the signal before the receiver. Here, the pre-emphasis circuit is optimized for each modulation rate, and is then kept constant for the varying transmission configurations of propagation distance.

Experimental validation of modulation rate and comparative analysis
Using the experimental setup depicted in Fig. 1a and described in the experimental setup section, we first evaluate the silicon modulator at varying modulation rates by examining the output modulation temporal response. This is accomplished by setting the clock rate to 5, 7.5, 10, and 12.5 GHz, electrically driving the silicon modulator with 5-, 7.5-, 10-, and 12.5-Gb/s NRZ OOK data. The electro-optic response of the silicon modulator then encodes the incoming light with the electrical data. Once that data signal leaves the chip, the eye diagrams of the optical signals are measured (Fig. 2a-d). The resulting eye diagrams show clear openings, and degrade as we drive the modulator at higher modulation rates, resulting from electrical carrier lifetime limitations as well as the transient response of the microring resonator and the closing of the temporal window. The pre-emphasis circuit to enable carrier injection and extraction is optimized separately for each modulation rate.
For each modulation rate configuration of the silicon modulator, we perform BER measurements of the resulting optical data signal leaving the chip (Fig. 3a). We first observe error-free operation (defined as having BERs less than 10 -12 ) for each configuration. Subsequently, BER curves are recorded for the 5-, 7.5-, 10-, and 12.5-Gb/s modulation rates. The experimentally-measured BER curves confirm the signal integrity degradation as the modulation rate is increased.
Using the experimental setup depicted in Fig. 1b and described in the experimental setup section, we compare the performance of the silicon modulator with a commercial LiNbO 3 Mach-Zehnder electro-optic modulator at varying modulation rates. This is accomplished by setting the clock rate to 5, 7.5, 10, and 12.5 GHz, electrically driving the LiNbO 3 modulator with 5-, 7.5-, 10-, and 12.5-Gb/s data. The electro-optic response of the LiNbO 3 modulator then encodes the incoming light with this data. The resulting optical data signal is then inserted into the silicon chip, bypassing the microring resonator. Once the data signal leaves the silicon chip, the eye diagrams of the optical signal are evaluated (Fig. 2e-h). The resulting eye diagrams show clear openings, and degrade as we drive the modulator at higher modulation rates, resulting mostly from the closing of the temporal window. The modulation bias is optimized for each configuration. For each modulation rate configuration of the LiNbO 3 modulator, we perform BER measurements of the resulting optical data signal leaving the silicon chip (Fig. 3a). We observe error-free operation for each modulation rate, and subsequently record BER curves for the 5-, 7.5-, 10-, and 12.5-Gb/s modulation rates. Again, the experimentally-measured BER curves confirm the signal integrity degradation as the modulation rate is increased. To draw a system-level functional comparison between the two modulators, the BER curves measured for the LiNbO 3 modulator are set as the back-to-back cases for the BER curves measured for the silicon modulator at each modulation rate. The resulting power penalties of the operation of the silicon modulator compared to the LiNbO 3 modulator are 1.2, 1.65, 3.15, and 5.4 dB for 5-, 7.5-, 10-, and 12.5-Gb/s modulation rates, respectively, at the BER of 10 -9 (Fig. 3b). Much of these power penalties may be further improved with more optimal pre-emphasis configurations for each modulation rate, as well as closer integration of the electrical driving circuit with the silicon modulator [1].   3. Experimentally-measured system-level performance characterizations of the silicon microring resonator electro-optic modulator for modulation rates between 5 and 12.5 Gb/s. (a) Bit-error-rate curves for an optical signal encoded using the silicon modulator, as well as a commercial LiNbO3 Mach-Zehnder electro-optic modulator, for 5-, 7.5-, 10-, and 12.5-Gb/s modulation rates. (b) Resulting power penalty associated with the operation of the silicon modulator compared to the LiNbO3 modulator for modulation rates between 5 and 12.5 Gb/s.

Experimental evaluation of long-haul transmission
Using the experimental setup depicted in Fig. 1c and described in the experimental setup section, we study the signal integrity degradation induced by the chromatic dispersion effects caused by varying propagation distances through SSMF, resulting from the induced chirp in the silicon modulator. We first transmit a 10-Gb/s modulated signal through SSMF lengths of 0, 1, 2, 5, 10, 15, 40, 60, and 80 km, inducing proportional amounts of chromatic dispersion. The eye diagrams of the resulting optical signals are evaluated for each configuration (Fig.  4a-i). The eye diagrams show clear openings, and remain relatively unchanged for optical fiber lengths up to 15 km (Fig. 4a-f). The eye diagrams subsequently begin to lose temporal window and display increased noise after 40-km propagation distances, as dispersion effects become more distinct (Fig. 4g), displaying noticeable degradation at 60 and 80 km (Fig. 4h,i).
To evaluate this dependence of the signal integrity degradation on modulation rate, the silicon  modulator is subsequently evaluated at a 12.5-Gb/s modulation rate for the 0-and 80-km optical fiber lengths (Fig. 4j,k). After this 80-km transmission, the eye diagram of the resulting optical signal is evaluated (Fig. 4k). Again, the eye diagram shows noticeable degradation from the induced chromatic dispersion.
We then quantify the signal integrity degradation caused by the chromatic dispersion effects induced by long-haul propagation using BER measurements at varying propagation Fig. 6. Summary of experimentally-measured results for varying propagation distances through standard single-mode optical fiber, induced chromatic dispersions, and bandwidth-distance products, of an optical signal encoded using the using the silicon microring resonator electrooptic modulator for modulation rates between 10 and 12.5 Gb/s. distances through SSMF. We perform BER measurements at the 10-Gb/s modulation rate for each propagation distance configuration (Fig. 5a). For each configuration, error-free operation is initially observed. BER curves are then recorded for the 0-, 1-, 2-, 5-, 10-, 15-, 40-, 60-, and 80-km propagation distances (Fig. 5a). Setting the configuration bypassing the SSMF as the back-to-back case, the measured power penalty is recorded for each propagation distance (Fig.  5b). Compared to the 0-km propagation distance, the measured power penalties remain constant at 0 dB for all propagation distances up to 40 km, and are 0.6-and 2.5-dB for propagation distances of 60 and 80 km, respectively, at the BER of 10 -9 (Fig. 5b). Using the chromatic dispersion for a given propagation distance characteristic of the SSMF employed in these measurements, which is about 17 ps/(nm-km) at the 1565-nm operating wavelength, the measured power penalties remain constant at 0 dB for all induced chromatic dispersions up to 680 ps/nm, and are 0.6 and 2.5 dB for induced chromatic dispersions of 1020 and 1360 ps/nm, respectively, at the BER of 10 -9 .
We then record BER curves at the 12.5-Gb/s modulation rate for the 0-and 80-km propagation distances (Fig. 5a). Once again, setting the configuration bypassing the SSMF as the back-to-back case produces a 2.5-dB power penalty. The measured power penalty for each bandwidth-distance product configuration is then determined, producing power penalties that, for the 10-Gb/s modulation rate, remain constant at 0 dB for all bandwidth-distance products up to 400 Gb-km/s, and are 0.6 and 2.5 dB for bandwidth-distance products of 600 and 800 Gb-km/s, respectively. For the 12.5-Gb/s modulations rate, the data signal incurs a 2.5-dB power penalty for the 1000 Gb-km/s bandwidth-distance product. These experimental results are summarized in Fig. 6.

Conclusion
Using BER and power penalty characterizations, we have demonstrated a silicon microring resonator electro-optic modulator with error-free transmission for modulation rates up to 12.5 Gb/s. We observed the effects on signal integrity with a varying modulation rate, quantifying the relationship between modulation rate and system-level performance. We then performed a system-level comparative analysis of the silicon modulator using a commercial LiNbO 3 Mach-Zehnder electro-optic modulator.
Furthermore, we have also experimentally demonstrated error-free long-haul transmission of optical signals modulated using the silicon modulator, obtaining a bandwidth-distance product up to 1000 Gb-km/s with an 80-km propagation distance and 12.5-Gb/s modulation rate. We characterized the effects of the induced chromatic dispersion on signal integrity, measuring less than a 1-dB power penalty for optical signals modulated at 10 Gb/s and propagated through 60 km of SSMF. We also reported the resulting power penalties for varying propagation distances up to 80 km, induced chromatic dispersions up to 1360 ps/nm, and bandwidth-distance products up to 1000 Gb-km/s. Quantifiable performance metrics extracted from experimental validation of silicon photonic devices aid in determining the functionality that these devices perform in large-scale photonic network architectures. We demonstrate that this silicon modulator is truly a versatile silicon photonic device, capable of enabling high-performance transmission for a wide range of short-, medium-, and long-haul applications.