Simultaneous RZ-OOK to NRZ-OOK and RZ-DPSK to NRZ-DPSK format conversion in a silicon microring resonator

: Simultaneous RZ-OOK to NRZ-OOK and RZ-DPSK to NRZ-DPSK modulation format conversion in a single silicon microring resonator with free spectral range equal to twice the signal bit rate is experimentally demonstrated for the first time at 41.6 Gb/s. By utilizing an optimized custom-made microring resonator with high coupling coefficient followed by an optical bandpass filter with appropriate bandwidth, good conversion performances for both modulation formats are achieved according to the converted signals eye diagrams and bit-error-rate measurements.


Introduction
All optical conversion between different modulation formats is required to introduce enhanced flexibility in future optical networks. Format conversion from return-to-zero (RZ) to non return-to-zero (NRZ) signals is an essential example for interfacing different parts of a future ubiquitous transparent optical network. In the past, significant efforts have been dedicated to on-off keying (OOK) RZ-to-NRZ conversion using nonlinear optical loop mirrors (NOLMs) [1], active injection locked lasers [2], or semiconductor optical amplifier (SOA) based devices [3,4] with relatively complex configurations. Single and multi-channel format conversions based on simple passive linear filter devices [5][6][7][8][9][10] have been demonstrated, however only applied to on-off keying modulation. On the other hand, because of its improved receiver sensitivity with balanced detection and superior transmission properties [11,12], differential phase shift keying (DPSK) has received special attention over the past decade. Recently, single [13] and multi-channel [14] RZ-to-NRZ format conversion for DPSK have been experimentally demonstrated using a delay interferometer (DI) with half bit delay. However the dimensions of such devices may remain prohibitively large, which would hinder their potential for integration.
Silicon photonics has received considerable attention lately due to its inherent advantages including compact size and compatibility with microelectronics fabrication processes. Silicon microring resonators (MRRs) are versatile ultra-compact devices that have been widely used for all optical signal processing [15,16].
Multiple channel RZ-OOK to NRZ-OOK format conversion based on a single silicon MRR has already been successfully demonstrated [10]. In this paper, we realize simultaneous RZ-OOK to NRZ-OOK and RZ-DPSK to NRZ-DPSK format conversion at 41.6 Gb/s based on an optimized silicon microring resonator design. We show that the use of an MRR for format conversion is compatible with DPSK modulation. We analyze the effect of the coupling coefficient of the MRR and the bandwidth of the following optical bandpass filter (OBPF) on the conversion results and then experimentally demonstrate simultaneous conversion for both OOK and DPSK formats at 41.6 Gb/s. Clear converted signals eye diagrams and bit-error-rate measurements show the good conversion performance of our scheme.

Operation principle
The principle of the format conversion is the linear filtering process at the through port of an MRR [8]. The simulated eye diagrams of RZ-OOK and RZ-DPSK signals at 41.6 Gb/s are shown in Figs. 1(a) and 1(b) respectively. The notch filtering function of the MRR is used to transform the spectrum of an RZ signal to the spectrum of an NRZ signal by suppressing specific spectral components. The spectrum of the input RZ signals (blue) and the transfer function of the MRR with FSR of 83.2 GHz at the through port (green) are shown in Fig.  1(c). When the free spectral range (FSR) of the MRR is designed to be twice the signal bit rate, RZ-OOK (respectively DPSK) can be converted to NRZ-OOK (respectively DPSK), at the expense of some amplitude ripple, as shown in Figs. 1(d)-1(f), which can be efficiently reduced by an additional optical bandpass filter (OBPF), as can be seen in Figs. 1(g)-1(i). The method has already been successfully demonstrated for RZ-OOK to NRZ-OOK conversion [10], but never so far for RZ-DPSK to NRZ-DPSK. Thanks to the frequency periodic filtering characteristic of the MRR, simultaneous conversion of both OOK and DPSK (carried on   The amplitude ripple and the Q value of the converted NRZ signals (following interferometric demodulation and balanced detection in the case of DPSK signals) depend on the properties of the MRR and the OBPF applied for amplitude ripple mitigation. As shown in Figs. 2(a) and 2(b), when the full-width at half-maximum (FWHM) of the OBPF is 80 GHz, the amplitude ripples of the converted signal are reduced with increasing power coupling coefficient of the MRR κ 2 . Meanwhile, Figs. 2(c) and 2(d) show that the Q value of the converted signal (respectively demodulated signal) is only affected by the bandwidth of the OBPF when it is reduced to 32 GHz. In order to generalize these observations, we have numerically evaluated the amplitude ripple and the Q value of the converted signal as a function of MRR power coupling coefficient and OBPF bandwidth, as shown in Figs. 3(a)-3(d). The amplitude ripple is defined as P ar /P ave , where P ar is the peak-to-peak power deviation of the high level rail in the converted eye diagram and P ave is its average power, as shown in Figs. 1(d) and 1(e). Fortunately, the trends of the variations of the ripple and Q value as a function of MRR power coupling coefficient and OBPF bandwidth are found to be similar for OOK and DPSK. For good conversion performance, low ripple and high Q values are required. For a fixed OBPF bandwidth of 125 GHz (~1 nm), those quantities have been simulated as a function of the MRR power coupling coefficient. As can be seen in Fig. 3(e), when the power coupling coefficient increases, the ripple decreases while the Q value becomes larger. An MRR with power coupling coefficient of 0.9 results in ripple lower than 0.8 and 0.5 and Q values higher

Device design and fabrication
After having assessed the requirements on the MRR power coupling coefficient for optimum RZ-to-NRZ conversion of both OOK and DPSK formats, we focus on the design and fabrication of a suitable silicon MRR. The 3-D full vectorial film mode matching method (FMM) and coupled mode theory (CMT) [17] are used to design the MRR.

MRR design
An MRR with FSR of 83.2 GHz and high power coupling coefficient is designed. As represented in Fig. 4(a), a silicon-on-insulator (SOI) wafer with a top silicon layer of 250 nm and buried silicon dioxide of 3 µm is used as the platform for MRR fabrication. The width and height of both straight and bend waveguides are 470 nm and 250 nm, respectively. Polymer (SU8-2005) is chosen for the top cladding layer. The waveguide propagation loss is assumed to be 8.2 dB/cm [10]. The structure and mode profile of the silicon waveguides are illustrated in Fig. 4(a). Figure 4(b) shows the corresponding effective index for the TM 0 mode as a function of wavelength. The group index is calculated to be 3.9005 at 1550 nm. The radius of the MRR is 147 µm, corresponding to an FSR of 83.2 GHz. Figure 4(c) shows the power coupling coefficient versus the dimension of the gap between the ring and the straight waveguide. We can see that a gap of around 100 nm induces a coupling coefficient higher than 0.9, as required for good conversion performance.

MRR fabrication
The MRR was fabricated on a SOI wafer with top silicon thickness of 250 nm and buried silica of 3 μm. Electron-beam resist ZEP520A was spin-coated on the wafer to create a 110nm thick masking layer. The MRR structure was defined in the resist layer by electron-beam lithography. Then the patterns were transferred to the top silicon layer by inductively coupled plasma reactive ion etching (ICP-RIE). Due to the linewidth reduction of about 30 nm during the fabrication process, the waveguide width and coupling gap are designed with dimensions of 500 nm and 70 nm respectively. Figures 5(a)-5(c) show the structure of the device after fabrication. The radius of the MRR is 147 μm with waveguide width of 470 nm and coupling gap of 100 nm, as desired. To decrease the coupling loss to and from the device, a silicon nano-taper, depicted in Fig. 5(c), was adopted. A layer of 3.5 µm polymer (SU8-2005) was spin-coated on the chip. The nano-taper was defined by UV lithography and directly formed by developing. Figure 5(d) shows the measured transmission spectrum of the MRR. A low insertion loss of 8 dB is achieved with an FSR of 83 GHz and an extinction ratio (ER) of 25 dB, which corresponds to a power coupling coefficient of 0.9, as designed.  The experimental setup for simultaneous RZ-to-NRZ format conversion of OOK and DPSK is shown in Fig. 6. Continuous wave light at 1549.35 nm is modulated in a Mach-Zehnder modulator (MZM) pulse carver driven by a 20.8-GHz radio frequency clock and another MZM driven with a 41.6-Gb/s pseudo-random pattern originating from a bit-pattern generator (BPG), resulting in the generation of a 33% RZ-DPSK signal at 41.6 Gb/s. Meanwhile, a 33% RZ-OOK signal centered at 1551.36 nm is generated at the same bit rate using another set of two MZMs. The pseudorandom binary sequence (PRBS) length is 2 31 -1 for both channels. The OOK and DPSK signals are combined in a 3-dB coupler and then amplified by an erbium-doped fiber amplifier (EDFA) to compensate the insertion loss of the modulators. Before being injected into the MRR, the states of polarization of the signals are adjusted to the TM mode of the MRR with a polarizer (Pol.) sandwiched between two polarization controllers (PCs). The converted NRZ-OOK and NRZ-DPSK signals after the MRR are filtered by an OBPF with 3-dB bandwidth of 1 nm and finally detected in a pre-amplified receiver. A 1-bit fiber DI followed by balanced detection in a pair of 45-GHz photodiodes is used for DPSK detection, while the OOK signal is detected using a single 45-GHz photodiode. The quality of the converted signals is analyzed using an optical spectrum analyzer (OSA), a 70-GHz sampling oscilloscope and an error analyzer (EA). Firstly the performance of the format conversion is evaluated with single channel operation by switching one of the two channels (OOK or DPSK) off. The central frequency of the input RZ signal is tuned to the center of two notches of the MRR transfer function at the through port. Figures 7(a) and 7(b) show the spectra transformation during the format conversion process for OOK and DPSK, respectively. We can see that specific spectral components of both RZ-OOK and RZ-DPSK are suppressed effectively after the MRR and their spectra are successfully transformed to the spectra of NRZ signals. However, the spectra are not significantly affected by the 1-nm OBPF. The same conclusion can be reached by observing the eye diagrams shown in Figs. 7(c)-7(e) for OOK and Figs. 7(f)-7(h) for DPSK. The amplitude ripples of the converted NRZ signals are not reduced significantly using the OBPF. This is because the bandwidth of the OBPF is relatively large while the power coupling coefficient of the MRR is sufficiently high so that the ripples of the converted NRZ signals remain small, in agreement with the numerical analysis of Figs. 3(a) and 3(c).  For the demonstration of simultaneous format conversion, the OOK and DPSK channels are both turned on. The center frequency of the OBPF is tuned to the OOK channel firstly and then to the DPSK channel prior to detection. Note that a frequency-periodic filter would be used in practice. It can be seen that simultaneous format conversion is realized successfully for both channels. The corresponding spectra of the input and output signals after the MRR are shown in Fig. 8(g). The spectra of the two-channel RZ signals are transformed to those of NRZ signals simultaneously. Thanks to the high MRR power coupling coefficient, small ripples are obtained in the converted signals.

Experimental results
Figures 9(a) and 9(b) show the results of BER measurements for both single channel and simultaneous two-channel format conversion. It can be seen that although there is a hint of error floor, error-free performances are obtained for both OOK and DPSK channels after format conversion (the widely accepted definition of error-free corresponding to BER values below 10 −9 is adopted here). In the two-channel case, the available power at the receiver was limited to −22 dBm for the converted NRZ-OOK signal, but the BER performance is strictly identical to the single channel case up to the maximum available power, corresponding to a BER of 3.5 × 10 −9 . There is nearly no cross talk between the two channels. A small power penalty is measured for DPSK format conversion, while a comparatively larger power penalty of 7 dB, similar to earlier reports [10], is shown for OOK. Compared with the MRR we used in the experiment, the cross section and the power coupling coefficient of the MRR in the simulations above remained unchanged. However, since the bit rate of the input signal is 100 Gb/s instead of 41.6 Gb/s, the radius of the MRR needs to be changed to 61 μm. Using the method introduced in Section 2, we can optimize the parameters of the MRR and the OBPF to get the best conversion performance.
Silicon chips may introduce large polarization dependences of the group index and optical coupling because of their strong confinement. The MRR we used in this paper is also polarization sensitive. However, using a polarization diversity circuit with a single microring resonator (MRR), for instance as proposed in [18], polarization independent format conversion could also be realized.

Conclusion
We have demonstrated, simultaneous RZ-to-NRZ format conversion for both OOK and DPSK signals in a well optimized custom-made silicon microring resonator with FSR of 83 GHz and power coupling coefficient of 0.9. Good converted signal quality is obtained, as attested by eye diagrams and BER measurements at 41.6 Gb/s.