1. Introduction

The recent rapid growth of data traffic leads to application of advanced modulation formats such as a quadrature phase-shift keying (QPSK) format to broadband optical-fiber networks. The QPSK signal is generated using an in-phase (I) quadrature (Q) modulator consisting of a nested Mach–Zehnder (MZ) waveguide, which includes two sub-MZ interferometers to generate I and Q components [1]. Most of the IQ modulators used in current optical-fiber networks are made of lithium niobate (LN), in which phase modulation is achieved by the Pockels effect in a waveguide of a few centimeters in length [2]. The large footprint of the LN IQ modulator limits cost reduction, and therefore, a new technology for size and cost reduction of modulators is desired for extensive use of the advanced modulation formats in a wide range of optical-fiber networks.

Small-footprint low-cost MZ modulators based on high index-contrast silicon waveguides are promising to meet the demands. Optical modulation in silicon waveguides is realized by carrier plasma dispersion which allows length reduction of phase-shifter waveguide down to a few millimeters. Silicon modulators are fabricated using CMOS-compatible fabrication processes, which are suitable for low-cost modulators. Silicon modulators consisting of a single MZ waveguide have been studied in on–off keying (OOK) formats; optical modulations of 50 Gb/s or higher have been reported [3–7]. QPSK using a silicon modulator consisting of nested silicon MZ waveguides was firstly confirmed at 20 Gb/s (10G baud) [8]. QPSK modulation at 50 Gb/s (25 Gbaud) was thereafter reported [9]. Polarization division multiplexing QPSK (PDM-QPSK) was achieved by using dual silicon IQ modulators monolithically integrated with a PDM waveguide [10]. Long-haul transmissions over 2400 km using silicon IQ modulators were demonstrated [11,12]. In addition to QPSK, 16 quadrature amplitude modulation (QAM) was realized in silicon IQ modulators [13,14]. These various applications were investigated at a baud rate of 28 Gbaud. In commercial 100-Gb/s transmission systems, baud rate higher than 28 Gbaud is crucial in order to take advantage of powerful third-generation forward error correction (FEC) schemes [15]. 64-Gb/s (32-Gbaud) QPSK was reported using a low-loss silicon IQ modulator [16].

Based on the baseline design of the silicon IQ modulator reported previously, we further improved high-frequency performance using short phase shifters, which have the same modulation efficiency by using a doping compensation method. In this paper, we describe the key characteristics of the improved IQ modulator. The design of the silicon IQ modulator is described along with its fundamental characteristics in Section 2. In Section 3, its modulation performances are presented in two modulation formats, namely 44.6-Gb/s differential QPSK (DQPSK) with differential detection and 64-Gb/s QPSK with coherent detection.

2. Design and optical characteristics of the IQ modulator

Figure 1 shows the top-view photograph of the silicon IQ modulator, which was fabricated on an 8-inch silicon on insulator (SOI) wafer using CMOS-compatible processes. The right schematic illustrates the layout of optical waveguides with their functions. The modulator consists of a nested IQ MZ interferometer (MZI), which is connected to input and output mode-field converters (MFCs) for lensed fiber coupling through 90-degree bend silicon nanowire waveguides. The nested IQ MZI has two sub-MZIs for I and Q components, respectively, including silicon rib-waveguide RF phase shifters connected to straight coplanar-waveguide electrodes. The electrodes are connected to input and output contact pads at the edge of the modulator chip with a minimum length without an electrode bending. This design minimizes unnecessary RF propagation loss caused by redundant electrode sections and bends on the silicon wafer. A thermo-optic (TO) phase shifter allows phase adjustment to sustain a π/2 phase difference between the I and Q sub-MZIs for stable QPSK modulation. These device blocks were monolithically integrated in a footprint as small as 3.5 × 2.9 mm².

 

Fig. 1 Top view photograph of the silicon IQ modulator and schematic of optical waveguides.

Download Full Size | PDF

The silicon rib waveguides in the RF phase shifters have a rib width of 500 nm, a rib thickness of 220 nm, and a slab thickness of 95 nm [16]. A lateral pn junction was formed in the rib waveguides using a doping compensation method as the cross-section shown in Fig. 2(a) [7,17]. Both sides of the silicon rib were made to be intrinsic by compensation doping to reduce carrier concentration, and thereby optical loss due to free carrier absorption was reduced. Figure 2(b) shows simulated characteristics of the phase shifter, which were calculated using an abrupt junction model with uniform dopant density equal to each of P and N regions. The left vertical axis indicates phase shifter length which produces π phase shift under RF modulation voltage of 8 V_pp with DC reverse bias of 5 V. The right vertical axis indicates optical losses of the phase shifters. The phase shifter length is reduced by increasing dopant density, while the optical loss increases conversely. The phase shifter length and the optical loss of the phase shifter are in so-called trade-off relation. However, the doping compensation method provides lower optical loss with keeping the same phase shifter length or shorter phase shifter length with keeping the same optical loss. We designed phase shifter with a length as short as 3 mm for higher speed performance in comparison with the phase shifter reported previously.

 

Fig. 2 (a) PN junction with dopant compensated region and (b) its characteristics as a phase shifter simulated under the following conditions: Doping density of each p and n region (N_A and N_D) is uniform and equal to one another; an abrupt junction and 8 V_pp driving voltage with DC reverse bias of 5 V are assumed.

Download Full Size | PDF

Phase shift and optical loss of the silicon rib-waveguide phase shifter with a 3-mm length were measured using a test sample of an asymmetric silicon MZ waveguide. Fig. 3(a) shows phase shift and optical loss of the phase shifter with respect to DC reverse bias voltage. V_π is obtained as 7.5 V, and optical loss is as low as 2.7 dB at 0-V reverse bias and 1.9 dB at 5-V reverse bias. The same V_π has been achieved with the shorter length of 3 mm as for the phase shifter of the 4-mm length reported previously, and therefore, phase shifter efficiency is higher with the present modulator than with the modulator reported previously [16].

 

Fig. 3 (a) phase shift and phase shifter (WG) loss versus DC reverse bias, and (b) E/O response under DC reverse bias.

Download Full Size | PDF

Electro-optic (E/O) response of the phase shifter was measured up to an RF frequency of 40 GHz; Fig. 3(b) shows the E/O response versus RF frequency in comparison with that of the previous phase shifter. The wavelength of the laser input to the IQ modulator was 1550 nm. 3-dB optical bandwidth ( = 6-dB electrical bandwidth; where an optical modulation amplitude becomes a half of that at 0.2 GHz.) of the rib-waveguide phase shifter is wider than 30 GHz under 5-V DC reverse-bias. An increase in the optical bandwidth by 6 GHz is achieved in comparison with the previous phase shifter.

The passive fiber-to-fiber loss of the IQ modulator without RF modulation signals was obtained with the phase at the TO phase shifter adjusted to maximize transmittance at wavelengths of 1530, 1550, 1590 and 1610 nm. A DC reverse bias of 5 V was applied to all the RF phase shifters. The passive optical loss at each wavelength is as low as 9 dB, including a 1.5-dB MFC coupling loss per facet to a lensed fiber.

3. DQPSK/QPSK using the silicon IQ modulator

In this section, performances of the silicon IQ modulator in the two modulation formats are presented.

3.1 DQPSK at 44.6 Gb/s

One of the major applications of IQ modulators is DQPSK transmission in 40-Gb/s transmission systems. Transmission performance of the silicon IQ modulator was investigated in DQPSK at 44.6 Gb/s.

The measurement setup is illustrated in Fig. 4. A precoded 2³¹–1 pseudo-random bit stream (PRBS) was generated in a pulse pattern generator (PPG) and output as differential electrical signals of DATA and inverted DATA at a baud rate of 22.3 Gbaud, which corresponds to a bit rate of 44.6 Gb/s extensively used in 40-Gb/s optical transport networks [18]. Both the DATA and inverted DATA signals were amplified to 8 V_pp with driver amplifiers and applied to the rib-waveguide RF phase shifters through the input contact pads of the traveling-wave electrodes using RF probes. A common DC reverse bias of 5 V was applied to the RF phase shifters from the output contact pads of the traveling-wave electrodes through bias-tees with 50-Ω terminators. TO phase shifters were adjusted to sustain a π/2 phase shift between I and Q components in each MZI .Output optical DQPSK signals from the silicon IQ modulator were transmitted through standard-dispersion single-mode fibers (SMFs) and dispersion compensating fibers (DCFs).

 

Fig. 4 Measurement setup for DQPSK transmission with the silicon IQ modulator and a receiver for differential detection.

Download Full Size | PDF

As for the receiver side of the measurement setup, differential-detection configuration based on 1-bit delay lines and balanced photodetectors (PDs) were employed for demodulation of received DQPSK optical signals after the optical-fiber transmission. The bit error rate (BER) of the demodulated signals was measured, and dispersion tolerance was obtained as a function of the optical signal-to-noise ratio (OSNR) in 0.1 nm noise bandwidth.

Figure 5(a) shows BER characteristics in the back-to-back configuration. The BER is lower than 10⁻³ for OSNRs of 16.3 dB or higher. The silicon IQ modulator is, thus, capable of 40-Gb/s transmission, because enhanced forward error corrections (FECs) ensure error-free transmission with a BER of 10⁻³ in 40-Gb/s transmission systems [19]. In comparison with the previous phase shifter, 1.7-dB reduction is achieved in OSNR penalty for a BER of 10⁻³.

 

Fig. 5 (a) BER for the 22.3G DQPSK and (b) path power penalty.

Download Full Size | PDF

The dispersion tolerance is another important characteristic for 40-Gb/s transmission systems, because large tolerance alleviates control of residual dispersion in dispersion managed fiber links. Path power penalty plotted in Fig. 5(b) was obtained as an excess OSNR required for the optical-fiber transmission at a BER of 10⁻³ with respect to the OSNR in back-to-back configuration. The path power penalty was less than 2 dB in the range of accumulated chromatic dispersion from −96 to + 107 ps/nm. BER performance in DQPSK transmission was simulated [20]. By scaling results in the simulation to the ones at the baud rate in the measurement, an OSNR at 22.3 Gbaud is estimated as 13.4 dB for a BER of 10⁻³ and a dispersion tolerance of ± 154 ps/nm in a 2-dB power penalty. The difference in the path power penalty between the measurement and the simulation may be due to residual frequency chirping in the output optical DQPSK signals from the silicon IQ modulator. The residual frequency chirping was caused by imbalances of driving conditions such as a phase adjustment in each MZI arm and an electrical signals applied to the phase shifter. The imbalances will be reduced if the phase and the applied electrical signals are calibrated.

3.2 QPSK at 64 Gb/s

For 64-Gb/s QPSK, the silicon IQ modulator was setup in a similar manner to that shown in Fig. 4, except the baud rate of 32 Gbaud and the electrical signals (DATA and inverted DATA) without differential coding. The modulated optical signals were measured in back-to-back configuration using a coherent detection scheme, which consists of a 90-degree optical hybrid, high-speed PDs with trans-impedance amplifiers (TIAs), and a 80-GS/s real-time oscilloscope, as shown in Fig. 6. For the coherent detection scheme, a light beam from the external cavity tunable laser diode (LD) in Fig. 4 was split for a local oscillator (LO) for homodyne detection.

 

Fig. 6 QPSK measurement setup.

Download Full Size | PDF

The constellation-diagram measurements at different input laser wavelengths were performed with TO phase shifters manually tuned to keep the phase state for QPSK, while the amplitude of the DATA and inverted DATA signals were not changed. Figure 7(a) shows constellation diagrams with LD wavelengths of 1530, 1550, and 1610 nm with an OSNR of 30 dB. At the all wavelengths, BERs less than 10⁻⁶ were confirmed. Therefore, 64-Gb/s QPSK is achieved in a broad wavelength range of C and L bands using the silicon IQ modulator.

 

Fig. 7 QPSK modulation results at 64 Gb/s: (a) constellation diagrams at different LD wavelengths, (b) a constellation diagram and spectrum with a FIR digital filter, and (c) BERs with and without the filter.

Download Full Size | PDF

Remaining signal distortion observed in the constellation diagrams can be compensated using digital signal processing [14, 21]. The constellation diagram in Fig. 7(b) was obtained using a 15-tap finite impulse response (FIR) digital filter for the QPSK signals at a wavelength of 1550 nm as shown in Fig. 7 (a). The low frequency peak is suppressed with the FIR filter as presented in the optical modulation spectra in the right graph in Fig. 7(b).

BER characteristics were investigated and are shown in Fig. 7(c). BERs less than 10⁻⁴ without an error floor is confirmed at all the input laser wavelengths. The baud rate as high as 32 Gbaud is crucial to the 100-Gb/s transmission accommodating a powerful third generation FEC [22]; BER threshold for error free transmission is 10⁻² with the FEC [15]. The OSNR required for BER of 10⁻² was 11.6–11.8 dB without the FIR filter. They are the lowest value among the reported QPSK modulations at various bitrates for 100-Gb/s transmissions using silicon IQ modulators. Moreover, its value was improved to 10.7–10.9 dB by adopting the FIR filter. It is only 2 dB different from the theoretical value [23].

4. Conclusion

A low-loss high-speed silicon IQ modulator was designed, fabricated and characterized in two modulation formats of DQPSK and QPSK. The straight traveling-wave electrodes design allowed RF propagation to silicon rib-waveguide phase shifters in a nested Mach–Zehnder structure without degradation of RF performance. The fabricated modulator using improved phase shifters with 3-dB optical bandwidth of 30 GHz has a low passive optical loss of 9 dB in C and L bands. For DQPSK transmission at 44.6 Gb/s, a dispersion tolerance from −96 to 107 ps/nm for a 2-dB path power penalty was realized with an OSNR of 16.3 dB. Moreover, in 64-Gb/s QPSK modulation, BERs of 10⁻² was confirmed with OSNR of 11.6–11.8 dB in C and L bands. This is the highest baud rate among the silicon IQ modulators reported and performances sufficient for 100-Gb/s transmission.