Demonstration of a directed optical encoder using ... - OSA Publishing

2 downloads 0 Views 522KB Size Report
Yonghui Tian,1 Lei Zhang,1 Ruiqiang Ji,1 Lin Yang,1,* and Qianfan Xu2. 1Optoelectronic System Laboratory, Institute of Semiconductors, Chinese Academy of ...
October 1, 2011 / Vol. 36, No. 19 / OPTICS LETTERS

3795

Demonstration of a directed optical encoder using microring-resonator-based optical switches Yonghui Tian,1 Lei Zhang,1 Ruiqiang Ji,1 Lin Yang,1,* and Qianfan Xu2 1

2

Optoelectronic System Laboratory, Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083, China Department of Electrical and Computer Engineering, Rice University, 6100 Main Street, MS-366, Houston, Texas 77005, USA *Corresponding author: [email protected] Received July 13, 2011; revised August 24, 2011; accepted August 27, 2011; posted August 29, 2011 (Doc. ID 150990); published September 22, 2011 We propose and demonstrate a directed optical logic circuit that performs the encoding function from a 4 bit electrical signal to a 2 bit optical signal based on cascaded microring switches. The four logic input signals control the states of the switches, while the two logic outputs are given by the optical power at the output waveguides. For proof of concept, a thermo-optic switching effect is used with an operation speed of 10 kbps. © 2011 Optical Society of America OCIS codes: 130.3120, 130.3750, 230.4555, 250.5300.

Directed logic is a novel logical paradigm that employs the optical switch networks to perform the logical operation. Compared to the traditional logic, the directed logic has many potential advantages such as higher operation speed, lower latency, etc. [1–6]. A microring resonator (MRR) based on a silicon-on-insulator (SOI) is an attractive structure owing to its outstanding performances, such as compact size, ultralow power consumption, and a complementary metal oxide semiconductor (CMOS)compatible process. Therefore, the directed logic using the MRR-based optical switching networks is easy to realize large-scale integration and low-cost manufacture in a high-volume CMOS-photonics foundry. Recently, many directed logic circuits based on MRRs have been proposed [6–10], and some of them have been demonstrated including XOR/XNOR [8], OR/NOR, AND/ NAND [9] gates, and an optical decoder [10]. In this Letter, we propose and demonstrate a directed logic circuit based on cascaded microring switches, which performs the encoding function from a 4 bit electrical signal to a 2 bit optical signal. The proposed circuit composed of three tunable MRRs and three waveguides is shown in Fig. 1(a). Monochromatic continuous optical wave with the working wavelength of λw is modulated by the electrical pulse sequences I 1 , I 2 , I 3 , and I 4 applied to MRR1, MRR2, MRR3, and MRR4, respectively. We use the logical 1 and 0 to represent the high and low levels of the voltage applied to the MRRs. The optical power at the output port Y 1 or Y 2 defines the logic output. Logical 1 is obtained when the signal light appears at the optical output port Y 1 or Y 2 , and logical 0 is obtained when the signal light is absent at the optical output port Y 1 or Y 2 . Note that the proposed encoder is not a priority encoder but a common encoder. It is active when and only when one of the input electrical signals (I 1 , I 2 , I 3 , and I 4 ) is at the high level. Therefore, only four input combinations (1 0 0 0, 0 1 0 0, 0 0 1 0, and 0 0 0 1) are possible. No matter whether MRR1 exists or not, the input electrical signal combination of 1 0 0 0 corresponds to the output optical signal combination of 0 0. That is to say, whether MRR1 exists or not, the device can work well. Therefore, we regard MRR1 as a dummy device. 0146-9592/11/193795-03$15.00/0

Each MRR acts as an optical switch and the status of the switch is controlled by the voltage applied to it, as defined below. The MRR is on-resonance at λw when the applied voltage is at the high level. The light is directed to the drop port. The MRR is off-resonance at λw when the applied voltage is at the low level and the light passes through the MRR without disturbance and appears at the through port. According to the above definition, the working principle of the device is as follows: the optical power is at the low level at the output ports Y 1 and Y 2 (Y 1 ¼ 0, Y 2 ¼ 0) when I 1 ¼ 1, I 2 ¼ 0, I 3 ¼ 0, and I 4 ¼ 0; the optical power is at the high level at the Y 2 port and at the low level at the Y 1 port (Y 1 ¼ 0, Y 2 ¼ 1) when I 1 ¼ 0, I 2 ¼ 1, I 3 ¼ 0, and I 4 ¼ 0; the optical power is at the high level at the Y 1 port and at the low level at the Y 2 port (Y 1 ¼ 1, Y 2 ¼ 0) when I 1 ¼ 0, I 2 ¼ 0, I 3 ¼ 1, and I 4 ¼ 0; the optical power is at the high level at both the Y 1 and Y 2 ports (Y 1 ¼ 1, Y 2 ¼ 1) when I 1 ¼ 0, I 2 ¼ 0, I 3 ¼ 0, and I 4 ¼ 1. In order to illuminate the principle of the device clearly, the logical truth table of the device is summarized in Table. 1. We find that the proposed architecture can perform the encoding function from a 4 bit electrical signal to a 2 bit optical signal. A thermo-optic modulating scheme is adopted for the proof of concept since it demands a less complex device

Fig. 1. (Color online) (a) Architecture and (b) micrograph of the device (CW: continuous wave, EPS: electrical pulse trains). © 2011 Optical Society of America

3796

OPTICS LETTERS / Vol. 36, No. 19 / October 1, 2011 Table 1. Logical Truth Table of the Device

I1

I2

I3

I4

Y1

Y2

1 0 0 0

0 1 0 0

0 0 1 0

0 0 0 1

0 0 1 1

0 1 0 1

layer structure and consequently yields easier fabrication steps. The device is fabricated on an 8 in SOI wafer with a 220 nm top silicon layer and a 2 μm buried SiO2 layer. The micrograph of the device is shown in Fig. 1(b). 248 nm deep UV photolithography is used to define the device pattern and an inductively coupled plasma etching process is used to etch the top silicon layer. The rib waveguide has less sidewall area compared to the strip waveguide of similar dimensions, which can reduce the scattering loss on the waveguide sidewall and thus the transmission loss compared to the strip waveguide of similar dimensions and a similar fabrication process [11]. Therefore, the bus waveguide and MRR are formed with a submicron rib waveguide with a width of 400 nm, a height of 220 nm, and a slab thickness of 70 nm. Finite element method calculation shows that the waveguide only supports a TE-like fundamental mode, which agrees with the experimental results [9,10]. The radius of each microring is 10 μm, and the gaps between the straight waveguides and the ring waveguides are 450 nm. Deposited as a separate layer is 1:5 μm thick SiO2 . After the waveguides are fabricated, the titanium microheaters with the thickness of 120 nm are fabricated on the top of the MRRs, which are employed to tune the MRRs through the thermo-optic effect, and the aluminum trances with 50 μm in the width are formed to connect the pads and the microheaters. Note that we do not fabricate dummy MRR1 and only fabricate the pad and microheater of MRR1 where the logic signal I 1 can be applied [Fig. 1(b)]. We use an amplified spontaneous emitting source, an optical spectrum analyzer (OSA), and four tunable voltage sources to test the static response of the device. Broadband light is coupled into the input port of the device through a lensed fiber, and the output light is fed into the OSA through another lensed fiber. Four tunable voltage sources are used to drive the four microheaters above the MRRs, respectively. When the MRRs are heated up, the refractive index of the silicon increases and the resonant wavelengths of the MRRs shift to the longer wavelength. The static response spectra of the device at the output ports Y 1 and Y 2 are shown in Figs. 2 and 3. In principle, any wavelength at the right side of the peak located at 1550:88 nm can be chosen as the working wavelength. However, in order to achieve a sufficiently large extinction ratio and the least power consumption, the working wavelength is chosen to be at the 1552:15 nm. The optical power at both the output ports is very low there [Figs. 2(a) and 3(a)] when the voltage applied to MRR1 is 0:5 V and the voltages applied to other MRRs are 0 V (I 1 ¼ 1, I 2 ¼ 0, I 3 ¼ 0, and I 4 ¼ 0.). Although the MRRs are designed to have the same structural parameters, they have slightly different resonant wavelengths, which is mainly due to the limited manufacturing

Fig. 2. (Color online) Response spectra of the device at the output port Y 1 with the voltages applied to MRR1, MRR2, MRR3, and MRR4 being (a) 0.5, 0, 0, and 0 V, (b) 0, 1.68, 0, and 0 V, (c) 0, 0, 2.44, and 0 V and (d) 0, 0, 0, and 3:12 V.

accuracy. Therefore, we observe two resonant peaks in Figs. 2(a) and 3(a). Clearly, the peak at 1548:84 nm is from MRR4, the peak at 1549:48 nm is from MRR3, and the peak at 1550:88 nm is from MRR2. Figure 2(b) is similar to Fig. 2(a) and there is also a dip at λw in Fig. 2(b) (representing 0) when the voltage applied to MRR2 is 1:68 V and the voltages applied to other MRRs are 0 V. Apparently, MRR2 does not affect the static response spectra at the Y 1 port [see Fig. 1(a)]. The resonant peak of MRR3 shifts from 1549:48 nm to λw (representing 1) when the voltage applied to MRR3 is 2:44 V and the voltages applied to other MRRs are 0 V [Fig. 2(c)]. The resonant peak of MRR4 shifts from 1548.84 to 1552:15 nm (representing 1) when the voltage applied to MRR4 is 3:12 V and the voltages applied to other MRRs are 0 V [Fig. 2(d)]. There is a dip at λw when the voltage applied to MRR1 is 0:5 V and the voltages applied to other MRRs are 0 V [Fig. 3(a)]. The resonant peak of MRR2 shifts from 1550:88 nm to λw when the voltage applied to MRR2 is 1:68 V and the voltages applied to other MRRs are 0 V

Fig. 3. (Color online) Response spectra of the device at the output port Y 2 with the voltages applied to MRR1, MRR2, MRR3, and MRR4 being (a) 0.5, 0, 0, and 0 V, (b) 0, 1.68, 0, and 0 V, (c) 0, 0, 2.44, and 0 V and (d) 0, 0, 0, and 3:12 V.

October 1, 2011 / Vol. 36, No. 19 / OPTICS LETTERS

Fig. 4. (Color online) Signals applied to (a) MRR1, (b) MRR2, (c) MRR3, and (d) MRR4, (e) the result at the output port Y 1 , and (f) the result at the output port Y 2 of the device.

[Fig. 3(b)]. There is also a dip at λw when the voltage applied to MRR3 is 2:44 V and the voltages applied to other MRRs are 0 V [Fig. 3(c)]. Obviously, MRR3 does not affect the static response spectra at the Y 2 port [Fig. 1(a)]. The resonant peak of MRR4 shifts from 1548:84 nm to λw when the voltage applied to MRR4 is 3:12 V and the voltages applied to other MRRs are 0 V [Fig. 3(d)]. Hereto, we have analyzed all eight static response spectra for four logical combinations. The results indicate that the device can implement the encoding function from a 4 bit electrical signal to a 2 bit optical signal. The dynamic performance of the device is further characterized (Fig. 4). Monochromatic light with the working wavelength of 1552:15 nm is coupled into the input port of the device. Four binary sequence non-return-to-zero signals at 10 kbps are converted to four electrical analog signals first and then applied to the four MRRs, respectively (The high levels are 0.50, 1.68, 2.44, and 3:12 V for MRR1, MRR2, MRR3, and MRR4, respectively, and the low levels are 0 V for all MRRs). The light at the output ports is fed into a detector. The electrical signals transformed by the detector and the four electrical signals applied to the four MRRs are fed into an oscilloscope for waveform observation. Clearly, the device performs the encoding function from a 4 bit electrical signal to a 2 bit optical signal correctly (Fig. 4). Note that the power level is the same for logical 0 but different for logical 1 in different cases, which mainly results from the different functions of MRR4 in different cases. When

3797

MRR4 is off-resonance at λw (I 4 ¼ 0), all the input light passes through MRR4 and is further directed either to the output port Y 1 (Y 1 ¼ 1) by MRR3 (I 3 ¼ 1) or to the output port Y 2 (Y 2 ¼ 1) by MRR2 (I 2 ¼ 1). When MMR4 is on-resonance at λw (I 4 ¼ 1), MRR4 behaves as a power splitter. Half of the input light is directed to the output port Y 1 (Y 1 ¼ 1) and the left half is directed to the output port Y 2 (Y 2 ¼ 1). Therefore, the power level for logical 1 in the second and third logical combinations is twice that in the fourth logical combination. The similar phenomena also can be observed in the static response spectra (Figs. 2 and 3), where the power level for logical 1 is −14 and −11:5 dB at the output ports Y 1 and Y 2 when MRR4 is on-resonance, respectively, and the power level for logical 1 is −9:7 and −9:3 dB at the output ports Y 1 and Y 2 when MRR4 is off-resonance, respectively. In conclusion, we have proposed and demonstrated a directed logical circuit that can implement the encoding function from a 4 bit electrical signal to a 2 bit optical signal. This work has been supported by the National Natural Science Foundation of China (NSFC) under grant 60977037 and the National High Technology Research and Development Program of China under grant 2009AA03Z416. References 1. J. Hardy and J. Shamir, Opt. Express 15, 150 (2007). 2. H. J. Caulfield, R. A. Soref, and C. S. Vikram, Photon. Nanostr. Fundam. Appl. 5, 14 (2007). 3. H. J. Caulfield and S. Dolev, Nat. Photon. 4, 261 (2010). 4. A. I. Zavalin, H. J. Caulfield, and C. S. Vikram, Optik 121, 1300 (2010). 5. H. J. Caulfield, in Opt. Supercomputing. Lecture Notes in Computer Science (Springer, 2009), pp. 30–36 . 6. R. Soref, Adv. Optoelectron. 2011, 627802 (2011). 7. Q. Xu and R. Soref, Opt. Express 19, 5244 (2011). 8. L. Zhang, R. Q. Ji, L. X. Jia, L. Yang, P. Zhou, Y. H. Tian, P. Chen, Y. Y. Lu, Z. Y. Jiang, Y. L. Liu, Q. Fang, and M. B. Yu, Opt. Lett. 35, 1620 (2010). 9. Y. H. Tian, L. Zhang, R. Q. Ji, L. Yang, P. Zhou, H. T. Chen, J. F. Ding, W. W. Zhu, Y. Y. Lu, L. X. Jia, Q. Fang, and M. B. Yu, Opt. Lett. 36, 1650 (2011). 10. Y. H. Tian, L. Zhang, R. Q. Ji, L. Yang, P. Zhou, J. F. Ding, H. T. Chen, W. W. Zhu, Y. Y. Lu, Q. Fang, L. X. Jia, and M. B. Yu, Opt. Lett. 36, 3314 (2011). 11. W. Mathlouthi, H. S. Rong, and M. Paniccia, Opt. Express 16, 16735 (2008).