We report a broadband digital electro-optical switch, based upon a multi-stage Mach-Zehnder lattice design in silicon-on-insulator. A digital switching response is demonstrated, engineered through apodization of the coupling coefficients between stages. The digital switching behavior results in crosstalk lower than −15 dB for drive-voltage noise levels in excess of 300 mVpp, which exceeds the noise tolerance of a conventional single-stage Mach-Zehnder switch by more than six-fold. In addition, the digital design enables a larger maximum ‘on’-state extinction (below −26 dB) and lower ‘on’-state free-carrier-induced insertion loss (less than 0.45 dB) than that of the single-stage switch. The noise-tolerant, low-crosstalk switch can thus play a key role within CMOS-integrated reconfigurable optical networks operating under noisy on-chip conditions.
© 2011 OSA
High-bandwidth, low-power optical interconnects, integrated in close proximity to digital logic and memory are increasingly being considered a viable interconnect strategy for enabling further scaling of the processing power of massively parallel computing systems. Short-range optical-interconnect approaches based on silicon photonic components have drawn considerable interest from the research community over the past ten years, owing in part to the promise of seamless integration with CMOS circuits [1–4]. On the inter- and intra-chip interconnect levels, optical network-on-chip (ONoC) architectures capable of providing reconfigurable communication paths between the processor cores and memory systems on a chip multiprocessor have been proposed [5,6].
A key device for an ONoC based on wavelength-insensitive optical routing scheme, is a broadband and noise-tolerant optical switch, capable of routing an optical data stream encoded in multiple parallel wavelength channels [7–10]. In , we proposed and analyzed an ultra-broadband digital switch design, based on a multi-stage Mach-Zehnder lattice (MZL), with strongly improved noise tolerance as compared to a single-stage Mach-Zehnder (MZ) switch. Such an improved noise tolerance not only helps to enable robust switching operation in an on-chip environment, which is known to be prone to substantial supply-voltage noise , but it also helps to improve the thermal stability of the ‘on’-state of the device, cancelling out any fluctuations of the injected-carrier density induced by temperature variations in the forward-biased p-i-n diode phase shifters .
In this paper, we report an experimental demonstration of the digital MZL-switch concept proposed in . Robust crosstalk levels of lower than −15 dB are demonstrated in an apodized four-stage MZL switch, for drive-voltage noise levels as high as 300 mVpp. In contrast, a single-stage MZ switch is shown to have a drive-voltage noise tolerance of only 45 mVpp. Also, greater ‘on’-state extinction levels (below −26 dB) are demonstrated as well as lower free-carrier-induced ‘on’-state insertion loss (less than 0.45 dB). As the improved switching performance comes at the expense of increased power consumption (compared to a MZ switch), we will briefly discuss the outlook for low-power operation of these digital switches.
2. Design and fabrication of a Mach-Zehnder lattice switch in SOI
The design of a broadband digital electro-optic switch based on a Mach-Zehnder lattice was extensively discussed in our earlier paper . The MZL switch consists of a series of balanced interferometric stages, in which each lower branch of the stages includes a forward-biased p-i-n diode phase shifter, as illustrated in Fig. 1(a) for the specific case of a four-stage lattice. In such multi-stage structures, a digital switching response originates from properly engineered multiple-path interference at the output ports, and can be realized by optimizing the coupling coefficients κi of the individual coupling sections. An optimized MZL switch design is expected to result in lower ‘on’-state crosstalk, lower ‘on’-state insertion loss, as well as a drastically improved tolerance to thermal and drive-voltage noise, as compared to a conventional single-stage MZ switch .
Here, we report the switching characteristics of a fabricated four-stage MZL switch, shown in Fig. 1(b). The switch is fabricated on a silicon-on-insulator (SOI) wafer using a subset of processing modules from a standard IBM front-end CMOS process flow . The cross-sectional dimensions of the silicon waveguides are 500 × 220 nm2. Each of the four p-i-n diode phase shifters in the bottom (active) branch of the stages are 50-μm long. The top branch contains shorted dummy p-i-n diodes, which are included to ensure balance of optical losses between branches. The coupling coefficients κi are apodized according to a Hamming window function (κ1 = 3.2%, κ2 = 13.1% and κ3 = 20.5%), resulting in respective directional-coupler lengths of L1 = 4.1 μm, L2 = 8.5 μm, and L3 = 10.8 μm. The device occupies a total footprint of 160 μm x 75 μm.
The expected switching characteristics of the fabricated MZL switch are obtained by simulation [11,13] as a function of the carrier density N injected into the p-i-n diode phase shifters, and are shown in Fig. 2(a) . The switching response of this four-port structure is given by the four transmittance curves Tij(λ) = |Sij(λ)|2, where Sij(λ) are the complex transfer functions of the optical field from input port ai to output port bj, with i, j = 1, 2. The Hamming apodization of the coupling coefficients results in a switching response where any sidelobes in the T12 transmittance are suppressed to well below −20 dB. As a result, if the switch is biased to an ‘on’-state level sufficiently above N = 2.8 × 1018 cm−3 (ΔnSi ~7.3 × 10−3), it is expected to have a substantial tolerance against noise on the injected-carrier density. In addition, at a larger bias of N = 5.6 × 1018 cm−3 (ΔnSi ~1.34 × 10−2), T12 extinction levels as large as −30 dB can be obtained. The insertion loss of the T11 transmission path is predicted to remain under 0.4 dB throughout the full operating range of the switch despite the substantial optical absorption loss (on the order of ΔαSi ~350 dB/cm) induced by the free carriers injected into the waveguide core. However, the T22 transmission path suffers from increasing insertion loss with carrier density, as the free-carrier absorption (FCA) is highest for this path. As such, the T22 transmission path is not as useful in practical applications, and the MZL switch is preferably used as a 1 × 2 switch.
Figure 2(a) illustrates the principal of digital performance demonstrated by the MZL design. When switched sufficiently deep into the ‘on’-state, the T12 (T11) transmittance will robustly remain low (high), even when appreciable variations in the injected free carrier density (or equivalently the p-i-n diode drive voltage) are introduced.
For comparison, the simulated switching response of a balanced, single-stage MZ switch with a one 200-μm-long p-i-n diode phase shifter and two 50% directional couplers is shown in Fig. 2(b). In contrast to the digital behavior of the MZL, the dual-path interference in the MZ switch type results in an analog switching response, with only limited tolerance against noise in the ‘on’-state, which is reached at N = 1.4 × 1018 cm−3 (ΔnSi ~4 × 10−3). In addition, the ‘on’-state T12 extinction is limited to −21 dB, while the ‘on’-state T11 insertion loss is as high as 0.8 dB, as 50% of the input light is sent through the lossy free-carrier-based phase-shifter (ΔαSi ~90 dB/cm).
Finally, as both the MZL not the MZ switch designs are built from balanced interferometric stages without any built-in phase delay (in the ‘off’ state), they can be considered optically temperature insensitive, under the assumption that any temperature difference between the upper and lower branches in each stage is insignificant. Temperature variations within the footprint of the device (160 μm x 75 μm for the MZL) would obviously still affect the obtained crosstalk and extinction ratio, as they would induce a non-intentional phase delay. However, the impact of these effects can be minimized through layout optimization and further reduction of the MZL device footprint.
In conclusion, when considering optical switches appropriate for on-chip CMOS-integrated photonic networks, in which supply-voltage noise and thermal variations are common [11,12], the multi-stage MZL switch has several advantages over the traditional MZ design.
3. Measured switching characteristics
3.1 Optical bandwidth
The optical bandwidth of the fabricated MZL switch is characterized by coupling TE-polarized light from a broadband LED source to one of the input ports of the switch and analyzing the intensity spectrum of the transmitted light at both output ports. This analysis is performed first without an applied bias voltage, in order to obtain the transmission spectra T11(λ) and T12(λ) for the ‘off’-state, shown in Fig. 3(a) . The intensity spectra measured at the respective output ports are normalized against the sum of the intensity spectra of both output ports, with the input signal at the same input port. As a result, the obtained T11(λ) and T12(λ) spectra do not include ‘off’-state optical losses of a passive nature, such as those due to optical scattering or bending. The additional passive insertion loss of the MZL switch is separately measured to be 1.25 ± 0.25 dB at 1479 nm. As seen in Fig. 3(a), the obtained ‘off’-state crosstalk is lower than −15 dB over an optical bandwidth ΔλBW of 40 nm centered on a wavelength of 1479 nm. The poor ‘off’-state extinction of T11 is the result of a small fabrication-induced optical path mismatch between the nominally identical phase-delay sections, which arises from variations in waveguide width and etch depth.
Subsequently, the same measurements are performed while applying a continuous-wave voltage VD that maximizes the T12 extinction at 1479 nm (VD = 1.3 V). The resulting normalized ‘on’-state T11(λ) and T12(λ) spectra include only the insertion losses from FCA loss. These spectra are also shown in Fig. 3(a). The best-case ‘on’-state crosstalk is lower than −26 dB at 1479 nm and lower than −20 dB over an optical bandwidth of more than 80 nm around a wavelength of 1479 nm. The excess ‘on’-state T11 insertion loss is approximately 0.45 ± 0.2 dB for the MZL switch, which agrees very well with the simulations presented in section 2.
For comparison, similar measurements are performed for the MZ switch, the results being shown in Fig. 3(b). A crosstalk of less than −20 dB is obtained at 1530 nm for both ‘on’ and ‘off’ switching states. The excess ‘on’-state insertion loss due to FCA loss is 0.8 ± 0.2 dB. The additional passive insertion loss of the MZ switch is measured to be 0.9 ± 0.25 dB at 1530 nm. Lower than −15 dB crosstalk is obtained over an optical bandwidth of 40 nm centered on a wavelength of 1530 nm. As such, the MZL and MZ switches have essentially the same optical bandwidth ΔλBW. This bandwidth is determined by the wavelength sensitivity of the employed directional couplers . For applications requiring wider optical bandwidths, wavelength-insensitive couplers, each consisting of an individual MZ structure, could be utilized [7,11].
The blue shift of the optimum operating wavelength of the MZL switch (1479 nm) with respect to that of the MZ switch (1530 nm) results from additional coupling in the waveguide bend sections which are immediately adjacent to the straight directional-coupling sections in each stage. This additional coupling was not accounted for in the original MZL design, but is found to be substantial for the employed partially etched waveguides (see  for a cross-sectional image) and the implemented bending radius of 6.5 μm.
3.2 Intrinsic switching response
Under steady-state biasing conditions, the constant current flowing through the p-i-n diodes leads to self-heating effects, which partially cancel out the free-carrier electro-optic effect and complicate the device analysis . In order to assess the intrinsic, free-carrier-induced switching response of the fabricated four-stage MZL switch, fully decoupled from thermo-optic effects, a series of time-resolved transmittance measurements are made. A drive signal consisting of 80-ns long pulses with a base voltage of 0 V and a peak voltage VD, and a 2-% duty cycle is applied to the switch. Light from a transverse-electric (TE) polarized continuous-wave laser operating at a wavelength of 1.48 μm is launched into one of the input ports of the device. The transmitted light is collected at the output port of interest, and subsequently amplified using an erbium-doped fiber amplifier (EDFA) followed by an optical bandpass filter with 1 nm bandwidth. The time-varying optical signal is then detected using a high-speed detector, and the resulting waveforms are recorded on a high-speed oscilloscope. In these waveforms, the transmitted optical power level is evaluated 60 ns after arrival of each pulse, at which point steady-state electrical conditions of the measurement setup are reached.
In order to compare the measured switching response with the simulated transmittance spectra in Fig. 2(a), the output optical power levels measured for the T12 and T21 transmission paths are normalized against the maximum value of the transmitted optical power, measured for each transmittance path at zero applied voltage. For the T11 and T22 transmission paths, the obtained optical power levels are normalized against the maximum power level measured for these transmission paths within the considered range of applied voltages (0-1.6 V). A direct comparison of the output power levels for the T11 and T22 transmission paths on one hand and the T12 and T21 transmission paths on the other hand proves difficult in these time-resolved measurements, owing to a difference in optical gain provided by the EDFA. The difference in gain originates from the large difference in average optical power fed into the EDFA for both sets of transmission paths, owing to the low duty cycle of the driving signal, in combination with saturation mechanisms in the EDFA. Therefore, the relative offset between and the T12 and T21 transmission curves, and the T11 and T22 transmission curves is estimated from their ‘on’-state transmittance values measured using continuous-wave bias voltages as discussed in section 3.1. This approach leads to a worst-case estimation for the intrinsic, FCA-related insertion loss of each switching state. Again, only optical losses originating from FCA losses are represented using this normalization scheme.
The resulting response of the MZL switch is shown as a function of applied peak voltage in Fig. 4(a) . It can be seen that in the ‘off’-state, obtained for voltages below 0.6 V, the obtained crosstalk level is lower than −18 dB, whereas the T12 and T21 insertion losses are relatively flat at approximately 0 dB. For voltages exceeding 1.16 V, the T12 and T21 transmittance are suppressed to a level lower than −15 dB. For voltages between 1.18 V and 1.28 V, the extinction of the T12 transmittance is larger than −20 dB. While a switching sidelobe is apparent for VD = 1.37 V, the T12 extinction is below −16 dB for voltages up to 1.6 V, and crosstalk levels in this voltage range are lower than −15 dB. For voltages higher than 1.18 V, the excess T11 insertion loss remains relatively flat and is lower than 0.45 dB. The T22 insertion loss increases monotonically in this voltage range as predicted by simulation, due to increasing FCA optical losses.
For comparison, the switching response of a single-stage MZ switch characterized using the same driving conditions, but measured at 1.53 μm wavelength is shown in Fig. 4(b). In the ‘off’-state, the crosstalk levels are lower than −20 dB, whereas the T12 and T21 insertion loss are relatively flat at approximately 0 dB. The ‘on’-state is reached at VD = 1.1 V (estimated N ~1.4 × 1018 cm−3), with a T12 and T21 extinction of more than −20 dB. With increasing applied voltage, the oscillatory analog switching response of the MZ switch becomes apparent.
The general shape of the switching response of both the MZL as well as that of the MZ switch show excellent qualitative agreement with the switching response calculated as function of injected-carrier density shown in Fig. 2. For the MZL switch, the higher level of the measured switching sidelobe (−16 dB) as compared to the simulation (−22 dB) can be explained in part by a deviation of the fabricated coupling apodization from the ideal simulated case. This deviation originates from additional coupling in the curved waveguide sections adjacent to the straight directional coupling sections, as previously discussed in section 3.1.
From the obtained voltage-dependent switching curves in Fig. 4, the voltage-noise tolerance of both switches can be estimated. For the ‘off’-state, the voltage-noise tolerance ΔVoff is on the order of 700 mV (peak-to-peak) for both MZ and MZL switches, owing to the rectifying behavior of the forward-biased p-i-n diode . For the ‘on’-state, the noise tolerance ΔVon depends on the considered ‘on’-state bias voltage Von. For the digital (MZL) switch, a trade-off can be made between Von and noise tolerance ΔVon, with a high bias voltage typically resulting in a higher voltage-noise tolerance, as evident from Fig. 4(a). As an example, if we specify the ‘on’-state voltage to be near the center of the first dip in T12 transmittance (Von = 1.23 V, estimated N ~3 × 1018 cm−3), the peak-to-peak voltage-noise tolerance ΔVon for obtaining at least −15 dB and −20 dB extinction of T12 is respectively 160 mV and 100 mV. At Von = 1.31 V, ΔVon for obtaining −15 dB extinction of T12 can be as high as 300 mV. In contrast, the noise tolerance of the MZ switch at Von = 1.1 V for the −15 dB and −20 dB extinction levels is only 45 mV and 25 mV, respectively. The data in Fig. 4 illustrates that the noise tolerance of the MZL switch is more than a factor three larger than that of the MZ switch. This advantage comes at the expense of an increased ‘on’-state power consumption of 13.9 mW for the MZL switch at Von = 1.23 V, versus 4.7 mW for the MZ switch. The measured device characteristics of both the MZL and MZ switch are summarized in Table 1 .
4. Direct noise-tolerance measurements
In addition to the estimates made possible by Fig. 4, a more explicit demonstration of the noise tolerance of the MZL switch is made by applying a noisy drive signal to the switch and recording the resulting time-resolved T12 transmittance waveform. The applied drive signal consists of 150-ns-long pulses with peak voltage VD and a low duty cycle (<1%), combined with pseudo-random voltage noise generated by a 500 MHz 27-1 PRBS signal with variable peak-to-peak noise voltage Vpp, as illustrated in Fig. 5(b) .
The resulting T12 transmittance waveforms, recorded for the MZL switch with VD = 1.25 V and voltage noise levels in the range of 0-300 mVpp, are shown in Fig. 5(a). In the ‘off’-state, i.e. from time t = 0 ns to t = 20 ns, the T12 transmittance is essentially insensitive to the noise level. This results from the rectifying behavior of the p-i-n diode, which minimizes the resulting noise of the injected-carrier density, as is also apparent from the switching response shown in Fig. 4(a). For the ‘on’-state however, the situation is vastly different. It can be seen from Fig. 5(a) that the worst-case T12 extinction reduces significantly with increasing noise amplitude Vpp, from more than −20 dB at Vpp = 0 mV to only −7 dB at Vpp = 300 mV. The results of these noise tolerance measurements are summarized in Fig. 5(c), where the worst-case T12 extinction is plotted as a function of applied voltage noise Vpp. For VD = 1.25 V, the worst-case T12 extinction for the MZL increases monotonically with increasing Vpp. Better than −15 dB extinction of T12 is obtained for Vpp < 140 ± 14 mV, which is in excellent agreement with the switching curves in Fig. 4(a). The same measurements are repeated using an ‘on’-state voltage Von = 1.5 V. For this ‘on’-state voltage, the worst-case T12 extinction is relatively flat at about −15 dB, even for noise levels as high as 300 ± 30 mVpp. The lower extinction level at this larger bias results from the T12 switching sidelobe apparent in Fig. 4(a). However, the sidelobe-degraded performance can be avoided in future designs in which optical coupling in the curved waveguide sections is appropriately compensated. Figure 5(c) also shows the worst-case T12 extinction measured for the MZ switch. As expected with reference to Fig. 4(b), the noise tolerance is much lower, with only about 50 ± 5 mVpp noise budget for obtaining −15 dB T12 extinction.
5. Bit-error-rate measurements
Finally, in order to demonstrate the ability of the MZL switch to route high-speed data streams, even in the presence of substantial voltage noise, a series of bit-error-rate (BER) measurements are performed, using a 40 Gbps 27-1 PRBS input optical data stream. First, the switch is kept at zero bias voltage, and the data stream is fed into the a1 input port. The transmitted data stream is collected at output port b2, and a BER curve is recorded for the T12 transmission path. Subsequently, the switch is biased (continuous-wave) to VD = 1.3 V, and a BER curve is recorded for the T11 transmittance path. Finally, the switch bias is maintained at 1.3 V, while a substantial amount of voltage noise (200 mVpp) is added in the form of a 2 GHz 231-1 PRBS signal. The results are shown in Fig. 6 . The ‘off’-state and the ‘on’-state without noise have essentially identical BER curves. The BER curve recorded for T11 transmission in the presence of 200 mVpp voltage noise shows a power penalty of only 0.7 ± 0.2 dB with respect to the T11 BER curve without added noise. These results indicate the ability of the switch to route 40 Gbps data streams with low error count, even in the presence of substantial voltage noise.
The switching characteristics of the MZL and MZ switches presented in sections 3, 4 and 5 agree very well with the simulated switching performance presented in section 2, and more generally confirm the design methodology for digital switches presented in our earlier paper . However, the experimental data obtained here enables us to refine the p-i-n diode model, which in  was assumed to be an ideal diode, featuring an infinite carrier lifetime and negligible series resistance. These assumptions led to an underestimation of the noise-tolerance of both the MZ switch (estimated to be only 5 mVpp for −20 dB crosstalk in ) and the MZL switch.
In the present work, the noise tolerance measured for the fabricated MZ switch is as high as 25 mV for obtaining less than −20 dB crosstalk. This higher voltage-noise tolerance results from the voltage drop across the series resistance of the diode, which in turn increases with the amount of current flowing through the device. This current is inversely proportional to the free-carrier lifetime, which itself strongly depends upon the carrier recombination rate in the intrinsic region of the p-i-n diode, primarily resulting from surface trap states and other defects in the silicon lattice . As such, a short carrier lifetime in the diode actually helps to improve the voltage-noise tolerance of a carrier-injection-based interferometric optical switch. However, a short carrier lifetime also results in higher power consumption and the associated parasitic self heating of the diode, which can strongly degrade the switching performance for long ‘on’-state durations . Therefore, carrier-injection-based switches preferably make use of p-i-n diode phase shifters with long carrier lifetimes, which can be obtained by electrically passivating the surface of the diode’s intrinsic volume . The electro-optic behavior of such a well-passivated p-i-n diode is expected to more strongly resemble that of an ideal diode, in which case, adoption of a digital design as presented in this paper will be even more important for obtaining switches with sufficient voltage-noise tolerance.
Finally, even with the relatively high power consumption of about 14 mW currently obtained for the MZL switch, a low switching energy per bit could still be obtained for a WDM-encoded data stream. Assuming that a 15 × 40 Gbps WDM data stream (i.e. 15 channels with coarse 2 nm spacing) is simultaneously switched by the device, the switching energy per bit is only 23 fJ/bit.
We have demonstrated noise-tolerant operation of a broadband silicon electro-optic switch, based upon a multi-stage Mach-Zehnder lattice design. The multi-path interference at the output ports of the switch is engineered to produce a digital switching response, which results in a voltage-noise budget as high as 300 mVpp for obtaining lower than −15 dB crosstalk levels. This is more than six times better than a conventional single-stage MZ switch. A 40 Gbps data stream can be switched error-free by the device in the presence of 200 mVpp voltage noise, with a power penalty of only 0.7 dB. Other attractive features of the MZL switch are the higher obtainable ‘on’-state extinction levels and lower FCA-induced ‘on’-state insertion loss. The demonstrated switching characteristics are highly desired for switches that route network traffic at the nodes of proposed optical-networks–on-chip.
This work was supported in part by the DARPA APS Program, under contract HR0011-08-C-0102. The views expressed are those of the author and do not reflect the official policy or position of the Department of Defense or the U.S. Government. Distribution Statement “A” (Approved for Public Release, Distribution Unlimited). The authors would like to thank Daniel. M. Kuchta for assistance with measurement-automation software. The authors also gratefully acknowledge the efforts of the staff of the Microelectronics Research Laboratory (MRL) at the IBM T. J. Watson Research Center, where the devices were fabricated.
References and links
1. W. M. J. Green, S. Assefa, A. Rylyakov, C. Schow, F. Horst, and Y. A. Vlasov, “CMOS integrated silicon nanophotonics: enabling technology for exascale computational systems,” presented at SEMICON 2010, Chiba, Japan, 1–3 December, 2010.
2. S. Assefa, W. M. J. Green, A. Rylyakov, C. Schow, F. Horst, and Y. A. Vlasov, “CMOS integrated silicon nanophotonics: enabling technology for exascale computational systems,” in Proceedings of the Optical Fiber Communication Conference, (Optical Society of America, 2011) paper OMM6; See also http://www.research.ibm.com/photonics.
3. A. V. Krishnamoorthy, R. Ho, X. Z. Zheng, H. Schwetman, J. Lexau, P. Koka, G. L. Li, I. Shubin, and J. E. Cunningham, “Computer systems based on silicon photonic interconnects,” Proc. IEEE 97(7), 1337–1361 (2009). [CrossRef]
4. C. Batten, A. Joshi, J. Orcutt, A. Khilo, B. Moss, C. W. Holzwarth, M. A. Popovic, H. Q. Li, H. I. Smith, J. L. Hoyt, F. X. Kartner, R. J. Ram, V. Stojanovic, and K. Asanovic, “Building many-core processor-to-dram networks with monolithic CMOS silicon photonics,” IEEE Micro 29(4), 8–21 (2009). [CrossRef]
5. A. Shacham, K. Bergman, and L. P. Carloni, “Photonic networks-on-chip for future generations of chip multiprocessors,” IEEE Trans. Comput. 57(9), 1246–1260 (2008). [CrossRef]
6. J. Ahn, M. Fiorentino, R. G. Beausoleil, N. Binkert, A. Davis, D. Fattal, N. P. Jouppi, M. McLaren, C. M. Santori, R. S. Schreiber, S. M. Spillane, D. Vantrease, and Q. Xu, “Devices and architectures for photonic chip-scale integration,” Appl. Phys., A Mater. Sci. Process. 95(4), 989–997 (2009). [CrossRef]
7. J. Van Campenhout, W. M. J. Green, S. Assefa, and Y. A. Vlasov, “Low-power, 2 x 2 silicon electro-optic switch with 110-nm bandwidth for broadband reconfigurable optical networks,” Opt. Express 17(26), 24020–24029 (2009). [CrossRef]
8. A. Biberman, H. L. R. Lira, K. Padmaraju, N. Ophir, M. Lipson, and K. Bergman, “Broadband CMOS-compatible silicon photonic electro-optic switch,” in Proceedings of Conf. Lasers Electro-Optics (CLEO)2010, paper CPDA11, May 2010.
9. M. Yang, W. M. J. Green, S. Assefa, J. Van Campenhout, B. G. Lee, C. V. Jahnes, F. E. Doany, C. L. Schow, J. A. Kash, and Y. A. Vlasov, “Non-blocking 4x4 electro-optic silicon switch for on-chip photonic networks,” Opt. Express 19(1), 47–54 (2011). [CrossRef] [PubMed]
10. B. G. Lee, C. L. Schow, A. V. Rylyakov, J. Van Campenhout, W. M. J. Green, S. Assefa, F. E. Doany, M. Yang, R. A. John, C. V. Jahnes, J. A. Kash, and Y. A. Vlasov, “Demonstration of a digital CMOS driver co-designed and integrated with a broadband silicon photonic switch,” J. Lightwave Technol. 29(8), 1136–1142 (2011). [CrossRef]
11. J. Van Campenhout, W. M. J. Green, and Y. A. Vlasov, “Design of a digital, ultra-broadband electro-optic switch for reconfigurable optical networks-on-chip,” Opt. Express 17(26), 23793–23808 (2009). [CrossRef]
12. E. Alon, V. Stojanovic, and M. A. Horowitz, “Circuits and techniques for high-resolution measurement of on-chip power supply noise,” IEEE J. Solid-state Circuits 40(4), 820–828 (2005). [CrossRef]
13. R. A. Soref and B. R. Bennett, “Electrooptical effects in silicon,” IEEE J. Quantum Electron. 23(1), 123–129 (1987). [CrossRef]
14. J. Van Campenhout, W. M. J. Green, X. Liu, S. Assefa, R. M. Osgood, and Y. A. Vlasov, “Silicon-nitride surface passivation of submicrometer silicon waveguides for low-power optical switches,” Opt. Lett. 34(10), 1534–1536 (2009). [CrossRef] [PubMed]