Open Access
Add to BibSonomyAbstract
In this paper, a high-speed reconfigurable card-to-card optical interconnect architecture based on hybrid free-space and multi-mode fiber (MMF) propagation is proposed. The use of free-space signal transmission provides flexibility and reconfigurability and the MMF extends the achievable interconnection range. A printed-circuit-board (PCB) based integrated optical interconnect module is designed and developed and proof-of-concept demonstration experiments are carried out. Results show that 3 × 10 Gb/s reconfigurable optical interconnect is realized with ~12 cm free-space propagation and a 10 m MMF length. In addition, since air turbulence due to high temperature of electronic components and heat dissipation fans always exists in typical interconnect environments and it normally results in system performance degradation, its impact on the proposed reconfigurable optical interconnect scheme is also experimentally investigated. Results indicate that even with comparatively strong air turbulence, 3 × 10 Gb/s optical interconnects with flexibility can still be achieved and the power penalty is <0.7 dB.
© 2013 Optical Society of America
1. Introduction
The electronic transistors are continuously scaling down into the deep sub-micrometer range following the Moore’s law [1]. With the smaller-sized transistors, the operation speed becomes higher and more transistors can be densely integrated on a single chip. Therefore, the computing capability supported by a single chip has increased considerably [2]. Furthermore, the multi-core architecture has been widely used for electronic devices and systems, especially for high-performance computing and in data centers [3, 4]. Consequently, broadband interconnects between chips, cards and racks are highly required [5–7]. Si photonics and nano-photonics technologies have been widely studied and sustained improvement in multi-channel on-chip and chip-to-chip interconnection has been demonstrated [8–10]. However, for the card-to-card and rack-to-rack interconnects, the capacity is still highly limited. Conventionally, electrical cables are utilized for interconnections between cards and racks. Nevertheless, for high-speed operation, copper based cables are impractical and encounter several fundamental limitations, such as the electric power consumption, heat dissipation, transmission loss and latency, as well as electromagnetic interference [11].
The use of optical interconnects have been proposed to replace the copper based electrical cables and realize high-speed card-to-card interconnections [12–17]. Most of the reported optical interconnect architectures are based on the usage of parallel short-range optical links, such as the multi-mode fiber (MMF) ribbon based schemes [12–14]. In particular, F. Doany et al. recently proposed and experimentally demonstrated an MMF ribbon based optical interconnect scheme with data throughput of up to 1 Tb/s employing 48 parallel interconnect channels in conjunction with a holey CMOS chip [14]. In addition, the polymer waveguide based card-to-card optical interconnects have also been studied and the polymer waveguide can be embedded inside the printed circuit boards (PCBs) [15–17]. C. Schow et al. have demonstrated an aggregate data rate of 240 Gb/s using polymer waveguides fabricated using a conventional low-cost PCB technology [16].
However, both the MMF ribbon and polymer waveguide based schemes are inherently point-to-point and non-reconfigurable, so they are not suitable for dynamically interconnected electronic cards that require flexibility. Several reconfigurable card-to-card optical interconnect architectures based on free-space signal propagation have been proposed and investigated [18–22]. In these architectures, the modulated optical beam directly propagates in the free-space until the final destination card and since no waveguide is utilized, the optical signal can be switched along different directions in the air via a link-selection block. Consequently, significant flexibility can be achieved for the communications between different electronic cards. The link selection block has been realized using liquid crystal on silicon or Opto-VLSI processor and gigabit scale optical interconnect with flexibility has been experimentally demonstrated [18–20]. In addition, recently a 3 × 3 10 Gb/s reconfigurable optical interconnect scheme employing MEMS-based steering mirrors as an efficient link selection block through signal reflection has been experimentally demonstrated [21, 22]. Nevertheless, these free-space propagation only based architecture can only achieve very limited range, due to the Gaussian beam divergence property.
In this paper, we propose a high-speed reconfigurable optical interconnect architecture based on the hybrid free-space propagation and MMF propagation [23]. A proof-of-concept integrated reconfigurable 3 × 3 10 Gb/s interconnect module is demonstrated with a transmission range exceeding 10 m attained through the use of hybrid free-space and MMF propagation. In addition, in typical interconnect environments, air turbulence always exists due to the high temperature of electronic components or due to the heat dissipation fans, which mix the hot air with cold air [24, 25]. The air turbulence normally leads to effects such as signal scintillation, beam broadening, and beam wander, thus affecting the system BER performance and degrading the receiver sensitivity [26–28]. In this paper, the impact of air turbulence on our proposed reconfigurable free-space card-to-card optical interconnect architecture is also investigated. Results show that the BER performance of proposed interconnect scheme is degraded in turbulent environments and compared with the scenario without turbulence, the receiver sensitivity at BER < 10−9 suffers a power penalty of ~0.7 dB in comparatively strong turbulence.
2. Proposed reconfigurable card-to-card optical interconnect architecture based on hybrid free-space and MMF propagation
The architecture of proposed reconfigurable optical interconnect based on hybrid free-space and MMF propagation for high-performance computing and data centers is shown in Fig. 1. A dedicated optical interconnect module is integrated onto each electronic card (typically a printed circuit board or PCB) and the optical modules on different cards are connected with MMFs due to their higher coupling efficiency and easy handling in comparison with single mode fibers. Inside the optical transmitter module, a VCSEL array is used in conjunction with a collimating lens array to generate digitally-modulated collimated Gaussian beams. A MEMS-based steering mirror array is employed to adaptively guide the optical beams, thus providing reconfigurablity and flexibility. Through appropriate beam steering, the optical signals are coupled to MMF collimators and transmitted to the destination cards. In front of the fiber collimators, MEMS mirrors are also employed to minimize the incident angle and to maximize the coupling efficiency. At the receiver side, the modulated optical signals exiting the MMF collimators are appropriately steered with MEMS mirror elements as well and focused onto the corresponding PD elements. To realize the channel reconfiguration, a dedicated low-speed transmitter with a larger beam divergence is utilized to cover multiple fiber collimators, hence, the reconfiguration information can be broadcasted to all destination cards. With analog steering mirrors being used, the transmitted optical beam can be dynamically steered to different cards thus realizing reconfigurable optical interconnects.
Fig. 1 Architecture of proposed reconfigurable card-to-card optical interconnects with hybrid free-space and MMF propagations.
Compared with previously proposed and demonstrated reconfigurable optical interconnect schemes solely based on free-space signal transmission, the combinational usage of both free-space and MMF signal propagation has the capability of significantly extending the maximum achievable interconnection range, avoiding possible signal blocking problems inside the rack, improving the bit-error-rate (BER) performance, as well as connecting electronic cards located in different racks. This is because in the free-space only based optical interconnect architectures, Gaussian beam divergence is typically the major limitation since it results in low collected signal power and substantial inter-channel crosstalk [29]. In addition, VCSEL and PD arrays operating at the same wavelength are adopted because it (i) is cost-effective; (ii) eliminates the need for complex circuitry for the precise control of the wavelength of the VCSEL elements; and (iii) increases the aggregate bit rate.
3. Demonstration experiments and discussions
3.1 Experimental setup
The proposed reconfigurable card-to-card optical interconnect architecture based on hybrid free-space and MMF propagation was experimentally demonstrated using the setup shown in Fig. 2. In the experiments, a PCB-based integrated optical interconnect module was designed and fabricated using the standard FR4 process. Specifically, a 1 × 4 VCSEL array at 850 nm band, the corresponding VCSEL driver circuits (4 packaged drivers), a 1 × 4 PD array and 4 trans-impedance amplifier (TIA) chips were integrated onto a single small-size PCB. Two micro-lens arrays were then aligned and mounted on top of the VCSEL array and the PD array to, respectively, collimate the VCSEL beams and focus received optical beams onto the active windows of the PD elements. Each of the micro-lens arrays was attached to an XYZ translational stage, and the distance between the VCSEL/PD plane and the lens was manually tuned to be equal to the focal length of the micro-lenses for minimum divergence.
Fig. 2 Experimental setup (not to scale) for demonstrating the proposed reconfigurable card-to-card optical interconnect architecture based on hybrid free-space and MMF propagations.
Due to device limitations, separate MEMS mirror chips were used instead of arrays proposed before to provide reconfigurability and flexibility. The chips had < 5 ms point-to-point switching time and > 96% reflectivity by coating. These MEMS mirror chips were attached to XYZ translational stages as well and were used to switch the optical beams to various MMF collimators by changing the voltage applied to the actuators. The size of MEMS mirror chips used in the experiment was larger than the pitch of VCSEL, micro-lens and PD arrays. Therefore, as shown by the inset of Fig. 2, only three channels were used (the third VCSEL and PD elements were not used). In front of the MMF collimators, MEMS steering mirrors were also employed to minimize the incident angles for maximum coupling efficiency. In real applications the positions of optical interconnect modules are always pre-known. Therefore, the corresponding steering angles of MEMS mirrors are fixed and can be pre-set. However, a control algorithm is still needed to overcome impairments such as rack shaking and this is currently being developed. Finally, after propagating in the MMF, at the receiver side the signals exiting fiber collimators were dynamically steered to the active windows of corresponding PD elements for signal detection.
In the experiments, an 850 nm VCSEL array with a 250 µm pitch was used and wire-bonded onto the PCB. The average divergence angle of the VCSEL beams was ~17° and varied slightly among the 4 elements of the array. The 3-dB electrical bandwidth of the VCSELs was ~9.6 GHz. The VCSEL and PD micro-lens arrays had a pitch of 250 µm, a clear aperture of ~236 µm, and a focal length of ~656.5 µm. The PD array also had a pitch of 250 µm. Each PD element had an active aperture diameter of 60 µm and a responsivity of ~0.61 A/W at 850 nm, and was wire-bonded onto a TIA chip. The 3-dB bandwidth of the TIA was ~12.6 GHz and its differential trans-impedance was ~5 kΩ. The fiber collimators were designed for 62.5µm-core-diameter MMF (10 m) coupling, had a collimated beam divergence of ~0.4°, and a total diameter of 5 mm (no isolator was used). Six collimators were used at the transmitter side and three were employed at each of the receiving cards and the distance between adjacent collimators was chosen to be 10 mm. The MEMS steering mirror chips and fiber collimators were fixed at the same height and the horizontal distances between the optical interconnect modules and the MMF collimators were 8 cm and 4 cm for the transmitter and receiver, respectively. These additional MEMS mirrors in front of fiber collimators were used mainly due to the limited collecting angle of collimators and the comparatively large spacing between adjacent collimators. For most fiber collimators, if no additional MEMS mirrors were used, the incident angles of optical signals would have exceeded the collecting capability and insignificant signal power can be coupled into the MMF. Note that the reconfigurability of the system allows adequate steering of a certain collimated signal at the initial steering stage (to minimize the incident angle for each VCSEL signal), however, for the signals radiating from other VCSELs, the incident angles still exceed the fiber collimator collection angular range and cannot be coupled into the MMF. The use of a comparatively large VCSEL-to-MMF distance at the transmitter was necessary because of the limited maximum steering angle of the MEMS elements and the large channel spacing between fiber collimators, and hence, to ensure that each MEMS element can steer its beam to the all the MMF collimators.
3.2 Experimental results and discussions
During the measurements, the bit rate for each channel was set to 10 Gb/s and all VCSELs were directly modulated with on-off-keying (OOK) format with 231-1 PRBS data. To demonstrate the concept of reconfigurable optical interconnects architecture, two scenarios were considered. First, as shown in Fig. 2, VCSEL 1 was connected to card 3 and VCSELs 2 and 4 were connected to card 2. The measured BER versus the transmission power from VCSELs is shown in Fig. 3(a). It is clear that with the same radiation power, channel 2 performs better than others and this is mainly due to the higher coupling efficiency to the MMF. In addition, with a transmission power of ~1.4 mW, error-free operation (BER of 10−9) is achieved for all channels. It should be noted that the BER requirement in interconnects can be even higher and forward-error-correction (FEC) codes can be used to improve the system performance at the cost of overhead (7% or 20%) [30, 31]. Figure 3(b) shows the measured BER versus the received optical power (obtained by changing the output power levels of the VCSEL elements) for the three working channels. It can be seen that channel 1 has the best receiver sensitivity and the sensitivity for channels 2 and 4 is slightly worse. This is because the PD elements of channels 2 and 4 are both located on card 2, leading to inter-channel crosstalk.
Fig. 3 Measured BER for configuration 1. Bit rate = 10 Gb/s for each channel. Channel 1: VCSEL 1 to PD 2 of Card 3; Channel 2: VCSEL 2 to PD 4 of Card 2; and Channel 4: VCSEL 4 to PD 1 of Card 2 (channel numbers are based on the original VCSEL element number). (a) BER versus the VCSEL transmission power; and (b) BER versus the received power (reprinted from [23]).
In the second configuration, VCSEL element 2 sent data to fiber collimator 4 at the transmitter side, which was interconnected to card 2, and then the data was transmitted to PD element 2 for detection. The other two VCSEL elements were connected to card 3 (VCSEL element 1 sent data to fiber collimator 2 at the transmitter side and the data was detected with PD element 1 of card 3; VCSEL element 4 sent data to fiber collimator 1 at the transmitter side and the data was detected with PD element 2 of card 3). The measured BER versus the transmitted optical power and the receiver sensitivity are shown in Fig. 4(a) and Fig. 4(b), respectively. It can be seen that error-free operation has again been achieved for all three channels and channel 1 still has the best receiver sensitivity, demonstrating the reconfigurablity and flexibility of proposed optical interconnect architecture. In addition, compared with the results shown in Fig. 3(b), channel 4 has worse receiver sensitivity. This is because in the first configuration, channel 4 signal was detected with PD element 1 of card 2 and PD elements 1 and 4 of card 2 were used, while in the second configuration, channel 4 signal was detected with PD element 2 of card 3 and PD elements 1 and 2 of card 3 were used. Therefore, in the second configuration channel 4 had much stronger inter-channel crosstalk, leading to degraded receiver sensitivity.
Fig. 4 Measured BER for configuration 2. Bit rate = 10 Gb/s for each channel. Channel 1: VCSEL 1 to PD 1 of Card 3; Channel 2: VCSEL 2 to PD 2 of Card 2; and Channel 4: VCSEL 4 to PD 2 of Card 3 (channel numbers are based on the original VCSEL element number). (a) BER versus the VCSEL transmission power; and (b) BER versus the received power (reprinted from [23]).
To investigate the impact of inter-channel crosstalk on the proposed card-to-card optical interconnect architecture, the worst-case scenario was also considered, where all signals were sent to card 2 and detected with PD elements 1 to 3. The measured BER with respect to the VCSEL transmission power and receiver sensitivity results are shown in Fig. 5(a) and 5(b), respectively. It is clear that the BER of all channels under this configuration is worse than the previous results, showing the impact of inter-channel crosstalk. In addition, it can be seen from Fig. 5(b) that channel 1 and channel 4 have similar receiver sensitivity while the channel 2 receiver sensitivity is worse. This is mainly because that channel used the PD element 2 on card 3, which was in the middle of PD array and had two adjacent crosstalk sources while other receiving elements only had one adjacent crosstalk source, making it vulnerable to crosstalk. Furthermore, comparing the results shown in Figs. 3-5, it can be concluded that even under the worst-case scenario, in the proposed reconfigurable optical interconnect architecture the power penalty due to inter-channel crosstalk is still <1 dB. This comparatively small power penalty can be attributed to the usage of MEMS steering mirrors, since the inter-channel crosstalk signals do not strike the mirror with the optimum incident angle and most of the crosstalk power is steered out of the active window of corresponding detection element.
Fig. 5 Measured BER for configuration 3 (the worst-case scenario). Bit rate = 10 Gb/s for each channel. Channel 1: VCSEL 1 to PD 1 of Card 2; Chanel 2: VCSEL 2 to PD 2 of Card 2; and Channel 4: VCSEL 4 to PD 3 of Card 2 (channel numbers are based on the original VCSEL element number). (a) BER versus the VCSEL transmission power; and (b) BER versus the received power.
4. Impact of turbulence on the proposed reconfigurable card-to-card optical interconnect architecture
4.1 Experimental setup
In typical interconnect environments, air turbulence always exists due to the high temperature of electronic components or due to the heat dissipation fans, which mix the hot air with cold air [24, 25]. For free-space based optical interconnects, the air turbulence normally leads to refractive index fluctuations along the optical path and results in effects such as signal scintillation, beam broadening, and beam wander [26–28], thus affecting the system BER performance and degrading the receiver sensitivity. In the reconfigurable optical interconnect scheme proposed here, free-space signal propagation is also used to provide the desired the reconfigurability and flexibility. Therefore, the impact of air turbulence on the proposed system architecture is also experimentally investigated and the experimental setup is shown in Fig. 6. Atmospheric turbulence was intentionally introduced into the interconnect architecture with heaters and electrical fans. The temperature of the hot air flow from the heater was ~60 °C. The heater and fans were placed at almost the same height as the interconnect link and the horizontal distance between the link and the turbulence source was ~40 cm to emulate moderate impairment. To emulate the comparatively strong air turbulence, the distance was reduced to ~20 cm. Due to devices limitations, we were not able to measure more parameters characterizing the air turbulence. However, since this experiment was carried out to demonstrate the feasibility of proposed reconfigurable optical interconnect architecture in real situations, our emulation of air turbulence is still reasonably practical.
Fig. 6 Experimental setup for investigating the impact of turbulence on the proposed optical interconnect architecture.
4.2 Results and discussions
The measured receiver sensitivity under both emulated moderate and comparatively strong turbulences is shown in Fig. 7(a) and 7(b), respectively. The link configuration is shown by Fig. 6 where all three working channels were detected with the receivers integrated on card 2 (PD elements 1-3). It can be seen that channel 2 always has the worst receiver sensitivity, consistent with the results shown in Fig. 5(b). In addition, the power penalty in receiver sensitivity under moderate turbulence is ~0.4 dB and for comparatively strong turbulence, the power penalty is ~0.7 dB. Furthermore, it is clear from Fig. 7 that even with fairly strong turbulence, the proposed reconfigurable card-to-card optical interconnect scheme based on hybrid free-space and MMF propagation can still achieve BER performance much better than 10−9.
Fig. 7 Receiver sensitivity for configuration 3 (the worst-case scenario). Bit rate = 10 Gb/s for each channel. (a) Under moderate turbulence; and (b) under strong turbulence.
5. Conclusions
In this paper, a novel reconfigurable card-to-card optical interconnect architecture based on hybrid free-space and MMF propagation has been proposed. The scheme has been shown to be capable of providing reconfigurability and flexibility through free-space signal transmission and the achievable interconnection range has been extended by MMF connections. Up to 3 × 10 Gb/s data transmission has been experimentally demonstrated with ~12 cm free-space propagation distance and ~10 m MMF length. BER better than 10−11 has been achieved and the reconfigurability has been demonstrated through three different link configurations. Experimental results have shown that even in the worst-case scenario, receiver sensitivity (at a BER of 10−9) better than −11.88 dBm has still been realized.
In addition, the performance of proposed reconfigurable optical interconnect architecture under air turbulence has also been experimentally investigated. Both moderate and strong turbulences have been emulated in lab conditions and the power penalty in receiver sensitivity has been shown to be ~0.4 dB and ~0.7 dB, respectively. In addition, it has been shown that even with comparatively strong air turbulence, BER better than 10−11 has still been achieved.
Finally, the emulation of air turbulence in this paper only considered direct air flow and simple temperature change. More realistic emulation of turbulence, where more complex air flows such as eddies are considered, and more detailed investigation deserve further attention. Furthermore, the effect of rack shaking in real applications should also be studied in the free-space based optical interconnect schemes.
Acknowledgment
This work was supported in part by NICTA and by the Department of Industry, Innovation, Science, Research and Tertiary Education (DIISRTE). NICTA is funded by the Australian Government as represented by the Department of Broadband, Communications and the Digital Economy and the Australian Research Council through the ICT Excellence Program.
References and links
1. T. C. Chen, “Challenges for silicon technology scaling in the nanoscale era,” in Proceedings of European Solid State Device Research Conference (ESSDERC, Athens, Greece, 2009), pp. 1–7.
2. J. S. Walling and D. Allstot, “CMOS powers toward system-on-chip integration,” IEEE Microw. Mag. 12(1), 6–16 (2011). [CrossRef]
3. V. S. Nandakumar and M. Marek-Sadowska, “A low energy network-on-chip fabric for 3-D multi-core architectures,” IEEE Emerging and Sel. Topics in Circuits and Systems 2, 266–277 (2012).
4. R. Kalla, B. Sinharoy, W. J. Starke, and M. Floyd, “Power7: IBM’s next-generation server processor,” IEEE Micro 30(2), 7–15 (2010). [CrossRef]
5. S. Kamil, L. Oliker, A. Pinar, and J. M. Shalf, “Communication requirements and interconnect optimization for high-end scientific applications,” IEEE Trans. Parallel Distrib. Syst. 21(2), 188–202 (2010). [CrossRef]
6. A. Vahdat, M. Al-Fares, N. Farrington, R. N. Mysore, G. D. Porter, and S. Radhakrishnan, “Scale-out networking in the data center,” IEEE Micro 30(4), 29–41 (2010). [CrossRef]
7. N. Ilyadis, “The evolution of next-generation data center networks for high capacity computing,” in Proceedings of Symposium on VLSI Circuits (Honolulu, Hawaii, 2012), pp. 1–5. [CrossRef]
8. M. A. Taubenblatt, “Optical interconnects for high performance computing,” J. Lightwave Technol. 30(4), 448–457 (2012). [CrossRef]
9. M. Heck, H.-W. Chen, A. W. Fang, B. R. Koch, D. Liang, H. Park, M. N. Sysak, and J. E. Bowers, “Hybrid silicon photonics for optical interconnects,” IEEE J. Sel. Top. Quantum Electron. 17(2), 333–346 (2011). [CrossRef]
10. K. Kim, “Electronics and photonics convergence on Si (CMOS) platform,” in Proceedings of 17th Opto-Electronics and Communications Conference (OECC, Busan, Korea, 2012), pp. 1–4. [CrossRef]
11. A. F. Benner, M. Ignatowski, J. A. Kash, D. M. Kuchta, and M. B. Ritther, “Exploitation of optical interconnects in future server architectures,” IBM J. Res. Develop. 49(4.5), 755–775 (2005). [CrossRef]
12. P. K. Pepeljugoski, F. E. Doany, D. M. Kuchta, L. Schares, C. L. Schow, M. B. Ritter, and J. A. Kash, “Data center and high performance computing interconnects for 100 Gb/s and beyond,” in Proceedings of Optical Fiber Communication Conference and National Fiber Optic Engineers Conference (OFC/NFOEC, Anaheim, CA, 2007), pp. 1–3. [CrossRef]
13. L. A. Buckman-Windover, J. N. Simon, S. A. Rosenau, K. S. Giboney, G. M. Flower, L. W. Mirkarimi, A. Grot, B. Law, C.-K. Lin, A. Tandon, R. W. Gruhlke, H. Xia, G. Rankin, M. R. T. Tan, and D. W. Dolfi, “Parallel optical interconnects >100 Gb/s,” J. Lightwave Technol. 22(9), 2055–2063 (2004). [CrossRef]
14. F. E. Doany, B. G. Lee, A. V. Rylyakov, D. M. Kuchta, C. Baks, C. Jahnes, F. Libsch, and C. L. Schow, “Terabit/sec VCSEL-based parallel optical module based on holey CMOS transceiver IC,” in Proceedings of Optical Fiber Communication Conference and Exposition and the National Fiber Optic Engineers Conference (OFC/NFOEC, Los Angeles, CA, 2012), pp. PDP5D.9.
15. R. Dangel, C. Berger, R. Beyeler, L. Dellmann, M. Gmur, R. Hamelin, F. Horst, T. Lamprecht, T. Morf, S. Oggioni, M. Spreafico, and B. J. Offrein, “Polymer-waveguide-based board-level optical interconnect technology for datacom applications,” IEEE Trans. Adv. Packag. 31(4), 759–767 (2008). [CrossRef]
16. C. L. Schow, F. E. Doany, C. W. Baks, Y. H. Kwark, D. M. Kuchta, and J. A. Kash, “A single-chip CMOS-based parallel optical transceiver capable of 240-Gb/s bidirectional data rates,” J. Lightwave Technol. 27(7), 915–929 (2009). [CrossRef]
17. F. E. Doany, C. L. Schow, C. W. Baks, D. M. Kuchta, P. Pepeljugoski, L. Schares, R. A. Budd, F. R. Libsch, R. Dangel, F. Horst, B. J. Offrein, and J. A. Jeffrey, “160 Gb/s bidirectional polymer-waveguide board-level optical interconnects using CMOS-based transceivers,” IEEE Trans. Adv. Packag. 32(2), 345–359 (2009). [CrossRef]
18. C. J. Henderson, D. G. Leyva, and T. D. Wilkinson, “Free space adaptive optical interconnect at 1.25 Gb/s with beam steering using a ferroelectric liquid-crystal SLM,” J. Lightwave Technol. 24(5), 1989–1997 (2006). [CrossRef]
19. N. McArdle, M. Naruse, H. Toyoda, Y. Kobayashi, and M. Ishikawa, “Reconfigurable optical interconnections for parallel computing,” Proc. IEEE 88(6), 829–837 (2000). [CrossRef]
20. M. Aljada, K. E. Alameh, Y. T. Lee, and I. S. Chung, “High-speed (2.5 Gbps) reconfigurable inter-chip optical interconnects using opto-VLSI processors,” Opt. Express 14(15), 6823–6836 (2006). [CrossRef] [PubMed]
21. K. Wang, A. Nirmalathas, C. Lim, E. Skafidas, and K. Alameh, “Experimental demonstration of high-speed free-space reconfigurable card-to-card optical interconnects,” Opt. Express 21(3), 2850–2861 (2013). [CrossRef] [PubMed]
22. K. Wang, A. Nirmalathas, C. Lim, E. Skafidas, and K. Alameh, “Experimental demonstration of 3×3 10 Gb/s reconfigurable free space optical card-to-card interconnects,” Opt. Lett. 37(13), 2553–2555 (2012). [CrossRef] [PubMed]
23. K. Wang, A. Nirmalathas, C. Lim, E. Skafidas, and K. Alameh, “Experimental demonstration of reconfigurable optical interconnect based on hybrid free-space and multi-mode fiber propagation,” in Proceedings of Optical Fiber Communication Conference and Exposition and National Fiber Optic Engineers Conference (OFC/NFOEC, Anaheim, CA, 2013), pp. OTh1A.6. [CrossRef]
24. R. Rachmani and S. Arnon, “Server backplane with optical wavelength diversity links,” J. Lightwave Technol. 30(9), 1359–1365 (2012). [CrossRef]
25. R. Rachmani, A. Zilberman, and S. Arnon, “Computer backplane with free space optical links: air turbulence effects,” J. Lightwave Technol. 30(1), 156–162 (2012). [CrossRef]
26. L. C. Andrews and R. L. Phillips, Laser Beam Propagation through Random Media, SPIE Optical Engineering Press: Bellingham, Washington, 1998.
27. S. Bloom, E. Korevaar, J. Schuster, and H. Willebrand, “Understanding the performance of free-space optics,” J. Opt. Netw. 2, 178–200 (2003).
28. H. G. Sandalidis, T. A. Tsiftsis, G. K. Karagiannidis, and M. Uysal, “BER performance of FSO links over strong atmospheric turbulence channels with pointing errors,” IEEE Commun. Lett. 12(1), 44–46 (2008). [CrossRef]
29. K. Wang, A. Nirmalathas, C. Lim, E. Skafidas, and K. Alameh, “High-speed reconfigurable free-space card-to-card optical interconnects,” IEEE Photonics Journal 4(5), 1407–1419 (2012). [CrossRef]
30. T. Mizuochi, Y. Miyata, K. Kubo, T. Sugihara, K. Onohara, and H. Yoshida, “Progress in soft-decision FEC,” in Proceedings of Optical Fiber Communication Conference and Exposition and National Fiber Optic Engineers Conference (OFC/NFOEC, Los Angeles, CA, 2011), pp. 1–3.
31. T. Mizuochi, “Recent progress in forward error correction and its interplay with transmission impairments,” IEEE J. Sel. Top. Quantum Electron. 12(4), 544–554 (2006). [CrossRef]
