Conference Paper

28-Gb/s × 24-channel CDR-integrated VCSEL-based transceiver module for high-density optical interconnects

Authors:
  • Furukawa Fitel Optical Device
To read the full-text of this research, you can request a copy directly from the authors.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... In this work, in order for to meet the criteria that was defined earlier in terms of latency, bandwidth, power consumption, bandwidth density and footprint we will exploit a commercial Silicon-photonic (SiP) (second row of Table I) optical on board transceiver operating in single-mode for the optical interconnects used in dRedBox architecture. This particular transceiver is composed of eight spatially multiplexed optical channels as presented in Table I ( [27]). The use of such transceivers operating in single-mode as opposed to multimode allows for a scalable network operating based on optical circuit switching (OCS), since the use of single mode fiber allows for optical switches with higher port counts [27]. ...
... This particular transceiver is composed of eight spatially multiplexed optical channels as presented in Table I ( [27]). The use of such transceivers operating in single-mode as opposed to multimode allows for a scalable network operating based on optical circuit switching (OCS), since the use of single mode fiber allows for optical switches with higher port counts [27]. It also offers the longest reach or having the highest power budget while delivering 1x10 12 FEC free BER. ...
Article
Full-text available
Disaggregated rack-scale data centers have been proposed as the only promising avenue to break the barrier of the fixed CPU-to-memory proportionality caused by main-tray direct-attached conventional/traditional server-centric systems. However, memory disaggregation has stringent network requirements in terms of latency, energy efficiency, bandwidth, and bandwidth density. This paper identifies all the requirements and key performance indicators of a network to disaggregate IT resources while summarizing the progress and importance of optical interconnects. Crucially, it proposes a rack-and-cluster scale architecture, which supports the disaggregation of CPU, memory, storage, and/or accelerator blocks. Optical circuit switching forms the core of this architecture, whereas the end-points (IT resources) are equipped with on-chip programmable hybrid electrical packet/circuit switches. This architecture offers dynamically reconfigurable physical topology to form virtual ones, each embedded with a set of functions. It analyzes the latency overhead of disaggregated DDR4 (parallel) and the proposed hybrid memory cube (serial) memory elements on the conventional and the proposed architecture. A set of resource allocation algorithms are introduced to (1) optimally select disaggregated IT resources with the lowest possible latency, (2) pool them together by means of a virtual network interconnect, and (3) compose virtual disaggregated servers. Simulation findings show up to a 34% resource utilization increase over traditional data centers while highlighting the importance of the placement and locality among compute, memory, and storage resources. In particular, the network-aware locality-based resource allocation algorithm achieves as low as 15 ns, 95 ns, and 315 ns memory transaction round-trip latency on 63%, 22%, and 15% of the allocated virtual machines (VMs) accordingly while utilizing 100% of the CPU resources. Furthermore, a formulation to parameterize and evaluate the additional financial costs endured by disaggregation is reported. It is shown that the more diverse the VM requests are, the higher the net financial gain is. Finally, an experiment was carried out using silicon photonic midboard optics and an optical circuit switch, which demonstrates forward error correction free 10 − 12 bit error rate performance on up to five-tier scale-out networks.
... For data-center application, the integrated VCSEL array could be operated for > 100-Gbps applications [37][38][39][40][41][42][43][44][45][46][47][48] . Parallel interconnects of 25-Gbps multimode VCSELs in combination with parallel ribbon multimode fiber have offered a clear path on connecting the data center networks at up to 400-Gbps or more data rate transmission potentially. ...
... In 2015, Westbergh et al. proposed 2-D six-channel 850-nm VCSEL arrays to achieve 40-Gbps error-free transmission at 85 °C, which enabled an aggregate capacity of 240-Gbps over a single multicore fiber 42 . A 28-Gbps × 24-channel clock-and-data recovery (CDR) integrated VCSEL optical transceiver module with a data rate density of 1344 Gbps/inch 2 operated under relatively low link power was also proposed by Nagashima et al. 45 . In addition, Tsunoda et al. also developed a high-density multi-rate VCSEL transceiver with CDRs to suppress the jitter of optical link for approaching a 24 to 34-Gbps × 4 interconnect 46 . ...
Article
Full-text available
An effective heat dissipation of uncooled 400-Gbps (16×25-Gbps) form-factor pluggable (CDFP) optical transceiver module employing chip-on-board multimode 25-Gbps vertical-surface-emitting-laser (VCSEL) and 25-Gbps photodiode (PD) arrays mounted on a brass metal core embedded within a printed circuit board (PCB) is proposed and demonstrated. This new scheme of the hollow PCB filling with thermally-dissipated brass metal core was simulated and used for high temperature and long term stability operation of the proposed 400-Gbps CDFP transceiver. During one-hour testing, a red-shift of central wavelength by 0.4-nm corresponding temperature increment of 6.7 °C was observed with the brass core assisted cooler module. Such a temperature change was significantly lower than that of 28.3 °C for the optical transceiver driven with conventional circuit board. After 100-m distance transmission over a multimode fiber (OM4), the 400-Gbps CDFP transceiver exhibited dispersion penalty of 2.6-dB, power budget of ≧ 3-dB, link loss of ≦ 0.63-dB, mask margin of 20%, and bit error rate (BER) of <10−12 with maintained stability more than one hour. The developed 400-Gbps CDFP transceiver module employing low-power consumption VCSEL and PD arrays, effective coupling lens arrays, and well thermal-dissipation brass metal core is suitable for use in the low-cost and high-performance data center applications.
... For the next generation optical interconnect, a data rate of optical path will be 400 Gb/s and more. To support such a very high data rate, we demonstrated a clock data recovery (CDR)- integrated vertical cavity surface emitting laser (VCSEL)-based transceiver module together with a unique packaging structure to achieve a data rate density as high as 1344 Gb/s/inch 2 [6]. The total power consumption was as high as 9.1 W when operating all 24 channels. ...
Chapter
Full-text available
Spiking neural networks (SNN) has been widely used to solve complex tasks such as pattern recognition, image classification and so on. The neuromorphic processors which use SNN to perform computation have been proved to be powerful and energy-efficient. These processors generally use Network-on-Chip (NoC) as the interconnect structure between neuromorphic cores. However, the connections between neurons in SNN are very dense. When a neuron fire, it will generate a large number of data packets. This will result in congestion and increase the packet transmission latency dramatically in NoC. In this paper, we proposed a software-hardware co-exploration framework to alleviate this problem. This framework consists of three parts: software simulation, packet extraction&mapping, and hardware evaluation. At the software level, we can explore the impact of packet loss on the classification accuracy of different applications. At the hardware level, we can explore the impact of packet loss on transmission latency and power consumption in NoC. Experimental results show that when the neuromorphic processor runs MNIST handwritten digit recognition application, the communication delay can be reduced by 11%, the power consumption can be reduced by 5.3%, and the classification accuracy can reach 80.75% (2% higher than the original accuracy). When running FSDD speech recognition application, the communication delay can be reduced by 22%, the power consumption can be reduced by 2.2%, and the classification accuracy can reach 78.5% (1% higher than the original accuracy).
Chapter
This paper presents a 32 Gb/s low power little area re-timer with Phase Interpolator (PI) based Clock and Data Recovery (CDR). To further ensure signal integrity, both a Continuous Time Linear Equalizer (CTLE) and Feed Forward Equalizer (FFE) are adapted. To save power dissipation, a quarter-rate based 3-tap FFE is proposed. To reduce the chip area, a Band-Band Phase Discriminator (BBPD) based PI CDR is employed. In addition, a 2-order digital filter is adopted to improve the jitter performance in the CDR loop. This re-timer is achieved in 65 nm CMOS technology and supplied with 1.1 V. The simulation results show that the proposed re-timer can work at 32 Gb/s and consumes 91 mW. And it can equalize >−12 dB channel attenuation, tolerate the frequency difference of 200 ppm.
Article
This paper gives a detailed description of a >1.3-Tb/s VCSEL-based on-board parallel-optical transceiver module for high-density optical interconnects. The optical module integrates a 28-Gb/s × 24-channel transmitter and receiver into one package with a 1-in 2 footprint, thereby yielding a data rate density as high as 1.34 Tb/s/in 2 . A unique module design is developed to utilize the whole top surface for thermal dissipation and the bottom surface for optical and electrical interfaces. The heat dissipation characteristics are studied in simulations and experiments. A test system of 24-channel optical loop-back link is built to evaluate the performance when operating all 24 channels by 28.05-Gb/s 2 31 -1 PRBS bit streams simultaneously. With a total power consumption of 9.1 W, the optical module can be operated within the operating case temperature under a practical air cooling condition. When operating all 24 channels simultaneously, a total jitter margin at BER of 10 -12 is larger than 0.4 U.I. at a monitor channel. The measured jitter margin is consistent regardless of any channel operation.
Article
An optical backplane ecosystem is described and demonstrated that is capable of multi-Tb/s bandwidth and is based on embedded polymer waveguides, passive optical backplane connectors, and midboard optical transceivers with bandwidth up to 28 Gb/s per lane. These systems provide the highest bandwidth-density, lowest power consumption, while maintaining the signal integrity. Ecosystems built around this architecture will provide the bandwidth-density required for next generation fabric interconnect for storage, switching, and routing applications in future high capacity generations of Data Centers and HPC systems. To demonstrate the applicability of this technology, it was used to provide embedded optical connectivity within a functional data storage enclosure.