This PDF file contains the front matter associated with SPIE Proceedings Volume 9775, including the Title Page, Copyright information, Table of Contents, and Conference Committee listing.

Traffic to and from data centers is now reaching Zettabytes/year. Even the smallest of businesses now rely on data centers for revenue generation. And, the largest data centers today are orders of magnitude larger than the supercomputing centers of a few years ago. Until quite recently, for most data center managers, optical data centers were nice to dream about, but not really essential. Today, the all-optical data center – perhaps even an all-single mode fiber (SMF) data center is something that even managers of medium-sized data centers should be considering. Economical transceivers are the key to increased adoption of data center optics. An analysis of current and near future data center optics economics will be discussed in this paper.

In this paper we discuss design and characterization of silicon-photonics-based 100 Gbps (4×26 Gbps) transceivers for parallel single mode fiber communication. We also address some key underlying technologies including silicon photonics wafer processing, photonic device libraries, light source integration and packaging technologies.

Short-reach optical links such as those used in data centers pre-dominantly employ VCSELs together with laser- optimized OM4 and OM3 multimode fiber (MMF), mainly due to their reliability, energy-efficiency and low end-to-end system cost. The IEEE 802.3bm specification for 100Gbps Ethernet utilizes four parallel MMFs each operating at a serial data rate of 25Gbps. Due to the rapidly increasing internet traffic, the IEEE P802.3bs Task Force is working towards a 400Gbps Ethernet standard requiring a commensurate increase in the number of parallel fibers deployed. Using 16 parallel lanes, while feasible, is not the most efficient use of cabling. One solution to the data rate - cable density problem is the use of shortwave wavelength division multiplexing (SWDM) near 850nm. For example, employing four wavelengths separated by ~30nm (with an operational window of ~840-950nm) results in a four-fold increase in the per-fiber data rate. Furthermore, SWDM can be combined with the parallel solution to support 400Gbps with the same cable density as the current 100Gbps Ethernet solution using OM4 fiber. Conventional laser-optimized OM4 gives diminished performance at the longer wavelengths compared to 850nm. Shifting the OM4 optimization wavelength to longer wavelengths sacrifices the 850nm performance. In this paper, we present next-generation wideband multimode fibers (NG-WBMMF) that are optimized for SWDM operation using a novel design approach employing multiple dopants. We have fabricated and characterized a wideband MMF that is OM4 compliant over the 850-950nm wavelength window. BER measurements demonstrate that this next-generation WB MMF satisfies the pre-FEC requirement of 5 × 10^-5 even after transmission over 300m.

In this paper, we review our experimental and simulation work on 28- and 56-Gbaud PAM4-based 100Gbps and 400Gbps short-haul links, including 10, 40, and 80km transmission distances. We also discuss the system performances, limitations, and areas of improvement for these links and their future applications.

Faced with surging datacenter traffic demand, system designers are turning to multi-level optical modulation with direct detection as the means of reaching 100 Gb/s in a single optical lane; a further upgrade to 400 Gb/s is envisaged through wavelength-multiplexing of multiple 100 Gb/s strands. In terms of modulation formats, PAM-4 and PAM-8 are considered the front-runners, striking a good balance between bandwidth-efficiency and implementation complexity. In addition, the emergence of energy-efficient, high-speed CMOS digital-to-analog converters (DACs) opens up new possibilities: Spectral shaping through digital filtering will allow squeezing even more data through low-cost, low-bandwidth electro-optic components. In this work we demonstrate an optical interconnect based on an EAM that is driven directly with sub-volt electrical swing by a 65 GSa/s arbitrary waveform generator (AWG). Low-voltage drive is particularly attractive since it allows direct interfacing with the switch/server ASIC, eliminating the need for dedicated, power-hungry and expensive electrical drivers. Single-wavelength throughputs of 180 and 120 Gb/s are experimentally demonstrated with 60 Gbaud optical PAM-8 and PAM-4 respectively. Successful transmission over 1250 m SMF is achieved with direct-detection, using linear equalization via offline digital signal processing in order to overcome the strong bandwidth limitation of the overall link (~20 GHz). The suitability of Nyquist pulse shaping for optical interconnects is also investigated experimentally with PAM-4 and PAM-8, at a lower symbol rate of 40 Gbaud (limited by the sampling rate of the AWG). To the best of our knowledge, the rates achieved are the highest ever using optical PAM-M formats.

In this paper, we present a performance review of the state of the art PAM4 IC chipset implementation on the mitigation of optical and bandwidth impairments at 100Gbps and beyond. We experimentally demonstrate, for the first time, by employing newly developed PAM4 chipset and direct detection with commercial 25Gbaud devices, the link operates error free under KR4 FEC threshold at 1310nm wavelength(s) for up to 40km distance of standard single mode fiber. It is foreseeable that PAM4 signaling will soon become the viable and cost effective solution for intra and inter data center connectivity for 40/50GbE, 100/400GbE and beyond.

There is considerable interest in combining multi-level pulsed amplitude modulation formats (PAM-L) and forward error correction (FEC) in next-generation, short-range optical communications links for increased capacity. In this paper we derive new formulas for the optical power penalties due to modulation format complexity relative to PAM-2 and due to inter-symbol interference (ISI). We show that these penalties depend on the required system bit-error rate (BER) and that the conventional formulas overestimate link penalties. Our corrections to the standard formulas are very small at conventional BER levels (typically 1×10^-12) but become significant at the higher BER levels enabled by FEC technology, especially for signal distortions due to ISI. The standard formula for format complexity, P = 10log(L-1), is shown to overestimate the actual penalty for PAM-4 and PAM-8 by approximately 0.1 and 0.25 dB respectively at 1×10^-3 BER. Then we extend the well-known PAM-2 ISI penalty estimation formula from the IEEE 802.3 standard 10G link modeling spreadsheet to the large BER case and generalize it for arbitrary PAM-L formats. To demonstrate and verify the BER dependence of the ISI penalty, a set of PAM-2 experiments and Monte-Carlo modeling simulations are reported. The experimental results and simulations confirm that the conventional formulas can significantly overestimate ISI penalties at relatively high BER levels. In the experiments, overestimates up to 2 dB are observed at 1×10^-3 BER.

Datacenter traffic is exploding. Ongoing advancements in network infrastructure that ride on Moore’s law are unable to keep up, necessitating the introduction of multiplexing and advanced modulation formats for optical interconnects in order to overcome bandwidth limitations, and scale lane speeds with energy- and cost-efficiency to 100 Gb/s and beyond. While the jury is still out as to how this will be achieved, schemes relying on intensity modulation with direct detection (IM/DD) are regarded as particularly attractive, due to their inherent implementation simplicity. Moreover, the scaling-out of datacenters calls for longer transmission reach exceeding 300 m, requiring single-mode solutions. In this work we advocate using 16-QAM sub-cycle Nyquist-SCM as a simpler alternative to discrete multitone (DMT), but which is still more bandwidth-efficient than PAM-4. The proposed optical interconnect is demonstrated at 112 Gb/s, which, to the best of our knowledge, is the highest rate achieved in a single-polarization implementation of SCM. Off-the-shelf components are used: A DFB laser, a 24.3 GHz electro-absorption modulator (EAM) and a limiting photoreceiver, combined with equalization through digital signal processing (DSP) at the receiver. The EAM is driven by a low-swing (<1 V) arbitrary waveform generator (AWG), which produces a 28 Gbaud 16-QAM electrical signal with carrier frequency at ~15 GHz. Tight spectral shaping is leveraged as a means of maintaining signal fidelity when using low-bandwidth electro-optic components; matched root-raised-cosine transmit and receive filters with 0.1 excess bandwidth are thus employed. Performance is assessed through transmission experiments over 1250 m and 2000 m of SMF.

Triggered by the recent vast demands of the information bandwidth, high speed transmission networking system, such as 100G/400G, have been developed and installed with extremely high speed. In order to lead the proper concept/design of the leading edge optical components which is applied for these high speed networking equipment, de-fact standardization bodies such as IEEE802.3/OIF have made important roles in the industry. The demands of the bandwidth will still continue to increase with endless degree, and therefore the future migration towards higher speed transceivers such as 1T class, and also towards the smaller form factor transceivers is absolutely inevitable, and to be examined.

The bandwidth of information and communication technology (ICT) systems is increasing and is predicted to reach more than 10 Tb/s. However, an electrical interconnect cannot achieve such bandwidth because of its density limits. To solve this problem, we propose two types of high-density optical fiber wiring for backplanes and circuit boards such as interface boards and switch boards. One type uses routed ribbon fiber in a circuit board because it has the ability to be formed into complex shapes to avoid interfering with the LSI and electrical components on the board. The backplane is required to exhibit high density and flexibility, so the second type uses loose fiber. We developed a 9.6-Tb/s optical interconnect demonstration system using embedded optical modules, optical backplane, and optical connector in a network apparatus chassis. We achieved 25-Gb/s transmission between FPGAs via the optical backplane.

Emerging short-reach data center interconnect is a scenario wherein the capacity has to be maximized over point- to-point optical links without intermediate optical amplification. Most of the transceiver solutions are based on 100G modules with direct detection modulation. Although these legacy solutions are cost-efficient in a short- term, they are not scalable in a long-term, when the capacity x distance product will become more and more stringent. This paper addresses coherent optical solutions for emerging data center interconnect, with optical transmission reach being limited to around unrepeated 100 km. The main advantage of coherent solutions, when compared to legacy direct detection technologies, is the inherently improved spectral efficiency (e.g. 400 Gb/s channels in a 50 GHz grid) and receiver sensitivity provided with high baudrate (>40 GBd) transceiver modules. In this paper, two technological options for single-carrier optical 400 Gb/s modules are exploited for high capacity links over short reach scenarios: 43 GBd polarization-division-multiplexed (PDM)-64QAM, suitable for a 50-GHz grid; and 64 GBd PDM-16QAM, suitable for a 75-GHz grid. These two solutions are compared in terms of capacity allocated in C band (∼4 THz bandwidth), when considering 50 GHz (80 channels at 400G, 32 Tb/s) and 75 GHz (53 channels with 21.2 Tb/s) grids and back-to-back requirements in terms of optoelectronics (digital-to-analog and analog-to-digital converters, modulators, receivers etc.).

Discrete multi-tone (DMT) technology is an attractive modulation technology for short-reach application due to its high spectral efficiency and simple configuration. In this paper, we first explain the features of DMT technology then discuss the impact of fiber dispersion and chirp on the frequency responses of the DMT signal and the importance in the relationship between chirp and the optical transmission band. Next, we explain our experiments of 100-Gb/s DMT transmission of 10 km in the O-band using directly modulated lasers for low-cost application. In an inter-datacenter network of more than several tens of kilometers, fiber dispersion mainly limits system performance. We also discuss our experiment of 100-Gb/s DMT transmission up to 100 km in the C-band without a dispersion compensator by using vestigial sideband spectrum shaping and nonlinear compensation.

Intra-datacenter traffic is growing more than 20% a year. In typical datacenters, many racks/pods including servers are interconnected via multi-tier electrical switches. The electrical switches necessitate power-consuming optical-to- electrical (OE) and electrical-to-optical (EO) conversion, the power consumption of which increases with traffic. To overcome this problem, optical switches that eliminate costly OE and EO conversion and enable low power consumption switching are being investigated. There are two major requirements for the optical switch. First, it must have a high port count to construct reduced tier intra-datacenter networks. Second, switching speed must be short enough that most of the traffic load can be offloaded from electrical switches. Among various optical switches, we focus on those based on arrayed-waveguide gratings (AWGs), since the AWG is a passive device with minimal power consumption. We previously proposed a high-port-count optical switch architecture that utilizes tunable lasers, route-and-combine switches, and wavelength-routing switches comprised of couplers, erbium-doped fiber amplifiers (EDFAs), and AWGs. We employed conventional external cavity lasers whose wavelength-tuning speed was slower than 100 ms. In this paper, we demonstrate a large-scale optical switch that offers fast wavelength routing. We construct a 720×720 optical switch using recently developed lasers whose wavelength-tuning period is below 460 μs. We evaluate the switching time via bit-error-ratio measurements and achieve 470-μs switching time (includes 10-μs guard time to handle EDFA surge). To best of our knowledge, this is the first demonstration of such a large-scale optical switch with practical switching time.

Interconnect architectures based on high-Q silicon photonic microring resonator devices offer a promising solution to address the dramatic increase in datacenter I/O bandwidth demands due to their ability to realize wavelength-division multiplexing (WDM) in a compact and energy efficient manner. However, challenges exist in realizing efficient receivers for these systems due to varying per-channel link budgets, sensitivity requirements, and ring resonance wavelength shifts. This paper reports on adaptive optical receiver design techniques which address these issues and have been demonstrated in two hybrid-integrated prototypes based on microring drop filters and waveguide photodetectors implemented in a 130nm SOI process and high-speed optical front-ends designed in 65nm CMOS. A 10Gb/s powerscalable architecture employs supply voltage scaling of a three inverter-stage transimpedance amplifier (TIA) that is adapted with an eye-monitor control loop to yield the necessary sensitivity for a given channel. As reduction of TIA input-referred noise is more critical at higher data rates, a 25Gb/s design utilizes a large input-stage feedback resistor TIA cascaded with a continuous-time linear equalizer (CTLE) that compensates for the increased input pole. When tested with a waveguide Ge PD with 0.45A/W responsivity, this topology achieves 25Gb/s operation with -8.2dBm sensitivity at a BER=10-12. In order to address microring drop filters sensitivity to fabrication tolerances and thermal variations, efficient wavelength-stabilization control loops are necessary. A peak-power-based monitoring loop which locks the drop filter to the input wavelength, while achieving compatibility with the high-speed TIA offset-correction feedback loop is implemented with a 0.7nm tuning range at 43μW/GHz efficiency.