Bringing the World the Latest in Optical Communications News
ISSUE 37 | Q2 2024
A partnership made in heaven? | p10 AI/ML AND PHOTONICS:
COMMUNICATION GETS READY FOR QUANTUM: Advances in connectivity and security | p20
WHAT’S WRONG WITH THIRD-PARTY COMPONENTS? Could vendor lock-in be history? | p22
EPIC CEO INTERVIEW: Andy Stevens | p28
AI DRIVES 800GBE GROWTH REVOLUTIONARY PHOTONIC CHIP QKD READY FOR DEPLOYMENT
Media Partners to
CONTENTS
4
AI COMES TO PHOTONICS
Industry News
8
G-AI, Datacentres and PICs Michael Lebby
Welcome to the summer edition of Optical Connections. In the run up to ANGA COM, we’ve got a very wide range of topics, from AI in PICs to the advantages of deploying third-party optical components. We kick off however with Michel Lebby’s musings over what the legendary code-breaker and mathematician Alan Turning would think of current developments in computing and communications. Speaking of new developments, veteran journalist John Williamson takes a look at how AI is influencing breakthroughs in PIC design. Another rapidly evolving technology is PON, and Katia Safonova charts the its recent evolution. In addition, regular contributor Antony Savvas explains how QKD is gaining ground in telecoms networks, while George Ashwin extolls the virtues of using third-party components to maximise capital outlay and circumvent vendor lock-in. As more people work from home, remote bank branches are being closed, and the increasing popularity of online sales, rural broadband availability has never been more important. John Tarleton looks at ways of speeding up rural provision of what he argues is an essential public utility. While much of Europe and North America has hitherto relied on coax networks for broadband and television services, Dipl.-Ing. Thomas H. Ritz , explains the very different challenges that cable switch- off presents. Finally, our EPIC CEO Interview this issue is with Andy Stevens , who charts his progress from studying Physics and Applied Optics at the University of Edinburgh, to becoming the CEO at Quantifi Photonics, a New Zealand manufacturer of innovative photonics test and measurement systems. With the latest industry news and products, and a round-up of Optical Connections’ TOP Conference, OFC, and a look forward to ANGA COM, I hope you find the features interesting and thought-provoking.
10 PIC Development John Williamson 14 The Evolution of PON Katia Safonova 20 QKD in Telecoms Antony Savvas 22 Third Party Components George Ashwin 24 Bridging the Rural Divide John Tarleton 26 Cable Switch-Off in Public Networks Thomas H. Ritz 28 EPIC CEO Interview Andy Stevens 30 TOP Conference Wrap 32 OFC Wrap 33 ANGA COM Preview 35 Product Focus
Peter Dykes Contributing Editor
READ ONLINE/SUBSCRIBE: www.opticalconnectionsnews.com FOLLOW US @opconsnews EDITORIAL : editor@opticalconnectionsnews.com ADVERTISING: sales@opticalconnectionsnews.com DESIGN: Antonio Manuel
MEDIA PARTNERS
Optical Connections is published by NEXUS MEDIA EVENTS LTD Suite 3, Building 30, Churchill Square, Kings Hill, West Malling, Kent ME19 4YU t: +44 (0) 1732 752 125
www.opticalconnectionsnews.com
3
ISSUE 37 | Q2 2024
INDUSTRY NEWS
AI drives 800Gbe module growth
Datacom optical component revenue grew for the fourth straight quarter with the explosion of demand for optics used in AI clusters, according to the 4Q23 Optical Components Report from
speed Datacom optical component market is forecast to exceed US$10 billion by 2025. Coherent port shipments increased QoQ, but they were down slightly YoY. 400ZR/ZR+ shipments grew 25% QoQ, and coherent port shipments increased QoQ, but they were down slightly YoY. 400ZR/ZR+ shipments grew 25% QoQ. The report also notes that datacom revenue reached record levels, up 11% YoY, and module shipments grew rapidly in the last two quarters of 2023 as AI demand
accelerated. For all of 2023 however, total Datacom revenue fell by 4%, weighed down by poor performance at the start of the year, but after a sharp drop at the start of 2023, 400GbE port shipments recovered and grew over 50% YoY in 4Q23. Cignal AI also found that Huawei, Infinera, Acacia, and Nokia shipped Gen120P 1.2T high speed coherent ports for revenue in 4Q23, which is the first quarter of production shipments for this new technology. The company also forecast
an increase of 8% for 800GbE modules based on accelerating demand, and initial forecasts of 1600ZR modules were also added to the Optical Components report this quarter. “Datacom shipments, especially 800GbE optics, are ramping up fast and shipped units are forecast to reach eight million in 2024,” said Scott Wilkinson, lead analyst for Optical Components at Cignal AI. “Telecom is slowly recovering from a bottom in Q3, but no immediate reversal is in sight.”
research firm Cignal AI. Nvidia, Coherent, and Innolight lead in
800GbE Datacom module shipments for hyperscale AI applications. Acacia and Marvell lead in shipments of high- performance coherent interfaces based on large volumes of 400ZR pluggables. The high-
PRL, iPronics design ‘revolutionary’ telecoms chip
A team from the Photonics Research Lab (PRL) at the ITEAM in the Universitat Politècnica de València (UPV), together with the company iPronics, has designed and manufactured a revolutionary chip for the telecommunications sector, data centres, and infrastructure associated with artificial intelligence computing systems. The groups say the chip is the world’s first universal, programmable,
for communications, data centres, quantum computing, artificial intelligence, satellites, drones, or autonomous driving, among many other applications. The chip allows on- demand programming and interconnection of wireless and photonic segments of communication networks, avoiding the generation of bottlenecks that can limit both capacity and available bandwidth. This chip is already integrated into an iPronics product, the Smartlight, and is already being used by
Vodafone in testing phase.
efficient management of data flows in data centres and networks of artificial intelligence computing systems. Our next goal is to scale the chip to meet the needs of this market segment,” highlights Daniel Pérez-López, co-founder and CTO of iPronics. “It is the first chip in the world with these characteristics. It can implement the twelve basic functionalities needed in these systems and be programmed on demand, resulting in greater circuit efficiency,” says Capmany.
The development of the chip is the main result of the European project UMWP-Chip, led by researcher José Capmany and funded with an ERC Advanced Grant from the European Research Council. The work has been published in the journal Nature Communications. “For us, the development of this chip represents a very important step because it has allowed the validation of our developments applied to a growing problem, the
and multifunctional photonic chip, and it is particularly useful
4
www.opticalconnectionsnews.com
| ISSUE 37 | Q2 2024
INDUSTRY NEWS
QKD ready for deployment over 300km fibre links
Toshiba Europe Ltd. and Single Quantum B.V. have collaborated to validate and test long-distance deployments of Quantum Key Distribution (QKD) technology. Following extended validation testing of Toshiba’s QKD technology and Single Quantum’s superconducting nanowire single photon detectors (SNSPDs), the companies say they have extended the transmission range for QKD deployment over
fibre connections up to and beyond 300km. The QKD solution can be deployed over fibre networks,
loss of the fibre link). To provide extended QKD transmission, operators typically concatenate fibre links together with trusted nodes along the fibre route which house QKD systems that relay the secret keys. However, the use of multiple trusted nodes may not be practical or desirable along certain networks, such as marine fibre links, cross-border fibre optic links and terrestrial links in locations where suitable buildings cannot
be found, such as in remote areas. To address this challenge, Toshiba and Single Quantum have used Single Quantum’s SNSPDs to detect single photons accurately over higher optical loss fibre links, significantly extending the QKD transmission distance over a single fibre optic link in a compact, stackable and rack mountable solution. Toshiba has been steadily developing a QKD solution since the first trials in 2018.
either coexisting with conventional
data transmissions on deployed ‘lit’ fibres, or on dedicated quantum fibres. Toshiba says achieving longer distance QKD fibre transmission is challenging due to the attenuation of the quantum signals along the fibre length, (the optical
www.opticalconnectionsnews.com
5
ISSUE 37 | Q2 2024
INDUSTRY NEWS
Paratus taps Infinera for 800g Africa-Europe link
Ribbon, SURF successfully trial 800G over 1,650km
African network provider Paratus Group, has deployed Infinera’s GX Series and FlexILS on the provider’s recently completed 1,890-km Paratus express fibre link between Johannesburg and Europe, via Botswana to Swakopmund, where it connects with the Equiano subsea cable from Namibia to Lisbon and on to London and the rest of Europe. It supports services with a latency of 123 milliseconds and wavelengths up to 800G. Paratus’ new superhighway offers network operators an unparalleled opportunity for capacity and
Ribbon Communications has announced that in cooperation with SURF, the collaborative organisation for IT in Dutch education and research, it has successfully achieved 800G over a brownfield 1,650 km fibre optic link connecting research institutes including Nikhef with The Large Hadron Collider located on the CERN campus in Geneva. The trial used Ribbon’s transport solutions, including the Apollo TM800_2, which uses 5nm- 140Gbaud transmission technology to deliver capacity-reach optimised 800G transport; Apollo Open Optical Line Systems, which include hybrid EDFA-Raman amplifiers that maximise the capacity of SURF’s brownfield G655 and G652 fibre, and
data flow, efficient communications, and uninterrupted services. Paratus is the landing partner for the Equiano subsea cable, which offers diverse routing and geographically separated paths. Infinera says that deploying its solutions mitigates possible cable station faults and ensures the network remains intact and fully functional around the clock. “As a steadfast partner on the ground in Africa, Paratus offers unrivalled wholesale capacity solutions for network operators, as exemplified by our advanced technology from Infinera, our infrastructure, and our commitment to offering redundancy,” said Martin Cox, Paratus group chief commercial officer.
have a proven ability to carry third-party vendor wavelengths; and the NPT 2400 metro router, which is interoperable with SURF’s network and delivered 2x400GbE uplinks running EVPN services on top of BGP to 8x100G ports on that network. “We are proud to collaborate with our partner Ribbon in this successful and innovative trial, which pushes the boundaries of our current fibre and shows us what is technically possible with Ribbon’s equipment,” said Harold Teunissen, Director Network & Campus at SURF. “This trial signifies a crucial step forward as we gear up our network to cater to the future needs of scientific research and education in the Netherlands and beyond.”
redundancy where resilience and high-
speed performance are required. This guarantees
Nokia, nbn demo multiple PON speeds
Nokia says nbn Co, Australia’s largest wholesale broadband provider, has successfully completed what it claims is the world’s first live network demonstration of multiple next-generation
the growing demand for faster connectivity needed by premium users, enterprises, and Industry 4.0 for generations to come. Designed to support a full range of PON technologies, from GPON to 100G PON, Nokia’s
PON technologies. For example, nbn was able to show a combination of 10G, 25G and 50G, as well as 10G, 25G and 100G on the same fibre. This trial shows how operators can easily enhance 10G PON to symmetrical 25G PON and eventually evolve to 50G PON or 100G using the same passive and active fibre components. Jaimie Lenderman, Principal Analyst and Research Manager at Omdia, said, “PON has been the technology of choice for broadband operators for years. With the recent, rapid
global expansion of next-gen fibre access networks, the entire telecoms industry is
now taking note. Nokia’s demonstration showcases the strength of the PON technology today and the multiple pathways available to support the applications of tomorrow. The flexibility to choose
PON technologies. Leveraging Nokia’s
Lightspan platform gives operators the
fibre access platform, nbn was able to deliver 10G, 25G, 50G and 100G broadband speeds over its existing fibre network. The successful results demonstrate how operators and wholesalers can scale their fibre access network, addressing
choice and flexibility to optimise their network to their specific business case and needs. Using Nokia’s fibre platform, operators like nbn are able to add capacity to their networks, demonstrating the co- existence of multiple
between 10G, 25G, 50G, and eventually 100G and beyond
empowers operators to customise their networks based upon customer requirements and future service plans for residential, enterprise, and beyond.”
6
www.opticalconnectionsnews.com
| ISSUE 37 | Q2 2024
SPONSORED FEATURE INTEL
Intel shows OCI optical I/O chiplet co-packaged with CPU at OFC2024, enabling massive AI infrastructure scaling Explosive AI infrastructure growth is bringing high bandwidth density and low power performance requirements to compute architectures which optical connectivity is well positioned to support. At OFC2024 in San Diego, Intel demonstrated its advanced Optical Compute Interconnect (OCI) chiplet co-packaged with a prototype of a next-generation Intel CPU running an error-free link, giving the industry a look at the future of high-bandwidth compute interconnect.
A pplications using AI are increasingly being deployed and positioned to drive our global economy and influence the evolution of our society in general. Recent developments in large language models (LLM) and Generative AI have only accelerated that trend. Larger and more efficient Machine Learning (ML) models will play a key role in addressing the emerging requirements of AI acceleration workloads. The need to significantly scale future compute fabrics drives exponential growth in I/O bandwidth and longer reach in connectivity to support larger xPU clusters and architectures with more efficient resource utilization, such as GPU disaggregation and memory pooling. Electrical I/O (i.e., copper connectivity) supports high bandwidth density and low power, but only very short reaches of <1 meter. Pluggable optical transceiver modules used in current data centers and early AI clusters can increase reach but at cost and power levels that are not sustainable with the scaling requirements for AI workloads immediately ahead of us. A co-packaged xPU (CPU, GPU, IPU) optical I/O solution can support higher bandwidths with high power efficiency, low latency, and longer reach, which is what AI/ML infrastructure scaling requires. Intel has developed a 4 Tbps bidirectional fully integrated OCI chiplet based on Intel’s in-house Silicon Photonics technology, to address the AI infrastructure’s tremendous need for bandwidth and to enable future scalability. This OCI chiplet or tile contains a single Silicon Photonics Integrated Circuit (PIC) with integrated lasers and SOAs, an electrical IC, and a path to incorporate a detachable/re-usable optical connector.
and other System-On-a-Chip (SOCs) with high bandwidth demand. This first implementation paves the way toward providing multi-Terabit optical connectivity with a >4x improvement in shoreline density over PCIe Gen6, an energy efficiency of <3pJ/bit, <10ns (+TOF) of latency, and a reach greater than 100 meters. At OFC 2024, Intel demonstrated its first- generation OCI chiplet co-packaged with a concept Intel CPU running an error-free link over fiber (BER <10E-12 with a PRBS31 pattern). This first OCI implementation is a 4 Tbps bidirectional chiplet compatible with PCIe Gen5, supporting 64 lanes of 32 Gbps data in each direction over 10’s of meters, realized as eight fiber pairs each carrying eight DWDM wavelengths. Looking beyond this first device, the platform has line of sight to 32 Tbps chiplets. The single PIC in the current die-stack can support up to 8 Tbps bidirectional applications and it contains a complete optical sub-system, enabled by Intel’s unique capability of integrating DWDM laser arrays and optical amplifiers on the PIC, providing orders of magnitude of higher reliability than conventional InP lasers. These integrated Silicon Photonics chips are manufactured at one of Intel’s high-volume fabs in the US, which has already shipped more than 8 million PICs with over 32 million on-chip lasers embedded in pluggable optical transceivers for data center networking,
with industry-leading reliability. The on- chip laser technology enables true wafer- scale manufacturing, burn-in, and testing, which translates into high subsystem- level simplicity and reliability (e.g., there are no fibers connecting the External Laser Source and PIC) and manufacturing efficiencies. An additional differentiating advantage is that OCI uses standard, widely deployed single-mode fiber (SMF-28), without requiring Polarization Maintaining Fiber (PMF) like other technical approaches in the market. PMF has been rarely deployed, as system vibration and fiber wiggle can negatively affect its performance and associated link budget. Intel’s field-proven Silicon Photonics technology and platform can provide the highest performance and most reliable optical connectivity solutions to make ubiquitous AI possible. Additionally, Intel is well-positioned to offer a complete next-generation compute solution with its leading silicon, optical, packaging, and platform integration capabilities.
For more information: www.intel.com/siliconphotonics
The OCI chiplet can be co-packaged with next-generation CPU, GPU, IPU,
www.opticalconnectionsnews.com
7
ISSUE 37 | Q2 2024
MICHAEL LEBBY PICS
BY G-AI, DATACENTRES AND PICs? I was reminded of Alan Turing on my way to the 2023 European Conference on Optical Communications, writes Dr Michael Lebby , Chief Executive Officer, Lightwave Logic Inc. I visited Manchester U.K., and there I saw Alan Turing’s statue and plaque (pictured above). He was born in England in 1912, died in 1954, and has been credited as one of the most famous mathematicians. His work towards theoretical computer science not only helped address highly complex mathematical problems such as halting, but cracked cipher codes during World War II. WOULD ALAN TURING BE EXCITED
O ne wonders what Alan would have thought of the advancement of neural networks, machine learning and artificial intelligence in general. Perhaps, this pioneer of computer science would be at the forefront of artificial intelligence taking it to another level that we have not thought about. This article questions how Generative Artificial Intelligence (G-AI) has an impact on datacentres, and more specifically, the impact with integrated silicon photonics chips, otherwise known as photonics integrated circuits (PICs). The article also considers a new technology platform that is changing the way we look at silicon photonics and PICs in general – electro-optic polymers. These are organic, polymer materials that are enabling very high-speed optics with extremely low power consumption for optical networking in devices called modulators. Before diving into G-AI, we must consider what G-AI is and how
it fits into the internet or optical networks. G-AI is an electronically- based computing solution. It is an approach to increase computational processing, i.e. allowing semiconductor ICs to process data faster and more efficiently. As an industry, we utilize photonics to send information that is processed by MPUs and GPUs from source to destination using fibre optic cables. These fibre optic cables form the architecture for the internet and optical network. At the highest perspective, electronics does the computational processing, and is expected to continue to do so, while photonics helps convey huge amounts of generated information optically through fibre optic networks. Popular photonic components that form optical links or interconnects consist or lasers, modulators, photodetectors etc. These components are now becoming integrated into PICs chips, typically one PIC chip to send or transmit data, and another PIC chip to receive data. PIC chips are the optical engines for pluggable transceiver modules, which
are utilized in hyperscaler based switches, routers, and server equipment that make up a data centre. GROWING IMPACT OF G-AI IN RECENT YEARS While machine learning and neural networks have been a major focus for computational research over the past 3-4 decades, it has only been in the last ~2 years that most of us have become aware of G-AI. We have been educated to accept that G-AI will drive lots of traffic on the internet, mostly because of the users experimenting and figuring out innovative ways to drive new applications. Even without G-AI, we are driving the need for higher data rates and information from the use of dial-up modems in the 1980s and 1990s, to high speed multi-Gbps downloads today, mostly through our appetite to use video-based traffic. Within the last year it became apparent that G-AI and associated increased computational processing is likely to drive higher traffic levels on the internet.
8
www.opticalconnectionsnews.com
| ISSUE 37 | Q2 2024
MICHAEL LEBBY PICS
The capacity for optical signal information traveling down fibres needs to be increased; and the data centres, where the optical routing and switching takes place, need to deal with higher amounts of information. This places a further constraint for the industry as the increasing traffic also increases power consumption which is a well-known Achilles Heel for the data centre industry.
for the industry. A significant part of the design factor involves optical modulators. An optical modulator in general, switches and modulates light, and there are millions of these devices on the internet or optical network architecture today albeit at slower speeds and high voltage levels. This has led the industry to look for faster optical modulators that can operate at 200G with sub-1V voltage levels so that power consumption is significantly reduced. One of these solutions that is very exciting is the use of electro- optic polymers for polymer-based optical modulators. Polymers offer a heterogeneous solution for a PIC as the material is organic, and can be spun or dropped onto an existing silicon wafer in a standard silicon fabrication plant. Already, polymer modulators have enhanced and ‘turbo-boosted’ silicon photonic PIC platforms with less than 1V drive 200G lanes. Future optical modulator designs are planned to address 400G and 800G lanes. These results and the roadmap to higher speed lanes with polymer modulators exceed incumbent semiconductor technologies being used on the internet today and enhance emerging architectures with increased speed and lowered power consumption . With polymer optical modulator 3dB bandwidths that exceed 100GHz, which have been measured to over 250GHz, electro-optic polymers are positioned very well to enable next generation modulators over the next decade. In addition, polymer modulators with drive voltages below 1V, provide power savings that the hyperscalers are always looking for. ELECTRO-OPTIC POLYMER MODULATOR PICS Lightwave Logic Inc., sources and creates organic materials to create a class of electro-optic polymers called Perkinamine®. The company starts with its own proprietary designed organic chromophores and these are deposited onto a silicon chip with silicon slots to add an optical modulator function. The company’s polymer materials, and the integrated photonic silicon that are applied to them to are very reliable in performance which positions them well for displacing incumbent semiconductor technologies as well as other exotic material technologies being considered for high speed, low power modulators. Polymer modulator devices are
formed on a silicon-based chip roughly a few millimeters on each side where the Perkinamine® chromophore electro-optic polymer is deposited. These chips represent the ‘optical engine’ of a fibre optic transceiver. Electro-optic polymers have an inherent high-performance advantage which allows them to extend out in higher performance over future- generation products. It is as if Mother Nature is working with the technology platform and this creates a roadmap headroom for future generations of hyperscaler equipment. SUMMARY Would Alan Turing be excited with G-AI, data centres and photonics PIC chips today? I believe Alan would have been extremely excited. In an age where folks are designing technology solutions to match computational processing needs, I think Alan would have risen to the challenge and create not just a tool for code breaking ciphers, but more advanced tools to help with the demands of G-AI and growth in electronic computations, communication, and machine learning. From Alan Turing’s perspective, our future could not be brighter; we have grand challenges ahead, and we have the motivation to utilize our tool kits to meet those challenges. One example we discussed is electro-optic polymers – I’m sure there will be others, however, this is one excellent candidate to move the needle forward for all of us.
DATACENTRES AND THE INTERNET
As G-AI is integrating deeper within our daily activities with new applications to make us more efficient, we must look at the consequences as new tools to proliferate and introduce machine learning to the mass population. We are seeing hyperscaler data centres being upgraded today through increased capital expenditures in a fashion that the industry has not seen for a couple of decades. While this feels like the internet ‘bubble’ of 2000, we don’t know if the trend is bubble-like as we are in a growth stage. What we do know from history, is that in 2000, the markets for the internet bubble collapsed from poor growth and business, however, today we are seeing hyperscaler data centre companies already investing in updating their equipment with strong confidence. This effect is expected to become the start of a growth driver for electronic computational processing chips and indirectly, photonics PIC chips for optical interconnect over the next decades. With a photonics perspective, we can now see that data centre operators have ignored the assumed next incremental photonics bandwidth standard of 400 Gigabit per second, and are focusing on 800Gbps, 1.6Tbps (or 1600Gbps) and 3.2Tbps today. Only two years ago, many market analysist covering data centres and more specifically pluggable optical transceivers, were forecasting huge growth in 400Gbps as the main vehicle for client side hyperscaler traffic. Data centres were looking at 4 channel 100G lanes (for an aggregate of 400Gbps) and 8 channel 50G lanes (also for an aggregate of 400Gbps) as solutions to support 400Gbps traffic. Today, this has changed substantially: market forecasts for 400G are flat if not declining, and today’s focus is on 800G using 200G lanes, with 400G and 800G lanes expected soon. Creating a 4 channel 200G lane optical transceiver is a challenge
Dr Michael Lebby, Chief Executive Officer, Lightwave Logic Inc.
www.opticalconnectionsnews.com
9
ISSUE 37 | Q2 2024
JOHN WILLIAMSON AI/ML AND PHOTONICS
A PARTNERSHIP MADE IN HEAVEN? The meshing of Artificial Intelligence/Machine Learning (AI/ML) and Photonic Integrated Circuits (PICs) looks to be a productive courtship currently being made in technology heaven, writes John Williamson . Accordingly, a recent analysis from Coherent Market Insights (CMI), valued the 2022 global silicon photonics market at US$1,584.8 million growing to US$8,317.9 million by 2030. CMI believes two of the main drivers of this are the increasing adoption of SiPho in data centres and the not-unrelated integration of SiPho with AI and ML. AI/ML AND PHOTONICS:
A I/ML has a ravenous resources, and lower latency – all of which PICs could help furnish in spades. “The all-to-all connectivity requirements (of) AI workloads demand higher bandwidth density and lower cost than optics used in traditional front-end compute networks”, instances Manish Mehta, VP of Marketing and Operations, Optical Systems Division at Broadcom Inc. Christian Urricariet, senior director of Product Marketing, Silicon Photonics Product Division, Intel Corp, acknowledges that electrical I/O (i.e., copper trace connectivity) can support very high bandwidth density with low power, but only for very short distances of about 1 metre, which makes it increasingly inadequate for advanced AI/ML applications. He also allows that pluggable optical transceiver modules used in requirement for increased speed and bandwidth, higher performance computational current data centres and early AI clusters can increase the maximum reach but at cost, power density, and latency levels that cannot support the scaling requirements of the emerging AI/ML infrastructure. “Significantly scaling AI/ML network and compute infrastructure will drive exponential growth in I/O bandwidth
ARTIFICIAL SWEETENERS In reverse, AI/ML can aid in the
with more efficient resource utilisation such as xPU disaggregation and memory pooling,” remarks Urricariet. POWER POINTS Reducing AI/ML PIC power consumption is a big deal. “One of the challenges for building large AI clusters is power consumption,” states Dr. Vladimir Kozlov, founder and CEO of market intelligence house LightCounting. “Optics doesn’t consume as much.” In this context, Mehta adds that his company’s SiPho PICs, integrated with CMOS drivers and transimpedance amplifiers, and co-packaged on a common substrate with a core ASIC – for example switch ASIC or xPU - can deliver up to 70% power consumption savings versus traditional optical interconnects. Additionally, says Mehta, the solution can offer 6.4 Tbps in just two-times the silicon area of the company’s 400 Gbps PIC - an eight-times improvement in silicon area efficiency. Meantime, Intel’s first Optical Compute Interconnect (OCI) chiplet recently demonstrated, is a 4 Tbps bidirectional device, supporting 64 lanes of 32 Gbps data in each direction over 10s of metres. This is realised as eight fibre pairs each carrying eight DWDM wavelengths, and it includes the InP lasers heterogeneously integrated on the PIC. Its target energy efficiency is <3 pJ/bit, compared to current pluggable transceivers at about 12-14 pJ/bit.
germination of new PIC solutions. “The main way is laying out and designing components and circuits,” offers Dr. Adam Carter, CEO of Photonic Application-Specific Integrated Circuits (PASIC) chip producer OpenLight Photonics. Vikas Gupta, senior director of Product Management at GlobalFoundries, notes that while there has been significant progress in the ability to simulate photonic integrated chips, the available Electronic Design Automation (EDA) software for photonic simulations is comparatively immature when compared to EDA software available to simulate electronic integrated chips. “Photonic simulations, just like electronic simulations, tend to be multi- physics problems - electronic, thermal, mechanical - with the added complexity of photonic physics,” says Gupta, expanding on the theme. “AI/ML can change this dynamic by making these multi-physics problems computationally more efficient and accurate.” AI could also be relevant to solving some of the well-known shortcomings and complexities of the PIC industry supply chain. ”AI could help because it could help you to model your supply chain better,” reasons Carter. “It can model component shortages and things like that.”
density and more efficient power management, coupled with longer
reaches in connectivity to support larger CPU/GPU/TPU clusters, and architectures
10
www.opticalconnectionsnews.com
| ISSUE 37 | Q2 2024
www.opticalconnectionsnews.com
11
ISSUE 37 | Q2 2024
JOHN WILLIAMSON AI/ML AND PHOTONICS
UNEQUAL ATTRACTION? At this point in time, though, the
pace of transition to higher bandwidths. “New data centres supporting AI/ ML are going to be utilising a lot higher bandwidth as we go forward,” emphasises Carter.” And, the transition between the different speeds – going from 800G to 1.6T and to 3.2T - is going to be a lot faster than we’ve traditionally seen.” Gupta has an arresting metric here. “Generative Large Language Models are expected to grow logarithmically in the number of parameters, adding to the need for more memory and faster, low latency interconnects,” he comments. “For example, Generative Pre-trained Transformer (GPT) 4 will potentially have trillions of parameters as opposed to its predecessor GPT3’s 175 billion.” The responsiveness and co-ordination of the optical component industry may also be due an overhaul. “AI is going to really drive high bandwidth, high density optical interconnects. In the networks that are being deployed to do AI it’s a big challenge right now with the way that the component industry does things today,” argues Carter. “Things will have to change.” Greater degrees of component integration and optics co-packaging is on the PIC front burner to further reduce power consumption and minimise losses from the use of separate packages. “PICs are currently used mostly in traditional optical transceivers and installed in fabric switches such as Ethernet and Infiniband and so on. Most deployed PICs are 4-channel or 8-channel,” points out Mehta. “Progression to higher levels of integration, coupled with manufacturing in the mature silicon foundry and Outsourced Semiconductor Assembly And Test (OSAT) ecosystem is critical to enable optical interconnects to begin to compete with copper links on power and cost.” Mehta also contends that silicon photonics PICs offer the most effective pathway to high density integration of optical devices in silicon using the widely used silicon foundry and OSAT advanced packaging. Kozlov concurs. “A big trend that we
see is the transition from pluggable optical devices to co-packaged optics” he reports. “I think that in the future it will all converge on integrated solutions, and probably silicon photonics will be the dominant platform – an integrated platform for a variety of materials.” “In the near future, photonics technologies will be used to develop optical interposers – think optical network on a “chip” to enhance GPU networks and memory disaggregation,” predicts Gupta. “The next usage would be photonic compute chips.” For its part Intel is investing heavily in next-generation photonics wafer process technology, which targets >15% power reduction and >40% die area reduction with respect to its current devices. Urricariet discloses that, inter alia, it is also adding performance enhancements such as 200G/lane with high yield, increased laser power, improved optical coupling efficiency, and improved modulator figure of merit. COMMERCIAL BRAKES? The greater commercialisation of PICs has not been without its complications. Increasing integration has meant PIC designers approaching foundries for custom process technology and packaging solutions. Gupta suggests custom solutions are not commercially interesting for high volume foundries and there has been the lack of a common optical test approach. “The ‘customisation’ has been the biggest challenge on the commercialisation of PICs due to the hesitancy of large-scale foundries, OSAT and test vendors to engage in a fragmented market” he maintains. Reliability is another potential obstacle to much wider PIC deployments. “In addition to meeting the aggressive performance requirements in bandwidth density, energy efficiency, latency, and cost, emerging optical I/O solutions also need to offer very high levels of reliability, especially for applications that are not pluggable or easily replaceable in the field,” asserts Urricariet.
reciprocal attractions of AI/ML and PICs may not be equal. While acknowledging that a number of players are breaking new ground and using AI/ML to produce advanced optical element designs that may be counter-intuitive to the human mind, Kozlov judges that it’s early days yet for the commercialisation of such innovations. “I think at the moment it’s probably more that PICs are helping AI and ML systems to become more powerful systems than vice-versa,” he says. “Optics is aiding AI probably more than vice- versa.”
PIC ‘N’ MIX Different PIC platforms have their
particular functional, operational and application features. Although a number of others, including Lithium Niobate and Gallium compounds, are also in the mix, three of the most prominent platforms are SiPho, Indium Phosphide (InP) and Silicon Nitride (SiN). Among other characteristics, InP scores in the realisation of active optical components, SIN has a wide spectral reach and SiPho can leverage the scale economies of the huge global CMOS semiconductor industry. There seems to be some agreement that some of the various platforms complement, rather than compete, with each other. “SiPho needs InP because silicon doesn’t generate light,” observes Kozlov. “There’s always an InP laser next to a SiPho chip. Or even right on a SiPho chip.” “In our case they complement. We need the InP to do the gain medium inside the PIC in silicon,” agrees Carter. “There are companies doing InP circuits, but I think the main challenge there is going to be cost. AI networks in particular, with the volumes that are shipping, are going to be very cost- sensitive.” PIC-TURE THIS What’s up next on the PIC horizon as it relates to AI/ML? Increases in speed look to be required, both in the bandwidth capabilities of photonic elements and the
Manish Mehta VP, Marketing and Operations, Optical Systems Division, Broadcom Incn
Christian Urricariet Senior Director, Product Marketing, Silicon Photonics Product Division, Intel Corp
Dr. Vladimir Kozlov Founder and CEO, LightCounting
Dr. Adam Carter CEO, OpenLight Photonics
Vikas Gupta Senior Director, Product Management, GlobalFoundries
12
www.opticalconnectionsnews.com
| ISSUE 37 | Q2 2024
SPONSORED FEATURE COMMSCOPE
CommScope Supports Gigaclear Mission to Bring Rural Britain up to Speed When it comes to enjoying the benefits of broadband in the U.K., urban residents often fare better than those in more rural locations. Not only are more urban areas connected to full fiber broadband, but they are also likely to be served by multiple providers. This is usually not the case in rural communities, because rural villages and towns are often more difficult to connect—and therefore more costly for providers.
In 2010, Gigaclear Ltd., a provider of rural broadband connectivity, set about narrowing this digital divide. Its mission was simple: to take its unique network to rural, often hard-to-reach communities that might otherwise be left behind. With a commitment to using existing infrastructure such as poles and ducting wherever possible, Gigaclear set about extending its network. At, Gigaclear celebrated reaching two milestones: 500,000 homes and businesses ready for service (RFS) across 26 counties, and achieving its 100,000th customer. Currently, the company has set its sights on reaching one million premises by the end of 2027. Helping them on this journey to ensure rural communities can benefit from the many opportunities that fast, reliable broadband brings is CommScope, which is providing Gigaclear with a number of network solutions. EXTENDING ULTRA-FAST ACCESS TO MORE PLACES Like any large-scale deployment, Gigaclear’s rural full fiber rollout poses several challenges, not least of which are geographical ones. Because each environment served is different, the company would need to utilise a variety of network architectures, such as support for physical infrastructure access (PIA) and microducts. The solution would also need to be fully compliant with CP08 PIA rules and other regulations. Gigaclear is seeking to accelerate its deployment by minimising design, build and installation times. Since the number of skilled engineers in the U.K. is limited, the solution would also need to be relatively simple to configure and deploy. Gigaclear is committed to serving
its customers over the long term; so, to future-proof the deployment, its infrastructure would require the ability to scale and expand as needs evolved. A VERSATILE PORTFOLIO THAT’S FAST AND EASY TO OWN After evaluating a variety of options, Gigaclear determined that CommScope best met its needs—and its customers’ expectations. Offering a broad portfolio of products, CommScope is capable of providing solutions to support a wide variety of build topologies. The two industry leaders have worked closely together for more than five years to meet the specialised needs of a distinctive marketplace—and proactively plan for future needs. “Gigaclear and CommScope have forged a strong partnership over the course of several years,” said Jason Prince, lead fibre engineer for the Infrastructure Engineering Group at Gigaclear. “CommScope’s selection as our preferred partner speaks volumes about their unwavering commitment to delivering high-quality products. However, what truly distinguishes them is their proactive approach in listening to our unique needs and consistently pushing the boundaries of innovation to keep pace with our ever- evolving demands.” SURPASSING MAJOR MILESTONES WITH SCALABLE, FLEXIBLE ARCHITECTURES CommScope has a large portfolio of products covering all build types, working closely with clients to understand their build challenges and take a consultative approach when advising. As part of the deployment initiative, CommScope worked closely to provide Gigaclear with the solutions that fit better for each demography area. NOVUX® single fiber HST hardened terminals are an essential component for delivering dependable high performance. They are designed to adapt to Gigaclear’s specific application—sharing connection capabilities, a unique product information system, a mounting platform and customer configurability. They require fewer splices than aerial nodes, making them less costly and simpler. Built for harsh environments, these hardened terminals cannot be opened— helping to prevent accidental break of
fibers or other damage. The ability to scale up homes passed and homes connected over time was another key requirement. With Gigaclear’s growing network of half a million homes and business ready for service (RFS), flexible components like fiber splice closures and FACT fiber-optic panels and modules simplify scalability. This positions Gigaclear to continue growing the network. CommScope has proactively identified improvements in manufacturing locations throughout the world, offering Gigaclear better lead times and a reduced environmental impact. CommScope’s supply chain can also help compensate for inconsistent availability of materials, a valuable advantage for Gigaclear during their peak deployment periods. BOOSTING BUSINESS AGILITY To accelerate installation and build times, Gigaclear can choose from a variety of precabled/ solutions and innovative designs. These preconfigured solutions enable the provider to sidestep many cumbersome field installation processes and get more deployments done, faster. Building the solutions in advance also helps minimise installation errors, as connectorization is performed at CommScope factories in a controlled environment. These innovative products are designed specifically with engineers in mind, decreasing install time and saving costs. A CONTINUING RELATIONSHIP BASED ON INNOVATION Thanks in part to the vast array of options provided by CommScope, Gigaclear is confident it can extend its network to even more rural communities and achieve its target of one million homes passed in 2027. Both companies consider their partnership a springboard for growth, and CommScope is currently working with Gigaclear’s planning and engineering teams to identify new products that will help bring additional efficiencies to Gigaclear’s rural network rollout in the future.
Author: Jarrett Roumania FTTX Account Manager – UK & Ireland
www.opticalconnectionsnews.com
13
ISSUE 37 | Q2 2024
KATIA SAFONOVA PASSIVE OPTICAL NETWORKS
THE EVOLUTION OF PON TECHNOLOGIES Driven by more remote work, the demand for higher bandwidth is increasing. This applies to download speeds as well as upload speeds – the latter applies especially to residential areas because remote workers today need to upload larger amounts of data from their home offices than before the pandemic. An enormous challenge for network operators, who need to expand their networks faster and more efficiently to meet the demand. With standardised and interoperable components, installation is simplified, maintenance costs reduced and upgrades can be made quicker and easier. This leaves operators with a choice to make: Do they want their networks to be fibre rich or fibre lean? Do they want to install more optical fibres in the ground or utilise and maximise the ones they already have by multiplexing the existing fibres, asks Katia Safonova , Market Development Manager, Corning Optical Communications.
P assive optical networks (PON) are also evolving to satisfy the need for higher speeds and lower latency of the end users. But where are we at right now? And how can the evolution of PON help network providers to connect more homes faster and more cost-effective than before? The decision around which PON to use will have a direct effect on the revenue of the providers as users decide which contract to sign based on the speed provided for them. With the EU’s push towards the Gigabit Society targeting 100 Mbps for all European households by 2025 the previous PON
aren’t providing the necessary speed, making additional investments all the more pressing. THE MOVE PAST 10G TO 50G+ The growing demand for higher bandwidth connections has driven a need for continuous evolution in PON technologies. From 10G to 50G+ there have been many improvements over the last years. Let’s take a look at the existing PON technologies and where they are headed. GPON is an ITU domain with a single upstream and downstream wavelength, while the 10G-PON comes in the forms of XGPON-1 and XGS-PON, both of which are ITU standard and
have single upstream and downstream wavelengths. They are working with symmetrical or asymmetrical line rates and due to a lower upstream rate can use more cost-effective Optical Network Units (ONUs). XGS-PON works on the same wavelengths as XG-PON and 10 Gbps EPON. The 10G-PON has surpassed GPON in ports sold. This leads to commercial launches of 2-to-5Gig residential services in 2022. The forecast for XGS-PON deployments continues to ramp through 2026, as a growing number of operators around the world are moving forward with XGS-PON as their next-generation PON technology.
Figure 1: Understanding available spectrum for PON coexistence. (Image: Corning Optical Communications)
14
www.opticalconnectionsnews.com
| ISSUE 37 | Q2 2024
KATIA SAFONOVA PASSIVE OPTICAL NETWORKS
One of the reasons for the increased use of XGS-PON is the need for more upstream speed. With more people working remotely and the increased use of technologies like cloud computing and artificial intelligence more people are uploading more massive amounts of data than ever before. The previous GPON with an upstream of 1.2 GB/s can’t keep up with the demand. Instead, symmetrical PON with a higher upstream is the most sought after technology making every second new installed PON an XGS-PON in 2023. Additionally, operators tend to use the more recent XGS-PON
GPON + XGS-PON: External or MultiPON Module (MPM) with internal CEX. (Image: Corning Optical Communications)
because it is possible to have a split ratio of 1:128 without sacrificing the quality of the service, making it more cost-effective and resulting in a much higher ROI. GPON offers in direct comparison only a split rate of 1:64. In order to achieve the higher 1:128 XGS-PON split ratio an additional splitter is included before the OSP Cross Connect as shown in Figure 1. This example by a European operator shows how they combine GPON and XGS- PON to connect more people with high upstream and downstream rates without the need for more optical fibres. Another technology, the 10G-EPON (Ethernet PON), is an IEEE standard that uses the same wavelengths as XGS-PON. China has embraced Ethernet PON, as well as the North American CATV industry, due to its backward compatibility with DOCSIS via DPOE. Additionally, the 25G-PON is not an IEEE or ITU standard, but rather a MSA formed in 2020 and led by Nokia. With over 20 members now, the technology leverages the need for enterprise and mobility requirements based on mature 25G optics technology. WILL WE BE GOING DIRECTLY TO 50G? With the move past 10G to 50G+, several technologies have emerged, including 25G-EPON evolving from 10GEPON, 50G-PON, which requires expensive dispersion compensation with integrated Digital Signal Processors (DSPs) or 50G-EPON which doesn’t require DSP for dispersion compensation. Prior to 50Gbps-PON becoming available, 25Gbps-PON will achieve market traction between 2022 and 2026, primarily in enterprise and 5G transport applications. This technology will allow service providers to deliver high bandwidth and low latency for cloud services and applications. However, beyond 2026, the future of 25G-PON remains in question due to the split among vendors and standards bodies between 25G-PON and 50G-PON. The timeline for 50G-PON component and equipment availability is uncertain, with general availability likely to occur in 2026. Until then, service providers
will continue to rely on XGS-PON for their large-scale residential PON deployments and 25G-PON for their 5G transport and enterprise networks. As the deployment of 5G networks increases, service providers are looking for efficient ways to transport traffic from small cell units to the core network. 25Gbps-PON is expected to be a suitable technology for this requirement, especially for operators who have experience using standards for transport in existing LTE networks. With the high bandwidth and low latency that 25Gbps-PON can provide, service providers will be able to address the growing demand from large enterprises that rely on cloud services and applications. It is expected that cable operators will also partially rely on 25 Gbps EPON for the delivery of enterprise services, as well as the aggregation of their remote PHY and remote MACPHY nodes, especially those that are being used to deliver DOCSIS 4.0 services and bandwidth. However, and as mentioned before, the future of 25Gbps- PON beyond 2026 is uncertain due to a split among vendors and standards bodies between 25Gbps-PON and 50Gbps- PON. There is an industry consensus that 50Gbps-PON equipment could possibly be ready by 2024 or 2025. Still, general availability is more likely in 2026. Until then, service providers will need to continue using XGS-PON for large-scale residential PON deployments and 25Gbps- PON for their 5G transport and enterprise networks.
XGS-PON can use their own fibres in the same sheath. To achieve coexistence, an External or MultiPON Module (MPM) with an internal CEX can be used. This approach allows for a smooth transition to XGS-PON without disrupting the existing GPON service, which can continue to operate until it is phased out over time. Overall, the smooth migration to XGS-PON is crucial for service providers looking to upgrade their networks while minimising disruption and maximising efficiency. WHAT’S NEXT AND IN THE FUTURE? In conclusion, the deployment of PON technologies is rapidly evolving to meet the growing demand for high-speed broadband services. XGS-PON is currently the preferred technology for most operators, with trials already moving to production deployments. However, 10G-PON technologies are queued up to overtake GPON as the dominant PON solution, while the 50G-PON is expected to be the next major ITU PON standard. The future PON technologies will gain momentum after 2024, with 25G-PON expected to gain traction as operators seek future-proof solutions that are backward compatible with existing PON deployments. Ultimately, 50G-PON is expected to dominate in the long term due to its ability to deliver unprecedented bandwidth, and Chinese telcos’ adoption of the technology is expected to decrease equipment prices. While 100G-PON lab tests have started in 2022, the technology is not expected to see widespread adoption until after 2027. Overall, the PON industry is continuously innovating to meet the ever-increasing demand for faster and more reliable broadband services. This leaves providers in need to invest in PON today to ensure the needed speeds for the Gigabit Society and because users are deciding on their contracts based on factors like upstream and downstream speeds, low latency and the associated costs. And they will always decide in favour of the most attractive combination of price and service level provided.
NETWORK SCENARIOS: SMOOTH MIGRATION
Because of this, they have to make sure that the migration to XGS-PON is smooth. It can be achieved through two ways of deployment. The first is through Greenfield deployment, where XGS OLT’s are deployed on day one without any changes to the existing OCS business. This deployment scenario is most suited for competitive carriers (OTTC). The second deployment scenario involves using the existing ODN with spare feeder fibres. In this scenario, both GPON and
www.opticalconnectionsnews.com
15
ISSUE 37 | Q2 2024
Page 1 Page 2 Page 3 Page 4 Page 5 Page 6 Page 7 Page 8 Page 9 Page 10 Page 11 Page 12 Page 13 Page 14 Page 15 Page 16 Page 17 Page 18 Page 19 Page 20 Page 21 Page 22 Page 23 Page 24 Page 25 Page 26 Page 27 Page 28 Page 29 Page 30 Page 31 Page 32 Page 33 Page 34 Page 35 Page 36Made with FlippingBook interactive PDF creator