Monday, October 2, 2023

Broadcom debuts 200G/lane PAM-4 DSP PHY

Broadcom announced shipping of its 5nm 200G/lane optical PAM-4 DSP PHY featuring 200G/lane serial optical interfaces, which enable 800G and 1.6T pluggable optical transceiver modules.

The new Sian BCM85822 is optimized for the 1.6T OSFP-XD transceiver module design and would effectively enable 51.2T switching capacity in a 1RU rack to improve bandwidth density in hyperscale data centers. Further, the adoption of 200G/lane optical interfaces lays the foundation for the eventual deployment of 1.6T and 3.2T solutions with 200G/lane electrical interfaces, which are needed for cloud providers to support next generation switches and scale AI workloads.


  • Monolithic 5nm 800G (8:4) PAM-4 DSP with integrated laser driver
  • Delivers best-in-class module performance in BER and power consumption
  • Proven interoperability with Broadcom’s 200G EML
  • Compliant to all IEEE and OIF standards, capable of supporting MR links on the chip to module electrical interface
  • Supports optical modules from 800G to 1.6T
  • Supports IEEE compliant (128,120) Hamming Inner Code

“This latest generation of DSP operating at 200G/lane extends Broadcom’s leadership in high-bandwidth PHY and enables the lowest power optical modules in the industry,” said Vijay Janapaty, vice president and general manager of the Physical Layer Products Division at Broadcom. “This, in turn, lays the foundation for next generation switching platforms for AI/ML clusters and networks.”

#ECOC23: DustPhotonics unveils 800G Silicon Photonics chip

DustPhotonics introduced a single-chip 800G PIC (Photonic Integrated Circuit) for DR8 and DR8+ applications, providing 8 optical channels independently modulated at 100Gb/s for an aggregate bandwidth of 800Gb/s. 

The chip is designed into a compact 7.5mm x 7mm package, enabling it to be used in industry standard QSFP and OSFP style form factors.  The PIC includes on-chip lasers, incorporating DustPhotonics' patented L3C (low-loss laser coupling) technology, whereby off-the-shelf lasers from a variety of different manufacturers can be integrated with the PIC.

The device is suitable for reaches up to 2km in applications including hyperscale data centers and AI and Machine-Learning clusters.

At ECOC,, DustPhotonics is demonstrating this device in multiple configurations including a traditional 800GBASE-DR8 application, an immersion cooling application, and a reduced reach application. The immersion cooling demo showcases how this chip is suited to be immersed in a liquid coolant, since there is no free-space interface between the laser and PIC or at the fiber attach interface at the optical output of the chip.

For the reduced reach application, DustPhotonics is demonstrating a second, cost optimized version of the product suitable for transceivers or AOCs (Active Optical Cables) up to 100m, or for LPO (linear-drive pluggable optic) applications.

"We are seeing a lot of customer traction for this 800G application, and we are excited to showcase the broad range of applications that we support," said Ronnen Lovinger, CEO of DustPhotonics. "We are well-positioned for the next phase of the company which is to scale into high-volume manufacturing."

The device is sampling to customers today and is expected to be in production by the first quarter of 2024.

#ECOC23: Avicena demos microLED transceiver in 16nm finFET CMOS

At this week's ECOC in Glasgow, Avicena is demonstrating its LightBundle multi-Tbps chip-to-chip interconnect technology.

Currently, high-bandwidth memory (HBM) modules must be co-packaged with GPUs because the GPU-memory electrical interconnect is limited to just a few millimeters in length. Conventional optical interconnects based on VCSELs or Silicon Photonics (SiPh) promise to extend the interconnect reach. However, they are challenged by the power, bandwidth density, latency, and cost requirements. 

Avicena says its microLED-based LightBundle interconnects provide much lower power and latency, much higher bandwidth density, and can achieve very low costs.

The LightBundle interconnect architecture is based on arrays of GaN microLEDs that leverage the microLED display ecosystem and can be integrated directly onto high performance CMOS ICs. Each microLED array is connected via a multi-core fiber cable to a matching array of CMOS-compatible PDs.

“As generative AI continues to evolve, the role of high bandwidth-density, low-power and low latency interconnects between xPUs and HBM modules cannot be overstated”, says Chris Pfistner, VP Sales & Marketing of Avicena. “Avicena’s innovative LightBundle interconnects have the potential to fundamentally change the way processors connect to each other and to memory because their inherent parallelism is well-matched to the internal wide and slow bus architecture within ICs. With a roadmap to multi-terabit per second capacity and sub-pJ/bit efficiency these interconnects are poised to enable the next era of AI innovation, paving the way for even more capable models and a wide range of AI applications that will shape the future.”

“We have previously demonstrated microLEDs transmitting at > 10Gbps per lane and a test ASIC in a 130nm CMOS process running 32 lanes at less than 1pJ/bit,” says Bardia Pezeshki, founder and CEO of Avicena. “Now we are bringing up our first ASIC in a 16nm finFET process with over 300 lanes and an aggregate bandwidth of over 1Tbps bi-directional at 4Gbps per lane. The ASIC measures less than 12mm2 and contains the circuitry for the optical Tx and Rx arrays, as well as a high-speed parallel electrical interface and various DFT/DFM functions like BERT, loopbacks, and Open Eye Monitoring (OEM). All key ASIC functionality has been verified and we are currently working on yield improvements for manufacturing scalability.”

Avicena plans to advance its LightBundle platform to enable interconnects with high-bandwidth density of multi-Tbps per mm2 in advanced CMOS process nodes. The low power, high density, and low latency of LightBundle is well matched to chiplet interfaces like UCIe, OpenHBI, and BoW, and can also be used to enhance system architectures that are limited by the reach of existing compute interconnects like PCIe/CXL, and HBM/DDR/GDDR memory links.

#ECOC23: Lumentum intros InP-Based 130+ GBaud Smart TROSA, 800G ZR

At this week's ECOC in Glasgow, Lumentum is showcasing its latest solutions, including:

  • 800G ZR+ and 0dBm 400G ZR+ Transceivers: Lumentum 800G ZR+ and 0dBm 400G ZR+ compact transceivers, which are crucial components for connecting data centers, enable IP over DWDM by directly plugging into switches and routers. These transceivers, available in QSFP-DD and OSFP form factors and utilizing Lumentum's new 130+ GBaud Smart TROSA, offer high output power and extended reach for metro and regional networks. Sampling of the transceivers will begin in December 2023, with official release expected in calendar year 2024.

  • InP-based 130+ GBaud Smart Transmitter Receiver Optical Sub-Assembly (TROSA): This advanced optical component enables data rates of up to 800Gbps on a single wavelength, allowing for a seamless transition to faster network architectures that can support increasing data traffic from emerging AI and ML applications. Lumentum will showcase this innovative technology within the Lumentum 800G ZR+ demo at their stand. This product is available for sampling now, complementing the nano-iTLA, CDM and ICR optical components that are already available and are currently being operated by multiple customers at up to 150 Gbaud.
  • M11 Series, Uncooled 980 nm Pump Modules: These modules are compact, have low power consumption, comply with Telcordia GR-468-CORE standards, and offer wavelength selection for optimal spectrum control with high power output. They are well-suited for small form factor and pluggable Erbium-Doped Fiber Amplifiers (EDFA), high bitrate and high channel-count EDFAs, and CATV distribution. The M11 series pump modules are now available for purchase.
  • CPO/ELS Technology: Lumentum's ultra-high power, 1310 nm distributed-feedback laser (DFB) will be demonstrated at over 400 mW optical power ex-fiber at 25°C. These new ultra-high power 13xx lasers will enable higher bandwidth for AI and ML applications by using co-packaged optics and external laser source solutions, as well as silicon photonics transceivers for the next generation of data centers.
  • OIF 400ZR+ Interoperability and CMIS demonstrations: At the OIF booth, Lumentum is participating in OIF's 400ZR Interoperability demonstration through the integration of its 400ZR+ QSFP-DD and OSFP transceivers. These transceivers enable high-speed data transmission over long distances. The demonstration aims to show how optical module manufacturers are reducing power consumption and complexity in data center interconnects, while promoting interoperability. 

"As the technology landscape evolves at a breakneck pace and the relentless need for data-intensive applications persists, we’ve designed our products to support the increased level of efficiency and performance that’s essential in this new era," emphasized Wupen Yuen, President, Cloud and Networking at Lumentum. “At ECOC, we’re showcasing cutting-edge innovations that draw upon our decades of photonic expertise, with solutions that empower our customers to stay at the forefront of this dynamic environment."

In addition to the solutions on display, Brian Smith, Lumentum Senior Principal Engineer, will deliver a Market Focus presentation on October 4 from 13:20 – 13:35 entitled “The 6th Generation of Fiber Optic Communications: Carrier and Spatial Division Multiplexing (CSDM).” Smith will provide his insights on the next era of optical networking, which Lumentum introduced at its LITE 2023 Investor Event in March of this year. 

#ECOC23: Semtech and Broadcom demo 200G/lane optical transceiver

Semtech is showcasing a 200G/lane optical transceiver featuring its latest FiberEdge 200G PAM4 PMDs and Broadcom’s latest generation 5nm 112GBd PAM4 DSP, EML and PD devices. These next generation of optical modules capable of bandwidths up to 1.6TB will drive the adoption of 51.2TB next generation switching platforms.

“The demonstration of a 200G/lane ready-to-adopt optical module validates the 200G/lane technology, of which Semtech has been an early enabler, confirming Semtech’s strong commitment to advanced technology and innovation,” said Nicola Bramante, senior product line manager for Semtech’s Signal Integrity Products Group. “The collaboration with Broadcom as active contributor to the 200G/lane ecosystem on both the DSP and the optics side has been instrumental to this important milestone towards the implementation of 1.6T and beyond, and ultimately towards the deployment of next generation 51.2T and a102.4T switches.”

“Broadcom is pleased to collaborate with Semtech on the development of the next generation of optical modules based on 200G/lane optics that use Broadcom's DSP along with Semtech's laser driver and TIA," said Khushrow Machhi, Senior Director, Marketing for Broadcom's Physical Layer Products Division.

#ECOC23: Semtech demos 100G/Lane Linear Pluggable

Semtech is demonstrating 100G/lane linear pluggable optical links featuring its latest FiberEdge 100G PAM4 PMDs.

The demo includes:

  • 800G SR8 LPO (Linear Pluggable Optics) module, featuring Semtech’s FiberEdge GN1848 quad VCSEL driver and GN1816 quad TIA, interoperating with latest switch ASIC technology, in Semtech’s booth #517
  • 800G SR8 LPO (Linear Pluggable Optics) module, featuring Semtech’s FiberEdge GN1848 quad VCSEL driver and GN1816 quad TIA, interoperating with latest switch ASIC technology and other LPO modules from multi-vendor ecosystem, in OIF’s booth #304
  • 800G SR8 DSP-based module, featuring Semtech’s FiberEdge GN1848 quad VCSEL driver and GN1816 quad TIA, interoperating with latest switch ASIC technology, in Semtech’s booth #517
  • Results from 800G 2xDR4 LPO module, tested offline with latest switch ASIC technology

“Semtech’s FiberEdge PMD portfolio includes high linearity devices that enable the linear compensation needed from the PMDs in a DSP-less datacenter environment,” said Nicola Bramante, senior product line manager for Semtech’s Signal Integrity Products Group. “The Linear Drive architecture demands close handshaking between PMDs and switch ASIC to ensure the success of LPO. This demonstration proves such interoperability for both multi-mode and single-mode datacenter applications.”

#ECOC23: Semtech and Coherent demo 200G per Lane driver with 200G laser

Semtech is showcasing its latest FiberEdge 200G PAM4 MZM driver driving a 200G DFB-MZ laser from Coherent Corp.

The joint demonstration at ECOC 2023 will include a differentially driven DFB-MZ InP laser from Coherent and Semtech’s PAM4 driver, demonstrating a clear path toward 1.6T and 3.2T optical transceiver deployments in data centers.

“We are very excited to continue our collaboration with Coherent as advanced lasers and optics get developed in the 200G per lane space,” said Nicola Bramante, senior product line manager for Semtech’s Signal Integrity Products Group. 

“This demonstration confirms once again the interoperability of state-of-the-art 200G lasers with Semtech’s latest 200G per lane FiberEdge developments, ensuring superior linearity and signal integrity in 200G per lane Mach-Zehnder modulator-based applications, both in indium phosphide (InP) and silicon photonics. This enables rapid and scalable deployment of next-generation high-performing and reliable switches in data centers.”

“Coherent continues to demonstrate its leadership in 200G per lane laser technology, which is a fundamental building block of 1.6T transceivers with up to 10 km reach,” said Dr. Sanjai Parthasarathi, chief marketing officer at Coherent.

U.S. Dept of Commerce opens funding for smaller CHIPS projects

The U.S. Department of Commerce’s CHIPS Program Office (CPO) issued its second Notice of Funding Opportunity (NOFOs) targeted toward the semiconductor supply chain. 

This NOFO will support smaller projects, which CPO defines as being under $300 million in total capital costs. Eligible applicants include those intending to use CHIPS funding to support the construction, expansion, or modernization of semiconductor materials and manufacturing equipment production capacity.

Applications will be evaluated using five criteria, the most significant of which is the economic and national security criterion, which encompasses three priorities: supply chain resilience; advancing U.S. technology leadership; and supporting vibrant U.S. fab clusters. 

Projects will also be judged on commercial viability, financial strength, project technical feasibility and readiness, and workforce and community investments. 

Vapor IO Launches Its “Monetize the AI Edge” Program

Vapor IO is kicking off a new"Monetize the AI Edge"initiative to help partners accelerate the availability of real-time AI by delivering bundled AI services via Vapor IO’s shared and neutral host infrastructure.

Tolaga Research has recently completed a report on the addressable market for edge AI across the 36 U.S. markets in which Vapor IO operates. Tolaga estimates that by 2028, the combined value of edge AI in those 36 markets will be approximately $180 billion.

Vapor IO has built a multi-vendor supply chain of pre-integrated, end-to-end AI services that can be bundled into other products or sold on their own.

“Forward-thinking channel partners have already recognized that real-time AI services are in high demand, and by signing up for Vapor's Monetize Edge AI channel program, they can accelerate their revenue goals,” said Cole Crawford, CEO and founder at Vapor IO. 

 The types of use cases include the following:

  • Public Safety: Cameras connected to the Kinetic Grid via fiber optics and wireless technologies use real-time AI to detect unusual or dangerous behavior in real-time and notify the appropriate authorities.
  • Video security: Cameras placed inside and outside of office buildings are monitored with real-time AI. The AI looks at the data on the Kinetic Grid to identify suspicious activity and automatically activate the proper response.
  • Retail: Retailers are upgrading their technology with cameras and AI to use services like real-time inventory tracking, loss prevention, and customer behavior tracking.
  • Manufacturing: Monitoring and quality control can be made more efficient and effective by using computer vision applications to detect anomalies and defects in real-time without human intervention.