To date, multiple system vendors have converged around Class 3-based solutions (Figure 1), recently announcing their next generation offerings. This industry convergence creates the benefit of economies of scale and broad industry investments into the technology used in this baud rate class, the same class being used for 800G MSA pluggable solutions.

Figure 1.  Acacia and other coherent vendors have announced Class 3 Terabit Era solutions.

Advancements Resulting in 65% Power-per-Bit Savings Over Current Competing Solutions
Doubling the baud rate from Class 2 to Class 3 in silicon was a significant engineering achievement, combining design advancements in high-speed Radio Frequency (RF) and Analog to Digital Converter (ADC) and Digital to Analog Converter (DAC) components plus well-designed co-packaging integration of silicon and silicon photonic (SiPh) components. These achievements led to Acacia’s successful 140Gbaud in-house capability that is being leveraged in the commercially available CIM 8 solution.

With high-volume shipments of multiple coherent Class 2 module products utilizing Acacia’s 3D Siliconization, this proven co-packaging integration solution provided the foundation for extending this capability to Class 3 140Gbaud implementation utilized in the CIM 8 (Figure 2). 3D Siliconization maximizes signal integrity by co-packaging all high-speed components including the coherent Digital Signal Processor (DSP) application-specific integrated circuit (ASIC), transmitter and receiver silicon photonics, and 3D stacked RF components into a single device that is manufactured in a standard electronics packaging house. Silicon technology has demonstrated cost and power advantages over alternative technologies, making it the material system of choice for these higher baud rates. These advancements enabling a doubling of the baud rate have led to a 65% power-per-bit savings of CIM 8 over current competing solutions that utilize alternative optical material systems. In addition, the size and power savings of this latest generation enabled the ability to house this 1.2T 140Gbaud solution in a pluggable form-factor.

Figure 2.  An example of 3D Siliconization used in the CIM 8 module, resulting in a volume electronics manufacturable high-speed single device larger than a quarter.

2^nd Generation 3D Shaping Advances Coherent Performance
The CIM 8 is powered by Jannu, Acacia’s 8^th generation coherent DSP ASIC. The design greatly expands on the success of the Pico DSP ASIC predecessor used in the widely deployed performance-optimized Class 2 AC1200 module (Figure 1). The AC1200 was the first module to introduce 3D Shaping, which provided finely tunable Adaptive Baud Rate up to 70Gbaud as well as finely tunable modulation up to 6 bits/symbol. The AC1200 had achieved record breaking spectral efficiency at the time of its introduction, as evidenced by a subsea trial over the MAREA submarine cable connecting Virginia Beach, Virginia to the city of Bilbao in Spain. Finely tunable baud rate helps maximize spectral efficiency in any given passband channel, converting excess margin into additional capacity/reach, and avoids wasted bandwidth due to network fragmentation.

Figure 3.  A popular feature is the fine-tunability of baud rate introduced by Acacia with the Class 2 AC1200; CIM 8 incorporates the same Adaptive Baud feature (as part of 2^nd Generation 3D Shaping) for Class 3 baud rate tunability.

The 5nm Jannu DSP ASIC in CIM 8 intelligently optimizes optical transmission using 2^nd Generation 3D Shaping with an increased Adaptive Baud Rate tunable range up to 140Gbaud, as well as finely tunable modulation up to 6 bits/symbol using enhanced Probabilistic Constellation Shaping (PCS). With 2^nd Generation 3D Shaping, the CIM 8 module can achieve a 20% improvement in spectral efficiency.

Terabit Era Solutions Provide Full Network Coverage
Class 3 technology not only ushers in the terabit era, but also enables full multi-haul network coverage as the high baud rate capabilities transport nx400GbE client traffic across a service provider’s entire network. Full network coverage is not only enabled by adjustment of the modulation, but also implies the capability to optimize for various network conditions which include overcoming transmission impairments.

Figure 4. CIM 8 1.2T, 1T, 800G, and 400G transmission constellations operating at Class 3 baud rates providing wide network coverage addressing multiple applications.

CIM 8 offers significant power-per-bit reductions as well as cost efficiencies for various optical network transport applications.

DCI/Metro Reaches
For transporting 3x400GbE or 12x100GbE client traffic with metro reaches in a single carrier, the CIM 8 is tuned to ~6 bits/symbol (equivalent to 64QAM, example constellation on left). Data center interconnect (DCI) applications would take advantage of this high-capacity 1.2T transport capability to tie data center locations together. This amounts to 38.4T per C-band fiber capacity.

Long-Haul Reaches

For transporting 2x400GbE with long-haul reaches, the CIM 8 is tuned to ~4 bits/symbol (equivalent to 16QAM, example constellation on the right). Wide 800G network coverage is achieved with the Class 3 140Gbaud capabilities enabling service providers to provide end-to-end 2x400GbE, 8x100GbE, or native 800GbE transport across their networks, covering essentially all terrestrial applications.

Ultra-Long-Haul/Subsea Reaches

And for ultra-long-haul/subsea reaches, the CIM 8 is tuned to ~2 bits/symbol (equivalent to QPSK, example constellation on the left). As with the previous scenarios, spectral efficiency with a wavelength channel is optimized by fine-tuning of the baud rate. These high spectrally efficient modes can carry mixed 100GbE and 400GbE traffic over the longest subsea routes in the world with lowest cost per bit. It’s worth noting that almost a decade ago, Acacia demonstrated SiPh capabilities for subsea coherent deployments. CIM 8 incorporates second generation non-linear equalization (NLEQ) capabilities to mitigate the non-linear effects of optical transmission especially for these ultra-long-haul/subsea links providing additional OSNR.

In all the above scenarios, the CIM 8 utilizes advanced power-efficient algorithms to compensate for chromatic and polarization dependent dispersion. In addition, the module accounts for coverage of aerial fiber network segments that require fast state-of-polarization (SOP) tracking and recovery due to lightning strikes. The SOP tracking speed of CIM 8 is double the speed of its predecessor. This fast SOP tracking feature can also be utilized for sensing applications.

Network Operators Achieve Record Breaking Field Trials with CIM 8
CIM 8 capabilities have already been put to the test as illustrated by multiple record breaking field trials across a wide range of applications. These include >5600km 400G transmission over a mobile carrier’s backbone network, 2200km 800G transmission over a research and education network, and >540km 1T transmission over a wholesale carrier’s network.

Acacia continues to demonstrate its technology leadership by leveraging mature knowledge in proven silicon-based coherent technology, producing the first shipping coherent solution to lead the industry into the Terabit Era with the 1.2T pluggable CIM 8 module. With the breakthrough capability of 140Gbaud transmission along with the advanced Jannu DSP ASIC using 2^nd Gen 3D Shaping and leveraging 3D Siliconization, network operators can support full network coverage for multi-haul applications, especially to support growing demands for nx400GbE and upcoming 800GbE traffic.

References:
Blog: Terabit Today: Maximize Network Coverage
Blog: How Industry Trends are Driving Coherent Technology Classifications
Blog Series: The Road Ahead for Next-Generation Multi-Haul Designs Part 1, Part 2, Part 3

 Over the last 10 years, we’ve seen rapid increases in capacity per wavelength by increasing modulation order from QPSK to 16QAM to 64QAM, as well as increasing baud rate supported by opto-electronic devices. However, beyond the coherent modulation order of 64QAM, the achievable performance isn’t sufficient to address target applications due to the reduction in reach.  As a result, increasing baud rates has been looked to as the primary means of increasing capacity per wavelength.  This requires innovative and cost-effective implementations to provide higher baud rate solutions and packaging advancements.  Opto-electronic integration and co-packaging are techniques that were discussed by Acacia’s Founder and Chief Technology Officer Benny Mikkelsen in his OFC 2019 Plenary talk and continue to be critical to support the ever-increasing need for higher data rates and smaller, cost-effective optical interfaces for cloud, access, and transport applications.

Opto-Electronic Integration and Co-Packaging Explained
These techniques are used to reduce components in size and power while also increasing functionality and performance of the solution. Opto-electronic integration generally refers to the process of integrating a wide range of optical functions on a single chip, such as the large amount of optical and opto-electronic functions being achieved in a photonic integrated circuit (PIC). While co-packaging is the ability to combine multiple chips into a single package which can be further integrated into a transceiver module. The main benefit is that it can then be manufactured as if it’s a single component with even more functions.

Opto-electronic integration, particularly through silicon photonics, enables the miniaturization of coherent transceivers. The benefits of opto-electronic integration can be seen in the below graphic, which shows how the size of a coherent transceiver was reduced significantly over a few product generations.  By leveraging these techniques, each new generation was able to raise the bar to increase capacity while reducing power and size.

Figure 1. Opto-electronic integration and co-packaging have enabled coherent transceivers to become significantly smaller over the last decade.

Three Main Benefits of Opto-electronic Integration and Co-Packaging

1. Reduced Power

It takes a massive amount of power to operate data centers, which is why sustainability ranks top on data center operator’s agendas. Opto-electronic integration and advanced packaging helps lower the power consumption of the coherent modules used for moving data across networks.

The benefit of having multiple devices packaged into one compact component means fewer interfaces and the ability to support higher speeds per lane. Electrical compensation of PCB routed high-speed signals, which consumes power, is essentially eliminated.  As an example, by co-design and co-packaging the trans-impedance amplifier (TIA) and driver chips with the silicon photonics-based PIC on the same substrate as the digital signal processor (DSP) ASIC, the DAC termination can be eliminated and can result in a 35 percent DAC power reduction.

Figure 2. Co-design and co-packaging of the TIA and driver chips with the silicon photonics-based PIC on the same substrate as the DSP ASIC eliminates the DAC termination and can result in a 35% DAC power reduction.

2. Reduced Size

Silicon Photonics
Using silicon as an optical medium and leveraging CMOS fabrication processing technology, silicon photonics allows tighter monolithic integration of many optical functions within a single device. While traditional optics systems used many discrete devices, silicon photonics allows all those devices to fit onto a single silicon chip reducing the size.  Silicon photonics has been a key enabler for achieving the tremendous size reduction in Figure 1.

Component Stacking
In component stacking, the DSP and PIC are tightly co-packaged on the same substrate, and the high-speed modulator driver and TIA components are stacked on the PIC which also reduces the size.  Component stacking is a process widely adopted in the electronics manufacturing process that is now being applied to opto-electronic technology manufacturing.

Co-packaging and Integrated Control IC
Size reductions are achieved by integrating functions and the control IC through co-packaging techniques. Smaller devices can translate into either more functionality within the same form factor and power consumption footprint or a smaller form factor with the same functionality and power consumption as the previous generation. For example, in the Acacia CFP2 form factor, the integration of multiple discrete control ICs into one integrated device led to a 500 percent reduction in board footprint.

3. Increased Capacity

Enhancing DSP and Increasing Baud Rate
With network capacity demands increasing, network operators are challenged with an ongoing need to deploy solutions that can keep up with these capacity demands while being power, size and cost efficient. High speed opto-electronic integration and advanced packaging can deliver high-capacity transport from the state-of-art DSP.

Increasing baud rate has always been an efficient way to enable more cost-effective optical networks by reducing the number of optics required to support a given transmission capacity. By doubling baud rate over previous generations, we can support twice the capacity per carrier over greater reaches than prior generations. This approach provides a simple, scalable path that supports higher capacity per carrier over the reaches needed for existing and new network architectures.

Acacia’s Implementation: 3D Siliconization
Acacia’s approach to co-packaging is called 3D Siliconization technology. This process utilizes highly scalable and reliable volume electronics manufacturing processes which applies 3D stacking packaging techniques to enable a single device to include all the high speed opto-electronic functions necessary for coherent transceivers. With 3D Siliconization, the high-speed RF interfaces are tightly coupled together, resulting in improved signal integrity for high baud rate signals.

Figure 3.  3D Siliconization improves signal integrity and performance via the reduction of electrical inter-connects, in addition to the benefits in cost, reliability, power, and size.

This device decreases footprint by including the DSP, PIC, drivers, and TIAs, and is manufactured using standard CMOS packaging processes that leverage the same reliability, cost, and volume scaling advantages.  This approach is utilized by Acacia’s 400G pluggable family and the 1.2T 140Gbaud Coherent Interconnect Module 8 (CIM 8).

That means 1) knowing what traffic granularity they need to transport and then 2) scaling the capacity in their transport network so that it aligns with the traffic granularity they need to move.

In addition, operators are looking for a common solution that can maximize network capacity over the widest network coverage. This can provide the ability to scale to higher capacity in a cost-effective way.

400GbE is Today’s Unit of Currency
Networks are evolving from 100GbE to 400GbE dominant traffic. As Telia Carrier stated, “400G: it’s here and huge!.” Network operators now need to figure out a way to build their networks to best support 400GbE traffic. True multi-haul solutions have emerged to meet this need because they were designed to be a flexible solution for DCI, metro, long-haul and subsea applications. However, a multi-haul solution is only effective if it can carry the different types of traffic it needs to transport while maximizing the value of existing infrastructure and reducing operational costs.

Multi-haul solutions maximize fiber utilization and simplify deployment.

Modulation Order Drives Design Decisions
Class 2 implementations utilizing 4 bits/symbol (~16QAM) and 60-64Gbaud have been standardized in the industry because they address a wide range of DCI and service provider metro applications as networks transition to 400GbE. Multi-haul implementations allow even greater reaches using approximately 4 bits/symbol with shaping and higher gain forward error correction algorithms. Class 2 multi-haul products transport 400GbE clients over the majority of network applications using 4 bits/symbol, with the ability to dial down the modulation format to 2 bits/symbol for the most challenging ultra-long-haul links.

Looking to Class 3 multi-haul products, efficient transmission of 400GbE traffic can best be achieved by doubling the baud rate to 120-128Gbaud. This enables 4 bits/symbol transmission supporting 800G line rates and 2 bits/symbol supporting 400G line rates. In high-capacity edge applications, these Class 3 products can support up to 1.2T line rates. By aligning client traffic granularity with the modulation orders that can best support network applications, identifying the preferred baud rate for Class 3 implementations becomes straightforward.

The industry has standardized on Class 2 implementations utilizing 4 bits/symbol (~16QAM) and 60-64Gbaud for various 400G implementations that address a range of DCI and service provider network applications.

Maximizing Network Application Coverage
A recently introduced Class 3 multi-haul solution, Acacia’s Coherent Interconnect Module 8 (CIM 8), can address transmission of multiple 400GbE client interfaces over virtually any network application, delivering 1.2T per carrier capacity for high-capacity DCI interfaces, 800G per carrier capacity over most optical links using 4 bits/symbol (~16QAM) modulation, and 400G per carrier over long-haul and subsea links.

Leveraging actual data from representative networks, the below simulation shows that the CIM 8 can effectively address subsea applications with 400G links, long-haul and metro with regional 800G (2x400G) and DCI and metro networks with 1.2T.  This means that the CIM 8 can provide efficient transport of 400GbE client traffic across the entire network, including 90 percent coverage using 800G (2x400GbE client traffic).

Utilizing data from actual service provider networks, Acacia’s CIM 8 can provide ~90% 800G network coverage compared to <20% for ~96Gbaud systems.

Scaling Efficiently and Cost-Effectively Today and Tomorrow
Network operators are being challenged more than ever before to scale their networks efficiently and cost effectively. Key to achieving this is knowing what kinds of traffic they are transporting and then building the right system that can most effectively transport that traffic.

As we approach Shannon’s Limit, further improvement will come from going to higher baud rates, but in a cost-effective way. As this article has discussed, having one solution that can be leveraged across all the various applications can enable an efficient and cost-effective solution for network operators looking to scale their networks today and in the future.

Network operators need to efficiently scale their networks to keep up with growing user bandwidth demands.  To meet this need, it is going to be increasingly important to develop next-generation solutions based on scalable technologies that are aligned with industry trends.

Trends Driving Coherent Optical Development

Coherent Moving to Shorter Reaches
As the industry moves to higher data rates, coherent is being adopted in shorter reach interfaces, which is increasing the share of coherent ports that are deployed in pluggable form factors. For example, the 400ZR specification targeted 80-120km reaches, and subsequently the OIF is now working on a project to define an 800LR that targets 10km and below. As this trend continues, it is likely that we’ll also see coherent move into the data center in future generations. These shorter reach interfaces tend to be higher volume than traditional transport applications, and as we look at this transition, it becomes more important to have interoperability and pluggable form factors that can plug directly into router interfaces.

Standardization is Key
The industry has been steadily moving to more standardization. 400G was a significant step forward and today we have a variety of standardized interfaces including 400ZR/ZR+ and Open ROADM – all of which are growing. These standardized interfaces are displacing both proprietary solutions in more traditional transport applications as well as direct detect solutions in shorter reach interfaces. Over time, most optical industry analysts agree that these standardized pluggable interfaces are expected to contribute a larger and larger portion of all the coherent ports in the industry.  Multi-haul solutions that address multiple applications will still be needed, but it will be important to align these multi-haul investments with standards.

Approaching the Shannon Limit
From a development perspective, incremental improvements in spectral efficiency are being made as we approach the Shannon Limit, but we need to find ways to grow network bandwidth more efficiently over time. In the past, we were able to scale the amount of data being transmitted over a single set of optics, while also increasing capacity on the fiber. Today, we are scaling as we increase baud rate, but the amount of data on the fiber is growing more incrementally. This is changing the way we develop products for the future and increasing baud rate cost-effectively is critical moving forward.

These three trends point to high-volume standardized solutions having a greater influence on next generation industry investment.

Coherent Technology Classifications

The below figure illustrates how coherent technology has evolved in response to growing bandwidth demands. Different baud rate classes are grouped based on technological capabilities, industry standardization (such as OIF, IEEE, Open ROADM, OpenZR+ MSA, CableLabs, ITU, IEEE) and common industry investments. Throughout this evolution, it has been critical that each successive class support similar reaches to the previous class.

Baud rate has doubled for each coherent technology class.

Class 1
There was a long period during the early days of coherent technology development with multiple investment nodes. At 30-34 Gbaud, there were 100G and 200G products and the industry made a significant investment at this stage. In this generation, there was widespread standardization of components such as modulators and receivers. However, standardized optical interfaces sacrificed significant performance compared to proprietary implementations, so they were not widely adopted.

Class 2
This is where the industry is today, and it is during this stage that standardized optical interfaces are being widely deployed for the first time. The first Class 2 products were proprietary interfaces supporting multi-haul applications in embedded module form factors. Later, 400G faceplate pluggables, driven by standardization efforts that drove heavy investments into products centered around 16QAM, 60+Gbaud per 75GHz channel transmission were introduced with strong industry adoption. When we migrated to Class 2, we doubled the baud rate and enabled an increase in the data rate.

Class 3
These products once again represent a doubling of baud rate compared to the class before it, with products migrating to 120-136 Gbaud. This approach of doubling baud rate is utilized in standards because it allows for this doubling of data rate using the same modulation format as the previous class. 4 bits/symbol was chosen for 400G standards because it supports a wide range of applications. When scaling from 400G to 800G, doubling the baud rate is the logical path forward.

Class 4
These products will continue the trend we saw in the earlier classes where the model is to increase baud rate but take additional steps to cover all applications. Like earlier trends, we expect the baud rate to also double from Class 3 to Class 4 products, while utilizing the same modulation order.

“These classifications basically mirror industry investment cycles, which is absolutely the right way to align coherent technology development around,” said Alan Weckel, Founder and Technology Analyst at 650 Group. “At the end of the day, operators need solutions to be cost effective and the best way to do that is to leverage the investment across all the various applications and benefit from higher volume production and scale. Increasingly coherent ports are moving to pluggable form factors and as we approach the Shannon Limit, further improvement in cost will come from going to higher baud rates, but in cost effective way.”

Acacia’s Class 3 Solution
Acacia’s recently announced Coherent Interconnect Module 8 (CIM 8)  is a Class 3 solution that can address transmission of multiple 400GbE client interfaces over virtually any network application, delivering 1.2T per carrier capacity for high-capacity DCI interfaces and 800G per carrier capacity over most optical links using 4 bits/symbol (~16QAM) modulation. This product supports 150GHz channels with double the capacity per carrier and longer reach than that of the previous class, providing a simple, scalable path that is compatible with the previous network architecture generation.

Acacia’s CIM 8 is the industry’s first 1.2 faceplate pluggable coherent module, powered by Acacia’s Jannu 5nm CMOS digital signal processor (DSP) ASIC. This solution delivers industry-leading performance with single carrier 1.2T operation and combines the Jannu DSP with 3D Siliconization packaging technology which includes the silicon photonics integrated circuit (SiPh PIC), high-speed modulator driver and transimpedance amplifier (TIA) in a single opto-electronic package. With 3D Siliconization, the high-speed RF interfaces are tightly coupled together, resulting in improved signal integrity for high baud rate signals. The high-density packaging as well as an advanced high-speed modulator design that enables the 140Gbaud performance as explained in this white paper.

The Future of Coherent Development

As the coherent technology classifications chart highlights above, the industry has undergone several large investment phases.  This has been good for the industry because it allowed vendors to have a significant ROI, worked well with network operator upgrade cycles, and avoided many small iterative steps in between. However, moving forward it is going to be important to align with these industry investments. As a result, we need to think about how to develop solutions that scale to high volume in a power-efficient and cost-effective way. Silicon photonic based coherent interfaces have proven to be successful in meeting these challenges generation after generation and we believe this technology can continue to help effectively meet bandwidth demands of the future.

As an example, Acacia’s high-performance Pico DSP-based 1.2 Terabit solution is currently deployed in well over one hundred networks around the globe and has been adopted by three of the four largest hyperscalers. In 2020 alone, Acacia shipped more than 30,000 Pico-based ports as customers increasingly recognized the competitive benefits that high-performance, flexible coherent transmission solutions can provide.

Network Transmission Flexibility Benefits

Multi-haul coherent solutions like the Pico DSP-based 1.2 Terabit solution are software configurable transponder modules that provide various transmission capacities and reaches. By varying the modulation order and baud rate settings, it can provide flexible options for service providers.

A multi-haul coherent solution addresses a range of network applications.

Balancing Modulation Order and Baud Rate

A common method of increasing throughput of a coherent channel is to increase the modulation order. However, this may result in a reduction in reach due to reduced optical signal to noise ratio (OSNR) tolerance for the higher modulation orders. Alternatively, the baud rate could be increased while maintaining a lower modulation order which provides additional capacity per channel with minimal sacrifice to reach.

Maximizing the channel capacity using continuously tunable baud rate can convert unused spectrum into usable capacity, with the goal to fill up the available channel bandwidth. However, as discussed in this whitepaper, increasing baud rate provides minimal improvements in fiber capacity once the transmission is well-matched to the channel.

Increasing transmission baud rate proportionally increases the transmission spectrum and can be used to convert unused spectrum into useable extra capacity.

With transmission flexibility, service providers can configure their client traffic with the following flexible options:

• 12×100 GbE or 3×400 GbE with 64 QAM modulation for DCI edge applications
• 8×100 GbE or 2×400 GbE with 16 QAM modulation for metro/regional and long haul
• 4×100 GbE or 1×400 GbE QPSK for the most challenging terrestrial and submarine networks

To ensure a smooth migration from 100GbE to 400GbE, it’s important to have a solution that can efficiently transport either type of traffic, or a combination of both, without restrictions on performance and functionality.

Increase Performance and Reach with 400GbE Long Haul

With 400GbE becoming the “common currency” for high-capacity Ethernet transmission, it’s important to have a solution that can support this traffic over long distances. For service providers supporting 400GbE traffic they can use Acacia’s Pico DSP-based solution and choose from various configurations, as previously mentioned, including the option to combine two 400GbE client signals into an 800G 150 GHz channel for transmission over their metro and long-haul networks.

Leveraging Acacia’s Pico DSP-based solution service providers can combine two 400GbE client signals into an 800G 150GHz channel for long haul transmission.

Migrating from 100G to Higher Speeds Just Got Easier

Multi-haul coherent solutions enable network operators to easily migrate from 100G traffic toward 400G and higher speeds to deploy new and exciting applications and services. Networks utilizing flexible coherent transmission can provide support for growing client traffic across the entire network—from DCI edge, metro, long-haul, and all the way to submarine. Learn more in this video.