The story of optical communications is one of increasing speed. Optical modules cannot be built without transceivers, but in complex, interlocked spaces such as telecommunications, these devices need to be interoperable. As a result, the development of optical transmission technology has paralleled the development of packaging. For example, the CFP8 is the latest addition to the Compact Pluggable (CFP) Multi-Source Agreement (MSA) family of packs. Designed to accommodate 400 Gb/s optical modules, the CFP8 will enable a single line card to transmit data at speeds up to 3.2 Tb/s.
Before starting the discussion, it must be noted that the packaging protocol only defines the size of the packaging and determines the size of the packaging box. These protocols specify plugs and connectors, but do not extend to optical and electronic components that perform tasks such as electro-optical conversion, modulation, detection, etc. This aspect of the design is determined individually by each component supplier. Depending on the size, the same package can accommodate a 100 Gb/s transceiver, for example, consisting of one channel operating at 100 Gb/s or four channels operating at 25 Gb/s.
Communication occurs under a number of constraints, including speed, panel density (packaging density within the rack), and distance. Panel density is a real concern - there's only so much space in a standard rack size. In addition, the tighter the package of the optical module, the more difficult it is to dissipate heat. Heat can degrade the performance and longevity of electronic devices and optoelectronic components such as lasers, a key concern. That said, speed is a top priority for service providers operating long-haul and metro networks.
10g optical module
Let's start with the 10G optical module package up to the recent major data rates for long-range networks. After a long transition from line cards to transponders to XFP, today's dominant optical module form factor for 10 Gb/s is SFP+, providing a good balance between performance and panel density. The next major jump in transfer rates (leaving the 40 Gb/s detour for now) is 100 Gb/s. This advance requires the introduction of coherent optics, which exploit optical phase and polarization to increase bit rates.
Coherent optics technology provides access to 100 Gb/s and beyond. Because it involves higher-order modulation schemes, it is not only fast, but also scalable. This is also a more complex approach and requires more components. The problem is that the SFP+ form factor is not large enough to accommodate coherent optical transceiver components. Following the early 10GB/s path, line cards were followed by transponders, followed by the development of compact pluggable (CFP) packages, as specified by the CFP MSA. The CFP package is 82mm wide, which provides more space for components, but it also takes up more space than previous modules, reducing data density. While it's not ideal, service providers need to push speeds up to 100 Gb/s, and they're willing to make concessions.
Compared with long-distance transmission, data communication is different under certain constraints. While speed is important, in data centers (and increasingly in central offices) the sheer number of devices and interconnects crowded together makes panel density and thermal management far more important than sheer speed. Because the distance is significantly shorter, performance metrics such as signal-to-noise ratio (SNR) can be relaxed. As a result, the datacom industry initially standardized around the Quad Small Form-Factor Pluggable (QSFP) package. These optical modules are much smaller than their CFP counterparts. They can't handle all the components needed for coherent optical transmission, but data communications don't need those speeds—at least not in the short term.
Optical module
Service providers do demand higher data rates over time. In terms of priority, the metropolitan area network is a hybrid between long-range and data communications. Fast operation is essential, but the huge core routers required for metro switching require maximum panel density. In response to demand, the CFP group developed the CFP2 form factor, which measures 41.5mm wide, half the size of the CFP. Essentially, it offers twice the data density (hence the increased numbers) of its predecessor.
Meanwhile, going back to data communications, Web 2.0 companies are starting to require their data center interconnects (DCIs) to run at 100 Gb/s, even when not inside a data center. While vendors can squeeze the necessary components into QSFP packages by taking advantage of faster ASICs, their form factors are still not suitable for coherent optics. In response, CFP participants began working on developing CFP4 MSAs. As the numbers show, the CFP4 package is about half the size of the CFP2 and again doubles the data density. This brings us to today and the migration to 400 Gb/s data rates.
The CFP8 package size is specifically designed for 400 Gb/s transmission. The width is 40mm, almost the same as the CFP2. However, in terms of performance, it is very different. According to the standard, it supports up to 16 differential channels in each direction. It is designed to achieve four times the data density of CFP2.
CFP8 optical modules do not use coherent optical technology. Instead, they use conventional non-return-to-zero (NRC) on-off signaling or another amplitude modulation technique called four-level pulse amplitude modulation (PAM-4). Internal implementations vary: CFP8 optics can deliver 400 Gb/s data rates over 16 lanes running at 25 Gb/s or eight lanes running at 50 Gb/s. The important point is that this package provides reasonable density while providing the required speed for metro and long-distance service providers.
At the same time, next-generation QSFP-packaged optical modules for data communication applications are also in use. Known as QSFP Double Density or QSFP-DD, it maintains the form factor but incorporates stacking connectors designed to support the migration to 100 Gb/s and higher data rates. In the long run, the industry may focus on a completely different approach, Octal Small Form Factor Pluggable (OSFP), but that's another topic.
Meanwhile, 400 Gb/s CFP8 optical modules have entered the market. This approach significantly increases line card data density from 800 Gb/s to 3.22 Tb/s. It is based on proven technology and provides a realistic path to 400 Gb/s per optical module in the short term.
