OIF document aims to spur line-side innovation

The Optical Internetworking Forum (OIF) has completed defining the CFP2-ACO (analogue coherent optics) module, used for coherent-based optical transmission. The industry body's CFP2-ACO Implementation Agreement document has been developed to help optical component vendors bring innovative line-side products to market more quickly.

The CFP2-ACO. Source: OIF

The pluggable CFP2-ACO houses the coherent optics, known as the analogue front end. The components include the tuneable lasers, modulation, coherent receiver, and the associated electronics - the drivers and the trans-impedance amplifier. The Implementation Agreement also includes the CFP2-ACO's high-speed electrical interface connecting the optics to the coherent DSP chip that sits on the line card.

One historical issue involving the design of innovative optical components into systems has been their long development time, says Ian Betty of Ciena, and OIF board member and editor of the CFP2-ACO Implementation Agreement. The lengthy development time made it risky for systems vendors to adopt such components as part of their optical transport designs. Now, with the CFP2-ACO, much of that risk is removed; system vendors can choose from a range of CFP2-ACO suppliers based on the module's performance and price.

Implementation Agreement

Much of the two-year effort developing the Implementation Agreement involved defining the management interface to the optical module. "This is different from our historical management interfaces," says Betty. "This is much more bare metal control of components."  

For 7x5-inch and 4x5-inch MSA transponders, the management interface is focused on system-level parameters, whereas for the CFP2-ACO, lower-level optical parameters are accessible given the module's analogue transmission and receive signals. 

"A lot of the management is to interrogate information about power levels, or adjusting transfer functions with pre-emphasis, or adjusting drive levels on drivers internal to the device, or asking for information: 'Have you received my RF signal?'," says Betty. "It is very much a lower-level interface because you have separated between the DSP and the optical interface."

The Implementation Agreement's definitions for the CFP2-ACO are also deliberately abstract. The optical technology used is not stated, nor is the module's data rate. "The module has no information associated with the system level - if it is 16-QAM or QPSK [modulation] or what the dispersion is," says Betty.

This is a strength, he says, as it makes the module independent of a data rate and gives it a larger market because any coherent ASIC can make use of this analogue front end. "It lets the optics guys innovate, which is what they are good at," says Betty.

Innovation

The CFP2-ACO is starting to be adopted in a variety of platforms. Arista Networks has added a  CFP2-ACO line card to its 7500 data centre switches while several optical transport vendors are using the module for their data centre interconnect platforms. 

One obvious way optical designers can innovate is by adding flexible modulation formats to the CFP2-ACO. Coriant's Groove G30 data centre interconnect platform uses CFP2-ACOs that support polarisation-multiplexed, quadrature phase-shift keying (PM-QPSK), polarisation-multiplexed, 8-state quadrature amplitude modulation (PM-8QAM) and PM-16QAM, delivering 100, 150 and 200 gigabit-per-second transmission, respectively. Coriant says the CFP2-ACOs it uses are silicon photonics and indium phosphide based.

Cisco Systems uses CFP2-ACO modules for its first data centre interconnect product, the NCS 1002. The system can use a CFP2-ACO with a higher baud rate to deliver 250 gigabit-per-second using a single carrier and PM-16QAM.

Ian BettyThe CFP2-ACO enables a much higher density line-side solution than other available form factors. The Groove G30, for example, fits eight such modules on one rack-unit line card. "That is the key enabler that -ACOs give," says Betty. 

And being agnostic to a particular DSP, the CFP2-ACO enlarges the total addressable market. Betty hopes that by being able to sell the CFP2-ACO to multiple systems vendors, the line-side optical module players can drive down cost.

Roadmap 

Betty says that the CFP2-ACO may offer the best route to greater overall line side capacity rather than moving to a smaller form factor module in future. He points out that in the last decade, the power consumption of the optics has gone down from some 16W to 12W. He does not foresee the power consumption coming down further to the 6W region that would be needed to enable a CFP4-ACO. "The size [of the CFP4] with all the technology available is very doable," says Betty. "But there is not an obvious way to make it [the optics] 6W."     

The key issue is the analogue interface which determines what baud rate and what modulation or level of constellation can be put through a module. "The easiest way to lump all that together is with an implementation penalty for the optical front end," says Betty. "If you make the module smaller, you might have a higher implementation penalty, and with this penalty, you might not be able to put a higher constellation through it."

In other words, there are design trade-offs: the data rates supported by the CFP2 modules may achieve a higher overall line-side rate than more, smaller modules, each supporting a lower maximum data rate due to a higher implementation penalty.

"What gets you the ultimate maximum density of data rate through a given volume?" says Betty. "It is not necessarily obvious that making it smaller is better." 

Could a CFP2-ACO support 32-QAM and 64-QAM? "The technical answer is what is the implementation penalty of the module?" says Betty. This is something that the industry will answer in time.   

"This isn't the same as client optics where there is a spec, you do the spec and there are no brownie points for doing better than the spec, so all you can compete on is cost," says Betty. "Here, you can take your optical innovation and compete on cost, and you can also compete by having lower implementation penalties."