In June, II-VI and ADVA announced a joint collaboration on a QSFP28-based 100Gbps coherent pluggable optic (aka 100ZR) to be released next year. For the past several years, Cignal AI has been forecasting 100ZR optics demand for enterprise and network edge applications. This is the first announcement of a product designed specifically for this market, and it holds the promise of mass deployment of 100Gbps at the network’s edge.
- II-VI’s initial announcement introduced its new DSP, named “Steelerton”, a low-power (2W typical) DSP designed to be used in QSFP28 modules.
- The follow-up joint announcement from ADVA and II-VI introduced a QSFP28 module based on the Steelerton DSP and an ADVA optical front end. The module, which is scheduled for sampling at the end of this year and production availability in the second half of 2023, supports industrial temperature specs and incorporates automatic wavelength tuning. The announcement also clarified that II-VI and ADVA jointly shouldered the development of the DSP.
The primary applications featured in both announcements are 5G mid-haul and backhaul. 5G radios require one or more 25Gbps interfaces to carry eCPRI data, and aggregating that data quickly scales to 100Gbps and beyond. Direct detect 25Gbps and 50Gbps solutions exist, but they are limited to about 15km reach. For networks with longer reach requirements (40-80km), the only current solution faster than 10G tunables are 400Gbps coherent pluggables or relatively expensive 100Gbps pluggables based on the same 400Gbps DSPs and form factor. A lower-cost, high-temperature tolerant 100ZR solution is ideal for this application.
Hundreds of thousands of 10Gbps DWDM ports are deployed annually, mostly at the edge of the network, and a simple upgrade path to higher speeds for those ports is needed. 10Gbps optics operate over 100km, and the direct detect solutions (25Gbps, 50Gbps, and 100Gbps) cannot meet those reaches. While not all spans are long, operators at the edge of the network typically prefer one solution that can meet all network spans, rather than a mix of options. Applications include enterprise handoffs, DSLAMs, FTTH OLTs, and CMTSs. Some of this equipment is already deployed with QSFP28 slots, which makes a QSFP28-based upgrade solution easy to introduce.
CableLabs studied the need to replace multiple 10Gbps link aggregation interfaces with a 100Gbps coherent interface on devices like CMTSs, and it concluded that there is a benefit to a low-cost coherent access solution. The resulting CableLabs Coherent Termination Requirements Specification is built on IEEE standards for 100Gbps coherent Ethernet transmission. This application – replacing 10Gbps DWDM in the access network – and CableLabs endorsement of the demand for a coherent solution both were the catalyst for Cignal AI’s original larger 100ZR forecast.
Cignal AI has long argued that there is serious demand for a low-cost 100ZR solution. However, the only solutions offered to date have been scaled down 400ZR modules in the larger 400G QSFP-DD format. Those modules have had limited commercial success, only in enterprise networks and 100Gbps coherent backbones in areas where 400Gbps is overkill. Those solutions’ price points and power have kept them from meeting the requirements at the edge of the network where the demand is greatest. In order to gain wider deployment, the price of a 100ZR module must be only 3-4 times the price of a 10Gbps module, or somewhere around $1500.
A solution meeting the requirements for QSFP28 (3.5W officially, although ADVA and II-VI argue that 5W is acceptable) would address a large, embedded base and huge potential application space at the edge of the network. Since 100Gbps interfaces are not currently widely deployed at the edge of the network, one could argue that the QSFP-DD format is adequate for 100ZR, as new boxes can be built to support QSFP-DD. Competitors like Cisco are sticking with the QSFP-DD module format for the additional power headroom that it affords. ADVA argues that QSFP28 is being designed into hardware now, is being deployed in current edge devices, and is required to meet the overall system power requirements at the edge of the network. ADVA’s position is significant, as the company is being acquired by ADTRAN, which has extensive experience in access networks.
ADVA is not building optical modules; it is purchasing them from II-VI. The agreement between ADVA and II-VI is not exclusive. II-VI will sell its own 100ZR module on the open market and it could partner with other optical module vendors. A broader set of suppliers benefits both ADVA and II-VI as it legitimizes the technology, but it especially benefits II-VI, which could provide DSP technology to multiple vendors. ADVA gets early access to the modules and the ability to start designing in 100ZR solutions – which will be even more critical when the company becomes part of ADTRAN.
II-VI (Finisar) has a long history of successful internal silicon development and vertical integration. But this is the first coherent DSP developed by II-VI, and the company has plans for more solutions in the future. The roadmap is not public, but II-VI currently has external sources (Marvell) for 60Gbaud (400ZRx) and PAM4 (datacenter) DSPs. II-VI could either work to vertically integrate those solutions or it could develop other innovative DSPs (higher baud rate, lower power, etc.).
In Cignal AI’s OFC 2022 Report, we speculated that a new DSP vendor (IPG or Effect Photonics, for example) might see the 100ZR market as an attractive entry point with few competitors. With the arrival of a massive competitor like II-VI, the market is no longer wide open, but there is still room for innovative solutions around 100Gbps coherent. Other DSP vendors – either small startups or established players – will likely enter the market in 2023 or 2024, especially if II-VI and ADVA have successfully met QSFP28 requirements for mass deployment.
Cignal AI’s Forecast
Cignal AI’s original forecast for 100ZR grew to several hundreds of thousands of units over three years. However, as commercial solutions failed to materialize outside of downscaled 400ZR, the forecast was pushed out and significantly reduced. The revised forecast is for a few modules to start shipping at the end of this year and grow to about 25k modules shipped in 2026. Should II-VI and ADVA (and competitors) capitalize on the pent-up demand for 100Gbps coherent at the edge of the network with a working product at a sub $1500 price, Cignal AI’s forecast will adjust back to several hundreds of thousands of modules by 2026.