ECOC is one of the most significant conferences in optical networking worldwide. This year’s conference was held in Gothenburg, Sweden, and the event was well attended by the engineering community, as the conference trends heavy on technology and light on marketing. We at Cignal AI like ECOC because it gives us another opportunity to converse with the folks directly responsible for charting the technical roadmap of upcoming optical systems.
In this first of several ECOC briefs, we share some of our findings on the changing trends in coherent optics. Subsequent briefs will explore China and client optics.
- CFP2-ACO Outlook – The ACO market is going away; what this means
- Silicon Photonics – Searching for the killer app
- Inphi ColorZ – Moving downmarket to grow volume
- 400G ZR – Examining investment risk and rewards
- Acacia CFP2-DCO – Market Feedback
CFP2-ACO Outlook & Roadmap
Demand for CFP2-ACO optical modules remains exceptionally strong because of Cisco and Ciena’s vigorous consumption. These two companies are supplying equipment utilizing these modules to Verizon, which is pursuing its ongoing metro WDM buildout. However, Cisco is the largest consumer based on the recent success of its NCS1000 compact modular system. Cignal AI’s recently released 1H17 Optical Applications report estimated demand for this system at nearly 10k ports.
Ciena’s major win at Reliance Jio relied on the 5430 system and consumed a significant amount of 100G CFP2-ACO modules. We estimate that the CFP2-ACO market is roughly 45k units in 2017, with potential upside depending on the level of success of the Cisco product in the market.
But this appears to be where the CFP2-ACO story ends. The universal consensus among all component and equipment vendors we’ve spoken to is that the ACO format will not survive past its existing 200G footprint. Equipment companies that designed the current generation of 200G CFP2-ACO systems decided that integration and interoperability challenges are too great a risk for 400G and 600G operation. We could not find one equipment or component vendor committed to making 400G ACO a reality now. That is not to say that several didn’t try.
Multiple component vendors have already developed the core Indium Phosphide technology to enable 400G ACOs. Fujitsu has built a high-performance Lithium Niobate based 400G ACO that is nothing short of an engineering marvel. Developments such as these will not go to waste, as the components originally designed for use within the ACO will find a new life when re-purposed for discrete designs, MSAs, and DCO designs.
Huawei, for example, is building its own CFP-DCO modules using in-house and external DSPs from NTT Electronics. The internal components of a CFP2-ACO provide the integrated optics within their DCO design instead of discrete components.
While CFP2-ACO development and equipment design are winding down, the vendors who built the enabling technology for the next generation are in a strong position to enable 400G in other form factors. Many existing systems will continue to ship with 100/200/250G ACOs for years to come. Oclaro supplies at least 80% of the ACO market according to Cignal AI estimates. Given the now static nature of this market, we do not foresee increased competition.
Silicon Photonics – Acacia & Elenion
We met with two silicon photonics vendors focused on the coherent optics market; Acacia and a newer company, Elenion. Both companies have developed Silicon Photonics (SiPho) TOSA/ROSAs WDM optics. It is well known that Acacia uses its SiPho technology in some of its modules and is continuing to invest R&D in the area.
Elenion, on the other hand, has emerged from stealth mode at OFC earlier this year. The company is wholly owned by Marlin Equity Partners, who also co-owns Coriant. Coriant currently purchases the SiPho optical components from Elenion and then builds two different CFP2-ACOs in-house rather than buying this technology off-the-shelf.
The first Coriant ACO design entering production now is a high-performance metro solution engineered to displace conventional designs from companies like Oclaro. The other ACO version is a lower performance variation suitable for shorter reach applications and eliminates the tunable filter and amplifier on the transmitter. Coriant is betting this product will give their Groove system an advantage over other Compact Modular systems in shorter reach applications.
Elenion is the only alternative supplier to Acacia for SiPho components for coherent optical interfaces now. In 2018, NTT Electronics will bring a comprehensive solution to market and increase the total number of suppliers to three.
There has been debate and hype over the killer telecom application in SiPho, but we have yet to see it. The existing equipment designs using SiPho do not appear to have material advantages over market alternatives.
That said, things might change with the next generation of CFP2-DCO modules, as Acacia’s new product has been universally hailed for setting the bar for power & performance. We can’t determine yet if this is a result of superior DSP design vs. the use SiPho. The next battleground is 400G ZR, and SiPho has the potential to be a crucial component in this challenging environment.
Inphi ColorZ Expands
Inphi is ramping up production of the 100G direct-detect ColorZ module, and it shipped 10k units in the first six months of 2017. Cignal AI’s estimate for total ColorZ production in 2017 is 25k units, and we’re starting to think that this estimate is conservative. Inphi expects ColorZ to be a $50M+ product in 2017. For the market to expand this much, we believe the pricing will need to drop below $2k/unit.
Inphi is moving in this direction. During ECOC, the company announced the ColorZ-LITE; a lower cost version of the ColorZ which is designed for applications less than 20km. It isn’t clear if this is a different physical device or just a lower spec’ed version of the existing ColorZ.
The inherent downsides of direct detect technology are minimized in this sub-20km application, which was illustrated in a compelling demonstration by InPhi. The ColorZ-LITE can reach 20km without external amplification and with a fixed dispersion compensation unit at the receive side. This shorter reach application eliminates much of the complexity in getting the technology to work on longer spans, and it is well-suited for fiber constrained DCI applications in campus environments.
We think coherent 400G ZR technology will eventually displace the direct-detect 100G WDM market, but Inphi is endeavoring to fill the short-term need with the ColorZ-LITE.
400G ZR & Cable Labs
The number of companies claiming to develop 400G ZR DSPs keeps growing, and as many of 5 companies have now thrown their hat into the ring. Cignal AI’s 1H17 Optical Applications report is the first to estimate the size of this market, but even the most optimistic estimate of its potential does not support five suppliers. We outlined this conclusion in our OFC report on the subject (read it here).
The RFPs for coherent 400G ZR modules come with a pricing expectation of sub $2k – this is pricing on RFPs received by component vendors. This is an exceptionally challenging figure considering this price is roughly the cost of a 100G ColorZ direct-detect solution from Inphi today. It is hard to imagine that in 4 years the same $2k buys 4x the bandwidth in a coherent format – but this is what the cloud & colo companies are asking for.
Momentum is building towards finding a way to make 400G ZR work in the QSFP28 DD format. The component vendors we spoke to are not counting on being able to use the more generous specs of the OSFP format as they are concerned this format will not take flight. Cisco and others are actively working with component vendors to identify technical compromises to deliver 400G in a traditional QSFP28 DD format. This format would also neatly open the possibility of 100G operation in the same form-factor.
A lower speed, higher performance 100G derivative of 400G ZR with greater un-amplified reach remains what we believe to be the holy grail product for traditional non-DCI networks. It is also a way to increase potential return on what is a high-risk component investment – the tens of millions of dollars spent designing and building the 400G ZR DSP.
The Cable MSOs are actively looking for a roadmap in this area, and CableLabs is specifically looking for 100G coherent in the edge of the network. We couldn’t pin down a precise volume figure for their needs, but the price target is 40km for sub $1k. Again, this would represent exceptionally challenging pricing, but the volume for this application as well as those at traditional telecom operators could outstrip DCI demand by 2021.
400G ZR development efforts cannot be justified solely with demand from a handful of cloud & colo companies. A broader market is needed, and the use cases are emerging.
Based on our conversations, the Acacia 200G CFP2-DCO is the darling device of 2H17. It is unrivaled in performance and power consumption. Inphi announced a 200G DSP designed for this application a few days before ECOC, but people we spoke to were skeptical was that the DCO modules designed with it could match Acacia’s product.
Cignal AI is not aware of many announced systems using the Acacia CFP2-DCO (aside from Infinera), but we expect success with the module at switch and router companies like Cisco, Juniper, and Arista. Cisco and Juniper already use Acacia products and have mostly abandoned in-house DSP and optics design. It’s a safe bet Acacia will win these accounts with this product if they have not already.