Optical components

Will Coherent DSP Control Reshape the Optical Vendor Landscape?

The advent of coherent transmission revolutionized the scaling paradigm for optical transmission speeds, ushering in electronic Digital Signal Processors (DSPs) as the key enabler for increasing the capacity in both metro and long-haul WDM networks.

Whereas past advances in wavelength capacity relied on laser sources, modulators and detectors operating at ever-faster speeds, DSPs and the associated complex modulation coding they implement are now the primary drivers for increasing network capacity. As optical transmission speeds move to 400Gbit/s per wave and beyond, the growing importance of coherent DSPs opens up the possibility of a major shake-up of the optical vendor and industry landscape, based on two important and emerging dynamics.

The first is the burgeoning ability to package and concentrate all the required technology for scaling wavelength speed (optics and DSPs) into low power, industry-standard, long-reach pluggable modules without meaningful sacrifices in network performance. Led by standards developments such as the Optical Internetworking Forum (OIF) proposal for 400G ZR (a simplified, low-cost standard for Ethernet connectivity), this finds a receptive home with hyperscale data center operators and packet switch vendors seeking to deploy WDM transport optics directly into packet platforms in a "plug and play" fashion, thereby removing the need for dedicated optical transport systems and the associated additional expenditure.

An early example of this opportunity was the widespread deployment by Microsoft of ColorZ 100Gb/s pluggable optics in the QSFP28 module format for direct connection of routers and switches over open line systems on short campus/metro distances, thereby eliminating the need for any intermediate optical transport systems. (See Microsoft Drops a Data Center Interconnect Bombshell.)

In parallel, the move to pluggable WDM modules operating at 400Gbit/s and beyond fundamentally requires the use of coherent technology. This applies even in metro applications, where, up to now, wavelength speed could be scaled up to 100G with 40-80km reach using non-coherent techniques such as PAM4 modulation, but will not scale to 400Gb/s and beyond. (See Inphi Debuts Data Center Interconnect Gamechanger.)

This capability will increasingly be implemented into Digital Coherent Optics (DCO) modules, that package both the coherent DSP and high-speed optics into a self-contained industry standards compliant module.

Significant industry activity is currently centered on developing the next generation of low-power DSPs using 7nm silicon node technology, which will enable them to be integrated with high-speed optics into standards compliant 400Gbit/s WDM DCO pluggable modules, with faster speeds expected in the future. This will make DSPs a fundamental building block required across a broad range of high volume network applications, and a critical technology for vendors that seek to become or remain dominant players in the optical ecosystem.

For optical suppliers, be they systems vendors or component manufacturers, in-house DSP technology therefore confers important benefits, including tighter control of required features, specifications and roadmap, and, very importantly, the ability to procure DSP chips at their manufacturing cost. The flip side is that, in order to fully monetize these benefits, suppliers need to amortize the R&D investment required to develop a new DSP across large deployment volumes. Given that the R&D investment required for a new DSP now runs to several tens of million dollars, the financial viability of an in-house DSP capability is heavily leveraged on maximizing volumes, and favors large and/or rapidly growing suppliers that can achieve the revenue and profitability required to fund this operational expense.

For systems vendors or component/module manufacturers without DSP expertise, comes reliance on external DSP technology or chip suppliers. These vendors must then procure DSP chips at price points that can be an order of magnitude or more than the manufacturing cost. This increases the cost of goods for a given optical product offering, thereby limiting profitability, and may create strategic dependency on external suppliers for a key technology required to maintain product leadership.

These dynamics create the conditions for a fundamental re-shaping of the optical supplier landscape. Interestingly, the complexity and relative newness of coherent DSPs has led to the concentration of DSP technology and know-how to a short list of key suppliers. The supply chain for these critical technological building blocks currently splits into a market landscape of "haves" and "have nots."

In the camp of the "haves," major optical systems vendors such as Ciena Corp. (NYSE: CIEN), Cisco Systems Inc. (Nasdaq: CSCO), Infinera Corp. (Nasdaq: INFN) and Nokia Corp. (NYSE: NOK) have built in-house DSP technology capabilities, which has allowed them to play a major role so far in the deployment of coherent networks at operating at 100Gbit/s and now 200Gbit/s per wave. (It's unclear whether Huawei is a "have," or whether it only sources DSPs from external component suppliers.)

Also in this category are the specialist DSP vendors such as Acacia Communications Inc. , Inphi Corp. and NTT Electronics Corp. (NEL) , which can act as a source of supply of this critical technology to the "have nots." Other suppliers in the DSP ecosystem include Broadcom, ViaSat, Macom and others that can provide key intellectual property such as ADC/DAC interfaces, DSP algorithms and ASIC design expertise.

System vendor Ciena seems to want to get ahead of the threat to transport system revenues, and is capitalizing on its DSP leadership to create new products, revenues and customer opportunities with integrated DSP/optics modules built with partners. (See Ciena Sets Its WaveLogic Free.)

Ciena seems to see the potential upside from this strategy, and has stated the objective of generating up to $50 million of annual revenues from this move within three years. Another system vendor, Infinera, seems to be exploring similar strategies to monetize integrated Digital Coherent Optics (DCO) products.

The significance of DSP technology also plays out in a resurgence of vertical integration as a way to develop differentiated business models. Like Acacia and Inphi, both Ciena and Infinera are able to complement their in-house DSP expertise with vertical integration around high-speed optics, which could enable them to offer vertically integrated DCO module products in direct competition to component/module vendors.

Infinera's capability to build high-speed photonic integrated circuit (PIC) optics has been an integral part of the company's business model since inception. Highlighting the importance of vertical integration, Infinera has touted how in-house PIC and DSP technology underpins the company's strategy for achieving financial synergies and cost savings in its deal to acquire Coriant. (See Infinera's Integration Situation Needs More Explanation.)

Meanwhile, Ciena's move to vertically integrate optics may have been as unnoticed as it was astute, and centers around the acquisition of optical component manufacturer TeraXion in early 2016, which brought in-house a high-speed optics capability. Nokia, whose wide deployments of DSP-enabled transport systems were recently upgraded with PSE-3 technology, remains a dark horse in this space. (See Nokia Pushes Optical to the Limit.)

In the camp of the "have nots" are optical suppliers without in-house DSP capabilities. This group includes transport system vendors and optical component/module vendors seeking to offer integrated Digital Coherent Optics (DCO) or pluggable coherent 400Gbit/s WDM modules. The range of component vendors without in-house DSP technology is vast, and includes suppliers of high-performance optics used in coherent 100G/200G networks, such as Lumentum/Oclaro, Finisar and II-VI Photonics, and suppliers of datacom optical modules such as Applied Optoelectronics, Innolight, Intel, Source Photonics and others. These system and components vendors must rely on external suppliers offering DSP chips and expertise to build competitive coherent optic products.

While moves by system vendors to explore DCO module product offerings may be viewed as a threat to component and module vendors, it also represents an opportunity for further consolidation by these latter companies. This year's announced acquisition of Oclaro by Lumentum made headline news, and served to highlight the importance of size and scale to ensure ongoing competitiveness. (See Lumentum Sweeps Up Oclaro in $1.8B Deal.)

As coherent optics move to 400Gbit/s speeds and pluggable DCO module formats, a significant opportunity exists for further M&A and merger activities in the component industry. This will drive major players to seek further consolidation of both high-speed optics and DSP technologies under one roof to offer competitive DCO module products, and derive the necessary cost reductions and profitability provided by vertical integration.

These dynamics also create important opportunities for the specialist DSP vendors. Independent DSP suppliers can leverage the importance of their technology for scaling to 400Gbit/s speeds and beyond to usurp traditional optical component vendors for supplier dominance, and/or become strategic suppliers to optical system or component vendors that lack DSP technology by providing a critical technology required for ongoing competitive product offerings.

At the same time, these DSP experts may become important players in M&A activity, where combinations would seek to pair their core technology with either optical component or systems businesses to leverage the benefits of vertical integration.

The veterans of the optical industry may find it ironic that near-term, the path to continued capacity scaling will primarily be enabled through innovation in electronics, compounded by a return to vertically-integrated business models. At the same time, the optical industry is no stranger to radical and sudden changes, often through unintended consequences. So as electronics become a key enabler for network scaling and Moore's Law makes its way into the optical ecosystem, continued technical innovation at breakneck speeds may serve as a catalyst for an ongoing, pervasive and broad realignment of the current optical ecosystem.

In other words, stay tuned!

Serge Melle is a senior leader of strategy, product marketing, business development, M&A and alliances. You can visit his LinkedIn profile here.

Page 1 / 2   >   >>
Serge-Melle 1/23/2020 | 1:34:00 PM
coherent M&A starts....

coehernt DSP leadersip is worth at least $2.6 billion in Cisco dollars.



more to come?


brooks7 9/24/2018 | 2:10:37 PM
Re: have, have not, not sure, printable calendar maybe? @partick.c,

Comcast does have a studio....NBC/Universal.


opticalnetworkingspecialist 9/23/2018 | 5:47:02 AM
Re: Huawei OptiXtreme oDSP This article is so biased it looks like sponsored by Infinera, at a time their shares are going down the sink. Not saying it's sponsored, just saying it looks sponsored (written by an Infinera Exec for 16 years and who probably still keeps stong ties), it feels sponsored (the verbage is identical to Infinera's Marketing department), and it tastes sponsored (makes Infinera look great at a time all other indicators make it look otherwise).

Having said that, it's true that DSP is central to Optical Networking, but like many other things. So this article is not enough of a smoke screen to cover for the very worrying trend in INFN's share price. INFN should focus less on Marketing and more on soul searching.
opticaljunkie 9/21/2018 | 5:18:55 PM
Re: Huawei towncalendars OptiXtreme printable calendar oDSP All the 3 letter acronyms sounds great, what exactly is the osnr performance of the Huawei device/module? 

0.1W/G on the DSP doesn't sound that great either. Wasn't Acacia claiming CFP2-DCO 200G at 20W for the whole module??

That CMS stuff looks like it's made to accommodate old line system that cannot recognize nyquist lambdas
[email protected] 9/21/2018 | 2:00:55 PM
Re: Huawei OptiXtreme oDSP SeniorR&57845 -- Thanks for the interest, but this is not a sponsored column.

Serge used to work at Infinera, but doesn't any longer, so he has detailed knowledge of that company.


SeniorR&57845 9/21/2018 | 5:40:33 AM
Re: Huawei OptiXtreme oDSP I think it's important to mention that the author of the article is actually employed by Infinera. Which explains why there is a pitch on Infinera's PIC in an article on DSP and also why Infinera is put in a positive light of having a DSP, whereas they might as well just be buying the 600G DSP from the two available merchant sources in the market.

So I suggest that Light Reading explicitely states that this article is not written by an independent analyst, but is in fact sponsored by Infinera.
mvissers 9/21/2018 | 4:04:57 AM
Re: Huawei OptiXtreme oDSP Serge,

Have you seen any other DSP vendor refering to "Channel-matched shaping (CMS)" and "Optical-Layer AI Neurons" and to "high-performance oDSP" and "new-generation power efficient oDSP"?

As james0075 indicated - "According to OVUM's data, Huawei has shipped over 127k 100G/200G ports in year 2017, which is more than Ciena+Nokia+Infinera(118K in total)." - Huawei is well positioned concerning the following consideration in your article:
"Given that the R&D investment required for a new DSP now runs to several tens of million dollars, the financial viability of an in-house DSP capability is heavily leveraged on maximizing volumes, and favors large and/or rapidly growing suppliers that can achieve the revenue and profitability required to fund this operational expense.".

When you follow Huawei activities in ITU-T SG15 and OIF, then you can see that Huawei is pushing very hard to get standardized 100G, 200G and 400G FlexO coherent interfaces for 80km (metro-edge, metro-aggregation) and 450km (metro-core), using Staircase FEC for 100G and CFEC/CFEC+ for 200G/400G FlexO interfaces. These FlexO interfaces can also be grouped to support PHY bonding and wavelength bonding, so that larger OTN OTUCn signals can be transported; e.g. 1T OTUC10 carried via five 200G FlexO coherent PHYs/wavelengths. G.709.3 (frame format, FEC) and G.698.2 (optical parameters) are the associated standards.

These 100G, 200G, 400G FlexO coherent interfaces are intended to become available in QSFP-DD-DCO/OSFP-DCO type module for 80km version and CFP2-DCO module for 450km.

These FlexO coherent modules are likely to support on their system side (electrical interfaces) OTN (FOIC, G.Sup58) and Ethernet (AUI, 802.3) protocols. Ethernet signals will be mapped into FlexO (with or without intermediate OTN functionality). Such FlexO coherent modules can be plugged into line cards in OTN equipment and in PKT switch equipment. This dual deployment is critical to prevent DSP R&D investment going through the roof.

I will address these developments in one of the panel sessions in ECOC on Sunday afternoon.
John Wick 9/21/2018 | 3:14:29 AM
Re: Huawei OptiXtreme Smartcric on web site oDSP The information contained within the links you sent references 600G/wave, 16nm finFET and 40T fiber capacity; these specs all happen to also correspond to those for the Acacia AC1200 DSP that is currently being introduced to the market.  
james0075 9/21/2018 | 2:01:28 AM
Re: Huawei OptiXtreme oDSP According to OVUM's data, Huawei has shipped over 127k 100G/200G ports in year 2017, which is more than Ciena+Nokia+Infinera(118K in total).

If Huawei is buying from any supplier on the market, they must be SUPER RICH now. I don't think Acacia or any supplier on market would have this kind of capacity. 

But you are right, if you consider Hisilicon as an "external" DSP supplier to Huawei. 
Serge-Melle 9/20/2018 | 3:34:12 PM
Re: Huawei OptiXtreme oDSP thank you for posting those links; very useful and on-topic info.

what remains unclear is whether the Huawei oDSP was developed *in-house* (ie:  Huawei engineers coded the DSP algorithm, designed the ADC/DAC, did the chip layout, etc), or whether Huawei's DSP capability was implemented using an externally-sourced part from a DSP vendor.

The information contained within the links you sent references 600G/wave, 16nm finFET and 40T fiber capacity; these specs all happen to also correspond to those for the Acacia AC1200 DSP that is currently being introduced to market.  

So it is also feasible (and IMO likely) that Huawei's oDSP feature is enabled through an external DSP supplier.  Happy to hear if you have more info on any in-house DSP development work at Huawei?
Page 1 / 2   >   >>
Sign In