The current state of-the-art for data center intraconnect, up to 2km, is 100G but that is not enough. The largest data center operators, the ICPs, want 400G. At OFC 2018, the industry stepped up with a multitude of announcements and demonstrations, covering the breadth of the 400G ecosystems. The optical components and semiconductor community offered modules, discrete lasers, and communications semiconductors. To date, Ovum has tracked the following:
AOI 400G QSFP-DD DR4 and FR4, PAM4 DML and EML
Broadcom Tomahawk 3: 56Gbps PAM4 SerDes, 400G gearbox
Finisar 400G QSFP-DD AOC VCSEL based
Inphi 400G PAM4 DSP for QSFP-DD and OSFP, TIA and driver
Lumentum 400G modules: QSFP-DD FR4, DR4 and OSFP FR4, 100G PAM4 EML
Macom EML driver for 53GBaud PAM4, PAM4 VCSEL driver and TIA
MaxLinear 400G PAM4 DSP with integrated laser driver
NeoPhotonics 53 GBaud EML with integrated driver, PIN PD, TIA
Oclaro 400Gbps QSFP56-DD FR4, 53GBaud TOSA and ROSA, gearbox IC, PAM4 EML, DFB-MZ PIC
Source Photonics QSFP-DD 400GBASE-LR8
Sumitomo QSFP-DD 400G-FR4
The 400G market is in its infancy. Substantial commercial revenues are still some way off in the latter half of 2019. There will be different consumptions models: Some ICPs may buy discretes and task contract manufacturers to assemble, while others will buy complete modules. The industry will work feverishly to reduce cost leveraging photonic integration and optimized manufacturing processes. The ICPs may throw further curve balls, recalling the cost-optimized 100G CWD4 OCP with relaxed specifications for cost efficiency. 400G will be a highly competitive contest. May the lowest cost per bit providers win.
At OFC 2018, Microsemi has introduced OTN3.0. OTN1.0 supported the 10G lineside era, while OTN2.0 stepped up to 100G. Meanwhile, OTN3.0 needs to support a much more complex era of variable lineside speeds from 100G to 600G. In addition, the client side has evolved to 10GE, 25GE, 100GE, 200GE, 400GE, FlexE, and FlexO. And network operators would also like a power consumption reduction.
Microsemi's OTN3.0, DIGI-G5, is a 1.2Tbps chip, supporting the client-side options noted above, and connects to optical transceivers and coherent DSPs via 56G PAM4 SerDes. The DIGI-G5 supports core OTN switching and network-edge applications. Wirespeed OTN encryption is included.
OTN began with a small set of CSP proponents. As lineside speeds rose and SONET/SDH faded, OTN gained further favor as the switching and aggregation mechanism of choice. With greater flexibility on the line and client sides, OTN is the essential client-to-line mapping mechanism. The OTN application space has broadened, so much so, that dare ICPs consider OTN?
On March 12, 2018, Fujitsu announced 1FINITY T500 and T600 series bandwidth-variable transport blades, its next-generation family of optical transponders. Fujitsu will use the latest coherent DSP technology developed by its partner NTT Electronics (NEL). The advanced programmable DSP technology will support over 100 modes of operation, including variable client and network rates. On the lineside, the data rate, the baud rate, the modulation format, and the forward error correction are all adjustable to enable network operators to optimize capacity and reach with one transport solution. Transport rates can be programmed for 100G up to 600G per wavelength. Additionally, the solution will support both C-band and L-band transmission, enabling up to 76.8Tbps per fiber pair.
The 1FINITY T500 and T600 series blades will be able to autonomously communicate optical parameters to Fujitsu's Virtuora Network Controller for capacity and wavelength adjustments based on service needs. This will enable bandwidth optimization, optical margin analysis, and failure prediction for next-gen control.
Coming out of the gate with L-band support will give Fujitsu an edge in two target markets: the L-band markets of the world and the ICP community. The ICPs with bandwidth pressures and fiber constraints will seek to deploy solutions that maximize spectral efficiency over a fiber pair. Fujitsu's solution has the horsepower to play in the subsea theater, a potential new market.
March 9, 2018, NEL announced its 64GBaud high-performance coherent DSP. The new DSP supports flexible line capacities enabled by variable baud rates and modulation formats. It can support 100Gbps/λ by 32GBaud QPSK and 600Gbps/λ by 64GBaud 64QAM. The DSP can support distances of tens of thousands of kilometers for potential subsea applications and support over 30Tbps for 120km applications. The new 64GBaud high-performance DSP is the second offering in NEL's third-generation CMOS coherent DSP family. It leverages Broadcom's cutting-edge 16nm CMOS finFET high-speed mixed-signal technologies. NEL is currently sampling the new DSP with its early access partners and customers, and is planning its production release in the second half of 2018.
The major optical system vendors pursue different strategies for sourcing coherent DSP technology. Some rely on full in-house supply, some a mix of in-house and merchant market supply, and some solely the merchant market. A strong, highly competitive merchant market accelerates innovation. A very strong merchant market can even swing the pendulum from in-house dominant to merchant dominant. Additionally, all vendors want a minimum of diverse supply and not be sole-source vulnerable. NEL is a key supplier in the ecosystem. The complexity of DSP development means that it is never "one and done." Rather, the market depends on suppliers with years and decades of multigeneration technology experience. NEL will carve out a significant position in the market with its 64GBaud–L-band coherent DSP.
On March 12, 2018, Infinera announced its Infinite Capacity Engine 5 (ICE5), an integrated 2.4T optical engine. ICE5 includes Infinera's fifth-generation photonic integrated circuit (PIC) and its latest FlexCoherent digital signal processor (DSP). The 2.4T is delivered via four wavelengths operating at 600G. The ICE5 solution operates at 66GBaud and utilizes 64QAM. ICE5-based system solutions will be announced later in 2018, for general availability in early 2019. The baud rate and modulation formats can be tuned to optimize around different capacity-reach scenarios. Layer 1 encryption support is built in. With ICE5, Infinera is targeting ICPs, CSPs, and mobile and cable operators.
Infinera has also introduced ICE6, which is expected to reach the market in product format 24 months after ICE5. ICE6 is a discussion about technologies of the future. For ICE6, technologies under consideration include 100GBaud, 800G wavelength, and constellation-shaping CS-1024QAM. Infinera advanced the position that integration becomes more critical at ever-higher speeds, as the inter-device interconnect paths must become shorter and more robust. ICE6 will have the DSP and the PIC co-designed and co-packaged into a tightly integrated module.
ICE4 achieved GA in 2017. ICE5 is targeted for late 2018–early 2019 and ICE6 24 months after that. Infinera is working toward shorter cycle times for the development of its optical engines to increase the pace of innovation. Infinera has a strong relationship with its ICP clients. ICE4 with L-band support and ICE5 when ready will play well into the demanding ICP category.
An Achilles' heel of subscale companies is an overreliance on one market or technology. If the one main market sours, smaller companies can be very vulnerable. The strategic answer is to diversify revenues from a regional market, technology mix, and buying cycle perspective. At its heart, Lumentum's acquisition of Oclaro is intended to strengthen the new combined company by diversifying revenue streams across technologies and markets. The larger company will be able to bring more R&D firepower to bear and increase the pace of its R&D cadence.
With a larger resource pool, Lumentum intends to support the telecom transport and transmission and datacom businesses, while ramping resources applied to the emerging and promising 3D sensing market. Lumentum sees a huge pipeline of 3D sensing opportunities and needs to rapidly redeploy design and fabrication resources to maximize its market position. The combined company will have transport strength from Lumentum's side and coherent transmission strength from Oclaro. The datacom market will be well served by Oclaro's strong DML, EML, and ASIC portfolio.
The Lumentum forecast of 5–6% gross margin improvement post acquisition, due to scale improvements, is music to the financial community's ears. The financial community was quick to positively endorse the merger. Lumentum has made a bold and positive move to strengthen the company for technology transitions to come.
Over the years, many vendors in the PON ecosystem have been hoping for convergence of EPON and GPON standards. The hopes for convergence at 10G PON were dashed, although there was some coming together around the optics. Then there was hope that the next next-gen standards would merge EPON and GPON together at 25G, or 50G, or 100G, as we discussed following OFC 2017. While OFC 2018 has just begun, there appears to be very little discussion around EPON and GPON convergence. Instead, WDM-PON has reappeared as a potential solution for 5G mobile fronthaul (MFH).
Earlier this year, China Telecom's research lab focusing on PON technologies sponsored a meeting with optical and equipment vendors to discuss the various PON technology options for support of 5G MFH. It appears that China Telecom is seriously evaluating the applicability of WDM PON for MFH. WDM PON is not new, it was considered for FTTH, but failed to materialize as a low-cost access solution. However, its disadvantages for FTTH are much less important for the less cost-sensitive but highly latency-sensitive MFH application. China Telecom's interest in WDM PON would be sufficient to reignite WDM PON given the massive amount of MFH equipment needed for extensive 5G networks. Rather than convergence, the commercial flavors of PON are likely to expand in the near future.
On March 7, 2018, Coriant announced the Coherent Multiservice Sled for the Coriant Groove G30. Coriant has optimized the platform for density, supporting 9.6Tbps of client and lineside capacity in a compact 1RU form factor. Coriant also has achieved a power consumption of 0.16 watts per gigabit. On the high-performance side, the new solution features 600Gbps per wavelength enabled by a 16nm coherent DSP running at 70GBaud. The DSP is programmable with variable Baud rates, FEC, and modulation format options.
Coriant is attacking the space and power challenge with gusto by leveraging its integrated photonics capabilities. Its solution squarely targets the requirements of the internet content provider (ICPs) community. Coriant has 45 Groove customers and its installed base is growing.
The programmable coherent DSP era is well underway. Programmable 1.0 could be consider the 100G–200G solutions that utilized 8QAM–16QAM. Programmable 2.0 coherent DSPs can support 100G up to 600G, utilizing very high GBaud rates. The ICPs are under tremendous bandwidth pressure, while fiber is constrained. The ICPs are highly motivated to deploy solutions that meet their demanding requirements. The programmable 2.0 DSPs will catalyze the purpose-built DCI sector and the market will see a notable upwards inflection point in early-2019. Coriant is well positioned for this market and technology transition.
Europe, Middle East & Africa team - +44 (0) 207 017 7700
Asia-Pacific team - +61 (0)3 960 16700
US team - +1 646 957 8935
Already an Ovum client? Login to the Knowledge Center now