Industry Spotlight: Infinera’s Robert Maher on Embedded Coherent Engines, Pluggables and More

July 24th, 2023 by · Leave a Comment

The optical industry’s job is never done, as there is always a need for more speed and bandwidth. I remember clearly when Infinera burst onto the scene in the mid-2000s with new Indium Phosphide technology and photonic integrated circuits to challenge the status quo. There’s been a lot of change since then, and plenty more still to come.  With us today to talk about Infinera’s viewpoint on where we are and where we are going in the world of optics and bandwidth is Robert Maher, CTO of Infinera’s Optical Modules and Coherent Solutions.

TR: What is your background and how did you land in the CTO slot at Infinera?

RM: I grew up in Dublin, Ireland, where I completed my bachelor’s degree in electronic engineering. At the end of my engineering degree, I felt like I needed to know a bit more, so I did a Ph.D. in optical communications. That was really the starting point, as I was completing my Ph.D., I was traveling to conferences like OFC, ECOC, and such. I would attend these events and remember when Infinera used to have a truck and they looked like a real leader in innovation in this space. After I finished my Ph.D., I did six years of research at University College London. I got to the point where the question was, should I be a lecturer, or should I move into the industry? Really, Infinera was the only company I considered, and so in 2016, I moved straight from London all the way to San Jose to work with Infinera. It’s been fantastic ever since.

TR: What does Infinera’s portfolio look like right now?

RM: ICE6 is our current embedded engine and has been on the market now for a couple of years.  It is very popular for metro, regional, long haul, and all the way to subsea applications. It was the first coherent engine ever to support symbol rates of over 100GBd, and today it has all these innovative features that enable us to provide the best-in-class performance at 800G, at the lowest cost and power per bit. It has set a lot of records in both the subsea and terrestrial spaces. The next generation of that embedded engine is really going to be ICE7, which is coming out next year. ICE7 will increase the symbol rates and the data rates with the sole purpose of driving down power and cost-per-bit. That’s what everybody ultimately cares about. ICE7 will be 1.2Tbps per channel, and then ICE8 with 2.4Tbps per channel. 

But of course, we’re now moving heavily into coherent pluggables. Our first fully vertically integrated 400G pluggable has been generally available since Q1 of this year. This is where we have XR optics, an innovation that Infinera has developed and is setting the groundwork for. Our 400G pluggable offers a small form factor (QSFP-DD and CFP2) and can serve traditional P2P transmission applications, like ZR and ZR+, but also opens a whole new landscape of point-to-multipoint. We see a whole host of applications at the edge of the network: talking to 5G radio towers, building capacity to business customers across PON networks, edge computing, or even looking at intra-data-center fabrics for AI clusters and things like that.  We’re the first product ever that can achieve coherent point-to-multipoint.

TR: Where does a coherent pluggable fit into a service provider’s actual deployment as compared to your traditional offerings?

RM: There are two different places. First, it can go straight into a switch router. Conventionally, if you have a chassis like ICE6, you have a router with grey optics going from the router down to the ICE6 line card. In this first case, which is the mode of operation used in DCI, you just take away the ICE6 line card and the grey optics and put the coherent pluggable directly in the router. The second application is to have a transport box that takes your pluggable. We develop and sell transport boxes that have a line card with two slots, and you can put your pluggable in there. It’s an intermediate option for customers who aren’t quite ready to go straight to using routers with coherent plugins. 

TR: What enables coherent pluggables to support point-to-multipoint?

RM: It’s a combination of new DSP technology, advanced photonic integration and software intelligence. You can have a DSP ASIC that supports this concept of digital carriers or digital channels, which we have used in ICE4 and ICE6, but purely in a point-to-point case to get better performance on a given line system. But with XR optics we can have these little channels which we generate in the digital domain. And we can send them to different places. So, if you have our 400G pluggable and you broadcast 16 digital subcarriers at 25G each, then you can have 16 25G endpoints where you can just sit there and listen and grab the subcarrier that’s destined for you. The beauty of this technology lies at the edge of the network that doesn’t use 400G or 800G, like a radio tower or dedicated service to a business customer. With a single pluggable able to talk to 16 endpoints, you take the aggregation equipment out of the network. In some places, it takes out real estate, and it avoids truck rolls. If you want to upgrade that endpoint from 25G today to 50G or 100G in the future, you just do it with software. What that means for the customer is a dramatic reduction in CAPEX and OPEX. Just as in any other part of the network, it all just comes down to cost and power consumption. And the closer you get to the edge of the network, the more critical that cost point becomes.

TR: In what way is this being standardized, and how does the OpenXR Forum fit into the picture?

RM: ZR is being standardized by the OIF. ZR+ is a small variation which went through the Open ZR+ forum. And then, of course, you have Open ROADM as well, which is another set of specifications. Typically, you’d have a single pluggable that would conform to all three of those standards with the same hardware. So that’s for interop transmission and typically just point-to-point. We put the OpenXR forum together to really drive point-to-multipoint standardization forward. It is not just focused on the hardware but also on the management aspect. We’ve seen pluggables offer huge benefits to our end customers, but one of the challenges has been management. For instance, IP over DWDM has been around for decades but has never really been that successful. Today, typically, you have an optical transport layer you manage, you have an IP transport layer you manage, and it all feeds off to an SDN controller. If you use a pluggable in a router, you lose the optical transport management, because you just have what the router tells you. If something goes wrong in the network, you may not be able to easily figure out what went wrong. A management specification is one of the first outputs the forum has created. This is something that we’re getting a lot of traction on because it is one of the biggest roadblocks to deploying coherent pluggables in the metro and long-haul from an operational standpoint.

Moving on from there, the next steps are obviously to standardize the optics and the DSP for point-to-multipoint. Infinera doesn’t want it to be a proprietary solution. We want it to be an open, disaggregated solution within an open ecosystem because that’s what our customers want.

TR: Open ecosystems are more popular now but are something vendors have not always been in favor of. How is that evolution going?

RM: It’s taken time and will take time. It started out with the optical line systems, which needed to be bookended with optical transport boxes from the same vendor. Customers in this fully locked-in situation find it difficult to onboard the latest technology, which is where they can get cost savings or power savings.  Infinera has fully embraced Open, both by opening our line system but also by deploying our transport equipment over open line systems. And I think with pluggables there’s a strong requirement in many places for interoperability and multiple vendors. I think the supply constraints we experienced during COVID, more than ever, kind of really highlighted the challenges with a single source, both for vendors making equipment and for customers deploying equipment. It has accelerated, and now I think the big vendors are pretty bought into open ecosystems for the pluggable space. The line cards are still proprietary, and I think they’ll stay proprietary for quite some time. 

TR: At what stage is the adoption of coherent pluggables in the wild today?

RM: I think 400ZR has seen huge volume already. 400ZR+ with 0dBm optical output power, where you really push the pluggables into the metro and regional space is just about to take off. I think pluggables will start to see a big movement into the metro and will become the dominant optical transport form factor in that space in the near future. The performance that Infinera can achieve with its 400G pluggable and our 800G pluggable to come, means that these technologies are going to be applicable to regional networks and even long haul. At OFC, we showed a 400G pluggable going over 2,400 kilometers of fiber — an enormous distance that covers most long-haul links in the United States and almost any in Europe. 

TR: Another thing we are seeing more of deployment of is L-band. What is driving that, and why now?

RM: Networks need more capacity, and one way of doubling the capacity of fibers is to use L-band. Most networks have been fiber rich until recently. We’ve supported L-band since our ICE4 product, which was released back in 2016. So, Infinera has had C-band and L-band line cards and systems for nearly 10 years. In that period, L-band deployments have been massively outweighed by C-band deployments, however we are starting to see more and more L-band deployments now with our current ICE6 technology. Most customers have C and L band ready line systems, so it is an easy way for them to upgrade capacity. Subsea networks have stayed completely with C-band, and I think it will stay that way because C-band is more power efficient in subsea. It is more power efficient for them to add another fiber than to do C+L.

But another advancement out there is Super C and Super L. The bandwidth of the C-band increased by 25% three times in the 2000’s, from 3.2THz to 4.8THz. The bandwidth has remained largely untouched for the past decade due to the emergence of coherent technologies. With coherent transmission becoming mature, fiber bandwidths will start to expand again. With Super-C, the bandwidth increases from 4.8THz to 6.1THz, and Super L will also offer somewhere in the region of 5.5THz to 6.1THz also.  So, you have close to 12THz in fiber bandwidth that can be exploited to increase the capacity of optical networks.

TR: What stood in the way of Super-C and Super-L before?

RM: It requires evolution in the line system infrastructure to enable it: wavelength selective switches, and amplification technologies. Once the optical line system is in place and deployed, the coherent transceivers can follow. From Infinera’s perspective, we own the design and development of our widely tunable lasers, which we can adapt to support Super-C and Super-L, but the line systems and the ROADMs of the world need to support it too, and I think that’s taking time. We’re starting to see that infrastructure in place now, and it’s something we support on our next-gen line system.

TR: What challenges lie ahead?

RM: You must continuously evolve the speed of your technology. ICE4 was a 32GBd system, which requires components to have an analog RF bandwidth of approximately 16GHz. At that time, that was difficult, but within 4 or 5 years, our ICE6 product was released at a symbol rate of 100GBd, a 3-fold increase in the speed of the RF electronics. Optics and electronics always need to support higher speeds, and that’s just going to continue in the future. You need to continuously innovate to support this, and it’s just as critical for the pluggables as it is for embedded transport solutions. So, you really need to ensure that your fundamental technology, your DSPs, your optics, your analog electronics can scale up to higher and higher speeds. And I think that’s where a lot of our focus is because driving speed is the best way to get to the lowest power and cost-per-bit. Our InP platform is particularly suited to high RF bandwidths, so we are confident that we can continue to scale in the future.

TR: How do you keep finding ways to drive that speed? 

RM: Vertical integration is absolutely critical as you go up in speed, which we’ve been talking about for as long as we’ve existed. But the reason why vertical integration is important is because it’s hard to make high-speed opto-electronics and ensure its robust with high performance. The only way you can get there is if you design every little element yourself and you know how each component operates to give you the best performance at the end. That becomes critically important as the speeds go up. I think that at the end of the day, the players that you’ll see in the pluggable space are going to be people who have the most vertical integration.

TR: Are there limits (like Shannon’s limit) that we are close to bumping up against?

RM: There are fundamental limits in DAC/ADC technology, but they’re quite a bit away. I think our Indium Phosphide platform is well-suited for these high speeds.  Silicon photonics came on the scene 10-15 years ago and made a big splash, but the modulator technology has met its fundamental limit or is soon to meet it and those vendors are looking for a new modulator technology to keep increasing speed.  With Indium Phosphide, we still have a nice runway in terms of the speeds that we can achieve. Shannon’s limit is more of a capacity thing: how much information you can send on a single fiber. I think we’ve really pushed the limits there with ICE6, but there’s still a way to go. I think we can still get 20%, 30% more capacity on the embedded engines as we go forward with ICE6 to ICE7 and ICE8. But clearly, a strong focus is power consumption, and cost is always a common denominator there.

TR: Thank you for talking with Telecom Ramblings!

If you haven't already, please take our Reader Survey! Just 3 questions to help us better understand who is reading Telecom Ramblings so we can serve you better!

Categories: Fiber optic cable · Industry Viewpoint · Telecom Equipment

Discuss this Post

Leave a Comment

You may Log In to post a comment, or fill in the form to post anonymously.

  • Ramblings’ Jobs

    Post a Job - Just $99/30days
  • Event Calendar