[00:00:12] Hello everyone and welcome back to the Cisco Optics Podcast where we talk about pluggable optics for networks. In the pluggable optics industry, every next generation is led by countless presentations and proposals at standards bodies meetings along with lots of lab development and testing behind the scenes. Right now, the industry has its sights set on 3.2T transceivers with 400Gb per electrical lane. All those meetings and all that testing are happening as we speak.
[00:00:39] Today, I am joined by John Calvin, Strategic Planner and Datacom Technology Lead at Keysight. John has been bridging the measurement science gaps of emerging telecom and datacom development efforts for 20 years. He has chaired multiple physical layer interoperability working groups and is currently serving as contributing member to IEEE 802.3, OIF-CEI, InfiniBand and PCIE development efforts.
[00:01:04] John holds a BSEE from Washington State University and his graduate level studies are in signal processing from Stanford University. In episode 65, we continue our conversation with John. We get into the tradeoffs between IEEE standards, MSAs, interoperability, and the race toward 3.2T Ethernet. And now join me as I talk with John Calvin.
[00:01:34] They can move quickly, they're smaller, and they're much more nimble. So that's why they exist. Is there a downside to being that nimble? Well, yeah. There is. The downside is that you wind up doing the same work five or six times. I mean, you could have MSA trying to do active copper cables, solving the same problem that the LPO people solved.
[00:01:58] And unless you have cross-pollination of people across these different multi-source agreements, you wind up reinventing the wheel. Or you encounter the same problem. And you'll find maybe novel ways to solve it. But you lose the benefit of a large team of people, which is what you get with IEEE, all trying to crack the same nut. And IEEE is very effective in that.
[00:02:23] But it's very systematic and not the fastest organization on the planet. So the risk with the smaller groups trying to get together to make their own, like you said, bespoke standards is that you'll have multiple groups trying to solve the same problem. They come up with different approaches to solve the same problem. And then you end up with still these many different groups.
[00:02:45] Whereas if you have a more concerted effort amongst the broader community, you'll end up with one solution. And then you get the benefit of a single standard where the supply chain ecosystem can benefit from having a single standard? The supply chain can focus on fewer alternative product definitions, number one.
[00:03:14] And it also solves interoperability problem better as well. If you have people focused on a bigger scope of interoperability, you can assure that the retime modules out there will interact with the low-power optics modules and vice versa. And that takes a degree of cooperation and collaboration across the MSAs and other SDOs or these are our standards organizations.
[00:03:44] So the other thing and then the other motivation behind MSAs is that they're increasingly today, there are geopolitical boundaries being put in place or force fields being put in place. You know, right now we're meeting up in Canada and IEEE tends to alternate between U.S. destinations and foreign destinations, be it Canada, Europe, or Asia.
[00:04:11] And they do that because, you know, many IEEE contributors can't travel to the United States, can't get visas to the United States. In fact, due to some of the regulations that are in place, we're given strict guidance that we can't even talk to these people unless it's within the confines of a standards organization.
[00:04:35] So the standards groups pull people together across the globe and overcomes a lot of these geopolitical boundaries, which is good, right? But once you break away and go into an MSA, they go back to their own country and they set up their own technology. And to be honest with you, there is a race on foot. There are people trying to define the next generation of 400 gig interface in the race to 3.2T. And it may very well not be from the United States.
[00:05:03] There may be others that reach that goal first. Okay. All right. Well, okay. So you mentioned 400 gig per lane before in 3.2T. I'm curious. So, I mean, you're in the thick of it right now at the meeting in Montreal. What are like some of the really difficult problems to solve that you're facing? Oh, boy.
[00:05:27] Well, you know, every time you double the bit rate, you take up more space in the channel, right? So you progress them.
[00:05:44] Remember, 802.3 CK was a 53 gigabaud PAM4 signal, which had a functional speed of 106 gigabits per second. And it operated at a fundamental, a signaling fundamental, down below around 53 gigahertz. As we advance to higher speeds, the Nyquist frequency is getting pushed up more.
[00:06:13] So as you push the Nyquist frequency up to 106 gigahertz, which is what's being discussed presently for 802.3 DV, which is potentially going to be, and I say potentially very carefully here, a 200 gigabaud PAM4 signal would have a Nyquist at 106 gigahertz.
[00:06:37] You push that frequency to the right, and the channel loss falls off pretty like a straight line, right? So where you had approximately 25 dB of loss at 53 gigahertz, at 106 gigahertz, you're now going to be at 46 dB of loss.
[00:06:59] And when you're talking about the channel loss, you mean the channel being like PCB traces or electrical interconnects? Ideally, yes. Increasingly, PCBs are being left behind in the dust.
[00:07:15] But we see, you know, TwinX-based or coaxial-based interfaces, various technologies that break off very close to the CERDEs and go to the MCBs in the back of an I.O. system. There's a couple technologies that are flyover cables, or there's a couple other technologies that have their own name.
[00:07:42] But the idea is to use TwinX cables, which have lower loss, and you don't have the issues associated with trying to go through a via. Or at least you minimize the via contribution in these sorts of signaling schemes because, to be honest with you, you know, vias are great filters, right? They filter everything, you know, above around 70 gigahertz. And we need to get signal content up around 106 gigahertz.
[00:08:10] So moving vias out of the pictures is one of the benefits of going to a TwinX-based transmission system. But are there people who are not so crazy about TwinX from the perspective, I'm thinking manufacturing, basically. Yeah, yeah. How is that perspective? Well, you know, there are people who are experimenting with high-end, you know, homogeneous board materials with high dielectrics.
[00:08:40] It's really expensive to put a server-class system, though, on a high-end board material. So really, you're better off. I mean, you've got tons of power that has to be delivered. You've got tons. You have it physically and structurally. Putting stuff on a fiberglass board still makes sense because you need to hold things in space. You need to, in these days, you have to route liquid cooling to everything and have to power delivery.
[00:09:10] So power is a big portion of that, and that's naturally all done through a fiberglass-based PCB. But when it comes to the actual signaling differential pairs, they're finding new ways to break the signals out on the top of the package or adjacent to the package. And this is where we've seen some of the new terms come into the vocabulary, like co-packaged copper or near-packaged optics or co-packaged optics.
[00:09:38] The whole idea is to reduce that length between the CERTES and the switch chip, which is, you know, this massive 50-millimeter chip, and the actual connection points that are part of your peripheral I.O. system.
[00:09:55] And this is a trade-off because, you know, that twin-axe that we're talking about is going to have 25, 27 dB of loss in it. And that's where the heat generation is coming from. There's acute sensitivity to lower power consumption. And those transmission lines are just resistors, right?
[00:10:20] You're losing the energy in those transmission paths, and it's going up as heat or radiated energy. And that's really the biggest problem that people are trying to solve. And they're doing solutions by making those interconnects shorter and shorter and shorter into the, you know, dozens of millimeters sort of distances now.
[00:10:43] And that's where the idea of putting, you know, optical engines right next to or on top of the actual switch chip is an attractive alternative for many of the people trying to find out how to consume less power and how to make the system's latency lower and basically still maintaining their I.O. scalability.
[00:11:07] So are flyover twin-axe cables an alternative to a CPO or NPO or CPC approach? Or are they put together in a single solution? Or are they solving the same problem or addressing the same problem? Well, you know, they address the same problem.
[00:11:31] But there's a unique aspect to some peculiar market dynamics going on here. I mean, there are companies out there that have CPO-based switches right now. They've had them for a couple of generations. They're not terribly popular because if you're a hyperscaler and you're going to buy somebody's CPO-based system today,
[00:11:55] that means you are captive to their ecosystem for modules and transceivers that might plug in there. And hyperscalers don't want that. They want to have their cake and eat it too. They want to be able to take their switch chip and then go out there and use market forces to get the optical transceiver vendors to give them the best deal. And that's a dynamic that's simply very, very strong in this business.
[00:12:23] So if you think about standard I.O. system where you have a row of OSFPs or QSFP connectors that are on the back of a switch, and you might have 16 or 32 sockets back there to plug in a module. And then on the inboard side of that, you'd have TwinAX going down to the PCB. That gives you, in effect, the best of a couple of worlds here.
[00:12:52] You have an electrical channel that's very well-tuned, doesn't have a lot of loss in it. And you can actually connect still to a module, a modular-based interface. And you can swap your modules out as you need. If you want to plug in a passive copper cable or a coherent module or a DR8 module, you can do that. If you want to swap vendors, you can do that too. So that kind of flexibility is really golden in this business.
[00:13:18] And the closer you get everything integrated onto the chip, the fewer options the end users have. And that's one of the force fields that's pushing back on some of these integrations. We really need a standardized form of this tight integration of optics next to the chips. And that's taking place right now. But it's going very, very slow. So what are some of the roadblocks that you've been up against recently?
[00:13:47] And we're talking about 400 gig per lane, right? Well, you know, our real roadblocks are ones that we have overcome. That was the second part of my conversation with John Calvin.
[00:14:12] Next time, we'll get into the signal integrity and connector challenges engineers face when pushing Ethernet to 400 gigabits per lane. Please follow the Cisco Podcast Network on Spotify or Apple Podcasts. Better yet, leave a review on Apple Podcasts. Remember, you're listening to the Cisco Optics Podcast, which is part of the Cisco Podcast Network, where you can find other great Cisco podcasts too. If you want to listen to us on a website, go to optics.podcastpage.io.
[00:14:42] We also have educational videos on YouTube. Just go to youtube.com and search on Cisco Optics. Thank you for listening. This is Pat Chow, Director of Technical Marketing Engineering at Cisco Optics. The next episode is part three of my conversation with John Calvin. See you there.