Huawei: 5G Is About Capacity, Not Speed

Paul Scanlan, CTO of the Huawei Carrier Business Group in Huawei Technologies, sat down with Semiconductor Engineering to talk about 5G, which use cases are attractive and why, and how that compares with previous wireless technologies.

SE: Where are you seeing 5G, and how do you see this rolling out both for sub-6GHz and millimeter wave?

Scanlan: 5G is a platform for transformation. The first obvious use case of 5G is it’s cheaper than 4G, 3G or 2G. If you have growth in a data market — and we have growth in every single country in the world in data, and the number is around 25% to 30% — 5G will always be cheaper than 4G or anything else. The reason is capacity. It’s not about speed. It’s not about latency. It’s about capacity, and 5G has 20 to 30 times more capacity than with 4G. If the equipment and hardware costs are similar number, obviously it’s better. Second use case is that it’s more efficient. If you keep building 4G, 3G or 2G networks and there’s growth, then you will double the carbon footprint for the mobile network. If you build 5G and you build even the worst case scenario, which is the slowest deployment, carbon emissions will more or less stay flat from now till 2025 and then start to drop. And I was on a webinar just a couple of weeks ago, and one of the comments was, ‘I don’t understand how that could possibly be. You need to have more sites and more densification for certain applications.’ Yes, but just think about the demand.

SE: One challenge has been trying to harness the millimeter wave signals commercially because they attenuate so quickly and are easily disrupted. Where do you see mmWave working?

Scanlan: So first, it’s cheaper. Second, it fixes your OpEx problem and it’s energy-efficient. You want to use the most efficient technology to keep the greenhouse gases under control as much as you can. The third part about the platform is it gives you opportunities for different deployment models. So before, a typical 2G, 3G rack was 1.8 to 2.4 meters tall. In a WiFi router size you have a couple of choices. You can have a 4T4R, which is similar to a 4G, but you have 100MHz of spectrum and you can use the new coding and framing and modulation of the 5G, so you get a better result. Or, with millimeter wave, you can have a 16T16R and get something about the size of a hard-cover book. If you take a conventional 64T64R in C band, which Huawei has, it weighs 25 kilos. Over here in China, we’re not thinking about deploying them on big towers. We’re putting them on lamp posts, integrating them into something that looks aesthetically pleasing. We’re putting them where the people are, where the transport is, where the traffic lights are, where the power is. All of those things are far more economic and purpose-driven than finding a site on a building or on a block of land, building a 70-meter tower, and radiating as much as you possibly can. That’s not what it’s all about.

SE: So how do you address all of the other B2B business models?

Scanlan: That’s the big opportunity for 5G. But in millimeter wave, you have a challenge. You’ve only got a couple hundred meters. You don’t want it to go through buildings because it doesn’t go through buildings. You don’t want it to go over 3 kilometers because it doesn’t go over 3 kilometers. So you don’t build it that way. The dominant traffic — 80% of all traffic — is indoors. And what do we use indoors? WiFi offload. There is no security and haphazard connectivity. You put in a 16T16R and you have 1 gig of bandwidth per operator at 16T16R. Or you’ve got 20 gigs-plus per person if you really need it. The scenario is hotspots in the right place — in hospitals, schools, inside manufacturing plants. That’s where you might want to consider things like millimeter wave. Deploying millimeter wave outdoors, unless you’re providing it all the way along every transport artery, makes very, very little sense. I think the U.S. was trying to do the right thing, but given it didn’t have C band spectrum to go and offer millimeter wave as competition to cable, it didn’t make a lot of sense. When the benefits for green and for efficiency and for OpEx reductions are not there, the opportunity is in transforming the industry, and that’s where the U.S. should have used its millimeter wave products.

SE: So you’re looking at a fixed-point distribution type of approach, rather than trying to follow around a typical mobile device user, right?

Scanlan: Correct: In rural America, where Huawei has a lot of equipment installed, our 6TR4 4G is running 5 miles to rural communities and delivering 647 megabits per second. You’re probably not getting 647 megabits per second from anybody. And that’s not even 5G. It’s 64 MIMO with a TDD (time division duplex) spectrum of 2.3 or 2.5 — you can even do it in 2.6. The key is spectrum, spectrum, spectrum. It’s the right spectrum allocated efficiently and effectively, and incentivizing operators to deploy across the whole geography and whole industries. That’s what you need to think of. Regulators often don’t necessarily have all the information at hand about what the latest products can really do or the efficiencies of telecom operating companies. Government wants industry transformation and economic growth. What does the telecom operator want? EBITDA. Profit, quarter-on-quarter earnings. If you look at health care and manufacturing in the U.K., all four operators are building 5G ‘here, here, here and here.’ So we did an overlay with all the manufacturers ‘there, there, there and there.’ They’re not in the same place. You pay $1.4 billion for spectrum, and you add another $0.5 billion for equipment and deployment to achieve an SLA that is 95% coverage of people. If you take a connected ambulance, it’s a great idea for saving lives. So you need throughput to an ambulance with low latency, and then great medical kit, analytics, and advanced predictive things so that when they get to the hospital everything’s on standby, ready to go. There’s all of this great technology. But you still get dropped calls, even on the main roads. So now you’re expecting that telcos to provide guaranteed coverage for this essential service to save lives? That’s not in the terms of the spectrum license. For the platform transformation, you really have to talk about facts and understand what 5G really does.

SE: How about the idea of distributed clouds?

Scanlan: I always call cloud a cloud, because the concept is that you put something somewhere, and somebody else can utilize it. So I post an app, you can use it anywhere. I offer you a service, and you can pick it up somewhere else. It’s not just a wrapper around a data center. But everybody talks about the MEC — multi-access edge computing. That’s a component that is required in order to achieve a very, very low latency — sub-1 millisecond. Huawei does about 0.3 milliseconds latency. With 5G, the only way you’re going to achieve that is if you’ve got some form of computing power very close to the edge. When you start to deploy things like this, and you want to put compute and storage closer, you’re trying to provide, for example, campus-to-campus, intra-campus and inter-campus based solutions. A lot of that traffic is local. So you may be providing patient information, because a lot of that tends to be more local. And then you’re trying to provide some computation to something that’s very latency-sensitive or delay-sensitive, like a robot, or perhaps cars. Cars are not essential, although connected and autonomous cars are quite different. And the use case for how 5G should or shouldn’t be used for the connected car is changing.

SE: It’s the same with robotics, right?

Scanlan: If you’ve ever had a robot walk toward you, the first thing you do is step back because you have 120 kilos of metal coming at you. And if you shake the hand of an articulated robot, what’s going to stop it crushing from your hand? Latency has to be very small. Now do I make a $1 million robot, which I can’t sell, or do I make a $1,000 robot and have all the compute somewhere else. To do that, I’ve got to have very low latency. Those sorts of interactive things can be done more quickly. So you want some of those sorts of things in this data center, which we’ll call an MEC. It’s just part of the core network that is provided closer to the radio. Every country will be different because of the deployment models. Every country will be different because of the amount of fiber they’ve got connecting the data centers, and the architecture, and the distances. If you’re trying to run a data center from New York and you want low-latency applications in California, you need to put a data center in California. In a small country, do they have 1 data center or 20 of these little tiny things?

SE: What impact will this have on smart phones?

Scanlan: The next smartphone just needs to be glass and a bit of battery. You don’t need any computation, provided the connectivity is there. The challenge we have is to stop everybody talking rubbish about 5G and to understand how to deploy it and where is its real economic benefit per industry? How do we then extract that? Telecom companies generally build a tower and they sell SIM cards. 5G came along three years ago and they couldn’t see what the business model would be. They said it’s just going to be faster. But suddenly they find it’s got a lot more capacity. It’s actually cheaper from an OpEx perspective and cost per bit per megahertz.

SE: How is this working out in China?

Scanlan: I have a 5G phone, and I’ve had it since the end of last year. I suddenly realized that something was different about it. The operator didn’t tell me they had just enabled 5G. I just noticed, especially when I was sending messages, that it was lightning fast. And when I was looking at certain sites in China, I decided to do a test. I’m getting 300MHz and it’s 5G. A smart operator used analytics to see what my behavior was, looked up my pricing package and said, ‘This guy we’ve got to keep. Keep the experience fantastic and he’s not going to go anywhere.” It’s very different thinking. The deployment models in China are automatic. When we were building hospitals, we were deploying full 5G networks in in 72 hours.

SE: So when you are deploying 5G, does it make a difference whether you’re deploying sub-6GHz or millimeter wave?

Scanlan: That requires a little bit of architectural thinking. Because the cell size is smaller, you have to consider things like speed, like if the if the objects are moving too fast. To get seamless millimeter wave, you need an awful lot of those cells, or you’re going to provide them in a hotspot. If you put a half-dozen 16 x 16 millimeter wave cells in a shopping mall or high-density traffic area, then the architecture really comes down to the transport layer. Everything becomes backhaul, backhaul, backhaul. That becomes your biggest challenge. If you’ve got more sites, because of the millimeter wave spectrum that you’re using, then you need to think about redundancy and loops and rings, and things like that. If sites go down, and you got very high traffic, a large portion of your traffic is out. So if you look at reticulation, how do you get there? That’s why the deployment model for millimeter wave along roadways would be fine. You can pick every few lamp posts, give one to AT&T, one to Verizon, one to Sprint, and you just keep going on like this. Fiber is already there. You’ve got the ducts. In most developed countries, there are ducts running through all the walls and lamp posts. There’s power there already — more power than you need. This is not a complicated scenario. And OpEx is the biggest cost for a telecom company. It’s 62% to 67%. OpEx, not CapEx. CapEx is in the range of 12%. A site is a rental.

SE: What do you see as necessary get to autonomous driving?

Scanlan: If you want autonomy, then you need to change the infrastructure. And you need to change the regulatory policy. We need all the infrastructure connected,. That means all the traffic lights, all the roadways —or at least those where the carriage way will only be autonomous — and then an operator has to be given the mandate to monetize it. Otherwise they will not build it. So if every car is going to give them $1 or $5 per month, what’s the incentive for a telephone company to have this very critical service running? The jury’s still out on autonomous driving, but it’s being piloted. It can work in some countries, on some roads, under some scenarios. Today, if you want to run millimeter wave-based autonomous driving, you would be crazy. But if you were running it between 700 to 800 MHz, maybe even 1,800 MHz, because there’s a lot of 1,800 infrastructure around the world — so you bring everything down to 4G and 1,800 — then it’s straightforward. And with 5G, that can operate in every frequency you’ve got.

SE: What about massive MIMO?

Scanlan: Massive MIMO is a different technology that’s been added and touted, but it’s not the fundamentals of 5G. We’ve been running massive MIMO on our radio for the equivalent fixed wireless access for 7 to 10 years already. We deploy it in 23 MHz, 25 MHz, even the old WiMax spectrum of 3.5 GHz. That’s why our products were really heavily integrated in the antenna technology into the filters.

SE: So putting those pieces together, how do you see all this unfolding for automotive?

Scanlan: There’s a couple of very innovative companies looking at a different type of transport, where the motors need to be close to the wheels. You don’t really necessarily need all that luggage space. And if vehicles don’t crash, you don’t need metal, so cars can be made of plastic. And if you think about the future, it’s not just about my car looking at the car in front. It’s my car being able to see through the car in front, and through the car in front of that one. That’s where 5G fits in, and it’s why you need a network-based technology as opposed to a WiFi sort of thing. The second reason why you might like 5G is if the vehicle could be done in a different way — basically a box on wheels, with people interacting inside it. You can carry on doing all the things you’re doing — and perhaps the windows are not glass necessarily, but they are multimedia — so with the throughput you’re getting, plus lower latency, you could have a lot of things inside the car that are connected. It becomes like a mobile home or a mobile office.

SE: Then who owns the cabin? Is it a BMW or an Apple?

Scanlan: If you think about what’s inside the car of the future, it’s wheels, drive train, motors, servers, storage, batteries, a power management system, mobile phone, IoT. None of that is made today by the car companies. That’s why Huawei partners with some of the automotive companies. The way you get to autonomous cars is you take people out of the driver’s seat. But now you’ve still got a problem, and this is where the opportunity really comes for 5G. Things like AI don’t work if you don’t have data. It’s the collection, storage, manipulation and analysis of data that really creates new use cases. I don’t need traffic lights. I walk out the door, walk down 100 meters, which is where the number of pedestrians exceeds the number of cars, and then you stop the cars and the pedestrians will go. You know that can work because you know where all the cars are coming and going, and where the people are coming and going.

SE: Where else does 5G fit in? And where doesn’t it work?

Scanlan: A telecom operator deploys infrastructure where the people are. That’s not necessarily where the farms are, or the tractors are, which is what we were discussing earlier. So how do they monetize it? Is it $1 a drone or $1 per tractor? There are a lot of good case studies that have been done in Europe. Take dairy farming, as an example. If you have lots of good data, then you know the right time for a cow to be milked to get the maximum production. We utilize data for fish farming in in the northern part of Norway, as well, where we’re putting high-definition cameras together with feeding and everything else. We’re integrating a lot of things to be able to detect when an animal is feeding or isn’t feeding to cut down on the waste. In the case of fish, we’re detecting lice and parasites. In the case of animals, we’re looking for diseases predominantly in order to improve yield. If you can do this autonomously, it’s faster and cheaper. In Norway, I call it facial recognition for fish, because they’re using cameras to actually identify them. The same concept could be extended into other types of agriculture. Some of those concepts are being used for fertilization of fields and application of pesticides, versus using crop dusters where the stuff goes all over the place in unknown quantities. With small drones, for example, the drone can be a delivery system for this stuff, but it’s also easy to have a camera and identify a whole bunch of other things. So you get this robot that can do more than just go along and pluck an apple or drop a bit of fertilizer or pollinate something. It becomes a lot more integrated.

SE: What does this require from a technology standpoint?

Scanlan: You need the antenna facing the orchard or the farm, and not where the people are. If you’ve got a farm that’s 10,000 acres, you need to think twice about using millimeter wave. If it’s 700 MHz, you’ve got a much better option. And remember, it’s not necessarily about the high throughput. With 700 MHz, you’ve got about 45MHz per operator. 5G is efficient when it’s 100, when there’s no guardband, and when you deploy massive MIMO, but you’re not going to get all the performance benefits you might think you’ll get.

SE: Can you scale down millimeter wave?

Scanlan: No, because you build radios for certain band. So for example, it’s fairly straightforward to offer everything for the U.S. spectrum of 1,700 to 3,500. You can package all that up in one box. It can cover all of that, and the antenna and everything won’t make it ridiculously large. We’ve actually already done this. The box is a little bit bigger, but it is 64 MIMO both ways, and it is offering 5G with all the frequencies inside. But remember, you can’t don’t get something for nothing. If the bandwidth only 20 MHz, you’re only offering 4 x 4 MIMO, so it’s only about 34% difference between 4G and 5G.

SE: One prominent use case for 5G is security. What are you seeing?

Scanlan: The reason you would want to use something like 5G is if you can put a 5G chip on the back of a charge-coupled device in a camera. So the concept of a connected high-definition camera requires a lot of bandwidth. And the reason you would think about security being a very, very good option — in fact, after efficiency and carbon, cameras are a good use case — is because of the uplink. The downlink is covered because the telco networks have been engineered 10 to 1. But this user-generated content is huge, and 4G doesn’t handle that very well. We can’t have higher speeds anymore, so we’re limited to about 75 Mbps, or maybe 150 Mbps. If you deploy a different option, you can have a couple of Gbps with 5G. But I also can slice the network and give security a guaranteed component. So think of a hospital where you can give patient monitoring its own slice. With 4G, you may have someone playing a game and chewing up all the bandwidth. But with 5G you can use slicing. You can do the same for security, and that’s why 5G is far more secure than 4G.

SE: But with all these connections, is the attack surface bigger?

Scanlan: No, because all those things become a lot more secure compared with what you’ve got today with IoT connectivity. And there’s so many standards from Zigbee Z-Wave, Sigfox LoRa, and then narrowband IoT, which is fitting under the 3GPP under LTE-M to give this this broad connection of Internet of Things, whether it’s low-speed or high-speed.

Related Material
From Cloud To Cloudlets
Why the intersection of 5G and the edge is driving a new compute model.
Winners And Losers At The Edge
No company owns this market yet — and won’t for a very long time.
5G Brings New Testing Challenges
Millimeter-wave and beamforming capabilities present the biggest testing challenges.
The Growing Challenges Of 5G Reliability
Rapid changes in next-gen wireless technology and standards are only adding to the complexity.
New Challenges In Testing 5G Devices
Complex hurdles in characterizing performance and optimizing signals.
What’s After 5G
The path to 6G will require some radical changes to both infrastructure and use models.
An Inside Look At Testing’s Leading Edge
FormFactor’s CEO peels back the covers on AI, 5G and HBM test issues.