Steven Cherry Hi, this is Steven Cherry, for Radio Spectrum.
Has there been any technology more widely talked about and yet still less understood than 5G?
Let’s step back a bit. In the mid-1970s, European telecom researchers came up with the GSM standard. By 1991 the first GSM calls were being made.
GSM was the first real cellular standard. Different phone calls could use the same frequency channel because GSM allocated the different signals into different time slots.
In the 1980s, researchers in the U.S. and elsewhere thought they could see limitations in the number of calls in that approach, and thought about alternatives, notably, what came to be called code-division multiple access. In a code division scheme, each call is encoded in a random or pseudo-random sequencing, and receivers at the other end are equipped to decode their own calls.
By the time this CDMA encoding came to be commercialized in the U.S., GSM was firmly established in Europe and, soon, Asia. Qualcomm and other companies persisted, in the belief they had the better technology. This schism, between time-division and code-division based phones, persisted until the fourth generation of phones; a scheme called LTE unified the two standards. I think my guest today would agree that it’s fair to say that LTE essentially uses code-division, wrapped around a bunch of other programming that makes it backward-compatible with GSM.
LTE, I think it’s also fair to say, was the last standard that was easily understood. It’s essentially a single air interface with, if not homogeneity at the base stations or the cellular devices, something like it. 5G will presumably have the same interoperability and compatibility, but across a wide variety of technologies.
The other enormous difference from earlier cellular standards is the number of frequencies in the radio spectrum that need to be managed. Each country freed up spectrum for cellular use different ways, and so LTE needed to range among something like 20 different frequency bands. 5G devices will operate across something like 10,000 bands, placing an enormous burden on antenna design and signal management—a burden that, unsurprisingly, is being met with AI.
Back in November we had a conversation with Shivendra Panwar of NYU about 5G and latency, a topic we’ll revisit here. And in December we had a preview with industry analyst Mark Gibson about the then-upcoming FCC auctions. So yet again, a podcast called Radio Spectrum is going to talk about the radio spectrum. We have with us one of the most qualified people on the planet for that task. John Smee holds dozens and dozens of patents in wireless technologies and is a Vice President of Engineering at Qualcomm. His designs and innovations started with CDMA and then LTE, and now 5G, as well as Wi-Fi.
John, welcome to the podcast.
John Smee Thank you very much, Steven.
Steven Cherry John, first, is there anything in that backstory you want to add or amend before we dove into 5G?
John Smee No, I think it‘s a good introduction to the cellular ecosystem and where we are today.
Steven Cherry Good. Well, John, one of the reasons 5G is so complicated is the use of millimeter wave bands. First, there‘s a lot more spectrum with millimeter waves to manage and the millimeter waves themselves are much harder to manage. One of Qualcomm‘s strategies is to tightly integrate a phone‘s modems. It‘s radio frequency, front end and to some extent, even the antennas themselves. Maybe you could say a word about each of these components in a cell phone and then what it was like to engineer this kind of integration.
John Smee One of the ways to look at it is millimeter is something ... it kind of brings fiber-like speeds to wireless. And to do that well, it‘s because, as you mentioned, there‘s wider bandwidth. And when we looked into designing a millimeter-wave system to have efficient system performance, we also used beamforming. So not only is the base station beamforming the signal to the user, but the device itself is also doing some of this tracking of these beams and forming these beams on the transmit side and the receive side.
And because of this approach to what we call beam-based mobility and the operation in millimeter wave, it does require and benefit from a tight integration across the antennas, through the RF front end and into the baseband processing. And so it‘s no surprise that millimeter wave is opening up huge multi-gigabit-per-second data rates, given the bandwidth itself, can approach a gigahertz. It‘s also something that when you‘re designing these systems and you look algorithmically at the trade-offs, then you have to recognize that what you‘re doing in baseband is also leveraging information that‘s come via the full transmit-receive data path. And that‘s one of the reasons that this optimization end-to-end through the antenna and RF subsystem is an important part of making millimeter wave power efficient. And in also delivering on these promises of super high data rates in an energy-efficient way.
Steven Cherry I want to get to the energy efficiency soon, but one of the big challenges in 5G is latency. As I noted, we had a podcast largely devoted to that. One problem is that millimeter waves don‘t pass through solid objects nearly as well as waves in the frequency bands we had been using for cellular. In fact, they can be stopped by the body itself. So paradoxically, even though 5G speeds—that is to say, data rates—are much higher, there could be added latency as a cellular network has to bounce a signal around those obstructions.
John Smee The latency from the propagation part is actually not a meaningful adder to the end-to-end system latency. So when we look at trying to deliver latency, whether it‘s ping times or end-to-end operation, the system is able to work with the beam management to turn some of these reflections into desired signal. So the part of the communication between the devices and the network, and even as the device is moving through the network and doing handoffs and other things, it‘s able to operate at incredibly fast speeds.
One of the interesting parts, when you look at 5G ... You know, we as engineers in particular for IEEE Spectrum, we often think in time-domain and frequency-domain, and we‘re able to move between those two frameworks. And so just as the system is a wideband system, it also is something that‘s narrowly divided in time into these slots or slot formats. And within a slot there‘s multiple OFDM symbols. And so the system is able to operate at high enough speeds that it‘s able to do things quickly enough, that it‘s able to form and adapt these beams, such that it can track the propagation.
And that‘s one of the things when we look at that—per your kind of introduction of the G‘s from the 2G and the 3G and 4G and now into 5G—what‘s amazing is the capabilities we have, whether it‘s silicon or RF or processing times. So putting all those things together is exactly a challenge that has been met with 5G millimeter wave. And so the system is literally able to create and form these beams fast enough that it can exploit this propagation. And the fact that, yes, different buildings have different reflection and different surfaces, whether it‘s reflectivity or blockages and every time something is blocked, it also creates a reflection. So that‘s the other side of the blockage part is that the signal is often being reflected and then harnessing those reflections is core to the millimeter wave overall system operation.
Steven Cherry Those obstructions that I mentioned can be, as I said, our own bodies and even our hand as we hold the phone, which can block parts of the antenna. That was even a famous problem a decade ago with the iPhone 4‘s antenna. I gather that this is a problem Qualcomm has been working on—and is solving it with AI?
John Smee That‘s right. So we apply intelligence in many different aspects of algorithms. And obviously, engineers who used to optimize things by hand would say that they were pretty intelligent as well. So the difference now is that when we look at adaptive algorithms and those algorithms‘ ability to basically self-optimize, as long as they have the right sort of KPIs, they‘re driving to.
And so when we look, for example, at a device that might have multiple antenna modules, we take what‘s called a spherical approach. So imagine it‘s like an orange or an apple. You‘re looking around the surface of the device and we focus on ways of using those different antennas and the different modules such that we‘re able to achieve a very good level of coverage in almost any direction for almost any phone orientation. And so the number of degrees of freedom for ways in which you can combine signals to create the appropriate performance is one of the areas where millimeter wave is an example of something that‘s substantially different from prior generations before it. And so it is something that we were researching for many years in our research labs and now obviously very happy to have in our products. And I have a twenty-eight gigahertz and thirty-nine gigahertz device in my hand as I‘m talking to you. Those are those optimizations and improvements that have gone into the 5G device.
Steven Cherry And this is all becoming ever more important with what‘s now, I think, a Qualcomm mantra—that is, video is the new phone call. It‘s also important for some other applications of 5G, notably the Internet of Things.
John Smee Yes. So video is something where the amount of data transferred both on downlink and uplink ... and I think now more than ever with people wanting to use video calls connecting with their colleagues, their friends, their family. And so being able to efficiently transport video end-to-end is one of the core underpinnings of 5G mobile broadband. And it‘s also something that when I look at what we‘re doing on our devices today. ... So when we go back to the early days of flip phones, the first smartphones, as they showed up later, it‘s amazing to think what we do today on cellular that earlier we would not have wanted to do on cell yet; we would have been concerned about our data plan; we would have been concerned about the effects of having a video call, whereas nowadays, whether it‘s streaming downlink movies. Participating in a live video call with a colleague or friend transporting video efficiently is one of the most important parts in these early 5G use cases.
At the same time, what makes 5G more expansive than prior Gs is that very early into its design and standardization, there was a focus on expanding into new industries and so bringing up IoT—Internet of things. We focus that on several aspects, such as the industrial IoT compared to Consumer IoT, or even how we can transform things like agriculture or logistics or transportation. So 5G really does have a wide level of applicability across these different industries. And that‘s where we see a lot of different companies working together to look at what‘s the best way that technology can solve a real need and add value.
Steven Cherry I‘m glad you mentioned agriculture. Cellular technology has in some very important ways closed the digital divide or I should say digital divides, because we need to think not just about the developed world versus the developing world—so, for example, farmers in China using their phones to sell into entirely new markets—but also urban versus rural, and not just in places like China, but right here in the U..S.
John Smee Yeah, the connectivity of 5G is something that can work across these geographies. So when we look at low-, mid-, and high-spectrum, we also look at the applicability to connecting the unconnected. So the fact that cellular now is in some sense replacing what used to be a necessary aspect of having wired connectivity to your home. And we know many people are benefiting from the fact that they have cellular connectivity and that that‘s able to bridge this last mile. So things like fixed wireless access, which previously was more of a niche market, is now something that‘s front and center in the 5G discussions. So when we look at the notion of being able to have that connectivity to homes, small businesses, rural communities, they can leverage from the fact that 5G is delivering these fiber-like speeds over wireless. And so that ability to leverage a network not only for cars driving by in terms of mobility scenarios, but then also providing that connectivity to homes and businesses. That‘s a great example of that unified 5G investment that pays dividends across a range of scenarios.
Steven Cherry I guess I don‘t want us to underestimate the problem. I mean, I can drive from Pittsburgh to Albany and encounter vast stretches where there‘s no connectivity at all, right on interstate highways—even going from New York City to Albany, you can lose your phone call right on the New York State Thruway.
Really anywhere in the US outside a major town, whether it‘s Woodstock or Wyoming, you can be lucky to have 3G. It seems paradoxical that 5G, a much more expensive set of technologies to implement, is going to solve that.
John Smee Well, 5G is going to—because it can operate across all these different frequency bands and even things like spectrum sharing—enable the upgrade of 4G frequencies into 5G frequencies.
We will see this continual transformation of the rollout of 5G into these other applications and scenarios. So there‘s a wide range of operators involved. If we look at the Competitive Carriers Association, a large number of carriers are also involved in providing connectivity in these different regions.
Steven Cherry One further solution to the rural-versus-urban digital divide is what‘s called fixed wireless connectivity, which I gather Qualcomm thinks is going to be pretty important in 5G. Maybe you could explain how that works.
John Smee With fixed wireless access ... so instead of just having a cell phone, talking to a base station, we can actually have what we call a CPE or “consumer premises equipment” is the old telco word for it.
And so what it means is the ability to have, for example, a fixed antenna and system could be on the outside of your home or small business. And it‘s able to communicate with a base station that could be hundreds of meters or kilometers farther away. And that ability to close the link over wireless with fixed wireless access, you can then leverage the fact that you can have Wi-Fi connectivity, for example, throughout the home. And so the ability to enable a larger number of people to have a competitive alternative to, for example, DSL or cable in the United States. The data rates of 5G make that very compelling to do on wireless, whereas historically that would have been more challenging to support such a large amount of capacity.
Steven Cherry Yeah, you know, there‘s a saying that what‘s old is new again. And we‘ve seen that as mainframes came and went only to see network computing rise in the nineteen nineties, only to submerge and reemerge as cloud computing ... Fifteen years ago, Spectrum had a feature story about fixed wireless in Malaysia, which at the at the time used a version of sort of 3.5G cellular.
John Smee I agree with the point that, as you say, what‘s all this new again. And it‘s something where these dichotomies that get challenged—what sort of information flows over wires, what sort of information flows over wireless, or as you say, even in the computing context, what is done on the device versus what is done in the cloud. So there are some examples where, you know, there‘s more of a hybrid emerging where, for example, we have the edge cloud showing up and the edge cloud is very important because it enables some of these low latency applications to be delivered towards these users.
So you‘re able to have the information that‘s processed at the edge cloud in communication with these devices at the edge themselves versus always having to go back all the way to the central cloud.
Or similarly, when we looked at the overall communications and computing paradigm, then we look at scenarios like industrial IoT. “We have to be very careful about reliability. Better put that over industrial Ethernet.” Well, one of the exciting things about 5G is the fact that things like Time-Sensitive networking, those industrial end to end systems are now incorporated into wireless. So we have the 5GCIA [5g-acia.org] an industry body that‘s focused, for example, on enabling 5G for industrial applications. And Qualcomm‘s researching very heavily, bringing that all the way through design, standardization and evolving [connec] scenarios to look at 5G in industrial IoT scenarios to do things that used to only happen on wires.
So it‘s interesting where you can have a scenario like fixed wireless access where you‘re changing in some sense, the business model—of the price points and who you‘re able to get serviced from—to a scenario like a factory or smart hospital where you‘re recognizing that the global volume of cellular and as those new applications are opened up, then all of a sudden there‘s the opportunity to serve another segment of the market in a way that they‘re definitely appreciating the fact that it‘s a mobile scenario. So whether it‘s an AGV [Automated Guided Vehicles] robot in a factory or the fact that you‘re even just reconfiguring your factory more frequently and you don‘t want to be reconfirming all of your Ethernet connectivity while you‘re doing that.
So I absolutely think it‘s an exciting scenario to be examining these foundational technology shifts of cloud and edge and computing that‘s either local or that‘s central. And the same thing with artificial intelligence. Absolutely there‘s going to be significant amounts of AI happening in the cloud. But I would say what‘s more exciting is the fact that we now have power-efficient AI on the device. So how we train devices, how the devices are able to locally process information—whether it‘s because of latency or security concerns—they can do processing, actually an inference at the device. And so that‘s a good example where as the applications are evolving, then how we exchange information and communications is also evolving. So that‘s why we view 5G and AI as quite interrelated because it goes into the fact that cloud processing edge cloud processing and device processing are also changing across use cases.
Steven Cherry There‘s a hope that new phones will have better battery life. And here again, there seems to be almost a paradox. 5G can easily use more battery life handling those millimeter waves, but you have some expectation that the modem-RF integration will help phones conserve power.
John Smee Yeah, modem-RF integration can help conserve power because it‘s about intelligently doing things efficiently. So if you‘re doing a high data rate communication, in many instances you‘re getting a large amount of data very quickly. And then it‘s a question of how quickly you‘re able to power off various components and then to power them up again just at the right time. So the way in which the system operates—that is when are you turning on and off various components and modules and sub modules—that‘s a key advantage to how you can recognize to get the high data rates and at the same time be in as low a power mode as much as possible whenever you can. So it‘s this fact that you have more controllability because you have a more intelligent modem-RF subsystem, but at the same time, you‘re able to do things more quickly and more adaptively and even predictively. And that‘s the part where AI comes in, to making sure these algorithms are operating as efficiently as they can while still delivering the actual KPIs that the user needs.
Steven Cherry We talked about time division versus code division, which puts a lot of pressure to get the coding scheme to be as efficient as possible. Two decades ago, researchers developed something called Turbo Codes, which got us remarkably close to the so-called Shannon Limit, named for the founder of Information Theory, Claude Shannon, and his idea that there was a theoretical limit to how much information could travel across a communication channel.
Now there‘s new research into coding schemes, including something called Polar Codes. How far along are we in new coding schemes? And weren‘t we already pretty close to the limit? Is there even room for dramatic improvement again?
John Smee We were relatively close to the limit with Turbo Codes. But what‘s interesting, when you ask in information theory, theorists, about how close they are to the Shannon limit, it really also comes down to, well, what scenario you‘re talking about. What code block size are you talking about? What level of reliability are you talking about? And one interesting thing, the asymptotic results for which Claude Shannon was famous, they‘ll be results, say, that are true with infinite blocklength, which means it takes infinite latency as in it takes a very long time to successfully decode this hypothetical signal.
So in a pragmatic sense, if you recall the excitement when Turbo Codes were discovered, it was about this parallel in serial concatenation of convolutional codes. And at the same time, something called low-density parity checker—LDPC—codes was also something that became very important for 5G, because it could enable incredibly high data rates very close to the Shannon limit. But in a way that was parallel, parallelizable in hardware. That is, it could have an efficient hardware instantiation so that you could decode gigabits per second with an LDPC code and still be very close to Shannon limits. So LDPC had advantages, for example, over Turbo Codes, depending on the code block size. And also as you start accounting for implementation complexity.
When we look at smaller code block size, for example, for the control channel, that‘s where Polar Codes have applicability. So Polar Codes are something that can deliver efficient link-level performance for a control channel. And so that‘s a small amount of information that‘s telling us, what data rate am I sending you? Are you getting served in this particular time slot? So you want to be able to quickly decode the code block for the control channel, but it‘s a substantially different communication signal than the data channel. The data channel will be doing gigabits per second with retransmissions and hybrid ARQ [automatic repeat request], whereas the control channel, it‘s replacing something that‘s sending a small amount of information quickly to the device.
And so what‘s exciting about the field of channel coding and its intersection with 4G and now 5G and even our longer-term research already on 6G is that it‘s about how you bring together the mathematical improvements with the pragmatic aspects of, you need to implement it not only on base stations, but also on a range of devices. And whether it‘s a premium to your smartphone or a low power IoT wearable, we want to make sure that the technologies that get selected in the standard can be this wide range of use cases. So one of the important parts about channel coding and standardization is the fact that the system is scalable. And so that‘s one of the other aspects where with LDPC codes, we can scale it across code block sizes, across data rates, and even how we implement things like hybrid ARQ where we selectively retransmit information to hit those super-high-reliability targets.
Steven Cherry John, one thing that I think is clear about cellular is that there‘s always going to be another G. The boundaries are becoming less distinct. What‘s coming up not just next, but down the road? And is it fair to say that that will involve almost obliterating the distinction between the base stations and the endpoints? When will we call it 6G?
John Smee When we look at the notion of what‘s happening where, and these end point, I think that‘s a very good question because it forces this analysis of do we have a false dichotomy versus when do we have a richer ecosystem and also a word we call topology. So not only can phones talk to base stations, but phones can talk to phones and cars can talk to cars and base stations, can talk to base stations. And so some of that‘s already possible at 5G, where we have integrated access and backhaul in Release 16, enabling base stations to talk to each other as well as to devices. We also have things like cellular-V2X [vehicle-to-everything], where communication between vehicles can happen in low latency, high-reliability scenarios, leveraging 5G for a side link channel.
And so the notion of what‘s happening where and the transformation of the network and even the transformation of the devices and the devices participating actively in the network, I would say yes to your question that it does represent this evolution and I wouldn‘t necessarily use the word “obliteration,” but it‘s something that ... It‘s an ever-changing fabric of communications. And so this connectivity between the devices and the range of types of devices does start blurring those differences between the network side and the device side.
Steven Cherry Well, John, the thing I admire most about engineering. Is that you guys can look at something in nature, for example, millimeter waves and say, “there‘s a potential benefit here” and engineers will try and it may even take decades to figure out. I remember writing about ultrawideband in 2005, which, by the way, quoted a researcher as saying it was a same-room technology, but at least maybe we could use it for Wi-Fi.
What you and your colleagues at Qualcomm, and around the industry, have done and are continuing to do is truly remarkable. And we‘ve barely even touched on all the benefits it will bring. Thanks for all that hard work and thanks for joining us today.
John Smee Thanks a lot, Steven. My pleasure.
We’ve been speaking with John Smee, a senior engineer at Qualcomm about the difficult but highly rewarding development of technologies surrounding 5G cellular communications.
Radio Spectrum is brought to you by IEEE Spectrum, the member magazine of the Institute of Electrical and Electronic Engineers, a professional organization dedicated to advancing technology for the benefit of humanity.
This interview was recorded February 8, 2021 via Microsoft Teams. Our theme music is by Chad Crouch.
You can subscribe to Radio Spectrum on Spotify, Apple, Google, Stitcher, and wherever else you get your podcasts, or listen on the Spectrum website, where transcripts are also available. We welcome your feedback on the web or in social media.
For Radio Spectrum, I’m Steven Cherry.
Note: Transcripts are created for the convenience of our readers and listeners. The authoritative record of IEEE Spectrum’s audio programming is the audio version.
We welcome your comments on Twitter (@RadioSpectrum1 and @IEEESpectrum) and Facebook.