It’s become increasingly impossible to talk about spectrum policy without getting into the fight over whether 5G is a miracle technology that will end poverty, war and disease or an evil marketing scam by wireless carriers to extort concessions in exchange for magic beans. Mind you, most people never talk about spectrum policy at all — so they are spared this problem in the first place. But with T-Mobile and Sprint now invoking 5G as a central reason to let them merge, it’s important for people to understand precisely what 5G actually does. Unfortunately, when you ask most people in Policyland what 5G actually does and how it works, the discussion looks a lot like the discussion in Hitchhikers Guide To the Galaxy where Deep Thought announces that the answer to Life the Universe and Everything is “42.”
So while not an engineer, I have spent the last two weeks or so doing a deep dive on what, exactly does 5G actually do — with a particular emphasis on the recently released 3GPP standard (Release 15) that everyone is celebrating as the first real industry standard for 5G. My conclusion is that while the Emperor is not naked, that is one Hell of a skimpy thong he’s got on.
More precisely, the bunch of different things that people talk about when they say “5G”: millimeter wave spectrum, network slicing, and something called (I am not making this up) “flexible numerology” are real. They represent improvements in existing wireless technology that will enhance overall efficiency and thus add capacity to the network (and also reduce latency). But, as a number of the more serious commentators (such as Dave Burstien over here) have pointed out, we can already do these things using existing LTE (plain old 4G). Given the timetable for development and deployment of new 5G network technology, it will be at least 5 years before we see more than incremental improvement in function and performance.
Put another way, it would be like calling the adoption of a new version of Wi-Fi “5G Wi-Fi.” (Which I am totally going to do from now on, btw, because why not?)
I elaborate more below . . .
There are a bunch of important questions to keep in mind when evaluating what we ought to do about 5G as a policy question. (a) What exactly is 5G? (b) How does it compare to existing LTE? and, (c) How much are we being asked to pay for it in policy terms?
What Exactly Do We Mean By “G?”
“G” technically means “generation.” My favorite explanation can be found in this old Best Buy commercial. As a general rule, we use “G” to indicate a significant shift in capability, architecture and technology. For example, the shift from analog to digital voice in 2G, or the inclusion of limited data capability as an overlay to voice in 3G. The shift to 4G was marked by a shift to an all packet-switched data network in which voice is supported as one feature on the network. In addition, 4G turned out to be fairly homogenous for a variety of reasons I won’t get into now. Basically, after a brief flirtation by Sprint and a few others with WiMax, all the carriers ended up using LTE.
So the switch to 5G ought to mean a major boost in both technology and speed. And it will, eventually. But for now, it’s not so much a generational shift like the previous shifts but a modest transition over time. By that I don’t mean simply that we will see 5G networks operating with 4G cores for a long time. That’s always true. Carriers deployed LTE and still maintained (some to this day) 3G networks in parallel. That is necessary so that people and businesses can switch legacy equipment at a rational pace. What I mean is that the capabilities that are supposed to make 5G so awesome are not really that awesome right now, and won’t be for at least 5 more years.
What Makes 5G More Awesome?
Here is where it gets confusing. You can see a good tutorial on the network architecture here. But this represents a relatively recent change in how we talk about 5G. Originally, i.e., back in 2015, we were talking about millimeter wave as 5G, with nothing else going on in the lower frequencies counting as 5G.
O.K. What’s Millimeter Wave?
If you’ve been following 5G mania for the last several years, the big thing in 5G has been “millimeter wave” (“mmWave”). Millimeter wave uses very high frequencies (above 20 GHz) to get very high throughput (greater than 10 gigabits per second in the lab, according to this piece). This is possible because mmWave creates big blocks of contiguous spectrum necessary for transmitting lots of data. (Additionally, as this real short IEEE piece explains, no one else has built anything out in this space yet, so it is easier to develop.) Unfortunately, the higher up the frequency chart you go, the more energy it takes to push signal further without degradation and the more easily the waves deflect off objects or get absorbed by objects. It makes a good imprecise rule of thumb that you can get more information in the signal the higher up you go, but the more difficult it gets to propagate the signal for mobile use. Technology helps fix that, and you have some companies like Starry Internet that use the fact that the waves bounce off objects to send signal significant distances without line of sight (LOS=line of sight; NLOS=non-line of sight; impressive your friends with your technobabble!).
Anyhow, until about a year or so ago, pretty much what everyone meant when they said “5G” was mmWave for gigabit wireless. That’s what the FCC “Spectrum Frontiers” proceeding is all about, and it’s why AT&T and Verizon spent many billions of dollars buying up legacy licensees in the 24 GHz and 39 GHz bands. The benefits of this technology are fairly straightforward — lots more faster better wireless broadband. The problem is that because of the physical properties of this spectrum, it requires (generally, for the most part) lots and lots of small cells bunched together (an architecture called “desnsification”). It also requires lots and lots of fiber, because a network is only as fast as its component parts, and you can’t have gigabit speeds if your backhaul is through copper. So you need nice big, thick fiber pipes to connect to your small cells to deliver gigabit capacity to your gigabit wireless last mile.
This has many possible uses, but primarily in urban areas. It can work with mobile stuff outside where small cells can reach, but not inside because the mmWaves don’t penetrate walls or windows. You can use it in urban areas to reach buildings where the landlord has an exclusive deal with the local cable company by sticking an antenna in the window, and try to offer a competing high-speed broadband service. But it doesn’t do much once you get out of areas where densification is not economically viable — like rural areas. Still the prospect of gigabit wireless has been enough to get everyone all excited and talk about how to preempt state and local governments so that carriers can rush out their small cell deployments (small being a relative term — these are still about the size of refrigerator) and lay down more fiber.
Another problem for mmWave is that the technology and business case are not really here yet. Nor is it likely to get here within the next few years. While carriers like to talk about their pilot projects, we are still many years away from seeing this technology widely deployed. Mind you, this requires advance work to make it happen, but if this were all there was we could focus on the mmWave stuff and not worry about things like the T-Mobile/Sprint deal, which does not involve any mmWave whatsoever.
So What Else Is 5G Other Than mmWave?
The big thing in 5G is hetnets.
O.K., I’ll Bite. What Are Hetnets?
One of the things that makes packet-switched networks like the Internet so useful is you can use them for a lot of different things and across a lot of different technologies. This means you can put together a bunch of different networks together and manage it as one network, or divide up capacity in a network and treat the different pieces as if they are separate. We call this sort of mixed network “hetnets,” short for “heterogenous” networks. (Heterogenous means composed of non-similar parts, if you were wondering.)
Mobile carriers have a lot of different blocks of spectrum in lots of different bands (a product of our allocating spectrum by auction over time). Until recently, the big challenge for carriers was how to take all the blocks of spectrum they have and use them to provide mobile voice and broadband But now, we have lots of other uses for wireless networks — such as internet of things (IoT). We have lots of different devices communicating with each other, which means a network has to support a very different kind of traffic than mobile voice/data. Whereas mobile voice/data is nomadic and bursty (meaning people wander around and go through periods of intense use followed by low use or no use). IoT, especially mobile IoT such as autonomous vehicles, operate differently. Often they may have tolerance for latency, but they might also be very latency sensitive. Additionally, portions of these networks may run fiber networks or WiFi networks and will need to work seamlessly with the pieces on the wireless networks.
The recent 3GPP “5G” releases, Release 15, are designed to make managing hetnets easier. First, they allow for network “slicing.” I.e., they make it easy to “slice” off network capacity to manage as a separate network. Second, Release 15 makes it easier to manage larger numbers of devices across multiple types of networks and frequency ranges. These two aspects combined make it easier and ore efficient to manage hetnets and operate IoT networks with lots of devices.
Finally, there’s massive MIMO. Massive MIMO isn’t specifically a 5G technology, but it gets tossed into the general 5G salad. For purposes of this blog post, all you need to know is that most wireless networks currently use MIMO, and that massive MIMO, as the name implies, means a significant boost in capacity at the cell site.
When T-Mobile and Sprint talk about “deploying 5G,” they are primarily talking about all this non-millimeter wave stuff. Mind you, T-Mobile and Sprint claim they will use densification for Sprint’s existing 2.5 GHz spectrum in the same way that AT&T and VZ will use mmWave. Sprint has large blocks of spectrum in the 2.5 GHz band. For years, Sprint has struggled with the fact that 2.5 GHz is sufficiently high-frequency that it has lousy penetration characteristics compared to the rest of the spectrum used for mobile. As a result, despite having big blocks for throughput, Sprint has suffered because it needs to build more cell sites to cover the same geographic area that other carriers cover with their lower-band spectrum. But now, Sprint can take advantage of the densification technology (whenever it becomes standardized and more generally available), but with spectrum that travels further and penetrates better than mmWave.
All of That Sounds Pretty Good. So Why So Down on 5G?
The question isn’t simply what does 5G do, but what it does that the current generation of wireless tech — LTE — doesn’t do. Other than mmWave (which we don’t have a standard for and which nobody really uses yet), LTE already does pretty much everything that the expanded definition of 5G does. Additionally, when looking at the T-Mobile/Sprint deal, the question isn’t simply whether merging offers some advantages. The question is whether the advantages to consumers offset the harms of going from 4 national competitors to 3. If Sprint and T-Mobile aren’t getting that huge a boost in capacity and capability as a result of the merger, and they could do it without merging, then why let them merge and have competition take the hit.
And this is where we get to the 5G Emperor being a bit more than naked but not nearly as awesome as the hype makes out.
First and foremost, mmWave band, the stuff that most people focus on with 5G, is many years away. The new 3GPP 5G standard doesn’t really address the mmWave stuff, except that it allows networks to incorporate mmWave into their hetnets.* That’s good in the abstract, but doesn’t do anything to bring us the gigabit wireless capacity the industry raves about and has all the policy folks in a lather about making sure we beat South Korea and Europe to deployment.
[Clarifying note, since I’ve had some folks check me on this. Release 15 does, indeed, operate on mmWave bands. But the additional technology to support densification and multi-gbps is still in pilot stage. Verizon is the only carrier with a serious deployment schedule for mmWave, and while I appreciate that they are generally first to deploy next generation technology (they were first major carrier to do FTTH with FIOS, first major carrier to deploy LTE) it always takes longer. All of which I tried to shorthand by saying that Release 15 doesn’t really, on its own, address the mmWave densifcation.]
Take away the mmWave stuff, and what you have looks like a modest, incremental improvement over existing LTE 4G technology. Industry reporter Dave Burstien has done a bunch of comparisons between existing LTE technology and incoming 5G in the non-mmWave bands. Burstien argues that existing LTE will drop latency down to 2 miliseconds or less by 2019, which is within the range promised by 5G over the next 5 years. Burstien also argues that 5G represents only a 25%-505 improvement in speed and capacity over existing LTE, which is a reasonable incremental improvement but hardly a game changer in the way that shifting from 3G to 4G was (which produced an approximately 400x increase in speed and capacity). All in all, Burstien argues that there is simply no material difference between existing advanced 4G LTE and 5G for the next 5 years or so. By which time we should be ready to move on to 6G Terrahertz beaming.
Other commentators who dig into the technology in the same thorough way as Burstien are a bit more charitable, but only a bit. Scott Marcus, a former FCC staffer now doing most of his policy work in Europe, argues that the big advantage of 5G is in the ability to customize the network for specific needs with each deployment. Marcus makes it clear that every generation has a cycle of hype followed by a cycle of coming back down to reality, and that “every generation has indeed been better, faster, and cheaper than the previous generation, but not to the degree initially hoped for.” (Italics in original.) Nevertheless, as Marcus points out, “the gap between hope and reality is even greater than usual” with regard to 5G. One does not have to conclude, as Fred Goldstein does, that 5G is “a spectrum eating monster that destroys competition” to argue that 5G is being way oversold by carriers as a means of pushing for policy goodies.
T-Mobile and Sprint acknowledge many of these limitiations in their Application to the FCC. The Applicants acknowledge that it will take 5 years to achieve the capacity they spend the bulk of the Application claiming justifies the merger. Digging into the attachments, the declaration of T-Mobile CTO Neville Ray puts an actual set of numbers on the percentage improvement — an increase of about 18% to 52% total capacity improvement. That’s in line with Burstien’s prediction as consistent with improvements in LTE rather than a radical gamechanger.
But what about comparison charts like this one that show a more dramatic bump from LTE advanced to 5G? Going from a theoretical top speed of 3 gbps for LTE to theoretical top speed of 10 gbps is nothing to sneeze at. Even discounting for the theoretical part, that still looks very impressive. Unfortunately, the emphasis here is on the “looks.”
One of the reasons to discount dramatic increase in the short term for 5G as compared to LTE is that massive MIMO works for LTE as easily as it does for any changes in protocol we are calling 5G. As massive MIMO accounts for a big chunk of the gain in capacity in the existing LTE spectrum, crediting this increase in capacity to 5Gis not a fair comparison. You get virtually the same increase by going from MIMO to massive MIMO whether you use 3GPP Release 14 (advanced LTE) or 3GPP Release 15 (5G NR).
Another factor that bumps up the theoretical speed estimates for 5G is the increased channel size from mmWave. But again, you can get dramatically increased throughput for contiguous blocks of spectrum for LTE. 5G has a faster base speed in large part because it starts with an assumption of a larger channel size. Additionally, the limitations on mmWave discussed above mean that this theoretical top speed should be discounted quite a bit when trying to determine the overall average increase in speed for anything but fixed, point-to-point LOS on a clear, non-humid day (water fade degrades signal in the higher frequency bands) (you can find a really old FCC paper explaining the propagation issues here).
That leaves flexible numerology, the trick that allows many more devices to connect to the network without significantly increasing latency, as the most important element of first generation 5G not found in LTE. that’s not nothing, but it isn’t exactly much to jump up and down about.
What About Rural? Does This Do Anything for Rural?
On it’s own, not really. Nothing about 5G changes the basic problem that the return on investment in rural areas is lousy because it has lower population density and usually more challenging terrain. As noted above, mmWave uses high frequency stuff that doesn’t propagate very well. Normally you might try solving this by allowing higher power for point-to-point in rural areas. Unfortunately, if you increase the power significantly to push signal further for this specific set of frequencies, you start to fry things (the signals get partially absorbed by living tissue rather than passing harmlessly through (which happens at lower frequencies), and the energy converts into heat). So mmWave is unlikely to help solve the big problem in rural areas of how to cover large geographic areas cheaply. Low-band spectrum does have much better propagation characteristics, but as discussed above the low-band uses of 5G don’t increase capacity that much more than existing LTE on the same frequencies.
So Is 5G A Total Scam?
Depends who you ask. I don’t think it’s a total scam, but I do think we are seeing the usual problem of fast-talking industry lobbyists stampeding lawmakers and regulators. (Insert Simpsons “Monorail Song” here.)
The wireless industry has a long history of whipping Policyland into a sense of crisis to get what it wants. Remember the Great Spectrum Crunch that obsessed Washington in the late 00s and early teens? While it was certainly the case that increased demand for wireless service meant finding more wireless capacity, it wasn’t exactly the crisis everyone claimed (at least not to judge by spectrum valuations in the broadcast incentive auction, which fell far short of the anticipated “crisis” pricing). Carriers faced with capacity challenges did what carriers have traditionally done when pushed to their limits. They found engineering solutions that allowed them to make more efficient use of their spectrum and meet increased demand.
Similarly, 5G offers some real improvements in wireless networks in the long term. But carriers have been working out their deployment plans based on market and technical realities, not based on some hypothetical race with Asian or European carriers to capture the 5G flag first. Even before the proposed merger with T-Mobile, Sprint announced plans to deploy massive MIMO in its 2.5 GHz band spectrum in multiple cities in 2018 and 2019. Verizon is already in early deployment in its mmWave spectrum. Companies like DISH and Ligado have plans to deploy IoT networks. As long as we make a reasonable amount of mmWave spectrum available via auction to encourage competitive deployment (since AT&T and VZ have most of the existing legacy licenses in the relevant bands), we don’t need to panic about losing out to China or S. Korea or Europe or whoever.
(I can’t help but notice that when it is something industry opposes, like government investment in fiber, we have no worries about falling behind even though countries like S. Korea that invest in fiber networks have had cheaper and faster wireline broadband for years. But when it comes to stuff the industry actually wants, like goodies for 5G, suddenly S. Korea deploying a 5G network means the end of American exceptionalism.)
Finally, keep in mind that the wireless market depends on global markets for economies of scale. Once carriers agree on a standard, it will get cranked out and deployed on a global basis. We moved out LTE in 700 MHz in 2009-10. The rest of the world did not take that long to catch up. Even if some other country did pull ahead in deployment briefly, we wouldn’t suddenly find ourselves technologically irrelevant.
5G actually will do some real things. Notably, the mmWave technology will dramatically increase speed and capacity of wireless networks, both fixed and mobile. But that will mostly impact urban areas where carriers can affordably run fiber and dense-pack small cells. Generally, 5G will allow existing low-band and mid-band spectrum networks to operate more efficiently and support more devices, increasing overall capacity and making wide-area IoT networks possible.
But the broad range of functionalities across multiple spectrum bands is not going to be the sort of magic game changer that industry keeps claiming it will be, and we won’t be reduced to technological irrelevance if some other country deploys something it calls a “5G network” before we do. We need to make policy decisions based on a clear-eyed cost/benefit analysis, not based on some industry-induced panic that if we don’t preempt local zoning laws, open spectrum, and allow every merger and acquisition right now then America will no longer be great again.
Stay tuned . . .