In the previous blog in our 5G series, we introduced the 5G usage scenarios and the acronym onslaught (eMBB, URLLC, mMTC) that makes it all more confusing than it should be. In this article, we’re diving into the wireless and radio frequency (RF) characteristics of 5G, which have enormous practical considerations that go beyond industry hype and marketing terms.
We also did a jargon white-belt intro blog, so hopefully you remember the two key acronyms: RAN and NR. The radio access network (RAN) is like a Wi-Fi access point and its antennas, and new radio (NR) is basically the brand for 5G RAN. Regarding the RAN, in this article, we’ll focus on key areas related to the radio layer, spectrum, and some comparisons with other wireless technology.
5G’s three usage scenarios (eMBB, URLLC, mMTC) are made possible in part by the variety of frequency bands that 5G puts to use. There are three separate swaths of spectrum for NR, informally known as low band (<1 GHz), mid band (1-6 GHz), and high band (>6 GHz). Historically, cellular focused on low band and mid band, so the high band itself is entirely new for mobile networks in 5G. As depicted in Figure 1, RF folks will recognize that each of these three bands have their own RF characteristics, with corresponding benefits and shortcomings.
Low band frequencies are great for wide-area coverage, which is well-suited for the blanket of macro cellular service that we rely on for mobile voice. This RF band has excellent range and penetrates through foliage and building materials quite well. Good coverage translates into more bars of service at longer ranges from the tower, and more green color on carrier coverage maps. That’s the good news.
But, the bad news is that low band also has scant bandwidth, which is the amount of spectrum, measured in hertz. Everything in wireless hinges on bandwidth. It is responsible for faster device connections and higher capacity; it’s analogous to more lanes on the highway.
In the end, low band is a one-lane highway, and there’s no room for more road. Low band doesn’t play much of a role in the ultra high-speed 5G data fairytale. Nonetheless, it still serves as a respectable overlay frequency for some data, as well as voice and IoT (the mMTC usage scenario), where many “things” need to connect at long range, but don’t need much bandwidth or throughput.
Meanwhile, mid bands are like Messi on Argentina’s World Cup team—they do most of the heavy lifting for broadband, but they can’t win all alone. Mid band is the real-world workhorse for LTE and 5G data. In all wireless endeavors, it’s common to think of maximum coverage as a good thing; in reality, coverage follows the Goldilocks principle—meaning, there’s such a thing as too little, and such a thing as too much. Too little coverage leaves a gap in service. Too much coverage prevents the same spectrum from being reused down the street, so there’s less aggregate capacity. Mid band (where Wi-Fi also operates) strikes that ideal mixture of coverage and capacity, enabling densification and driving net capacity.
Finally, we get to the most controversial—and new to cellular with NR—high band. High band has the exact opposite characteristics of low band: excellent bandwidth and raw throughput, but short range and weak obstacle penetration. If you squint hard and focus on the strengths of high band, and add warm water, you get a rich and foamy lather that’s great for marketing. You can print impressive maximum speeds on the box.
But the devil is in the details. And honestly, they’re not details, they’re basic properties of physics. High bands have very short wavelengths, measured in millimeters, which is why this spectrum (from 30-300 GHz) is known as millimeter wave (mmWave). Because of its radio characteristics, the effective range of mmWave spectrum is very short. Signals in this frequency band experience tremendous loss when passing through objects like foliage, building walls, windows, and even air.
On the flip side, mmWave is a largely untapped spectrum resource with few incumbent users in nearly every geography. The regulatory hurdles for its use are also low (ish), and there are contiguous bands of substantial spectrum available, which drive capacity and high speeds. In that light, mmWave has the potential to drive fast 5G connections where costs and deployment practicalities are manageable.
But that’s the problem. Because of its RF coverage challenges, mmWave radio deployments must be ridiculously dense (like every 500 feet). That kind of density requirement raises cost and implementation challenges because of the equipment costs, site acquisition, backhaul readiness, transport costs, and more. Urban and metro areas are a better fit because of user density, but with specific challenges. Rural mmWave is a complete non-starter, and the ROI in the suburbs is questionable too. But some of 5G’s world-beating claims may depend on it.
The map in Figure 2 should illustrate the point. This map displays the 5G coverage of a leading carrier in my area. The pervasive red coverage is standard low or mid band 5G service. The deep red shades show mmWave coverage, which is so special it deserves its own brand name, like 5G+ or 5G Ultra Wideband. Notice how mmWave coverage follows the street contours quite linearly? This picture tells the story of mmWave propagation physics better than all my inept words. Buildings and trees and air eat mmWave signals like bacon-wrapped jalapeño poppers. Beware the hype.
(Extend me the grace of a disclaimer. My comments are by no means a slight against this mobile carrier. Their 5G investment to date is aggressive and admirable. Nonetheless, physics are physics.)
In its multi-band approach, 5G is a collection of radio solutions used as a composite. No single frequency range has all the goods for all the usage scenarios, so customers and operators mix together solutions to solve a breadth of interests. In case it’s not clear, 5G is not mmWave. mmWave is not 5G.
One more term to add to your lexicon is massive MIMO (mMIMO). MIMO stands for multiple-input, multiple-output. Think of it as having many mouths and many ears, each with their own dedicated brain. Weird, I know. With more mouths, you can speak to multiple people at the same time (analogous to multi-user MIMO), or possibly talk in a way that’s clearer to your listener if they’re further away (analogous to beamforming). There are similar signal enhancement benefits on the receive side when you have multiple ears. MIMO has been around for many years, in both cellular and Wi-Fi, and it radically changes the reliability and capacity of wireless systems.
Massive MIMO is an exponential increase in the number of simultaneous transceivers doing MIMO magic. Cellular MIMO designs commonly use 4 or 8 radio transceivers (you might see this referenced as 4T4R or 8T8R, which indicates the number of transmitters [T] and receivers [R]). With massive MIMO, you could have 128-antenna (64T64R) systems. Due to its sophisticated signal processing complexity and physical size, mMIMO only makes sense as an infrastructure-side technology. As the technology matures, massive MIMO has a lot of potential to multiply the capacity of wireless spectrum and make better use of the available spectrum.
Speaking of spectrum, if you’re following cellular news, you’ll see headlines every week about auctions and spectrum, which make the wireless world go round. Regulatory leadership groups across the globe are trying to find underutilized bands of spectrum and reclaim them. By doing so, they generate revenue from auction funds, and also make the spectrum available to operators for 5G rollouts. Technology and connectivity enable the economy in a variety of ways, which benefits the country and its citizenry. Most spectrum auctions have strict legal contracts requiring the auction winner(s) to roll out services in a defined minimum time frame. You’ll see a tremendous amount of worldwide interest in C-Band (part of mid band) auctions, as well as mmWave activity. Unfortunately, there’s not that much low band spectrum to repurpose, so less activity there.
It’s worth noting that the first phases of auctions and spectrum gymnastics have primarily been focused on licensed use for mobile operators. A few countries, including the US, UK, Germany, France, Japan, Netherlands, and Finland (and probably others) have made some shared or lightly-licensed spectrum (like CBRS) available in mid-bands to fuel localized LTE/5G deployments by operators or enterprises that can’t (or don’t want to) afford the expensive spectrum. If you’re focused on enterprise network rollouts and evaluating LTE/5G in one of those geographies, you might focus your attention there first.
There’s a good chance I’ll write an entire blog on this topic later, but for now, let’s focus on the spectrum aspect of Wi-Fi. While operators rally behind this nonsense that “5G replaces Wi-Fi,” Wi-Fi keeps getting more attractive every day. By comparison with 5G, Wi-Fi has a similar diversity of frequencies at its disposal, all unlicensed. We always think of the classic mid-band Wi-Fi workhorse that is 5 GHz spectrum (with Wi-Fi 6), but there are also unlicensed Wi-Fi opportunities in low band (900 MHz with Wi-Fi HaLow) and high band (60 GHz with WiGig), not to mention the mostly-garbage-but-still-useful-for-IoT 2.4 GHz band. And we’re at the beginning of a load of new Wi-Fi services built around Wi-Fi 6E in 6 GHz.
Enterprises should see 6 GHz as a major confidence boost for Wi-Fi to address critical services while addressing capacity, speed, and latency gaps. Is Wi-Fi perfect for apps that require utmost determinism for application QoS and mobility on licensed spectrum? Probably not. There’s absolutely a place for the predictability of 5G in private spaces. Those places are somewhat limited right now, and several of those interests are net new; meaning, they don’t even displace existing tech, wired or wireless. They’re additive.
On the point of 6 GHz spectrum, consider for a minute that the CBRS PAL (priority access license) auction in the US took in ~$4.5B from winning bids (which goes to show the value of spectrum control for deterministic performance). That’s for up to 70 MHz of prioritized license in mid band spectrum. As shown in Figure 3, unlicensed 6 GHz band used by Wi-Fi 6E has up to 1200 MHz of mid band spectrum. That’s more than double the existing 2.4 and 5 GHz spectrum previously designated for Wi-Fi deployments, and almost 8x the total size of the CBRS band. And there’s over 500 MHz of 5 GHz mid band spectrum already available. Spectrum is massively valuable, and ~1700 MHz of total mid band spectrum for Wi-Fi is a ridiculous amount of unlicensed spectrum for enterprise use cases.
While we’re doing comparisons, we should recognize that the traditional complaints against Wi-Fi may be addressed in part by this new 6 GHz spectrum and new cellular-like protocols (e.g. OFDMA) in Wi-Fi 6. More spectrum covers a multitude of QoS shortcomings. Nonetheless, determinism is the core value of private cellular. While there is a refocus on QoS and latency within the Wi-Fi Alliance and IEEE, will it cut deep enough into Wi-Fi behavior? Not for all the use cases. Cellular offers strict wireless coordination, while Wi-Fi remains an opportunistic approach to QoS, and it’ll likely remain that way. So, both technologies play a role in future enterprise services.
Regarding spectrum usage, the unlicensed Wi-Fi model has an ease of use advantage over cellular, and the Wi-Fi model is well-proven and mature. It does not require extra licensing or spectrum access registration, and the industry is full of skilled engineers with years of hands-on deployment experience. You could even lump in other IoT-focused protocols and products here as well, like Zigbee, Z-Wave, BLE, and LoRa, which share the same benefits of unlicensed spectrum as Wi-Fi, but lack the mature device ecosystems and wealth of trained engineers.
From Wi-Fi, let’s divert briefly to a topic of interest across the mobile industry, even though it’s not the foremost enterprise concern. That is, OpenRAN and virtual RAN (vRAN). I said before in this series that you can think of RAN as the 5G version of a Wi-Fi access point (AP) and its antennas. However, let’s deconstruct it a little more. In Wi-Fi, we’ve grown accustomed to the architectural simplicity of an AP. Effectively, the AP contains everything, from internal antennas to RF front-end, baseband and digital signal processing, encryption, and decryption to backhaul format conversion.
In cellular, these functions are potentially spread across many boxes at different sites, which could be on the tower mast, in the equipment shelter, or at a central datacenter. There are various architectural models to centralize or distribute this functionality as well. Historically, operators buy all this equipment as an end-to-end RAN solution, provided by a single telecom supplier, running on purpose-built hardware. The RAN model of 3G and early versions of 4G is a bit like the WLAN controller model of the early 2000s—software running on purpose-built hardware.
But software is still eating the world1, so tech purchasers are wary of hardware-specialized anything. If you can virtualize functionality in software and run it on commercial off-the-shelf (COTS) hardware, do it. This is the impetus for vRAN, the virtualization of the historically hardware-integrated RAN. vRAN opens up cost savings on the hardware. More importantly, vRAN enables a whole new world of flexibility in terms of software operation and scale for RAN functions—with virtual, cloud, and other software tools.
Mobile operators are also wary of the deeply integrated end-to-end RAN of telecom suppliers. Today, most tier-1 operator networks are provided by one of three vendors: Huawei, Ericsson, or Nokia. As the argument goes, this kind of “lock-in”2 prevents price competitiveness and new market entrants and the innovation they bring. To solve this, OpenRAN is a new RAN specification that defines open interfaces between radio components. In theory, this will lead to better interoperability and promote flexibility in solutions.
One of the important by-products of both vRAN and OpenRAN is disruption. Architectures are like substrate. As they shift, what’s built on top also shifts. Consequently, the mobile industry as a whole is reviewing the technology stack for 5G and incorporating new best-of-breed products and platforms that improve service delivery. Even though RAN initiatives are in the headlines, this openness shift is valid for 5G end-to-end; it affects RAN and core, and includes transport elements, visibility and analytics, management, operations, and automation.
If you’re anything like me, most 5G RAN articles might as well be written in Cuneiform. C-band this, OpenRAN that, mMIMO here, mmWave there. It’s a lexicon zoo. If nothing else, I hope this article gave you a foothold on the RAN so you can approach other articles and industry news with confidence and context.
We covered a load of good stuff in this article—I’m suddenly self-conscious that this should’ve been two separate articles. Thanks for hanging in there.
Some quick takeaways:
This article is the third of a 5-part series. The next blog will focus on the business and deployment approaches to 5G in more detail.
1This quote (or something like it) is famously attributed to Marc Andreessen. If you have time for a 5-minute diversion, I think it’s fascinating to read about his technology optimism and predictive insight from ~10 years ago.
2I’m not a huge fan of the term “lock-in” because there’s some nefarious intent in it, which isn’t always justified. If you take the positive view, you could call it “validated end-to-end integration and design.” For complex systems like cellular, this can be a tremendous value, as proven by the model’s success. In OpenRAN designs, operators likely pay extra for integration services, testing, and validation, and there will likely be performance tradeoffs and some eventual finger-pointing when problems arise. I think of this topic a bit like CAPWAP (for you Wi-Fi folks). Just because there’s an “open” spec doesn’t mean interoperable nirvana follows.
This blog was originally authored by Marcus Burton, Architect, Cloud Technology