DeWi, 5G, And Tower Density

5G requires significantly more cell towers and antennas than 4G. Deploying hardware in so many places is prohibitively expensive for traditional wireless companies. DeWi is the solution to the challenges raised by 5G.

I’ve seen versions of the argument above a dozen times. It’s mostly wrong. Several bits of information are getting combined and confused by folks hoping to shine a positive light on DeWi.

5G Frequencies

Most 5G coverage relies on signals with frequencies between 600MHz and 6GHz. That’s the same frequency range used for 4G. Holding other factors constant, a 5G signal at a given frequency will travel similarly to a 4G signal using that same frequency.

One of the innovations coming with 5G is the possibility of using high frequency spectrum that 4G didn’t work with. So-called “millimeter wave 5G” uses spectrum around 30GHz.1 Millimeter wave 5G signals don’t travel far and are easily blocked by buildings and other obstacles. However, millimeter wave can offer incredible speeds. If you’ve seen 5G tests showing speeds of multiple gigabits per second, the tests probably relied on millimeter wave 5G.

Close to 0% of land in the U.S. is covered by millimeter wave. While networks may roll out more millimeter wave 5G over time, it’ll be supplemental to lower frequency service. Thanks to its poor propagation characteristics, millimeter wave 5G will never be the backbone of a nationwide network.

DeWi Frequencies

DeWi projects use small cells (radio nodes) that operate in the CBRS Band (around 3.5GHz). These CBRS small cells have limited coverage potential largely because they’re low power. “5G” isn’t to blame. No DeWi projects are even using 5G smalls cells. They’re all 4G LTE. While 5G small cells will become part of DeWi eventually, they’ll continue to have limited range unless power limits for CBRS hardware are changed.

But, But, But…

DeWi projects aren’t using millimeter wave 5G today, but could they eventually? I won’t rule out that happening, but a lot would have to change about the world before it’s a possibility. I’m not aware of any DeWi project that legitimately has millimeter wave service on its roadmap.

If we drop the focus on 5G, we might be able to rescue a bit of the argument I’m objecting to. In many places, cell networks are congested. There’s a case for “densifying” networks—i.e., adding capacity in areas that already have coverage but are short on bandwidth. Densification is possible with 4G or 5G technology. Short-range small cells might be helpful for networks looking to densify. Maybe DeWi projects can roll out that hardware in a uniquely cost-effective way.

Connectivity Where It Matters

Steve Sellers of had a good Twitter thread running through his takes on the Helium Mobile announcement. Arman Dezfuli-Arjomandi, host of the excellent The Hotspot podcast, responded with this tweet:

I like Arman’s point about rural areas—Helium Mobile may have the most potential in the places with the worst cell service.1 I’m concerned that Helium’s upcoming quality of service (QoS) requirements for cellular nodes could prevent deployments in some areas where they’d be most helpful.

Under the current plans, there are three QoS criteria cellular gateways will need to meet:2

  • 100Mbps or higher download speed
  • 10Mbps or higher upload speed
  • 50ms or lower latency

About 40% of Helium’s cellular nodes fail to meet these performance standards. Most people falling short will be able to pass the tests after making a few changes, like upgrading their internet service or swapping out lousy networking hardware. But that won’t be the case everywhere. In some places, no sufficiently powerful internet is available. Even with a plan that allegedly offers a 100Mbps download speed and a 10Mbps upload speed, QoS tests will tend to fail since internet service providers’ actual speeds generally fall short of their advertised speeds.

Fortunately, it’s easy to figure out what regions have lousy connectivity. The FCC regularly collects data about where service is available from cellular network operators and internet service providers through Form 477. That data is public.3

There are many ways Helium could leverage this data to adjust quality of service requirements in poorly connected areas. I’ll sketch out two simple options, though there are cases to be made for more complicated approaches.

Option 1: Selectively Drop QoS Requirements

Using the Form 477 data, it’s easy to find all regions lacking internet with speeds above a given threshold. Helium could eliminate the QoS requirements anywhere that doesn’t have internet options advertised to offer at least 200Mbps download speeds and 20Mbps upload speeds.

Helium could also eliminate the QoS requirements in areas that none of the three major cellular networks (Verizon, AT&T, and T-Mobile) cover. Chances are, this set of areas overlaps heavily with the areas lacking impressive internet service.

In the future, Helium could slowly introduce QoS requirements to these areas. Performance thresholds could be phased in gradually based on the highest quality internet available in each region according to future rounds of Form 477 data.

Option 2: “Starlink Or Better” Threshold

While I’m partial to initially eliminating QoS requirements in poorly connected areas, I can see an argument for keeping a minimum bar. After all, satellite internet is available everywhere in the US.

In poorly connected areas, the minimum performance standards could be set so that decent Starlink setups would generally meet them. For example:

  • 30Mbps or higher download speed
  • 5Mbps or higher upload speed
  • 100ms or lower latency

While this level of performance isn’t impressive, people often overrate what kind of speeds they need for a decent experience on a phone. Backhaul that meets these standards could support multiple phone users that are browsing, messaging, or streaming standard-definition video.

Want My Help?

I’ve worked with Form 477 data a fair amount. If anyone involved with Helium wants me to lend a hand, please reach out.

MNOs & Zero Marginal Costs

Here’s a 1GB test file. What does it cost your cell carrier if you download that file to your phone? Most likely, nothing.

Mobile network operators (MNOs) like Verizon, AT&T, and T-Mobile usually have zero marginal costs.1 Buying radios, renting tower space, acquiring spectrum licenses, and supporting customers is expensive. But once a subscriber has coverage, there’s typically almost no cost involved in letting the subscriber use incrementally more data.

People claiming MNOs will be eager to offload data on DeWi networks often miss this point. Sure, it will be awesome if DeWi networks get so big that an MNO can rent fewer towers or buy fewer radios. But DeWi won’t hit that kind of scale anytime soon. The benefits of DeWi, in the short term, will show up on the margins.

While data use typically doesn’t directly cost MNOs money, it does contribute to network congestion. When networks get congested, users get slower speeds. MNOs with coverage overlapping with DeWi networks might sometimes pay to offload data and avoid degradation in the quality of their subscribers’ service.

When considering DeWi’s potential for lending MNOs extra capacity, modesty is warranted. Network congestion isn’t ubiquitous. It’s rare in most areas, and even in congested areas, issues are likely limited to certain times of day. Further, $0.50/GB is a whole lot more than $0.00/GB. If congestion is only modestly reducing users’ speeds, an MNO might prefer to save money over spending big bucks to ensure the best possible service.

Non-Zero Marginal Costs

Data isn’t always free for MNOs. When subscribers roam on other MNOs’ networks or use services from a neutral host like Boingo, there may be substantial costs for each additional increment of data used. Helium and Pollen’s proposed $0.50/GB rate could be a bargain in comparison.

Additionally, none of this analysis applies to mobile virtual network operators (MVNOs). These carriers don’t own network infrastructure but instead piggyback on other companies’ networks. Mint Mobile and Consumer Cellular are examples. The agreements MVNOs have with their host networks tend to be private, but I’d ballpark the data rates somewhere in the range of $1.50-$6 per gigabyte.2

The Unknown Cost Of DeWi Data

Folks argue that cellular service from DeWi networks is more cost-effective than service from traditional networks. The standard rationale goes something like this:

Pollen and Helium plan to charge $0.50 per GB. The typical consumer pays much more than that for data transfer on traditional cell networks.

The rationale has problems. First and foremost, the $0.50 figure was decided by fiat and is barely connected to the actual costs of running a DeWi network. Pollen or Helium could have about as easily said data will be $0.25/GB. We can get a different view by looking at existing DeWi deployments. The median DeWi radio has probably been involved in less than 1GB of data transfer. Given the typical cost of a radio, we’re looking at a cost of data so far that’s way north of $1,000/GB on most setups.

The low ($0.50 per GB) price of data in early DeWi networks is largely enabled by investors’ speculation on the value of projects’ tokens. In a roundabout way, the investment subsidizes data purchases in projects’ early days.

Ultimately, DeWi’s cost-effectiveness will be determined by what cost advantages it can achieve over traditional wireless. Today’s supposed $0.50/GB peg doesn’t indicate much.

DeWi’s PoC Paradox

With its LoRa network, Helium demonstrated the viability of Proof of Coverage (PoC). The LoRa protocol is well-suited for PoC. Thanks to LoRa’s long range and good propagation through obstructions, a network of LoRa hotspots sending out signals can simultaneously serve as a network of coverage validators.

In my view, PoC is the biggest innovation that’s come out of the DeWi industry. Understandably, folks want to repeat PoC’s initial success in emerging DeWi cellular projects. It’s not going to be easy. Given current regulations, Helium and Pollen’s cellular networks can only use low-power small cells. To make matters worse, transmissions are limited to the relatively high-freuqency CBRS spectrum. At high frequencies, cell signals struggle to penetrate obstructions and travel long distances.

We’re left with something of a paradox. Proof of Coverage is great for incentivizing the creation of cell coverage. The hardware and spectrum the DeWi industry has at its disposal suck for creating coverage.