The first generation of WISPs, including Webpass and Netblazr, extended their range by operating as mesh networks. Each antenna in the mesh acts as a relay, capturing some data for the subscribers at that point and passing along other data to subscribers at points farther along the path. The more hops, the less bandwidth is left for the subscribers at each point—which puts a limit on the capacity of a point-to-point network.
Starry doesn’t work this way. Kanojia explains that its network is a point-to-multipoint system, meaning each base station communicates directly with many end points—typically, antennas atop multiple-dwelling structures such as condominium buildings.
Point-to-multipoint systems don’t suffer from the bandwidth dropoff problem, the way mesh networks do. But there is a real limitation on their range: the strength of the central signal. Many of Starry’s technical decisions, Kanojia says, have been aimed at addressing that problem.
“The core technological things that we have solved are how to do point-to-multipoint at relatively long range in a high-frequency spectrum,” he says.
Let’s examine those things one at a time.
The first part of the solution was to use active phased array technology for Starry’s base stations. A single antenna emits radio signals along repeating wavefronts that expand spherically. But when a row or a grid—in other words, an array—of antennas emits the same signal, the wavefronts combine to create a plane wave—in effect, a beam that’s much stronger (and propagates farther) than any single wavefront. By using a computer to subtly shift the timing or phase of the signal from adjacent antennas—that’s the “active” part—that beam can be steered in any direction, or in a series of directions. In this way, a single base station can aim a strong signal at multiple end points.
“The lay analogy you could make is if you look at an incandescent bulb, it’s just radiating energy in all dimensions,” Kanojia says. “As you walk away, it attenuates. But if you make the source coherent like a laser, it will go miles and miles longer. That’s the way we get range. And if you’re doing an electrically steered phased array, which is what we do, you can basically steer those beams that you’re forming in a sub-microsecond time frame, and that allows you to paint the entire landscape with the energy, one dot at a time.”
But it’s not that simple, of course. Starry’s second tactic is to use a method called MU-MIMO, which stands for multiuser multiple-input, multiple-output. If both the transmitters and the receivers have multiple antennas, MU-MIMO allows data to be sent to multiple devices simultaneously (up to eight, currently) on the same frequency.
MU-MIMO is already used on most home Wi-Fi routers, and Starry decided to use it for long-range communications too. “In radio-frequency communications there is probably one free lunch, and that is MIMO,” Kanojia says. “Essentially what you’re doing is taking the same frequency and spatially reusing it, meaning if you have sufficient separation between those two spatial streams, you can have essentially a new pipe being created.”
Finally, there’s the question of how the data is modulated: how many bits-per-second to try to cram into the frequency ranges available for outdoor wireless networking. That’s where orthogonal frequency division multiplexing, or OFDM, comes in. It’s a way of dividing up amplitude-modulated digital data by broadcasting signals at different angles to the direction of travel, to reduce cross-talk and errors. Or to put it more simply, it’s a way to jam more streams of data into the same channel.
Combine eight spatial streams through MU-MIMO with high modulation rates through OFDM, and you can suddenly deliver tens of gigabits per second to a particular location, Kanojia explains. Add active phased arrays, and you can deliver such speeds to many locations, and to relatively distant locations.
It’s a lot of trouble to go to—which is why it hasn’t been done before. Kanojia says Starry’s engineers weren’t even sure, at the beginning, whether they could build receivers that would convert the densely packed OFDM waveforms at the frequencies used outdoors (around 40 Gigahertz) down to home Wi-Fi frequencies (around 5 Gigahertz) without destroying the data.
But once they’d solved that problem and started filing patents, Starry’s team knew it would have a strong economic case. Long range means that there are more potential customers within the coverage area of each base station. “That reduces your initial capital expense and your operating costs,” Kanojia explains. “And your total expectation of how many customers you need to turn that thing into a cash-flow-positive installation goes down.”
Using OFDM and MU-MIMO, meanwhile, means that Starry’s network has enough capacity to absorb lots of growth. Kanojia calls it “future-proofing” the network. Even if data consumption doubled every year for the next four years, the network would still be only 50 percent utilized, he says. “It’s all about making sure you have sufficient capacity so that, God forbid, if you’re successful, you can actually serve your customers,” he says.
Only time will tell how future-proof Starry’s network really is. “It’s an ambitious promise, given the sort of growth we see in data consumption in broadband,” says Mark Lowenstein, a Boston-based wireless industry analyst. “Certainly a wireless carrier in the traditional LTE bands would never say that.” (LTE frequencies are those used for mobile broadband service, which is almost always data-capped.)
There’s one more ingredient to Starry’s recipe: the Starry Station. It’s a funny-looking triangular device with a screen that displays the speed of the network connection and the health of the router’s connections with each Wi-Fi device in the household. It comes with a Starry subscription, but costs $349 if you want to buy it separately.
Why throw in an expensive router that makes it harder to break even on each customer? Because half of all customer-support calls to ISPs are actually about Wi-Fi, Kanojia says.
“The goal in this was to extend our control all the way from the cloud to your home,” he says. “So if you’ve got a Roku, Apple TV, Xbox, once a minute we monitor the quality of the connection to these devices, and we can say, ‘Hey, we noticed your Roku is first-gen, so you may want to consider swapping that, in particular if you want to watch HD quality. Because otherwise, if we don’t do it, we get blamed for the [slow] buffering.”
Providing custom hardware is all in the service of “redefining what the experience should be,” Kanojia says, “so that we know exactly what’s going on and the customers are going to be happy.”
The company makes other devices too, including a variety of “Starry Point” receivers that residents of single-family homes or small buildings, like brownstones or triple-deckers, can mount on their rooftops or outside their windows. The Starry Points feed data to the Starry Stations in individual units via a building’s internal telephone or Ethernet wiring.
For now, the company is still focused on reaching large apartment and condominium buildings, where it can install a larger, more powerful Starry Point on the roof to serve the whole building. Kanojia says the plan is to make a big push to sign up customers in smaller residences starting in late 2018.
[This paragraph updated to clarify reason for the data rate—Eds.] Already, Starry’s 1-gigabit-per-second wireless signal is reaching buildings that house a collective 240,000 households, Kanojia says. (While the company could, in theory, offer that full gigabit per second to customers, Starry currently rate-limits its plan to 200 megabits per second.) The company doesn’t say how many of those households are actual customers. But it announced last week that it’s ready to push beyond its Boston beachhead, offering beta service in Washington, DC, and Los Angeles this month and 14 more cities by the end of the year.
Lowenstein notes that Starry’s expansion plans are swinging into motion a bit later than expected.
“If you go back and look at what they’ve said publicly, they’re taking quite a while to get to market in a meaningful way,” he says. “They are around a year behind where they said they would be. I would imagine that’s because they’re still tweaking the technology—they are pioneers doing this in the high-gigahertz band. And once they’ve solved the technology piece, what is the installation going to be like? That is the thing, from a cost perspective, that has been difficult for other companies in the same business.”
One new element of Starry’s strategy may help free it from the challenge
