Ask consumers in the developed world about their household Wi-Fi connection and they’ll likely tell you that lately it seems to be getting worse, not better. Some might even say, “It stinks!” Even the residents of the White House have Wi-Fi problems. In an interview with the BBC just before Super Bowl 50, First Lady Michelle Obama complained, “It can be a little sketchy. The girls are just irritated by it.”
The White House, along with homes inhabited by more than 80 percent of the United States and 50 percent of the worldwide population, are in urban areas where Wi-Fi connections are steadily getting worse. The reason would appear to be obvious: There are many more people—and things—using Wi-Fi than a decade ago, and the numbers continue to grow. Today, 6.4 billion connected devices are in use around the globe. By 2020, that will mushroom to 20.8 billion—that’s 2.8 mobile devices for every person on Earth. So certainly the wireless highways through which Wi-Fi traffic moves have gotten and will continue to get more crowded.
But the reason for the virtual traffic jam is not as straightforward as simply having more vehicles to accommodate: The roads themselves are causing conflicts. And the situation has been made worse by three changes in the market.
First, not only is every house on your block likely to have a router but quite a few will have more than one, and many communities are also served by public Wi-Fi networks. Second, an increased demand for speed demands wider virtual lanes on the Wi-Fi highway, which means there will be fewer of them. And finally, cellular operators are dumping traffic into the Wi-Fi spectrum—to stay with the roadway analogy, imagine all the people who had been commuting by train suddenly taking to private cars.
Here’s how Wi-Fi has become a victim of its own success—and what engineers can do to improve things.
Wi-Fi operates in what is known as the unlicensed spectrum. That is, though the Federal Communications Commission (FCC) generally demands licenses to use the airwaves in the United States, and national spectrum regulators like Japan’s Ministry of Internal Affairs and Communications do the same elsewhere, the regulators leave some frequency bands relatively open. Users do have to comply with technical requirements, including power limits, but they do not have to apply for specific permission. There are several of these bands out there, but home Wi-Fi networks primarily operate in the 2.4-gigahertz and 5-GHz bands, because these are the only available parts of the radio spectrum that have the range and bandwidth needed. The 2.4-GHz band works the best here: It easily penetrates walls and furniture, and signals generally travel farther at the same power level as they do in the 5-GHz band.
In the United States, in the 2.4-GHz band, the FCC has authorized roughly 80 megahertz for Wi-Fi use. The channels operating there under the IEEE 802.11 standard are 20 or 22 MHz wide, so you can fit in only three nonoverlapping channels: 1, 6, and 11. The situation is only slightly different in Europe, where 13 channels allow for still just three nonoverlapping channels at a time, and in Japan, where 14 channels allow for four nonoverlapping channels.
So when you scan for available networks on your phone or computer in the United States, if you can see more than three 2.4-GHz routers (a likely scenario for anyone not living in a rural environment), or if you see only three but any one of them is not on channel 1, 6, or 11, there is interference.
Signals in the 5-GHz band have a shorter range in the home, mostly because of the walls and furniture, but that band, which extends from 5.180 to 5.825 GHz, has 24 nonoverlapping 20-MHz-wide channels in North America, with five fewer in Europe and Japan. That’s a huge number of additional lanes on our crowded wireless highway. But roughly half of these channels—more in North America—are allocated for primary use by weather and military radar. Inserting Wi-Fi into this radar-priority spectrum requires special technology, so to date most consumer routers ignore these lanes. But they are important, and we’ll come back to them later.
In either band, we have a set number of channels that don’t interfere. As more and more routers come online, and more and more devices connect with them, interference becomes the norm. In the Wi-Fi world, when two conversations collide, all the devices go quiet and then try to talk again a little while later. The amount of time they wait is determined by an exponentially increasing time delay, known as a backoff. With more collisions, the backoff increases, and the Wi-Fi becomes slower and less reliable.
Today, congestion has gotten so bad in many regions that it has pretty much made the 2.4-GHz band unusable for transferring data at high rates. Several broadband-service providers (including AT&T, British Telecom, and Comcast) no longer use 2.4 GHz for video or voice, and almost all smartphone makers, including Apple, no longer recommend using their smartphones at 2.4 GHz. The latest, and fastest, variant of Wi-Fi, IEEE 802.11ac, provides for operations in only the 5-GHz band, although most Wi-Fi equipment includes both bands to accommodate older mobile devices.
So we have a limited number of lanes on the Wi-Fi highway in which your signals can travel without slowing down traffic overall. But it’s not just devices that are contributing to congestion. It’s the networks themselves.
Moving Wi-Fi communications from 2.4 GHz to 5 GHz initially helped with congestion but sacrificed coverage, so many consumers turned to simple boosting solutions—like network range extenders or mesh networks—to get Wi-Fi to every room in their homes. These extenders or mesh networks are placed at the range limits of a router where the transmission signals fade away. These devices listen to all transmissions, then rebroadcast them at a higher power level, sometimes on a different channel. Now you have even more Wi-Fi signals overlapping in the same frequency ranges.
Making matters worse has been the introduction of public hotspots—that is, places where Wi-Fi is available to the general public or some subset of users (like subscribers to a particular Internet service). In 2005, Spain-based Wi-Fi provider Fon Wireless pioneered the concept of community hotspots—that is, hotspots that piggyback on a private router—and they have become more and more common around the world. Today, Internet providers like AT&T, Comcast, and Verizon in the United States are quickly rolling out such hotspots, accessible to any of their subscribers, by piggybacking this functionality on the wireless gateways installed in their customers’ homes. Juniper Research, based in England, estimates that by 2017 one in three home gateways around the world will allow community access by incorporating a second network identifier and allowing some of the Wi-Fi spectrum available to that gateway to be shared, typically without the awareness of the people in whose homes the gateways are installed.
Wi-Fi congestion will soon get even worse, thanks to the mobile-phone carriers, which have exhausted much of their exclusive spectrum. These wireless carriers are planning to off-load mobile-data transmissions, as much as 60 percent within the next three years, onto the unlicensed spectrum used by Wi-Fi.
The technology for that is called either LTE-Unlicensed (LTE-U) or Licensed Assisted Access (LAA). It uses 4G LTE radios and routers to send and receive data via the same 5-GHz frequencies as Wi-Fi. While carriers downplay interference to Wi-Fi users, some organizations like Cable Television Laboratories, Google, and Microsoft claim that LTE-U and LAA will absolutely increase congestion on Wi-Fi channels and degrade Wi-Fi service. In the United States, Verizon and T-Mobile have begun trial deployments for LTE-U to determine its impact on Wi-Fi. Carriers in Europe and Asia are planning similar trials.
On top of all of this, the latest variant of the Wi-Fi standard—IEEE 802.11ac—is actually reducing the number of lanes on the radio highway.
IEEE 802.11ac satisfies a growing need for speed—speed to stream high-definition videos and to allow mobile devices to conserve battery life by transmitting at high rates for only a limited time. This form of Wi-Fi delivers 1.3 gigabits per second compared with 450 megabits for 802.11n, the previous generation.
To allow data to move that quickly, 802.11ac has to merge channels. In its highest-performance configuration, IEEE 802.11ac Wave 3, it combines the entire available Wi-Fi spectrum into two 160-MHz-wide channels. This merger means that only two pairs of devices can communicate on the widest of channels simultaneously without interfering. So if one of your neighbors, say, is using one of those two channels to watch a movie, and your other neighbor uses the other one, there may be nothing left for you. Suddenly, all those extra noninterfering channels that made 5 GHz an improvement over 2.4 GHz are gone.
With developments of this kind on so many fronts, Wi-Fi connectivity is likely to soon go from often annoying to completely broken. In 2013, Ofcom, the British national telecommunications agency, published a study, “The Future Role of Spectrum Sharing for Mobile and Wireless Data Services,” which predicted that Wi-Fi and mobile Internet airwaves could become critically congested by 2020—now just four years away.
To date, technical-standards developers and router manufacturers, who have worked hard to improve speed over the past 15 years, have all but ignored these issues. In particular, they haven’t addressed the fact that a widespread rollout of 802.11ac, with its ability to offer wider but fewer channels, will make the congestion problem vastly worse.
There is, however, a near-term fix. Remember that chunk of 5-GHz spectrum I mentioned, where radar has first dibs and which requires special technology to use for Wi-Fi? Today, the makers of consumer Wi-Fi routers are ignoring these channels. Opening up these other frequencies to consumers would make a huge difference.
This additional spectrum was made available for Wi-Fi traffic by the FCC and other regulators around the world in 2007. The regulators realized that radar—for example, the Terminal Doppler Weather Radar system, which warns of low-altitude wind shear at airports—is not located everywhere and does not run 24/7. So the Wi-Fi industry could move Wi-Fi communications into these frequencies, as long as the devices that use these channels implement a mechanism called Dynamic Frequency Selection (DFS) to stay out of the way of radar signals.
DFS acts as a high-speed traffic cop—when it spots a radar signal in one of these protected channels, it quickly shifts all Wi-Fi traffic to another lane. There are a few rules about how this traffic cop works: It must listen for radar for at least 60 seconds before declaring a channel free to use, then continue to listen while Wi-Fi traffic is on the channel. If the mechanism detects even a 1-microsecond radar pulse, the Wi-Fi transmitter must clear the channel within 10 seconds and stay off it for half an hour.
The vast majority of mobile devices introduced in the last three or four years have radios that can operate in these bands and have the required software to respond to instructions from what is called a DFS master. But they need that DFS master built into the router to tell them when it’s all right to use a radar-priority channel and when they need to move aside.
DFS-master technology isn’t trivial to implement. Radar pulses are hard to detect because they are very, very fast (each pulse lasts just about one half of 1 microsecond) and can be present at very low power levels (as low as –62 to –64 decibel-milliwatts). Incorporating radar-detection tools eats up bandwidth—as much as 17 percent—because a router must listen on a channel for a minimum of 60 seconds before concluding that it is clear for transmitting, and then continue to listen during and in between normal transmissions.
Currently, DFS-master technology is available in expensive routers, the kind that are typically installed only by large businesses. It is migrating to some lower-cost consumer-level routers in Europe and Japan. But both the expensive business versions and the cheaper consumer versions are not all that clever: When they detect radar, they quickly shift traffic back to a set default channel in the non-DFS part of the 5-GHz band—a crowded spot. And they don’t go back to using radar-priority channels until their users reboot their routers. In a business environment, that is often scheduled to happen daily, but in a residential setting, it might take weeks or months before the user realizes that the router is not performing well and needs to be reset. So even routers with DFS capabilities are staying out of those express lanes, at least most of the time.
Still, it’s in these DFS channels where the solution to Wi-Fi congestion lies. The trick is creating a less costly and more effective radar-detection technology. My colleagues and I at Ignition Design Labs, located in San Jose, Calif., think we’ve figured that out.
We have designed an enhanced router, called Portal, which incorporates a full-spectrum radio scanner and a CPU dedicated to radar detection and channel management alongside standard router hardware. The scanner continuously sweeps the entire 5-GHz band for radar as well as for Wi-Fi traffic and general interference. Making this detection system completely separate from the Wi-Fi sending and receiving radio solves a lot of the problems of current radar-detection technology, which shares the main processor and Wi-Fi radio for radar detection as well as communications.
In such standard hardware, the radio sees what is happening only within a single channel in the DFS band at a time, so a DFS master can monitor only one specific DFS channel at a time. And when a DFS master’s radio first monitors a DFS channel, the FCC requires that the radio refrain from transmitting on any channel for at least 60 seconds to make certain it does not interfere with the receiver looking for radar. To avoid this kind of shutdown, most of these radio designs look for an open DFS channel only when the router is reset.
A second radio dedicated to the task of sensing radar signals removes that barrier. It can also regularly scan through all the channels. So when a radar pulse is detected in a channel being used for a transmission, the system will know whether or not another DFS channel is clear of radar, in which case it can move data traffic over there instead of to a preset default channel. And when the DFS master does require the router to abandon a channel because it detects radar traffic, it can automatically go back and recheck that channel, after a 30-minute required waiting period, without having to shut down ongoing transmissions.
Meanwhile, the dedicated CPU can minimize the number of false radar alerts, reducing the times that Wi-Fi traffic has to vacate a channel. Today, when the processor in the router is handling a large number of Wi-Fi streams—you’re watching Netflix, your child is playing a game, other family members are listening to music or browsing Facebook—it doesn’t have the processing power left to analyze radio energy it detects in the protected channels to determine whether it fits the pattern of a radar signal. So it errs on the side of caution—if it detects any interference in its channel (which might just be Wi-Fi traffic from a neighbor’s router), it vacates the channel until the router is reset.
We’re trying to make the process of assigning channels even more intelligent by gathering information not only on radar but also on all sorts of interference, and sending that information, along with data on general Wi-Fi and radar-traffic patterns, to a cloud server; there, our software analyzes the data and makes adjustments to how our Portals behave: We call it network self-optimization.
With this system, we can determine the best channels for Wi-Fi devices to use in different places. For example, let’s say we know that at 8 p.m. in Europe, where DFS channels are already being used by consumers in a limited way, the default DFS channel, channel 100, gets very busy. We can then shift one user’s traffic to channel 132 and his neighbor’s traffic to channel 154. This kind of coordination can have a huge impact on the quality of Wi-Fi communications.
We have received approval for our technology from the FCC and will ship our first products later this summer in North America and late autumn in Europe. We are also working with some manufacturers of Wi-Fi equipment and Internet providers in anticipation that they will eventually incorporate our hardware into their Wi-Fi routers and gateways.
It is essential to get this kind of comprehensive, intelligent system for managing Wi-Fi resources out into the world before Wi-Fi becomes so unreliable as to be unusable.
Those of us who work in the communications industry didn’t manage the way Wi-Fi devices use the 2.4-GHz spectrum, and we essentially exhausted it. But we were able to move to the 5-GHz spectrum, so consumers didn’t really notice when 2.4 GHZ ran out. But when we exhaust the 5-GHz spectrum, we will be out of real estate, at least for the immediate future.
In the longer term, technology will be developed to move some of the traffic to other types of communications networks that are not compatible with current Wi-Fi. The FCC has several frequencies under consideration for possible spectrum reallocation, including small amounts at 5.9 GHz, 4.9 GHz, and 3.5 GHz. But this process of reallocating spectrum can take years, even decades. And all these frequency ranges encompass radar and other primary uses (like public-safety communications). So if these ranges are approved, they will also require intelligent use of DFS technology.
That’s why finding affordable, robust technology to allow us to use all the lanes on today’s wireless freeways is the only way out of the data traffic jam.