Blog Post

We need to think about spectrum allocation differently … now.

When we talk about data usage and internet-connected wireless devices, the figures are so massive that they feel abstract. What does half a billion internet-connected devices in the U.S. (as of early 2013) and 8.7 billion worldwide (as of late 2012) actually look like? What do 150 million SnapChats look like or 700 million tweets? The point is, to the majority of consumers, quantifying these things is not only impossible, but also inconsequential on a day-to-day basis. If your smartphone or tablet works, who cares.

And that’s the rub: they soon might not. The number of internet-connected wireless devices is very much quantifiable. Eight point seven billion definitely means something. (For starters, it exceeds the number of humans on the planet.) Also — and perhaps more importantly — the wireless spectrum in which these devices move data is quantifiable. It has physical limits. Thus the physical and economic implications of billions of devices competing for limited airwaves through which they can send and receive data are massive.

lots of tablets

Spectrum 101

But before we dive into those implications and explore potential solutions, let’s run through a quick Wireless Spectrum 101. Spectrum is the range of electromagnetic radio frequencies that carries data wirelessly—be it digital voice between cell phones and towers, digital television shows from broadcasters to your TV, or information from one computer to a router.

In the U.S., the Federal Communications Commission governs who (i.e., a wireless provider) can use what “slice” of the spectrum and for what purpose. For example, broadcast FM radio (often the easiest way to imagine the actual spectrum) occupies the sliver approximately between 88.0 MHz and 108.0 MHz. Similarly, mobile phones function throughout the 700 MHz to 2.6 GHz range, while unlicensed airwaves in the 2.4 GHz and 5 GHz slice carry Wi-Fi. Needless to say, the portion of the frequency allocation map below shows just how crowded (and limited) this place is. And we can’t make more of it. We can only hope to use it more efficiently.


By now the point has been made clear: a day is coming when the existing spectrum will be unable to keep pace with the ever-increasing number of internet-connected devices and the data the move. So what happens? Yes, the FCC could potentially shuffle some things around (think back to the analog-to-digital transition of television). But what about the other side: The mobile technology side?

Think differently about apps

What if we were to build mobile apps and devices that were “frequency intelligent?” Today’s devices are ‘locked’ to particular frequencies. They can use WiFi (2.4 GHz and 5 GHz), Bluetooth (2.4 GHz) and a variety of cellular network frequencies depending on the carrier. New work in Software Defined Radios (and not relying solely on hardware and crystals) could enable future devices to hop across frequencies as environment dictates.

Similarly, an app that moves a lot of hi-def video would be self-aware enough to know that it could use the software-defined radio and hop on a frequency that has less contention for bandwidth. Another example: an app that’s moving simple packets of data without the need for real-time responses would say to itself, “Even though there’s some 4G here, I don’t need that. I can use the 3G that’s available here instead.”

This, of course, has profound effects for developers building the next-generation of apps. In a world with heterogeneous networks, where we are constantly bathed in different radio frequencies with different capabilities, knowing what to use—and when—will ultimately deliver apps that perform better in more scenarios. Rather than simply accepting a crowded 4G network in downtown San Francisco when uploading your latest cycle ride, developers will need to build both front-end user experiences and also implicit backend network awareness to access the most appropriate network for the use.

Think differently about networks

Right now, software defined radios, programming interfaces, and wireless protocols that will allow for seamless switching between networks are still being defined for widespread use. But you can see mobile OS providers like Microsoft, Apple and Google becoming smarter about how they offload data from their devices, letting users pick which networks they wish to use to upload photos and videos.

Other radio innovations are coming too. Picocells (small base stations that use the same frequencies our devices speak to today) have proven very effective at increasing network capacity in dense usage areas. Used on a large scale, picocells could greatly reduce the data throughput burden on a large mobile network since the radios inside our devices could talk to a “tower” closer to their location.

Nokia Siemens Networks' conception of a heterogeneous network
Nokia Siemens Networks’ conception of a heterogeneous network

We could also look at bridging techniques. A few years ago I worked on a project that would turn a PC into an access point so others could use your PC’s internet connection to ride the net. Much like many people tether their computer to their phone, this solution could allow a host PC to use newly vacated “white space” airwaves (today residing at around 700 MHz) while the other devices use a low-power radio on a different frequency to talk to the host.

The team at Commotion is also working on bringing something similar to market. Additional trials are happening to see if devices can be made smart enough to ‘share’ federal spectrum beyond 700 MHz when those agencies aren’t using it while still giving them priority when they need it. Other organizations like Serval are working to use frequencies like those used by cordless phones to create mini-networks among many devices that can work even without any infrastructure in place. Ultimately these concepts rely on devices working together to balance the load of web requests, voice calls, and funny cat videos across all the available spectrum each device can access.

Is the spectrum’s ability to keep pace with devices and data a problem tomorrow? No. But if the rapid upticks witnessed over the last few years are any indication, it’s clear: we need smarter spectrum allocation and, more importantly, smarter use thereof.

Stefan Weitz is the Director of Search at Bing. He is obsessed with science. Follow him on Twitter at @stefanweitz. </em

6 Responses to “We need to think about spectrum allocation differently … now.”

  1. A nit: this article uses “device” and “app” a little too interchangeably. If we’re expecting apps (and by proxy, app developers) to know anything about which frequency or even access technology to use when multiples are available, or expecting it to take an active role in rerouting to avoid congestion, whether in spectrum or on the wired part of the network, we’re probably doing it wrong. The better model is for us to give app developers good language to articulate what their apps need for proper performance (latency, jitter, throughput, duration, timeframe) to the network and device, and for the network to communicate what it can offer in similar terms based on what it knows about capacity, topology, other demands, etc and then have the network and device make more intelligent decisions about what to do.
    This also requires a more seamless way to hand off between networks, a common authentication model, etc. Even if we got to some future utopia where spectrum was shared, the network infrastructure connected to that spectrum isn’t, and it all goes pear-shaped if my device has to wait to authenticate to a new/less-congested network before switching to it, especially if that authentication works like public WiFi today (with http redirects to a portal for credentials that requires direct user involvement). And of course this says nothing about the support for things like first-class mobility that doesn’t require a device to be tethered to a specific mobility anchor like a P-Gateway so that handoffs don’t trigger session resets when the IP addresses being used to send/receive traffic change from one network to the next. True multihomed mobility is *hard*, even if the radios are smarter unless we can get away from carrier-infrastructure-based mobility and move to something like ILNP.

  2. Roger Entner

    Network sharing is a great concept and the direction we will have to go in the future, but even if we just look ten years out, it is not going to happen. Obstinate existing license holders aren’t even the problem. The most significant problem is that all the existing and even planned transmission technologies for both licensed and unlicensed use need virgin spectrum. Shared spectrum is anything but virgin…

  3. Working on shared networks without infrastructure is shaky at best due to security issues (client devices talking to other client devices directly). A focus, not only on frequency availability, but also on power, is extremely important to the Capacity issue. A low power WiFi network, properly designed, can handle thousands of users in a high rise building, by making use of the spectrum available while minimizing interference. Use of TV White Space is better for backhaul infrastructure due to the frequency characteristics and power permitted, in addition to incompatibility with current WiFi devices.

  4. Shiggity

    Spectrum is an open and public resource for all to use and share. ANYTHING ELSE IS WRONG, PERIOD.

    When you understand the economics and math behind fiber optics and internet bandwidth in general, you may as well put an infinite economic value on spectrum. Spectrum space can grow in bandwidth with Moore’s Law and if we let it get monopolized it will stay that way forever and the US will fall behind in competitiveness because $ / bit of data will quickly be more important than $ / oil. If you don’t see that, you need to go look at Google’s stock price.