How software-defined radio could revolutionize wireless

6 July 2012 Timothy B. Lee In 1976, two shaggy-haired college dropouts founded a company called Apple to manufacture personal computers. The company’s prospects looked so poor that the third co-founder relinquished his 10 percent stake in the company for $800 that…

6 July 2012

Timothy B. Lee

In 1976, two shaggy-haired college dropouts founded a company called Apple to manufacture personal computers. The company’s prospects looked so poor that the third co-founder relinquished his 10 percent stake in the company for $800 that same year. It simply wasn’t clear why anyone would want the firm’s Apple I computer. It was so under-powered that it couldn’t perform many of the functions of mainframes and minicomputers that were already on the market. And most consumers had no interest in having a computer in their homes.

Today, of course, Apple is the world’s largest company by market capitalization. What was important about the Apple I wasn’t the meager capabilities of the original version, but the promise it held for rapid innovation in the coming decades.

Now, a company called Per Vices hopes to do for wireless communication what Apple did for computing. It is selling software-defined radio gear called the Phi that, like the Apple I, is likely to be of little interest to the average consumer (it was even briefly priced at the same point as the Apple I, $666.66, but has since been placed at $750). But the device, and others like it, has the potential to transform the wireless industry. This time, the revolution will depend on hackers enabled to manipulate radio signals in software.

The versatility of software-defined radio

Traditional radio chips are hard-wired to communicate using one specific protocol. For example, a typical cell phone has several different chips to handle a variety of radio communications: one to talk to cell phone towers, another to contact WiFi base stations, a third to receive GPS signals, and a fourth to communicate with Bluetooth devices. In contrast, software-defined radio hardware works with raw electromagnetic signals, relying on software to implement specific applications.

This makes software-defined radio devices tremendously versatile. With the right software, a single software-defined radio chip could perform the functions of all of those special-purpose radio chips in your cell phone and many others besides. It could record FM radio and digital television signals, read RFID chips, track ship locations, or do radio astronomy. In principle it could perform all of these functions simultaneously. Software-defined radio hardware also enables rapid prototyping of new communications protocols.

Software-defined radio will make it possible to use the electromagnetic spectrum in fundamentally new ways. Most radio standards today are designed to use a fixed, narrow frequency band. In contrast, software-defined radio devices can tune into many different frequencies simultaneously, making possible communications schemes that wouldn’t be feasible with conventional radio gear.

Most significantly, the widespread adoption of software-defined radio hardware could undermine the FCC’s control over the electromagnetic spectrum itself. Right now, the FCC largely focuses on limiting the transmission frequencies of radio hardware. But this regulatory approach is likely to work poorly for software-defined radio devices that aren’t confined to any specific frequency.

The effective deregulation of the airwaves could create headaches as careless hobbyists pollute frequency bands that have been reserved for other applications. But it’s also likely to usher in an era of unprecedented radio innovation as millions of people have the opportunity to experiment with technologies that, until recently, were the exclusive domain of well-funded industrial labs.

The pioneers

Software-defined radio has had political undertones since its inception. A decade ago, some early software radio enthusiasts became interested in the “broadcast flag” debate then raging in Washington. Hollywood wanted to force consumer electronics companies to detect and comply with metadata in HDTV broadcasts that would signal what consumers were allowed to do with television content.

Eric Blossom, founder of a software project called GNU Radio, hoped that implementing an HDTV receiver in software and releasing it as open source would demonstrate the futility of this approach. Even if the government forced his project to implement the broadcast flag, he argued, anyone could tweak the source to disable the broadcast flag code and then re-compile it.

The effort to build a software receiver for the ATSC television format was ultimately successful. “We would record samples off the air, then process them in our app in GNU radio, and you could watch the MPEG of Law and Order,” Matt Ettus, a contributor to the effort, told us.

Ettus said the hardware used to build the ATSC receiver “wasn’t something that someone else could go out and buy.” Also, it “wasn’t well set up for what we were doing.” It could only capture a narrow slice of spectrum: 100 kHz at most. That was enough for Law and Orderreruns, but Ettus believed better hardware would be needed to unleash the full potential of software-defined radio technology.

“To do more interesting things you need more hardware,” Ettus said. He wanted to capture a much wider range of frequencies. And he wanted other advanced features like the ability to handle multiple antennas simultaneously.


“I went a long time trying to convince somebody else to build this thing and nobody would,” Ettus told us. So in 2003, he began work on what became the Universal Software Radio Peripheral (USRP). In 2004, he quit his job as an engineer working on conventional radio products to focus on the USRP full-time, shipping his first unit on January 1, 2005.

Today, Ettus Research builds a range of devices specifically designed for software-defined radio. A working USRP system comes in three parts: the main USRP box, an RF daughterboard, and a computer. The daughterboard handles the actual reception of radio signals, and passes the analog signal to the main USRP unit. Ettus explained to us what happens from there.

“First it converts the analog signal to digital. Then the digital signal is sent to a field-programmable gate array. The FPGA does the high-speed processing and the user can modify it and put all sorts of interesting things in there. In the most basic configuration, the FPGA reduces the sample rate, does some frequency translation, and then sends that out over the interface” to the CPU.

The interface that connects the USRP to the computer is the main thing that distinguishes the various USRP models from each other. The cheapest model (costing $650) delivers the data to the user’s computer over a relatively slow USB link. The priciest model (costing $1700) has a gigabit Ethernet interface. In between, the company offers an “embedded” model that includes a built-in CPU capable of running a full Linux distribution, which allows it to function as a stand-alone device.

Each RF daughterboard is designed to receive a different range of frequencies. “We used to need a lot of different daughterboards to cover an interesting frequency range,” he told us. “When we first started, you could only get a couple hundred MHz with decent performance. But as technology advanced, we’ve gotten newer and newer daughterboards with wider range.”

Ettus said one of the most interesting applications for the USRP has been for open source cell phone telephony. Users have configured USRPs to provide GSM cellular service, deploying them “in a number of places, from Burning Man to small islands in the Pacific.” The hackability of the USRP makes it more versatile than traditional cellular gear, making it ideal for unusual environments.

Wireless security research is another key application for the USRP. For example, one research group used a USRP to discover security vulnerabilities in the wireless communications protocol of a commercially available pacemaker. “If you want to determine wireless security, you need to be able to send those devices interesting packets,” he said. “You need complete control of the packets you send, and you need to examine the received packets at a fundamental level.” Conventional wireless hardware can’t match the flexibility of software-defined radio hardware for this kind of application.

A new challenger

So far, the USRP has dominated the software-defined radio world. But that could change with the introduction of the Phi, that new device from the Y Combinator startup Per Vices. Ars talked to co-founder Victor Wollesen.

“The USRP is a great product. That’s what we actually started our prototyping efforts on,” he told us. “The USRP was the first device that brought the cost down to the point where enthusiasts and people in academia were able to start to play around with wireless.”

But he said the Phi has several key advantages. First, it’s implemented as a PCI Express card, providing up to 8Gbps of bandwidth to the CPU. Wollesen says that’s fast enough that the interface isn’t a performance bottleneck.

Second, the Phi is a single integrated package. The USRP is designed to be highly configurable. As we’ve seen, that means users must buy a separate daughterboard for the specific range of frequencies they’re interested in receiving or transmitting.

In contrast, Wollesen says his pitch to users is “here are the capabilities, here is the architecture we’re going to provide you.” The Phi is somewhat less versatile than the USRP, but the reduced complexity of the device makes it easier for less experienced users to get started.

Easy as Rails

Perhaps the most fundamental difference between the USRP and the Phi is the target market. Ettus is focused on serving relatively sophisticated customers. “You’re not going to see our stuff at Fry’s or Best Buy,” he told us.

In contrast, Per Vices hopes to bring the benefits of software-defined radio to the masses. He draws parallels to the Raspberry Pi and Arduino. Both are cheap, simple devices that have proven wildly popular with hobbyists.

At $750, the Phi costs about as much as the cheapest USRP-plus-daughterboard bundle offered by Ettus Research. It also boasts a much faster interface. And Wollesen is hoping to achieve dramatic price cuts in the coming years. “Price drops astronomically with scale,” he told us.

As the cost of the Phi comes down, he hopes many more people will become interested in experimenting with software-defined radio. “We want people to be able to design RF protocols with the same ease as building Rails apps,” he said.

Eventually, Wollesen envisions a future in which every home has a software-defined radio device. “We see our device being used as a center that’s able to take any sort of wireless signal, process, and re-package it. A universal router,” he said.

“If I were to tell you, you need one laptop for a Word document, another for e-mail, another for solitaire—that’s a ridiculous point of view. Yet that’s exactly what you see in wireless right now,” he told us. “We’re trying to advocate for an economical platform that does the same thing the first microcomputers did.”

Software-defined radio hardware, he said, offers “intrinsic capabilities that can be exploited by applications. Our device, by loading a piece of software on it can replace a router, a cell phone, a base station, or a garage door opener.”

Right now, most people probably couldn’t imagine why they’d want software-defined radio hardware in their homes. But people said the same thing about microcomputers in the 1970s.

Smarter spectrum use

Operating as a universal hub is one possible application for software-defined radio. Another is using the electromagnetic spectrum itself more efficiently.

Tom Rondeau runs the GNU Radio project, a job he took over from Eric Blossom in 2009. He also does research on software-defined radio technologies at the University of Pennsylvania. He explained to us how software-defined radio could lead to more efficient spectrum use.

Traditional radio communications occur in narrow, fixed frequency bands designated by regulators. For example, in the United States one block of spectrum is reserved for ATSC television broadcasts, others for AM and FM radio. This rather bureaucratic approach doesn’t necessarily make the most efficient use of spectrum. The FCC has effectively “lopped off blocks of land that nobody can touch,” Rondeau said.

He drew a comparison to the old debate between packet-switched and circuit-switched networks. A circuit-switched telephone network offers callers guaranteed bandwidth, ensuring a high quality of service. But for many applications this approach is wasteful, since the dedicated capacity sits unused most of the time. A shared, best-effort approach can lead to much more efficient use of the available bandwidth, Rondeau argued.

The same point applies to the electromagnetic spectrum. We’ve covered the “white spaces” proposals to allow unlicensed use of unused spectrum in the television bands. But the white-space devices the FCC ultimately approved are still relatively “dumb,” deciding which frequencies to transmit by consulting a centralized geographic database. Rondeau believes that sophisticated “cognitive radio” devices can share the electromagnetic spectrum much more efficiently by adapting to their environment and each other.

Spread spectrum

Wollesen agrees, and he’s excited about a related technique called “spread spectrum.”

“There’s a trade-off between the rate at which you can encode information, the power level at which it can be received, and the amount of spectrum you need,” Wollesen told us. He pointed to GPS as an example: GPS satellites are extremely far away and transmit a relatively weak signal. But because they transmit a tiny amount of data in a predictable format, GPS hardware is able to reconstruct the weak signal.

Wollesen told us that rather than transmitting data in a narrow range of frequencies at high power, a software-defined radio device can transmit data across an extremely wide range of frequencies at low power—so low that it’s imperceptible to the conventional radio devices that are operating on the same frequencies.

Wollesen believes that intelligent software algorithms will be able to pluck these faint signals out of the noise much more effectively than conventional radio hardware could manage. And because communication can occur at many frequencies simultaneously, the overall data transmission rate can be respectable even though the power level at any given frequency is low.

Disrupting the FCC

Given that software radio fits so poorly in conventional regulatory categories, we wondered how a device like the Phi could be legal at all. The answer is that FCC regulations—and comparable regulations in other countries—exempt laboratory equipment from many of the regulatory restrictions that apply to consumer radio hardware. The USRP and Phi both take advantage of that exemption. For example, the Phi manual includes a disclaimer that “this device is intended for engineering, research, or science laboratory use only. It is not for open office or residential use.”

Also, according to Wollesen, the rules are more permissive for devices that transmit at low power. “Most regulatory efforts focus on power levels,” he told us. The more powerful a device’s transmissions are, the more red tape regulators impose. “We’re aiming for our device to operate below those thresholds. Rather than speaking loudly at one frequency, we want to speak across multiple frequencies.” He hopes that keeping power levels down will “avoid a large amount of the regulatory framework.”

Sticking to low-power applications might be a good way to avoid regulatory headaches in the short run. But as software-defined radio technology becomes more popular, there will be a growing demand for software-defined radio devices that transmit at higher power levels. And that will pose a real dilemma for the world’s regulators.

Until now, regulators like the FCC have relied on the fact that conventional wireless gear is hard-wired to transmit only at certain frequencies. By inspecting devices before they leave the factory, regulators can ensure that each device stays in its assigned lane.

But applying this model to software-defined radio is problematic. The FCC might require device manufacturers to sell locked-down devices programmed to only transmit at FCC-approved frequencies. But such devices would be a juicy target for the jailbreaking community. And too much lockdown will limit the usefulness of the hardware.

Pirate radio 2.0

In the long run, then, the spread of software-defined radio hardware could erode the power of regulators to control the electromagnetic spectrum. This wouldn’t be an entirely new development. Unauthorized radio broadcasting has existed for as long as the airwaves have been regulated. For example, in the 1960s, radio stations floating in international waters blasted pop music into the United Kingdom where state-owned monopoly broadcasters only offered programming the authorities regarded as more wholesome. Pressure from these pirate radio stations hastened the introduction of legal popular music to the British airwaves.

Running offshore pirate radio stations was expensive enough that the authorities were eventually able to shut them down by passing laws that cut off the flow of advertising dollars. Controlling unauthorized transmissions could be much more challenging in a future where every wireless router and cell phone has a software-defined radio chip waiting to be jailbroken. The power of governments to control the airwaves could be undermined in much the same way that the Internet, and software like BitTorrent, has undermined the government’s power to control the distribution of copyrighted works.

“If you look at the Internet right now, 4Chan is not going away,” Wollesen said, referring to the website that gave birth to Anonymous and other troublemakers. “At the same time, it hasn’t spelled the downfall of the Internet. It’s an impetus to make the Internet more robust against these problems.” He told us that a similarly adaptive approach will be needed to deal with the new challenges that could be created by software radio.

Fortunately, software-defined radio technology will provide powerful tools to deal with a more anarchic electromagnetic spectrum. It will allow the development of smart transceivers with greater capacity to adapt to or cut through clutter.

More importantly, software-defined radio has the potential to dramatically accelerate the pace of innovation in radio technologies. Just as the PC ushered in a dramatic increase in the pace of software innovation, so the proliferation of software-defined radio hardware will allow many more people to build and experiment with new ways of communicating wirelessly.

Ars Technica