A “computer” used to be a system in and of itself. It was a device whose primary purpose was computing. We are all familiar with the form factors and metaphors of “computers” – mainframe, desktop, laptop, server… We in the embedded computing industry, however, have made careers out of putting computing technology into things that are not computers – airplanes, mobile phones, home appliances, industrial equipment – the list goes on and on. This is a natural, second-generation evolution of the technology. In some way, this is the concept we all think of as “convergence” – but it goes beyond that.
For several years, global positioning technology (GPS) has been primarily in GPS devices. These devices have proliferated in various form factors – designed for a wide variety of uses from aircraft and marine navigation to personal auto use, hiking, cycling, and even GPS-specific activities such as geocaching. Now, GPS is becoming an embeddable technology. Like the process of embedding computing, embedding GPS opens the door to a wide array of applications we wouldn’t consider in a stand-alone device.
The ability to embed GPS into our designs brings up the question: “When does it help to know where I am?” My personal first experience with a location-enabled mobile device was an old-generation Palm connected to an Omni-sky wireless modem. The killer-app for that configuration was the “find the nearest Starbucks” capability – charming, but hardly a feature that would compel customers to rush out and buy the device. Here in the Pacific Northwest United States, at least, we already have a much simpler solution to finding the nearest Starbucks. In any urban area, you simply move your eyes through about a 45 degree arc. You will usually locate at least two.
Mobile phones were among the first devices to approach embedded GPS technology. Spurred by the US Government’s E911 mandate – requiring mobile network operators to supply the location of subscribers making 911 calls, mobile handset providers started diving into the embedded GPS race somewhere around the year 2000. Many decided on location-finding technology other than GPS, later finding that the accuracy of these technologies did not meet the regulatory requirements. Those that pursued GPS ran into a plethora of problems of their own, including poor performance indoors and in urban areas, significant power consumption by the GPS chipset, diffuculty in integrating GPS antennae into the existing handset form-factor, and of course increased BOM and IP costs from integrating GPS technology.
While the E911 fiasco spiraled out of control – with providers and manufacturers continuing to miss regulatory deadlines and performance specs, the embedded GPS proponents continued to press ahead with development and improvement of the technology. With the market for GPS receivers in mobile handsets potentially amounting to hundreds of times the “GPS unit” market – there was significant incentive to push beyond the barriers that were obstructing embedded GPS adoption.
GPS technology companies pressed toward single-chip solutions – getting past the traditional separation between the RF and digital functions into separate devices. While this approach was sound for creating the lowest-cost GPS chipset for many embedded applications, extremely cost-sensitive applications like mobile phones demanded an IP-based solution. Driven by cost, form-factor, and power concerns, most mobile handsets were already reduced to two chips – an RF chip and a massive system-on-chip (SoC) device performing all of the digital functions from baseband signal processing to user-visible feature applications. Adding additional devices to integrate GPS was not an attractive option, and the challenging work to integrate RF and digital sections of GPS on one device was unnecessary. What mobile device developers needed was an IP-based solution that would add GPS capabilities to the existing mobile phone chipset – the RF portions to the RF device, and the digital portions to the SoC.
In addition, much of the dedicated hardware required by stand-alone GPS was already present in some measure in existing mobile handset systems – the most obvious of these being the processor and software required for GPS operation. If the handset’s existing architecture had enough spare processing power embedded, the GPS application could be run on that hardware without the need for additional IP. Unfortunately, GPS requires substantial signal processing that far exceeded the capabilities of typical mobile phone embedded processors. The challenge, then, was to offload enough of the signal processing tasks into dedicated hardware accelerators that the GPS application could be successfully run on a processor shared with the other mobile phone applications.
Just getting the technology into the mobile phone chipset, however, doesn’t solve the biggest challenges with GPS in mobile phone applications. Traditional GPS requires a clear line-of-sight to several satellites at once and considerable time to obtain the required sensitivity at any given location in order to orient itself. Most mobile phone use is indoors – where neither of these conditions is easily met. In order to tackle the first part of this problem, we have the advent of what is called A-GPS or “Assisted GPS”.
Assisted GPS gives the GPS receiver a big head start by providing location- and time-specific information on the location and identification of GPS satellites in view, as well as Doppler-prediction data. This information comes from what is called an “Assistance Server.” The Assistance Server generally has a GPS receiver with an excellent view of the sky, a strong satellite signal, and a very clear understanding of its own location. In the case of mobile phones, the assistance server can communicate with mobile phones in its area, supplying them with orbital information for GPS satellites. It can also compare partial or poor signals relayed to it by mobile phones with its own known good signals, giving it the ability to determine the phone’s position and relay that information to both the phone and (potentially) other clients such as emergency services trying to determine the phone’s location. AGPS is a broad category of technologies without a standard, and a variety of implementations abound. Data is exchanged between mobile devices and Assistance Servers using everything from GPRS to SMS services.
The second major challenge is improving the sensitivity of GPS receivers so that severely attenuated signals inside buildings can still be used for position determination. Recent generations of receivers, however, have incredible sensitivity performance – in some cases, almost obviating the need for AGPS. These high-sensitivity receiver chipsets vastly expand the possibilities for location-enabled applications in mobile devices.
Vendors like NXP (a spin-off of Philips) are now providing complete platforms for embedding GPS technology in mobile devices. Turnkey OEM solutions like NXP’s will likely account for most of the GPS embedded in handsets over the next few years. “The complexity of developing a convergent handset is extreme,” says Michel Windal, Operators and Partnerships Marketing Director with NXP. “We have extreme experience in things like antenna design – trying to reduce antennae and transceivers – putting things like Bluetooth, XM, and GPS on a single antenna. Only very sophisticated companies have the expertise and resources to design this themselves.”
With the infrastructure for location-based applications coming within reach of the masses of system designers, the only thing we need to determine is which location-based services consumers will actually pay for. For some of us, sitting in one Starbucks and sipping our 22-oz lattes – perhaps the ability to locate the next Starbucks within a radius that will allow us to traverse the distance to it before our current caffeinated drink runs dry – well, we just might have to do better.
