Directions 2008: Software-Defined Radio Role to Grow

December 1, 2007 By: Logan Scott GPS World

The growing need for ubiquitous positioning and navigation to service the anywhere, anytime needs of the consumer location-based services (LBS) markets will strongly influence future receiver architectures. In particular, the need to operate with a multiplicity of GNSS systems, but also with other positioning techniques such as Wi-Fi access-point mapping and radio frequency identification (RFID) tagged locations, points towards increased use of software-defined radio (SDR) architectures.

The trend accelerates when we take into account the myriad other communications needs of a mobile user: GSM, CDMA, WiMAX, UMTS, 700 MHz band, cognitive radio modalities, and so on. The availability of these other communications modes are essential to many LBS offerings, for example automated parking reservation systems. Thus we can expect consumer offerings to move towards integrated communications/navigation devices as opposed to standalone personal navigation devices (PND).

In the future, a mobile device may encompass upwards of 50 radio standards, and so the question becomes: will the navigation function be implemented using signal-specific, special-purpose hardware subsumed into a chip, or will it take place in a set of callable software routines, possibly executed on some combination of general-purpose digital signal processor (DSP) and SDR-targeted field-programmable gate arrays (FPGAs) along the lines of the Altera Cyclone III family?

The potential impact of cognitive radio (CR) on future radio architectures deserves special mention. Even in dense urban areas, active spectrum utilization typically falls below 5 percent at any given geographic location. Rising demand for wireless access and increasing spectrum congestion provide strong motivations for finding ways to dynamically allocate spectrum to achieve higher spectrum utilization. CR actively senses the RF environment and tries to find unused spectrum, so-called “white space” within which to establish a communication channel.

Sophisticated geographic/spectrum models and a knowledge of one’s position has great utility in this process. CR also has the potential to seek low-cost communications channels, using voice-over-Internet protocol (VoIP) via Wi-Fi or WiMAX if available and resorting to traditional cellular providers only as needed. Another interesting aspect of CR is its potential to exploit synergisms between the various radio interfaces. As an example, most assisted GPS (A-GPS) systems are based on a cellular subscription model where the cellular provider provides access to frequency disciplining and ephemeris data so one can operate indoors with impaired GNSS signals. A CR radio might achieve the same result, but without a cellular subscription: Wi-Fi could provide the ephemeris and simply listening to a GSM C0 channel could provide requisite frequency discipline, without a subscription. One thing is clear: for CR to be viable, CR radios must support a wide range of radio interfaces.

Anywhere, anytime navigation requirements will favor architectures that can deal with a large variety of signal types. Rather than rely on just one signal type for navigation, the savvy receiver will need to scan through the universe of potentially available navigation signal sources and dynamically choose those that are available at its current location. As an example of this, Figure 1 shows how selection diversity can enhance operation in indoor/dense urban environments. Here, we consider four design options: use only one GPS signal (e.g. L1 C/A), use three GPS signals (L1C,L2C and L5), use all the signals from GPS and Galileo combined, and finally, use all signals from all GNSS systems. The signal strength of individual signals varies considerably, mainly due to signal-blockage effects and constructive and destructive multipath interferences. Figure 1 shows the strength of the strongest signal and its associated probability based on a hybrid log normal shadow fading / Rayleigh fading propagation model. Referring to the red dashed line, if the receiver’s sensitivity is 2185 dBW, then using only one signal would leave 25 percent of the region uncovered, without positioning, whilst using all GPS and Galileo civil signals would yield close to 100 percent coverage. It should be clear that a receiver capable of scanning through the universe of GNSS signals will operate in more locations since there is a good chance that at least one of the signals will be strong enough.

FIGURE 1

Now add in the possibility of using Wi-Fi access point (AP) mapping. The basic idea is to listen to the unique Basic Service Set Identifier (BSSID) broadcast by most Wi-Fi APs and then look up the associated location in a database such as Wireless Geographic Logging Engine (WIGLE) which contains locations for more than 12 million APs. It sounds hokey but works surprisingly well in urban areas and indoor areas where there are large concentrations of APs. Because Wi-Fi APs are inherently short-range systems, simply hearing a particular one can give you a pretty good idea of where you are. With a little refinement using received signal strength indication (RSSI) for range estimation, accuracies in the 10s to 30s of meters range are readily achieved. Adding location-dependent RF fingerprinting and mapping techniques, accuracies can be better than 5 meters. Other researchers are looking at a similar concept, except using cellular base-station signals.Texas Instruments has a location engine offering based on Zigbee, expected to be the standard in intelligent building and lighting architectures. Rosum promotes the use of television signals for positioning, and then, of course, there are the myriad ultra-wideband (UWB), RFID, and pseudolite/repeater proposals. Micro-electromechanical sensor (MEMS)-based, low-cost inertial measurement unit (IMU) technologies also receive serious consideration for consumer navigation.

No one technology will likely cover all scenarios. A diverse set of sensor/RF technologies will be required to meet the consumers’ anywhere, anytime requirement. The notion of diversity processing is fundamental to how cellular telephony achieves its coverage levels. Similar thinking will necessarily be applied to positioning technologies, except with a diversity of systems as opposed to just RF diversity.

Receiving diverse signals will require flexible RF front-end transducer technologies. Within the cellular community, multiband front ends are already available to provide coverage in different regions of the world and for different radio standards. Qualcomm recently announced a 12-band RF transceiver front end (RTR6280) to provide global roaming for voice and data. The universal software radio peripheral (USRP) shown in Figure 2 provides a window into the future. Marketed by Ettus Research LLC and intended for the hobbyist/homeland security market, this board is interoperable with the GNU radio free-software (open source) framework for the creation of software-defined radios. Depending on the specific daughter boards added, it can cover anywhere in the DC to 2900 MHz frequency range. It sports four 14-bit, 128 Msamp/sec D/A converters for transmit functions and four 12-bit A/D converters running at 64 Msamp/sec for receive functions. Also, it includes an Altera Cyclone FPGA for high-speed RF processing functions. Within a 10-year time frame, it is not inconceivable that such a device would be available on one or two chips.

FIGURE 2

Many of the newer communications standards employ orthogonal frequency division multiplexing (OFDM) to mitigate dispersive channel effects. Examples include the Wi-Fi 802.11a/g/n standards, UMTS, and some versions of UWB. This is significant in that it implies a fairly beefy fast Fourier transform (FFT) processor to generate and receive signals. Within the GNSS community, there are two widely used approaches for dealing with low signal levels: massive correlator architectures and frequency domain, FFT-based approaches. With a high-performance FFT processor on board, the frequency domain approaches would tend to be favored.

In my experience, frequency domain processing has the further advantage that it is flexible and adaptable to diverse waveform types. Adding another reference waveform is not all that difficult, and the core frequency domain signal processing can be used with almost any waveform type. This can be a real advantage: we want to support reception of diverse GNSS waveform types. Finally, we should note that the RF connectivity of future wireless appliances will permit rapid updates, upgrades, and reconfigurations if we follow an SDR paradigm.

In conclusion, integrated consumer communications/navigation devices will need to support reception and transmission of diverse waveform types. The classic collection of special-purpose hardware elements approach towards architecting such a device is becoming problematic. In the longer term, I expect software-defined radio architectures to gain favor.