The creation of true software-defined-radio (SDR) systems has been made possible by the digitization of traditionally analog RF hardware. SDRs have been under development for many years. Now, they are beginning to move from the military to the commercial world. But what exactly are SDRs? Why are they important to wireless designers? Perhaps most importantly, what products will first be designed for SDR technology? These are just a few of the questions that will be answered in this report.
SDRs can be defined in many different ways. One definition comes from the SDR Forum (www.sdrforum.org), which is an international industry association for SDR technology. According to the organization's mandate, "SDRs are elements of a wireless network whose operational modes and parameters can be changed or augmented, post-manufacturing, via software." The U.S. Federal Communication Commission (FCC) has adopted a more formal definition. It states that a SDR is "a radio that includes a transmitter in which the operating parameters of frequency range, modulation type, or maximum output power (either radiated or conducted) can be altered. This is done by making a change in software without making any changes to hardware components that affect the radio-frequency emissions."
Like any complex technology that can be implemented in a variety of ways, the term "SDR" means different things to different people. Bob Plunkett, Director of Product Management for QuickSilver (www.qstech.com), notes that the Holy Grail of SDRs consists of a flexible bandwidth and air-interface system that can be tuned across a broad RF band. On the other end of the SDR definition are devices that use fixed bands and bandwidths, but provide a fully programmable baseband process.
To encourage a more common view of the term "SDR," the SDR Forum has created a five-tier description of software-radio categories. Starting with Hardware Radios (Tier 0), it climbs in capabilities all the way up to the Ultimate SDR (Tier 4). The simplest example of an SDR is Tier 1: Software-Controlled Radios. At this level, only the control functions are implemented in software. The baseband processing is performed with ASICs or fixed hardware. Early dual-mode cell phones fall into this category, as they use two hardware radios to support two different communications standards.
The Tier-2 SDRs, called reconfigurable SDRs, are most commonly used today. Mainly, these radios are implemented in cross-platform military operations. At this level, software is used to control a variety of modulation techniques: wideband or narrowband operation; security functions like hopping; and the waveform requirements of current and evolving standards over a broad frequency range.
In addition to being the most commonly used SDRs, Tier 2 devices constitute the most popularly understood definition of software radios. These systems consist of various processing technologies, such as ASICs, FPGAs, and DSPs.
The Tier 3 Ideal Software Radio has all of the capabilities of Tier 2 systems. Yet it also is the most advanced type of SDR that is achievable in the near future. Tier 3 eliminates the analog amplification or heterodyne mixing prior to digital-analog conversion. Its programmability extends to the entire system, with all analog conversions taking place at the antenna, speaker, and microphones.
The last category—Tier 4: Ultimate Software Radios—is defined by the SDR Forum for comparison purposes only. Ultimate Software Radios could theoretically switch from one interface format to another in milliseconds.
Whether they appreciate the different categories or not, many wireless designers are curious about software-radio architectures. They want to know how these evolving architectures will affect current RF and baseband designs. After all, SDR technology can be implemented in any device that requires a reconfigurable RF capability. Among these devices are next-generation base stations, military communications, public-safety radios, and telemetry systems.
For a straightforward example of software-embedded radios in telemetry systems, look to the ER900FHTRS transceiver from Low Power Radio Solutions, Ltd. (www.easyradiousa.com). The Easy-Radio software associated with this transceiver module is transparent to the baseband designer. As a result, it greatly simplifies the RF design portion of the system.
More generally, though, it is the reconfigurable nature of software radios that garners so much attention. This reconfigurability comes from the continued "digitization" of many radio-frequency functions that were formerly analog functions. Highly flexible SDR applications do require a change, however, in both the front end and baseband side of today's wireless devices.
Currently, most commercial SDR technology is implemented in base-station systems. Compliance with every changing communications standard can therefore be met with mere software updates. Very little—if any—hardware changes are needed.
For military operations, software-defined-radio technology greatly enhances interoperability between different organizations. A single SDR device can support multiple waveforms.
Many companies have aided in the advancement of technologies that were needed for SDR systems. For example, Analog Devices, Inc. (www.analog.com) believes that its "Flexible Radio Platform" is the first commercially viable approach to SDRs (FIG. 1). Scott Behrhorst, Wireless Infrastructure Marketing for High Speed Converters at Analog Devices, explains that their system-level design is built around ADI's wideband-radio front end. That front end uses ADI's RF/IF, dataconverter, RSP, and TSP technologies in conjunction with the TigerSHARC platform for digital baseband processing. This tight integration of the RF and baseband technologies yields a highly programmable radio system.
The high data rates needed for SDRs can create interface challenges for connected devices. Bob Sgandurra, Product Manager for Pentek (www.pentek.com), explains that wideband receivers can produce data throughputs of 300 MBps or greater. These rates can exceed the sustainable data rates of interfaces like the PCI computer bus. This is why Pentek has created a Velocity Interface Mezzanine (VIM) that provides users with multiple data pipes. Each pipe is capable of sustainable rates going up to 400 MBps. Currently, high-speed interfaces form the foundation for the SDR systems used in infrastructure-based VME chassis.
Off-the-shelf DSPs and FPGAs form the bulk of most existing commercial and military SDRs. For now, these applications are confined to cellular base stations and military platforms. In these relatively low-volume designs, off-the-shelf modules and existing development tools are critical cost factors. It's more practical to integrate FPGAs and DSPs, but these platforms are large and power hungry.
One field-programmable-gate-array vendor, Altera (www.altera.com), has succeeded in developing a custom system architecture for SDR applications. Its Mercury chips have been integrated with Echoteck's wideband transmitters and receivers to achieve input-output bandwidths greater than 100 MHz. Compared to the fastest application-specific-standard-product(ASSP) approach, Echotech claims that its FPGA implementation can run twice as quickly. In addition, ASSPs tend to be too expensive for current, low-volume software-radio products.
The performance and cost challenges related to custom silicon design are being addressed by the continuing improvements in reconfigurable technology. For such technology, the key tradeoffs shift from performance and design cost to flexibility and configuration time. For example, several recent startups offer reconfigurable platforms based on coarse-grained computation units. These units have a programmable interconnect structure. The interconnect and computation units can be reconfigured. But they require a large number of clock cycles.
During the reconfiguration, most other processing must be stopped. The latency associated with these reconfigurable systems forces designers to use the systems in a linear-dataflow manner. They process data at the input rates of the system, explains Quicksilver's Bob Plunkett. This processing-rate reduction combines with the overhead associated with programmable systems. Together, they result in designs that are effectively larger than those created using the existing ASIC/DSP approach.
Technologies are emerging that hope to overcome these limitations. One example is the Adaptive Computing Machine (ACM). According to Plunkett, adaptive computing systems can be reconfigured on the fly, at run time, and in as little as a single clock cycle. This capability saves the computation units from any linear-dataflow limitations. It also permits them to run at optimized processor speeds. These computational units can then be reassigned to perform multiple tasks—a key requirement for the needed multimode capabilities of SDR devices (FIG. 2). As a result, virtually any number of air interfaces and protocols could be implemented as software in ACMs. They could include, for example, cdma2000, W-CDMA, GSM/GPRS, Bluetooth, GPS, MPEG-4, and MP3 protocols.
But adaptive computing machines are not merely a future technology. Recently, AOI adopted Quicksilver's ACMs for its digital-imaging applications. Though this is not an SDR product, this partnership does endorse the technology. It helps to validate its low-power, low-cost potential in the handheld consumer market.
ACMs aren't the only promising reconfigurable technology, however. Check out Motorola's Reconfigurable Compute Fabric (RCF), which was created especially for SDR base-station applications. Arif Ahmed, Manager of Strategic Marketing for the RF and DSP Infrastructure Division of Motorola's Semiconductor Products Sector (www.motorola.com), notes that the signal-processing requirement for tasks near the RF front end is increasingly becoming MIPS-Intensive and Repetitive (MIR). Conversely, tasks that are closer to the network side remain Higher Complexity and Irregular (HCI). Ahmed also observes that while traditional DSPs are ideal for HCI tasks, they fall short when it comes to MIR tasks.
In response to this issue, Motorola's Reconfigurable Compute Fabric plans to address the needed programmability and flexibility requirements of MIR tasks in wireless-base-station software-radio systems. Essentially, an RCF core is an optimized array of compute elements or DSP cores. These elements are connected through a high-bandwidth interconnect fabric and high-speed local memory. A RISC processor controls all of the RCF cores (FIG. 3). Currently, Motorola is demonstrating an RCF test chip.
Several potential players are carefully watching developments like the RCF and ACM. Intel (www.intel.com), for example, is currently investigating a number of architectures for mobile devices. Dubbed "Radio Free Intel," these architectures are still in the research stage.
SDR technology can best be thought of as the next generation of RF systems. For today's SDR applications, much of the computational power is coming from improvements in baseband-processing designs and faster interface buses. Yet software radios are still limited to specialized RF chips that target particular frequency bands and waveforms. Eventually, advances in software-tunable RF technology will lead to truly cost-effective front-end designs. At that point, the commercialization of SDRs will really begin.