SDR goes to war

SDR (software-defined radio) is not a new technology. The industry has in the past made attempts to implement SDR, but these efforts have fallen short. What has sparked recent interest in SDR is the rise of innovative hardware and software technology that appears to offer enough flexibility, performance, and power efficiency at a price the market can bear. For example, wideband converters are now available that operate over a wide range of frequencies and can SDR goes to war process incoming waveforms based on a specific bit resolution, and high-frequency processors can work with signals directly, eliminating IF (intermediatefrequency) stages and reducing cost.

The appeal of SDR stems from an understanding that moving to a single hardware platform that can handle multiple radio technologies will not, as you might expect, cost more than individual radio designs that target the lowest cost for an application. The hope is that a versatile platform will introduce significant economies of scale, increase radio functions, and release the military from the bondage of proprietary implementations.

The cost efficiency of SDR derives from the premise that maintaining a single radio platform is less expensive than managing multiple platforms. Certainly, ASICs offer a cost advantage in highvolume applications over FPGA and DSP implementations. However, when you consider multiple platforms, the continually decreasing cost of FPGAs and DSPs and the reusability of software can offset the cost of developing and maintaining multiple ASICs.
<%@ LANGUAGE="VBSCRIPT" %>
<% Randomize: ord=int(rnd*1000000000) %>


ECONOMIES OF SCALE
The drive behind SDR in both military and commercial applications is to reduce cost. The military seeks to achieve this goal by attempting to reduce the total number of radios that an individual must carry. From a commercial perspective, SDR focuses less on supporting multiple radio technologies than it does on reducing design and development effort, increasing radio robustness through comprehensive upgrades, and enabling next-generation architectures to reuse IP (intellectual property) by building on previous implementations.

Currently, the military wants to support more than 30 protocols. These protocols use a number of both secure and nonsecure waveform types. Driving the promotion and advancement of SDR for military applications is the JTRS (Joint Tactical Radio System). The JTRS' SDR efforts aim to promote the availability of off-theshelf hardware and software components that enable a single implementation to handle multiple protocols' waveforms that also target the best power, performance, and cost.

The ability to support multiple protocols holds less value in the commercial world than in the military world. Often, less than a handful of protocols is competing within an application space. Two primary nonmilitary applications- cellular and safety-focus on SDR technology. In the cellular world, base stations that can support multiple radio technologies-such as both GSM (Global System for Mobile communications) and CDMA (code-division multiple access)-expand the capabilities of the cellular network and increase potential operating revenues. For safety applications, for example, service groups such as police and fire personnel use different radios; a universal radio would eliminate the need for these individuals to carry multiple radios to coordinate efforts.



These perspectives significantly alter how these industries have approached SDR. For example, a world phone that can operate on any band would have limited usefulness and an even smaller market share. The military seeks consolidation, whereas the commercial world seeks design simplification and efficiency.

SDR'S PROMISE
The ultimate goal of SDR is to create a radio in which frequency, IF, baseband, modulation, protocols, hopping frequency, and so on are programmable. A typical radio implementation fixes many of these characteristics, limiting the radio. True SDR strives to make it possible to modify any of these parameters.

Current implementations of SDR are still a long way from the ideal universal radio that marketing managers envision. Today, most radios use ASICs to implement a radio, and front ends target particular applications. Generic front ends covering multiple bands and frequencies that you could modify using software are still not here. Rather, operators must either send the radios back to the factory to reprogram the waveforms or, more appealing to those in the field, swap out an RF module on their own.

Another cost savings that SDR enables is the ability to dynamically reconfigure radio resources to match current system needs. Consider a military or commercial base station populated with a single type of reconfigurable board. You could configure this base station to support waveform allocations through software alone. For example, you could allocate the resources of various boards in a base station to handle 20 channels of Waveform A and 20 channels of Waveform B. In the event of equipment failure, such as a board that supports 10 channels of Waveform A becoming disabled, you could reallocate the remaining channels appropriately to support a 20/10, a 15/15, or another mix appropriate to the application and current waveform usage.

A fully dynamic and reconfigurable radio offers many benefits besides support of multiple bands. For example, a radio could use FEC (forward-errorcorrection) techniques to match the terrain in which the radio is operating. Additionally, reconfigurable radios can immediately take advantage of innovation, such as a more efficient waveform implementation or a robust security mechanism.

SDR also promises to minimize development time, as designers can reprogram a radio if a problem arises in the field or a protocol changes. For example, although protocols themselves may stay static, algorithms, waveforms, and modulation schemes continue to improve over time. With an SDRbased architecture, you can take advantage of these improments and innovations without redesigning the radio.

THE SCA FRAMEWORK
The keystone of the JTRS' mandate to promote SDR in military applications is the SCA (Software Communications Architecture) framework. The JTRS is developing this framework to break the link between hardware and software, which in turn will break the hold that proprietary implementations have on the military. The system implements functions using a radio's available hardware and software resources, which may differ from vendor to vendor.

The SCA sits above the operating system and is the glue between hardware and software (Figure 1). The goal is that any SCA-compliant waveform will operate on any SCA-compliant radio hardware. Such hardware/software interoperability is a major consideration in military applications; it is not trivial to swap out a radio in an F-15 fighter jet. SCA promises to effectively eliminate the difficulty of supplying new radios and upgrading any deployed radios in the field.

Today's SCA framework supports portability of waveforms a few steps shy of the transparent interoperability developers are seeking. One reason for this intermediate stage in the SCA's evolution revolves around the fact that the SCA framework primarily targets general-purpose processors, such as the Pentium or the PowerPC. However, to achieve sufficient performance and cost efficiency, most designs today employ specialized hardware based on DSPs, FPGAs, or both (Figure 2). The SCA effectively addresses portability for general-purpose processors, but it clearly misses the mark where specialized hardware is concerned. This shortfall is the weakest link in the SCA, because, if you cannot easily port software between platforms, you cannot achieve the required interoperability- or, more accurately, near or sufficient interoperability.

The JTRS is addressing the issue of specialized hardware. A major concern is that designers develop software in a linear and sequential fashion, whereas hardware achieves its highest efficiencies through parallel processing. How a developer defines functions has a major impact on the efficiency of the final implementation. Certainly, it is possible to port a C implementation of a waveform intended to run on a generalpurpose processor over to a DSP or FPGA. All you need to do is run the C code through an appropriate tool that converts the function into an FPGA implementation. This step, however, neglects to use the FPGA's primary architectural advantage: extensive parallelism. In fact, the only way to exploit this parallelism is to effectively rework the code, which transforms a simple port into an extensive redesign. At this point, it becomes questionable whether the FPGA actually increases cost.

This problem is difficult to solve. The most efficient code matches the architecture in use: FPGA, DSP, mixed DSP and FPGA, general-purpose processor, and so on. Although requiring software vendors to create such substantially different versions of their code is simply infeasible, a single generic implementation doesn't seem effective either.

SPECIALIZED HARDWARE
The JTRS attempted to address specialized hardware in Version 3.0 of the SCA spec. The hope was to create abstracted processing resources that you could assign to various tasks. Many developers, however, have described the approach as software-oriented and insufficient for capturing DSP and FPGA capabilities. Specifically, the SCA didn't clearly define the level of abstraction, forcing vendors to interpret the spec with the result that each implementation differs sufficiently to be effectively proprietary. Additionally, in some cases, the extra software necessary for implementing the abstraction was enough to completely erode any DSP and FPGA performance advantages. As a consequence, the JTRS has reverted to a previous version of the SCA.

DSPs and FPGAs are essential for creating efficient and costeffective SCA-based implementations, both in military and commercial applications. Generalpurpose processors are effective at efficiently implementing generalized tasks but are much less efficient than specialized hardware at implementing well-defined, computationally intensive processing algorithms, leading to greater power consumption and larger radio form factors-both undesirable results. The JTRS is seeking to leverage the advantages of specialized hardware without tying software to proprietary hardware implementations. After the JTRS was reorganized in 2005, its leadership decided to resolve this issue internally. Although, on the one hand, this approach limits industry participation in shaping the SCA standard, the decision prevents commercial forces from driving changes that will negatively affect how quickly the industry can mandate SCA to military suppliers.

Work is under way to enable the efficient use of specialized hardware in SCA implementations. For example, one of the primary sources of overhead in implementing SCA on an FPGA or DSP is the use of CORBA (Common Object Request Broker Architecture). CORBA is the middleware layer between the operating system and the SCA framework. It transmits data between software objects over a software bus. Implementing CORBA on a DSP or an FPGA is not straightforward. However, several vendors are developing ways to implement CORBA on specialized hardware that would make hardware look more like software without an unreasonable performance compromise.

SDR IN THE INTERIM
Getting SDR to the point at which radios are no longer proprietary platforms is a high priority for the military, and the JTRS continues with its mandate to make this goal possible in the real world, even if some military-radio suppliers are unenthusiastic about the prospect of widening their narrow playing fields to outside competition. SCA is the foundation of this reality.

In the commercial world, however, the SCA framework has a limited hold and perhaps even less appeal. Although many commercial applications could gain an advantage from an SCA-based foundation, the SCA framework perhaps bites off more of the interoperability problem than it can chew. Completely interchangeable hardware and software leads to a market in which the only differentiation is cost. There are few advantages to opening designs in the commercial world simply to invite competition that will drive down profit margins.

The commercial world thrives on innovation, and, in radio applications, hardware is a key differentiator. As it stands, the SCA framework inhibits the introduction of specialized hardware. However, as the SCA standard evolves, there will be much efficiency that you can leverage without adopting the formal standard. In fact, many commercial base stations deployed today have some software-defined basis, given that a significant number of functions, such as remote upload fixes to code, baseband processors that are programmable, and DSPs and FPGAs that can support a variety of waveforms, are software implementations. The base stations use just the bits and pieces, so to speak, that make the most sense.

There is a wide spectrum of adoption for SDR. The military is pursuing the ideal of total interoperability and interchangeability. The commercial world strives to balance efficiency, function, and cost. With such diverse goals and applications, it's almost as if they're dealing with two disparate technologies.

AUTHOR INFORMATION
Contributing Technical Editor Nicholas Cravotta covers digital and communications technologies, including networks, buses, and data security. He is currently developing a unique accessory for the video iPod.