The Coming of Age of the Software Communications Architecture

The idea of the "Software-Defined Radio" (SDR) has been circulating in industry and academia for almost 20 years, the term having been coined by Joe Mitola in a 1992 paper.¹ Since that time, many commercial and defense-oriented radio SDR products have been developed and released, most using basic technology. Until recently, however, most SDRs have used proprietary middleware to facilitate communications between their different radio componentsónot only within waveform applications, but also to higher layers in the communications protocol. These proprietary middleware components are often narrowly focused, resulting in rigid, monolithic radios that inhibit IP reuse, platform-independence among applications, or innovations that would significantly reduce cost and time-to-market. The Software Communications Architecture (SCA) is a software specification that tries to improve this situation, and focuses on the "software" part of a software-defined radio.

SCA was born out of requirements for the Joint Tactical Radio System (JTRS) program. It standardizes the middleware that governs the interoperation of software across all operating layers within SDRs, and ensures portability and modularity between SDR software components and hardware implementations. Thus, SCA-compliant waveforms can be assembled, loaded, run and networked into systems across radio sets. This interoperability facilitates IP re-use, lowers platform costs and development times, and lengthens the service life of platforms by improving their adaptability. SCA is realizing other secondary, indirect benefits. As an open middleware specification, it has helped create a stable SDR industry ecosystem, enabling third-party vendors to streamline development tools and provide additional middleware components. The overall result is an increase in SDR design efficiency.

Today, the SCA standard is proliferating beyond the original JTRS program into other US DoD programs and Mil/Aero SDRs around the world (such as the ESSOR initiative in Europe). This, along with continued evolution (such as the 'SCA Next' initiative under the auspices of the Wireless Innovation Forum), is evidence of the value SCA is adding to SDR design. SCA is enabling smaller and smaller form-factor radios, and is reaching beyond mil/aero applications into commercial telecom SDRs, such as the ETSI reconfigurable Radio Systems (RRS) effort.

Methodology for Developing SCA-compliant Radio Components
As stated, the Software Communications Architecture governs the structure and operation of software within an SDR, allowing waveform components and full radio applications to share a common control interface and signal path connections. SCA also provides common methods for radio applications to communicate across the middleware boundary. However, the development of waveform components for most SDRs still follows a very traditional design methodology when targeted to FPGA, DSP, or general purpose processors (GPP). While these methodologies produce highly efficient designs for a given target, they typically require additional manual development of interface code, as well as additional code development to revise a design if a new target device is identified. This manual process is time-consuming and prone to overdesign in order to meet given performance targets. Also, due to the disjointed nature of the various tools used in the implementation of waveform components, the final integrated system is not always fully optimized for maximum performance.

Figure 1 Design flow for SCA-compliant waveforms.

Improved methodologies now revolve around a "model-based design" paradigm. PrismTech's Spectra CX can be used to model the SCA design. For the functional design, products like Agilent's SystemVue allow waveform components to be developed quickly. Functional design starts with algorithmic construction of various signal processing functions, and then follows with mixed-signal performance analysis of baseband signal processing with realistic analog, RF and channel models, incorporating environmental waveforms and measurements as needed. SystemVue's ability to provide continuous verification throughout the design process, with a gradual transition to hardware measurements, insures that metrics are being met for performance and standards compliance. Finally, improved methodologies allow the developed, optimized, and verified IP to be quickly and conveniently targeted via a SCA-compliant waveform implementation using automatic code generation in C/C++ and/or HDL wrapper interfaces, for porting to any number of signal processing HW targets. Commercial tools enable a closed-loop, model-based design process for SCA-compliant waveforms that includes RF and measurements, as well as baseband and algorithmic validation, so that SDR waveforms can be (re)deployed faster, with greater confidence and portability, than ever before. Such a flow is depicted graphically in Figure 1.

Using this improved methodology allows developers of advanced SDR waveforms to be de-coupled from their final hardware targets, allowing them to focus on high-performance design of the waveform components and overall radio application. It also facilitates a quick and convenient way to prototype and re-target a design to a new hardware platform, leading to faster deployment of radios.

Hardware Prototyping and Measurement
Different SDR waveforms often place conflicting requirements on the radio hardware platform, especially in the Analog/RF domain. For example, a frequency-hopping FSK waveform may need fast local oscillator (LO) switching and settling, whereas an OFDM waveform may require low phase noise. While it may be possible to design a fast-tuning LO with low phase noise, it will likely be expensive and power hungry. An alternative to the one-size-fits-all approach is to use adjustable radio technology. In this example, the LO tuning parameters may be adjusted one way to optimize LO performance for the frequency hopping FSK signal, and adjusted another way when operating under an OFDM signal. Similar arguments can be made for the power amplifiers, internal drive levels and IF bandwidths in these analog designs.

While adjustable hardware can improve performance, lower cost and increase battery life, it also implies that the hardware settings may need to be optimized for each waveform. Adjustable hardware introduces another complication, in that each adjustable component geometrically increases the number of radio configurations. A radio with just three independently adjustable components, each with three settings, could have up to 27 usable configurations. While it may be possible to adjust radio parameters to reasonable values, simply based on the waveform characteristics, a large number of simulations and measurements will likely be required to optimize radio performance over large combinatorial sets of radio settings.

The flexibility of SDR baseband processing enables the use of test waveforms. Test waveforms are less complicated than regular communication waveforms allowing them to be developed more quickly, and with specific goals in mind. For example, using some of the design methodologies described earlier, one might quickly develop a simple 64 QAM radio, for the sole purpose of making BER measurements on prototype hardware. Test waveforms may also be designed specifically to make specific types of hardware measurements easier. For example, a test waveform that generates a multi-tone signal with a specific peak-to-average power statistic could be used while making adjustments to the radio hardware. It might also simplify intermodulation distortion measurements in manufacturing. These low complexity waveforms are usually less demanding on baseband resources, making them easier to implement, and highly portable if implemented within the SCA framework.

The RF performance of a software-defined radio is a function of both the hardware and software, which leads to diagnostic complexities. For example, a radio that exhibits excessive adjacent channel power (a spectral mask violation) might suffer from insufficient numerical range or resolution in the signal processing, or it could have an analog amplifier problem. A high bit error rate could indicate a lack of receiver sensitivity, a bug in the receiver's algorithms, or a timing problem that only shows up when running on the radio platform.

Most likely, unless an SDR is seriously broken, it is difficult to isolate individual mechanisms for a particular problem. More likely, there will be several contributing factors, and each 1 dB performance problem is actually a combination of several sub-dB root causes. Identifying these root causes is exacerbated if the same measurement must be performed at different points in the radio, using different measurement algorithms. Ideally, identical measurement algorithms should be used at all stages in the signal path, and during all stages of development. Consistent measurement algorithms in the software can be used throughout the signal chain and design and verification processes, eliminating uncertainties that interfere with troubleshooting and root cause analysis of SDR performance issues that often cross domains (see Figure 2). If instead, different measurement tools are used, then something as seemingly simple as a power measurement on a digitally modulated signal can give apparently different results, due to differences in the shape of the RBW filters and the methods used to average results. Fortunately, it is possible today to make identical and comparative measurements on any signal, whether simulated or real, digital or analog.

Finally, it is possible to combine design analysis with actual waveform HW using "hardware in loop" (HIL) or, using a new term, "OE in the loop" to stream live test signals into the waveform HW/Firmware as it runs in an Operating Environment (OE) on the target. With PrismTech's SpectraOE and Agilent's SystemVue, such a link has been developed allowing direct simultaneous simulation and HW "in the loop" validation. Figure 3 illustrates this for a simple QPSK modulator running inside the SpectraOE.

Conclusion
The Software Communications Architecture is no longer a "science experiment" for SDR development. It is now a technology ready for the mainstream, whose maturity is being driven by the commercial off the shelf (COTS) products available for SDR implementers. The COTS vendor community is making SCA development productive, robust, high performance and, most importantly, affordable. For the first time the full potential of SDR is being delivered in commercially viable platforms, and a new breed of development tools is allowing rapid development of new and complex waveforms and radio applications both in aerospace/defense and commercial applications. Using these development tools, designers are moving from concept to deployed waveforms in months rather than years, with extreme portability that allows for future HW upgrades.

Reference

1. J. Mitola, "The Software Radio," IEEE National Telesystems Conference, 1992 - Digital Object Identifier 10.1109/NTC.1992.267870.