FPGAs Help Software-Defined Radios Adapt
With the proliferation of wireless standards, future devices will need to support multiple air interfaces and modulation formats. Software-defined-radio (SDR) technology enables such functionality. It uses a reconfigurable hardware platform across multiple standards. Because of these capabilities, software-defined radio has been touted as the superior—although as of yet unattainable—solution for base stations. As field-programmable-gate-array (FPGA) technology and intellectual-property (IP) cores evolve, however, SDR is increasingly becoming a reality.
Software-defined radios require the support of multiple modulation formats or waveforms. To develop each of these waveforms, the designer must surrender significant design and debug time. He or she also must have expertise in waveform architecture, hardware, and software. Typically, the architecture will be split across a general-purpose processor (GPP), digital signal processor (DSP), and dedicated hardware (implemented in the field-programmable gate array).
Now, FPGAs can take on more of the computational tasks that are required in waveform implementations. Field-programmable gate arrays have associated tools that help the system designer quickly implement, simulate, and test each of his or her waveforms in the SDR system. These system-level tools are used to abstract the details of HDL coding into a higher-level, modular, block-architectural design. This step is aided by the ready availability of communications, DSP, and embedded-processor IP cores, which can be used in the functional design.
Lastly, the IP cores are integrated into the system-level tools. This integration, when combined with the interworking of the tools, allows the rapid handoff from the implementation phase to the simulation phase to the testing phase and back until the waveforms are complete. The rapid handoff therefore translates into a faster development cycle for the designer. Often, it decreases the time to final implementation by 50% or more.
Software-defined radio is the underlying technology behind the Joint Tactical Radio System (JTRS) initiative. This initiative focused on developing software-programmable radios that could enable seamless, real-time communication across the U.S. military services and with coalition forces and allies. The functionality and expandability of JTRS is built upon an open architecture framework, which is called the software communications architecture.
JTRS terminals must support the dynamic loading of any one of over 25 specified air interfaces or waveforms. Typically, those waveforms are more complex than the waveforms that are used in the civilian sector. FPGAs have the necessary processing power and flexibility to address such requirements. At the same time, they stay within the power budget allotted to the system.
With its increased flexibility and communications security, SDR also is well suited for applications outside of the military arena. Through a combination of in-the-field hardware and software reconfigurability, SDR technology must support multiple air interfaces and modulation formats—just as future wireless devices must support them. The ideal platform for implementing SDR consists of programmable logic in conjunction with state-of-the-art design software and a comprehensive portfolio of intellectual property. Low-cost, high-performance FPGAs are now delivering the capabilities that are required to cost effectively implement SDR applications. This article will describe an optimized FPGA-based SDR system.
FIGURE 1 illustrates the hardware partitioning of an SDR-based 3G base station. This type of base station can be reconfigured to support multiple standards. In order to reconfigure the entire system, however, an SDR base station would ideally perform all of the signal-processing tasks in the digital domain. Yet current-generation wideband data converters cannot support the processing bandwidth and dynamic range that are required across different wireless standards. As a result, the analog-to-digital converter (ADC) and digital-to-analog converter (DAC) are usually operated at an intermediate frequency (IF). Separate wideband analog front ends are used for subsequent signal processing to the radio-frequency (RF) stages.
Digital IF extends the scope of the digital signal processing beyond the baseband domain and out to the antenna—to the RF domain. This approach increases the flexibility of the system while reducing manufacturing costs. Compared to traditional analog techniques, digital-frequency conversion provides greater flexibility and higher performance (in terms of attenuation and selectivity). FPGAs provide a highly flexible and integrated platform to implement computationally intensive, digital IF functions including digital up- and/or downconverters. At the same time, they reduce the risk involved in introducing new techniques, such as DPD, CFR, and smart antennas.
Often, data formatting is required between the baseband-processing elements and the upconverter. Such formatting can be seamlessly added at the front end of the upconverter (FIG. 2). This technique provides a fully customizable front end to the upconverter. It also allows for the channelization of high-bandwidth input data. Custom logic or a soft-core embedded processor, such as the Nios processor, can be used to control the interface between the upconverter and the baseband-processing element.
In digital upconversion, the input data is baseband filtered and interpolated before it is quadrature modulated with a tunable carrier frequency. To implement the interpolating-baseband, finite-impulse-response (FIR) filter, one must first make speed-area tradeoffs. Such tradeoffs will uncover the optimal fixed- or adaptive-filter architectures for a particular standard. In addition, numerically controlled oscillator cores can be used. These cores will generate a wide range of architectures with very high performance and spurious-free dynamic ranges in excess of 115 dB. Depending on the number of frequency assignments that must be supported, multiple digital upconverters can be easily instantiated in a programmable-logic device.
Crest factors also must be taken into account. Third-generation (3G) code-division multiple-access (CDMA)-based systems and multi-carrier systems, such as orthogonal frequency division multiplexing (OFDM), exhibit signals with high crest factors (peak-to-average ratios). Such signals drastically reduce the efficiency of the power amplifiers (PAs) that are used in the base stations.
The third-generation standards and their high-speed mobile-data versions also employ non-constant envelope modulation techniques. Examples include quadrature phase shift keying (QPSK) and quadrature amplitude modulation (QAM). These techniques place stringent linearity requirements on the power amplifiers.
In the FPGAs that contain DSP blocks, DPD linearization techniques can be effectively implemented. Such techniques include both lookup-table and polynomial approaches. The multipliers in the DSP blocks can reach speeds up to 380 MHz. They also can be effectively time-shared to implement complex multiplications. When the FPGA is used in SDR base stations, it can be reconfigured to implement a DPD algorithm. That algorithm will efficiently linearize the power amplifier that is used for a specific standard.
On the receiver side, digital intermediate-frequency techniques can be used to sample an IF signal and perform channelization and sample-rate conversion in the digital domain. High-frequency IF signals (typically 100+ MHz) can be quantified using undersampling techniques. For SDR applications, remember that different standards have different chip/bit rates. As a result, non-integer sample-rate conversion is required to convert the number of samples to an integer multiple of the fundamental chip/bit rate of any standard (FIG. 3).
To support higher data rates, wireless standards are continuously evolving through the introduction of advanced baseband-processing techniques. Such techniques include adaptive modulation and coding, space-time coding (STC), beamforming, and multiple-input multiple-output (MIMO) antenna techniques. To support such computationally intensive algorithms, the baseband-signal-processing devices require enormous processing bandwidth. FPGAs have proven successful in such applications. For example, field-programmable gate arrays have been used in channel coding for HSDPA and beamforming.
In addition, the baseband components need to be flexible. They must enable the SDR functionality that is required to support the migration between enhanced versions of the same standard as well as the capability to support a completely different standard. When programmable logic is coupled with soft-core-processor and IP blocks, it also makes it possible to provide remote upgradability in the field. FIGURE 4 illustrates an example scenario in which FPGAs can be easily reconfigured to support the baseband transmit functions for either W-CDMA/HSDPA or IEEE 802.16a standards. Such support is garnered through the available IP functions, such as the turbo encoder, Reed-Solomon encoder, symbol interleaver, symbol mapper, and Inverse Fast Fourier Transform (IFFT).
THE TABLE gives a condensed list of waveforms and the estimated logic usage that must be implemented. It is important to note that there is a wide range in the size of logic that is needed for each of these waveforms. Some newly proposed waveforms significantly increase this logic.
By using hardware for waveform processing, one gains the computational efficiency of dedicated hardware over software. If these waveforms were implemented in software, they would require processors with hundreds of MIPS of processing power. Although such power is practical for fixed applications, it becomes an intractable problem for mobile or man-pack applications as required in the JTRS cluster 5 specification.
Often, SDR baseband processing requires both processors and FPGAs (FIG. 5). In this type of scenario, the processor handles system control and configuration functions. The FPGA implements the computationally intensive, signal-processing datapath and control. It thereby minimizes the latency in the system. To go between standards, the processor can switch dynamically between major sections of software. Meanwhile, the FPGA can be completely reconfigured, as necessary, to implement the datapath for the particular standard.
FPGA co-processors interface with a wide range of digital-signal and general-purpose processors. In doing so, they provide increased system performance and lower system costs. Meanwhile, the freedom to choose where to implement the SDR algorithms adds another layer of flexibility.
FIGURE 6 shows the results for a small portion of the digital-predistortion algorithm done as software only and as a software-controlled, hardware-acceleration coprocessor. For this application, a 10X improvement in speed was observed. Keep in mind that improvements are extremely dependent upon the application and the architecture. Clearly, however, raw potential is available to do SDR waveform processing in a combination of both software algorithms and hardware coprocessors. This potential shows the immense flexibility that is inherent to FPGAs.
Interest in software-defined radios is currently growing. This growth is being spawned by mandates from the military as well as the need to use available spectrum more efficiently. At the same time, the need for flexible solutions that utilize both software and hardware is increasing. When they're used in conjunction with soft-core microprocessors, IP cores, and hardware acceleration coprocessors, FPGAs succeed in offering one method of implementing flexible software-defined radios.