It's a brave new world for DSP instrumentation designers

As Albert Einstein said, "We cannot solve our problems with the same thinking we used when we created them." The wireless marketplace is like our expanding universe. Currently, a big bang of new RF technologies is underway, posing new challenges.

Among the difficulties faced are flexible high-resolution waveform generation, digitization, and analysis subsystems capable of manipulating RF signals along with digital down-conversion and tuning multiple regions of interest. Subsequently, new schemes are needed for demodulating these regions in real time. Moreover, the hardware must operate in harsh environments, exist in small form factor packaging, consume minimal power, and incorporate thermal management.

Existing systems employ arrays of dedicated DSPs working in tandem with an RF digitizer to provide the computational bandwidth needed to implement digital down-conversion and demodulation functions. Although effective, these systems are complicated and expensive because multiprocessor programming requires sophisticated process management and load balancing while avoiding race conditions and data bottlenecks.

Mainstream DSP devices' performance has effectively stagnated. System clocks on such devices are currently limited to 1 GHz, 800 MBps across a 100 MHz, 64-bit external bus. By contrast, instruction set optimizations, enhanced caches, 80-bit floating-point coprocessing, and multiple on-chip cores are commonplace in the x86 world, offering 3 GHz processing and 5 GBps external bus bandwidth.

In terms of software, the Intel Integrated Performance Primitives and Thread Building Blocks libraries support optimized native signal processing an order of magnitude faster than existing DSP devices. However, a desktop or industrial PC with general-purpose processors simply does not meet the portability, packaging, or environmental requirements of many embedded applications.

New COTS modules emerging

Fortunately, the PC market has responded to this dilemma with several small form factor Computer-On-Module (COM) standards well suited to create embedded, portable instrumentation.

The COM Express format has rapidly become the de facto standard among users requiring the utmost in reliability, scalability, and portability. A COM Express module essentially is a highly integrated SBC, as shown in Figure 1.

Figure 1 (click graphic to zoom by 1.4x)

COTS COM Express modules are available from reputable electronics vendors such as Kontron, RadiSys, DTI, and Advantech. The COTS COM Express PC architecture and related development tools easily interface with proprietary OEM carrier boards for application-specific deployments. Embedded COM Express modules offer designers numerous advantages, including:

• Faster time to market
• Better control over form, fit, and function
• Reduced development cost and risk
• Lower total cost of ownership through scalability

The cost and computation advantages of a COM Express PC compared to traditional chip-level DSP devices are enormous. Additionally, the PC marketplace offers excellent development and debugging tools featuring superior performance and low cost.

Although powerful computationally, a COM Express PC does not provide direct support for real-time RF analog signal acquisition or analysis. Moreover, the multicore x86 CPUs available in the foreseeable future will still not offer sufficient bandwidth to process full-rate RF signals. Consequently, some form of I/O and FPGA RF processing expansion is required.

Customizable board for embedded DSP designs

A carrier board capable of interfacing a COM Express module with I/O expansion components and other integrated peripherals is needed to create a rugged small form factor PC. This type of carrier board could be embedded within OEM equipment to create intelligent instrumentation, servo control, and RF processing nodes.

Just as COM Express provides a PC repackaged into a mezzanine card format, I/O cards in small, rugged form factors are available for embedded instrumentation. XMC modules (compliant with VITA 42) now support the new PCI Express full-duplex serial bus. The XMC cards feature mechanical and software compatibility with older PMC modules, plus dramatically enhanced throughput up to 64x faster than the older standard. Furthermore, PCI Express features guaranteed quality of service and point-to-point data flow capabilities for real-time applications.

Innovative Integration has developed a highly customizable carrier board to meet the stringent requirements of embedded RF processing. The carrier board is packaged into the company's new eInstrument product line as shown in Figure 2.

Figure 2 (click graphic to zoom by 1.4x)

The carrier integrates an Intel-based COM Express board with a pair of high-performance XMC sites. The primary site typically hosts an advanced XMC module, which implements RF front-end analog I/O and FPGA-based DSP capabilities. The second site is uncommitted and available for future expansion. Numerous vendors provide additional capabilities, such as Fibre Channel communications, auxiliary voice or ultrasonic band analog channels, or extra FPGA resources. USB, SATA, and video ports are also available for expansion, remote access, and field diagnostics.

The primary XMC site features eight 2.5 Gbps PCI Express I/O lanes between the host CPU and FPGA module, yielding 1 GBps sustained data transfer rates even under non-real-time operating systems such as Windows XP or Linux Suse 10.2. The secondary site features four I/O lanes and one-half the bandwidth. The sites interconnect via eight dedicated Xilinx RocketIO communication lanes dedicated to allowing implementation of algorithms in which large volumes of data can be shared between modules. Sustained inter-site data rates of 1.8 GBps are possible. Figure 3 shows a block diagram of how the COM Express card interfaces with the two XMC sites on the carrier card.

Figure 3 (click graphic to zoom by 1.5x)

Given the 1 GBps sustained throughput, site 0 has sufficient bandwidth for low-bandwidth digital down-converted baseband data plus plenty of additional bandwidth in reserve should it become desirable to capture or log raw intermediate frequency data in future applications.

The PCI Express standard also supports fast, random, asynchronous I/O access to peripheral registers on the XMC modules to accommodate operations such as filter coefficient uploads, digital down-conversion channel tuning, and myriad other software radio applications. Typically, CPU memory access will complete in less than 1 microsecond using a modern COM Express module.

All in the XMC/PMC family

Innovative's new X5 module product family employs the powerful Xilinx Virtex-5 FPGAs. These modules feature a variety of high-resolution analog I/O devices packaged as industry-standard XMC/PMC modules. These products combine up to four channels of wide bandwidth analog I/O plus an FPGA-based signal processing core capable of performing the real-time signal digitizing, data buffering, and signal processing required for RF DSP applications. Figure 4 shows a typical X5 PCI Express module.

Figure 4 (click graphic to zoom by 1.4x)

These modules support conduction-cooled operation per the VITA 20 conduction-cooled PMC mechanical specification. The standard logic continuously monitors temperature, ensuring excellent in-field reliability. Also, the modules comply with the green European RoHS standard.

Custom firmware for the FPGA can be built using IP cores and The MathWorks' MATLAB modeling language, yielding high performance and accelerated time to market for embedded applications. Custom Virtex-5 firmware builds on supplied VHDL source code to interact with the onboard analog devices, DDR and Quad Data Rate (QDR) memory pools, and PCI Express interface. The firmware works with PC-based software tools and C++ libraries, providing a comprehensive software development system for integration into host applications.

To guarantee optimal analog performance, a low-jitter sample clock drives the RF input circuitry on the XMC modules. The onboard clock circuit exhibits 260 fempto-second (fs) root mean-squared jitter and excellent thermal stability. Alternately, an external clock capable of driving a 50 ohm load may be supplied. An integrated clock driver preserves all peripherals' integrity.

The eInstrument carrier features an integrated GPS receiver and sample clock time base circuitry. Control logic on the carrier FPGA servo-locks to the epoch (1 pps) events produced by the GPS. Thus, COM Express PCs located throughout the world can start acquisition and sample synchronously to within 1 microsecond.

Faster, cheaper – and smaller

It's a brave new world for embedded DSP designers. It takes courage and imagination to risk the R&D costs necessary to keep up with emerging technologies. Designers must be mindful of Einstein's admonition to constantly find new solutions to the problems we create. As the challenges grow bigger, the solutions must become faster, cheaper – and smaller.

James Henderson is cofounder and president of Innovative Integration, based in Simi Valley, California, a leading manufacturer of data acquisition and signal processing electronics and software since 1988. Previously, he was a robotics and automation engineer for Northrop Grumman and Litton Industries. James earned his BSME from Valparaiso University of Indiana.