Modules accelerate use of radio systems software

The wireless marketplace is like our nascent universe. A veritable ‘big bang’ of new radio frequency technologies has emerged creating opportunities for solving old problems in innovative, new ways.

Flexible, high-resolution waveform generation, digitisation and analysis subsystems capable of manipulating RF signals in conjunction with down-conversion and tuning multiple ‘regions of interest’ are required.

Subsequent, real-time, multichannel demodulation of these regions using a wide variety of schemes is necessary. Often, such equipment must be portable and operate under harsh conditions creating tremendous challenges in packaging, power consumption and management.

Existing solutions employ arrays of dedicated digital signal processors working in tandem with an RF digitiser to provide the computational bandwidth needed to implement down-conversion and demodulation functions.

Though effective, this is complicated and expensive since multiprocessor programming requires sophisticated process management and load balancing while avoiding race conditions and data bottlenecks.

New solutions are emerging as highly modular devices that use the industry-standard, commercial-off-the-shelf COM-Express PC architecture and development tools in conjunction with PCI Express-based XMC mezzanine modules to create customisable RF processing block solutions.

The performance of mainstream DSP devices has effectively stagnated. The system clock on such devices is currently limited to 1 GHz or less, with bandwidth limited to about 800 MBps throughput from the common 100 MHz, 64-bit external bus. By contrast, the x86 architecture continues to evolve, adding instruction set optimisations, enhanced caches, floating point co-processing and multiple cores-on-a-chip with no abatement in sight.

Octal core processors featuring 3 GHz processing cores and 10 GBps external bus bandwidth are ubiquitous. Moreover, the Intel performance primitives support native signal processing on an x86 processor faster than existing DSP devices with the added convenience and accuracy of 80-bit floating-point capabilities.

However, a desktop or industrial PC does not meet the portability, packaging or environmental requirements of many embedded applications.

Fortunately, due to the immense popularity of the PC as a development and processing tool, the market has already responded with several small form factor PC standards which are well suited to creating embedded, portable instrumentation such as COM-Express.

This format has become a de facto standard among users requiring the utmost reliability, scalability, portability and computational performance.

COM-Express modules are commodity items readily available. The modules are small mezzanine types which are mounted onto a carrier board that is customised to meet specific application requirements.

Innovative Integration has developed a special carrier board to meet the stringent requirements of embedded RF processing.

Packaged into the company’s new eInstrument, a COM-Express module, I/O expansion modules and an array of integrated peripherals combine to create a small form factor, rugged PC which can be embedded within OEM equipment to create intelligent, autonomous instrumentation, servo control or RF processing nodes.

Figure 1: X5-210M PCI Express PMC module with RF digitising front-end.

Any mix of the standard peripherals typically found on a PC can be made available in an eInstrument-based system, including ethernet communications, disk drives, USB and SATA ports. Keyboard and video ports can be provided to aid in field diagnostics.

For instance, it is entirely feasible that an eInstrument assembly in a remote site in a foreign country could be accessed via the internet using remote desktop or VNC by tech support staff at the corporate offices to provide interactive support or software upgrades.

While the cost and computation advantages of a COM-Express PC compared with traditional chip-level DSP solutions are enormous, a critical side benefit of the COM architecture is its ability to use the existing body of development and debugging tools available for the PC.

Whereas Texas Instruments or Analog Devices are sole sources for compiler and debugger tools for their DSP devices, the PC market sports thousands of well-established providers offering sophisticated, useful, mature tools featuring superior performance and reduced cost.

Moreover, conventional desktop PCs may be used to host these tools, further accelerating and simplifying the development of COM-Express-based products.

Though powerful computationally, a COM-Express PC does not provide direct support for acquisition or analysis of RF analog signals. Moreover, the multicore x86 CPUs available now and in the foreseeable future do not offer sufficient bandwidth to process RF signals in real time.

Consequently, some form of I/O and processing expansion is required. Just as COM-Express provides a PC repackaged into a mezzanine card format, I/O cards are available in a small, rugged form factor suitable for use in embedded instrumentation.

This format is known as PMC (PCI mezzanine card). PMC modules support PCI or PCIe (PCI Express) bus communications, identical to that used in standard desktop PCs but packaged in a small, rugged format.

PCI Express, the successor to the ubiquitous PCI bus, is the ratified standard replacement for I/O expansion in the PC industry. PCIe features software compatibility with enhanced throughput up to 64 times faster than PCI. Additionally, the bus features guaranteed QoS (quality of service) and P2P (point-to-point) data flow capabilities, making it suitable for real-time applications.

The eInstrument COM-Express PC performs initialisation, supervisory control, user-interface as well as high-performance computational duties in RF processing applications.

PCI Express-based PMC modules provide I/O and digital signal processing expansion, as illustrated in the block diagram (see Figure 1).

Figure 2: Diagram of COM-Express-based eInstrument carrier card.

User interface devices, such as the keyboard, mouse and display, may be exposed as diagnostic ports or omitted entirely. Alternatively, access to embedded user interface controls such as buttons, graphical OLED displays, etc which are commonplace in embedded instrumentation can be provided.

Similarly, the USB and hard disk interfaces may be hidden, omitted in the initial design or presented as optional devices available only on designated systems.

The eInstruments-based PCs run standard Windows or Linux variants such as XP, Windows 7 or OpenSuse to fully use drivers available for existing PMC modules. These PCs are provided with C++ libraries that exploit high-performance signal processing features provided in the optimised Intel performance primitives library yielding DSP functionality and performance running on standard x86 platform.

Figure 3: The eInstrument embedded PC with modular DSP packaged into a 7 x 10 x 3″ chassis.

Two PCI Express XMC module sites are provided for I/O expansion. One is typically used to host the PMC module which implements the RF front-end analog input and output and FPGA-based digital signal processing capabilities. The second site is uncommitted and available for future expansion.

A bevy of PCIe-compliant PMC modules compatible with these sites is available to provide additional capabilities such as fibre channel ethernet communications, auxiliary voice or ultrasonic-band analog channels or additional FPGA resources.

Each eInstrument site features four or eight 2.5 Gbps PCI Express I/O lanes, which is essential to support sustained high-speed data transfers. Sustained module-host transfers at 1 GBps are readily achievable, even under non-real-time operating systems such as Windows XP or Linux.

Additionally, the two sites provide eight dedicated communications lanes to allow implementation of algorithms in which large volumes of data are shared between modules. Even accounting for lane inefficiencies, sustained inter-site data rates of 1.3 GBps are realised.

The PCI Express bus provides good, high-bandwidth connectivity between the COM-Express CPU and the PMC module. Given the 1 GBps sustained throughput, the interface has sufficient bandwidth for low-bandwidth down-converted baseband data plus plenty of additional bandwidth in reserve should it become desirable to capture or log raw IF data in future applications.

PCIe also supports fast random, asynchronous I/O accesses to peripheral registers on XMC modules, to accommodate operations such as filter coefficient uploads, DDC channel tuning and the myriad other operations typically required in software radio applications. Typical, individual slave-type accesses will complete in under 1 µs using a modern COM-Express module.

Innovative’s X5 module product family combines what is claimed to be the most powerful FPGA ever offered by Xilinx - the Virtex 5, with a variety of RF-speed, high-resolution analog I/O devices, packaged as PMC/XMC modules.

These products combine up to four channels of high-resolution analog input and/or output plus an FPGA-based signal processing core that is capable of performing the real-time signal digitising, data buffering and signal processing required for RF processing applications.

These XMC modules support conduction-cooled operation in accordance with the VITA 20 mechanical specification. Additionally, the standard logic includes provisions for continuous temperature monitoring.

Software may receive alert messages whenever the temperature exceeds a programmed warning threshold and the logic is configured to automatically shut down if the temperature exceeds a programmed failure threshold temperature.

This combination of thermal management ensures good real-world, in-field reliability.

Custom firmware for the FPGA may be built using standard IP cores and fully modelled under MATLAB, which facilitates high performance and accelerated time-to-market for embedded applications. Custom firmware for the Virtex 5 FPGA builds on the FrameWork Logic, provided by the vendor, to interact with the onboard analog devices, DDR and QDR memory pools and PCI Express bus interface.

The firmware works in conjunction with PC-based software tools and C++ libraries, providing a software development system for integration of the PMC with the host application.

To provide optimal AC performance, the high-speed RF analog input circuitry must be driven using a stable, low-jitter sample clock. The onboard clock circuit is a derivative of the company’s X3-timing module that has <100 femtoseconds RMS jitter for 6.25 to 1000 MHz clock range, long-term thermal stability and integrated clock drivers capable of simultaneously sourcing into 50 Ω loads on each of the XMC sites and external devices through the EXT CLK connector.

In some applications it is desirable to synchronise the module sample clock with world time as provided by a GPS circuit. To accommodate these requirements, the eInstrument carrier features an integrated GPS receiver and sample clock timebase circuitry.

Control logic embedded into the carrier FPGA servo-locks to the epoch (1 pps) output events produced by the GPS receiver, ensuring that eInstrument PCs in disparate locations across the world start acquisition and sample synchronously to within 1 µs.

X5 PMC modules are engineered to support RF signal processing applications with minimal external circuitry and with no modification of the X5 bus interface or PCIe back-end infrastructure.

For instance, the diagram shows the X5-210M functional block. Features of the module are its high-performance analog front end with four channels of 250 MSPS, 14-bit A/D, a processing core built around the Xilinx Virtex5 Pro FPGA and memory, sample rate clocking and synchronisation, and a high-performance packet-protocol PCI Express interface for system integration.

Figure 4: X5-210M block diagram.

The benefits are four channels of 14-bit, 250 MSPS A/D digitising high dynamic range and AC performance Virtex 5 SX95T FPGA down-conversion, multiple channels, four input sources 100 dB SFDR DDS with 48-bit tuning, 256 tap channel filter and 512 MB DDR DRAM data buffer. All data buffering is on-card 4 MB SRAM array storage coefficient storage memory that supports multiple vector sets. External clock input or onboard crystal onboard oscillator may be ultrastable at an application-specific frequency.

As with all X5 modules, the personality of the 210M FPGA is user programmable using either HDL or MATLAB using Xilinx system generator. Typically, the FPGA is modified to implement independent down-conversion channels, filters, FFTs and other operations which much be performed at IF frequencies within the FPGA to form the basis for baseline RF tuning functionality.

The eInstrument COM-Express PC performs initialisation, supervisory control, user-interface as well as high-performance computational duties in such RF processing applications. The MATLAB board support package for the X5-210M allows signal processing to be developed using MATLAB/SimuLink.

This is used to model the signal processing for bit-true, cycle-true design, which may then be directly tested using hardware-in-the-loop features for hardware testing.

This allows the signal processing to be developed at a high level, using Xilinx IP cores, and tested in the MATLAB environment. This technique reduces risk and shortens development time by allowing efficient and thorough verification of the signal processing from within SimuLink.

The signal processing logic core from MATLAB is then integrated into the FrameWork HDL for the final logic design.

The FrameWork package provided with the module provides the hardware interface and support functions such as the A/D interface, memory controllers, host data interface and controls. All standard logic features such as A/D interface, triggering, multi-queue data buffering, DDC control and PCI controller interface are provided as components which must be augmented with custom logic blocks usable in either SimuLink or Xilinx ISE, to form the foundation for the user application firmware.

Ordinarily, the basis for the desired, application-specific signal processing functions can be provided by the module manufacturer or from engineering firms specialising in developing IP.

Such firmware may implement customisable digital down-converters, optimised, high-resolution FFT processing blocks capable of operating at a sustained rate of greater than 100 MHz or other capabilities dictated by the application. This is delivered within an extensive training session in which the developed technology is transferred to the client engineering staff.

This is claimed to be the most cost- and time-effective development process. Armed with this infrastructure, engineering teams ‘hit the ground running’ and have little trouble modifying the exiting code to meet application-specific requirements.

In summary, ultra small form factor PCs allow the creation of a new breed of embedded instrumentation using COTS hardware to lower system cost and improved availability. COM-Express processor modules provide scalability in host processing power for current and future products.

Use of PCI Express PMC modules allows integration of very high-performance FPGA-based computational engines which can be dynamically loaded with customised firmware to address changing RF processing requirements and markets.

Innovative Integration

www.innovative-dsp.com