Is your Test System up to the task?
22 July 2009
As device complexity increases, reduce time to market by adopting a commercial-off-the-shelf FPGA platform.
As electronic devices continue to grow in complexity, new challenges arise as we strive to validate designs. Is it now time for us to take a step back and evaluate how we architect our test systems? The continued development of multicore processors has been empowering engineers to do more through the adoption of a software-defined approach to instrumentation. Other technologies are also emerging, such as FPGAs, which take this one step further; providing a re-programmable platform upon which hardware determinism and reliability can be achieved. In this article we will be exploring how FPGAs and Commercial-Off-The-Shelf (COTS) hardware can play an important role in the rapid development of next generation systems and custom instrumentation.
We live in a converging world where a phone is no longer just a phone and the cars we drive are controlled by computers rather than mechanics. Electronic devices are everywhere and it is inevitable they will continue to become ever more complex. For the last 50 years CERN have been working at the forefront of technology and the Large Hadron Collider (LHC) is a perfect example of this. The LHC is the world’s largest particle accelerator. It uses collimators to prevent particles from straying from their intended paths which requires an incredibly complex control system. To implement this, CERN used a design platform based on COTS Field Programmable Gate Array (FPGA) hardware from National Instruments. When discussing the adoption of the National Instruments FPGA platform, Roberto Losito, Engineering Manager at CERN, addresses challenges we all face today. “We selected a design platform that incorporated only the features we needed without adding unnecessary cost and helped us avoid creating our own software drivers to reduce the manpower required to complete the system.”
To address these challenges, innovation is a necessity. Not only are we required to keep up with the complexity of our Devices Under Test (DUTs), but many of us are faced with shrinking budgets and time to market pressures. Now is a good time to re-evaluate how we implement our design and test process, to take a step back and see how we can do things differently. Consideration of COTS products is an important part of this process, and can be used to get to first prototype much faster.
For a long time now there has been a general trend towards software-defined, or virtual instrumentation. For an engineer, this brings flexibility and re-use, meaning that what was an oscilloscope, could become a frequency counter or a video analyser. It is inevitable that over time a product is going to evolve, as new functionality is required. Using the virtual instrumentation approach, our systems, be it test or embedded, can evolve easily too, as flexible software allows us to adapt to any changes in specification. When you think about taking a COTS approach to hardware, other benefits can also be seen. Take CPUs as an example. Leading chip manufacturers invest billions in R&D each year on new product development. It would therefore be uneconomic to design our own. As off the shelf technologies, such as CPUs, advance we can simply upgrade a specific part of our system to immediately take advantage of the technology improvement.
The result of CPU development in recent years has led to the multicore technology we see today. Alongside virtual instrumentation this has helped us to create ever more powerful test systems. Parallel processing power means that applications can be run faster and therefore achieve higher performance and greater DUT throughput.
Let us focus on the test side of things for a moment. As our DUTs increase in complexity, regardless of how fast tests can be performed, the traditional approach of using static test vectors is not enough. The difficulty of fully verifying our DUTs using traditional ATE has led to an increased demand for so-called protocol aware test, the ability to test devices by emulating the real-world signals connected to them. This is important as the DUTs we are required to test become more like a black box with multiple components within it. It is now increasingly difficult to pump test vectors directly into each component, we need to talk to the DUT over the protocol it was designed to use. By creating an intelligent test system that can simulate the real world environment, we are able to achieve complete test coverage on our complex devices.
In order to do this kind of next generation test, we can leverage a COTS technology previously mentioned above, the FPGA. FPGAs are reprogrammable silicon chips that can be configured to implement custom hardware functionality without ever having to pick up a breadboard or soldering iron. Software-defined hardware brings obvious benefits: determinism, rapid response time, reliability and flexibility. But is an FPGA really COTS technology? Maybe yes, but on its own an FPGA still requires a board design. PXI is an industry standard modular platform based on PCI technology that combines high performance embedded computing with I/O modules that are connected by a dedicated timing a triggering bus. FPGA modules are available for PXI, removing the need for investment in board design.
Going back to the CERN example, this is the platform they adopted, enabling them to take advantage of powerful, customisable FPGA technology whilst minimising development time and costs.
Traditional software tools have made the programming of FPGAs a complex task, restricting access to this technology. It has been limited to those digital designers who have been trained in using a hardware description language. The advent of high-level design tools such as National Instruments’ LabVIEW has allowed users to directly program FPGAs without the need to be a digital design engineer. This enables engineers to know a single design tool, and be able to quickly implement FPGA-based systems. LabVIEW FPGA abstracts complexity, meaning time can be focussed on functional design and algorithm development rather than low level details such as how to connect the logic blocks on the FPGA to give the desired functionality.
The benefits of COTS FPGAs are clear, although there can sometimes be limitations if your I/O requirements fall outside the specification of what has been pre-built onto the board. National Instruments’ FlexRIO is an extension of the NI FPGA product family that addresses this problem by allowing much tighter control over the I/O. The FlexRIO architecture consists of two components; an FPGA module, and an adapter module.
The FPGA module is based on the PXI platform, and gives all of the same advantages other NI off-the-shelf FPGA based instrumentation brings. Being able to design your own hardware by simply writing a software application provides performance and reliability whilst helping to address cost and time-to-market concerns. The FlexRIO board features a Xilinx Virtex-5 FPGA with up to 128 MB of onboard DRAM. With it being on the PXI platform, it also provides high-speed data streaming and synchronisation to other components of your system.
The adapter module is what defines the physical inputs and outputs of an NI FlexRIO system and this is both interchangeable and customisable. Different adapter modules are offered by National Instruments, third-party vendors, or they can even be custom built to your exact requirements through the adapter module development kit and your own PCB design tools. This open, customisable signal front end allows the exact requirements of your test system to be met. Specific analogue-to-digital converters, digital buffers, connectors, and even specific channel counts can be designed to work directly with a LabVIEW-programmable FPGA target.
Rather than developing, validating and supporting a full custom design, you can now take commercially available platforms to quickly develop a complex test system. Adopting a flexible platform not only allows for future upgrades as test requirements become more demanding, but it also delivers a greater return on the investment in both time and money. The same platform used for test, could now also be used in design process. A FlexRIO board could be used to prototype any kind of digital device due to the processing power, ease of programming and I/O connectivity. Intellectual property created during product development could then be re-used in the test system. This would allow us to effortlessly add the intelligence needed to get complete test coverage.
Ultimately, combining a single graphical design tool that can target devices from multicore PCs right down to FPGAs, with COTS hardware, we are able to develop intelligent test applications which would typically require a complete custom design. This allows us to get these designs prototyped and deployed faster and at a lower cost.
Richard Silley is a Technical Marketing Engineer at National Instruments UK & Ireland.
Contact Details and Archive...
Related Articles...
Most Viewed Articles...