Commentary: August 2005

By John Pasquarette, National Instruments

To deal with escalating complexity effectively designers must shift their approach toward platform-based design. Platform-based design works across multiple domains of specialization seamlessly, and it minimizes data or code transformation. More importantly, platform-based design elevates physical test data from late in the process and applies it throughout the development process effectively. Every domain has specialized design tools. The challenge is that today’s designs contain disparate technology domains, and engineers must unite the different domains. For example, many designs incorporate multiprocessors with FPGAs or DSPs.

  Specializing in one target technology isn’t the answer anymore. Designers must mix and match different computing elements within design tools at a system level to more easily optimize cost, size,  performance, or development flexibility.

 

John Pasquarette, National Instruments

Too often, specialized tools may not solve enough of the problem to maximize design and implementation efficiency. Control engineers, for example, might a use a tool designed to simulate and optimize control algorithms. Or DSP experts might use signal-processing tools to design voice or data processing routines—but algorithms are only part of the problem. The next step is to totally re-implement the algorithm in C and add the error checking, I/O, user interface, and overall execution logic around the algorithms to complete the product. This requires a completely different engineer, or team, to develop.

Why can’t domain experts use a single design tool to develop the entire application, eliminating the translation step? Platforms such as my company’s—National Instruments—LabVIEW mix multiple domain-specific design approaches, such as control block diagrams, state charts, and interactive design tools, with a graphical programming language. The graphical language is intuitive enough for the domain expert to write the entire application as well as the specific algorithm. Because LabVIEW can target code to microprocessors, FPGAs, and real-time OSes, it can eliminate costly and error-prone middle steps.

It may be naive to think that one design tool can solve today’s complex design challenges, but, many designers desire this. Consequently,  companies spend a great deal of energy strategically choosing and   investing in an integrated toolchain to facilitate smooth reuse from design to simulation to implementation. Regrettably, companies often overlook the test element in this decision process.

This is a sizeable oversight because test plays a critical role in calibrating simulations. For example, road-load data collected from in-vehicle loggers and drive simulators provides realistic stimuli in design tests. Engineers can more easily and accurately measure startup transients in electronic circuits to drive simulations than they can simulating these conditions. Test engineers use test data to correlate simulation models and as inputs into a simulation.  Simulation, in turn, helps test engineers better optimize the test process.

In these examples, engineers balance physical test data with simulation tools to quickly verify if even the earliest prototype is on track. This approach yields a much stronger, tighter, and faster product development process.

Engineers can also use the open integration of a product like LabVIEW to exchange measurement test data. For example, LabVIEW exchanges data with SPICE electronic circuit simulators such as Multisim or PSPICE, with mechanical simulators from SolidWorks/COSMOS, and with algorithm design tools like MATLAB, Maple,and Mathematica.

The battle against complexity is one that engineers can win with a platform-based approach that balances the physical world of test data with the virtual world of design and simulation. With open, interlocking platforms between test and design, engineers can shift between the physical and the virtual easily. This leads to earlier fault detection,  faster design iterations, and better products.

John Pasquarette has an electrical engineering degree and joined National Instruments in Austin, Texas, in 1990. Send your thoughts on this commentary by clicking here. Please reference August 2005 Commentary.