The value of modern automobiles has become heavily influenced by their electronic content. Consequently, selecting the right electronic components and choosing the optimal design methodology is vital in developing a successful product. The flexibility of new components, such as FPGA devices, is intriguing. The potential of these devices, however, cannot be fully (and safely) utilized without incorporating the latest design and verification methodologies.
The widespread use of FPGA devices in automotive applications has not yet arrived, but trends show that their potential advantages have not gone unnoticed(1). The capacity of these devices to implement and integrate both software and digital-hardware functionalityon a single componentis very attractive. Challenges remain, such as ensuring that these devices are compatible with harsh automotive environments and are compliant with the exacting reliability requirements of the industry. The biggest challenge in utilizing FPGA devices, however, may be one of methodology.
Design methodologies for automotive applications must consider the complexities of mechatronic(2) systems. Even something as simple as an electronic throttle control system (see Figure 1) is a sophisticated combination of feedback control systems, analog and digital circuitry, multi-physics sensors and actuatorsall controlling an electromechanical physical device (the throttle body). The importance of unambiguous, verifiable system requirements to the success of an automotive electronic product cannot be overemphasized.
FPGA designers are familiar with HDL-based, requirements-driven design methodologies for digital electronics. But how can requirements be expressed for a system that, while it contains digital elements, is fundamentally non-digital? Fortunately, an executable HDL exists that extends the capabilities of the digital VHDL language with continuous time, differential and algebraic equations, multi-physics, transfer functions (both s and z domain), energy conserving analog circuit capabilities (like SPICE), statistical distributions for parametric variations, and functions expressed in software C code. This language is the IEEE Std. 1076.1 VHDL-AMS language. VHDL-AMS is the perfect language for providing continuity in design and verification at all levels: functional specifications; architectural partitions; and component implementations (see Figure 2).
The VHDL-AMS language standard was completed in 1999. The description of this language sounds ideal, so why aren't more designers using the language today? Simply put, implementing the standard has been very difficult technically. Now, however, after years of development, several different tool suppliers are providing simulators that can efficiently execute the VHDL-AMS language. The long-awaited promise of this language standard and the resultant methodology is now a reality.
Digital designers at major automotive suppliers, such as Magneti Marelli(3) , have confirmed significant benefits by using the VHDL-AMS language. Since VHDL-AMS is a pure superset of the VHDL language, the designer starts with all of the well-known benefits of HDL design and verification. Then, using the extensions provided by VHDL- AMS, the design can be thoroughly analyzed by incorporating the impact of the neighboring engineering disciplines: analog electrical engineering (Kirchoff's current and voltage laws), ADC, and DSP circuits; control system transfer functions; mechanical engineering (Newton's and Bernoulli's laws); and extensibility any other desired engineering or physics discipline.
To be specific, VHDL-AMS allows expression of simultaneous, nonlinear differential and algebraic equations in any model; the model creator need only express the equations and let the simulator solve them in time or frequency domain. Domain knowledge from any engineering discipline can be encapsulated in reusable libraries(4) that are accessible by any member of the design team. It is then possible for the digital developer to start with a clear, executable specification that incorporates all of the requirements (including non- digital) and to use the same specification as a virtual verification environment. Since VHDL-AMS supports the concept of component statistical distributions(5), it is also practical to verify that the digital design will operate in the non-deterministic environments common to automotive applications.
The VHDL-AMS language is an undiscovered asset for FPGA designers a powerful tool to define and verify requirements in a non-digital context.