User Stories
Texas Instruments Streamlines Research and Development with Simulink and DSP Tools
Download this user story 1502k
|
|
|
Challenge
With a new focus on specialty chip sets for DSP customers, R&D engineers at TI faced the market pressures and competition that are now a way of life throughout the semiconductor industry. To reduce these pressures, TI addressed one of the biggest hurdles in the development process—the gap between R&D and product development.
"We see the future of advanced DSP design in development tools like those from The MathWorks."Dr. Randy Cole
Texas Instruments
DSP Solutions Research and Development Center
Solution
TI decided that rapid prototyping environments, consisting of high-level software tools and real-time development hardware, would be essential if they were to accelerate deployment of new technologies for their expanding line of increasingly sophisticated DSP products.
Engineers at TI now incorporate MATLAB, Simulink, the Signal Processing Blockset, and the Signal Processing Toolbox into their algorithm design process. This new approach enables them to streamline the verification and delivery of application-specific algorithms to product development teams.
“Using DSP Workshop helps tie the two sides of the process together in a way that lets every engineer see the whole project from concept to testing to coding,” says Dr. Randy Cole, branch manager at TI DSP Solutions R&D Center.
Dr. Cole’s team is using MATLAB, Simulink, the Signal Processing Blockset, and Real-Time Workshop—along with TI development tools—in a rapid-prototyping approach based on interactive block-diagram simulation and automatic code generation. With these tools, DSP engineers can refine implementation details directly in the system model and produce real-time software prototypes without traditional programming.
By using MathWorks tools to design real-time algorithms and generate code for use on target DSP platforms, TI expects to dramatically shorten development cycles. “The MathWorks DSP Workshop is a promising tool suite for rapid DSP prototyping,” states Alec Robinson, Digital Audio Branch, DSP Solutions R&D Center. “It demonstrates a significant potential benefit to R&D engineers, who are developing sophisticated applications for advanced TI DSP product lines.”
TI engineers used MathWorks tools to create a real-time audio application. The software runs on a PC/C32 card from Blue Wave Systems. It feeds real-time stereo sound inputs through a series of sophisticated audio processing algorithms and delivers stereo output to a set of loudspeakers for real-time subjective listening tests. They developed and tested algorithm concepts in MATLAB and converted them seamlessly into a block-diagram model of the real-time implementation in Simulink and the Signal Processing Blockset.
The graphical, self-documenting algorithm design in Simulink becomes an executable specification for product designers. Using the Simulink model as a reference, software designers can work concurrently on implementation details and constraints, such as processor speed and precision.
Dr. Cole’s team used Real-Time Workshop to generate C code to run in their real-time environment, as well as TI software that supplied host-DSP communication and stream I/O support for their specific card’s A/D and D/A devices. The C code generated by Real-Time Workshop provides a solid basis for the optimized software ultimately embedded in the product.
Results
- Accelerated design. Using MathWorks tools, TI engineers created a real-time audio application in just one day. Traditional design methods would have taken weeks of DSP programming—particularly if iterations were required.
- Efficient time allocation. MathWorks tools helped R&D engineers create and evaluate their own real-time prototypes, enabling TI to shorten DSP software engineering time on product implementation and optimization.
- Reduced overhead. Because the entire DSP engineering team can use the same interactive MathWorks products to create, refine, and verify algorithms, they can better understand algorithm behavior and performance much earlier, reducing implementation errors from misinterpretation.
Store