As time-to-market pressures build, OEMs are looking to their suppliers to broaden support and help them succeed in their markets. As we have explored in previous columns, suppliers of DSP technologies to the real-time embedded markets have adopted a new paradigm to address this expanded opportunity, one that can be broken down into four elements: silicon, software, systems, and support.
In analyzing the ideal silicon element, we found that manufacturers should look for vendors that offer a broad portfolio of code-compatible products that span the continuum of performance, power, integration, and price, as well as complimentary system-level devices.
Programmability is paramount to success across these vectors. Hence, much of the focus of DSP vendors has shifted from hardware to software. We will now delve more deeply into this element of the new paradigm.
In less than 10 years the industry has driven DSP performance from about 50 mips to almost 10,000 mips. As performance has increased exponentially, so has the complexity of the design process.
To put this into perspective, one need only map the progression of applications software, which has gone from hundreds of lines of code to hundreds of thousands. It's easy to see why OEMs are turning to vendors for help. They need access to easy-to-use software that simplifies the design process, freeing them to spend valuable time and development dollars writing code that truly differentiates their products.
This point is clearly illustrated by the explosion of programmers for new DSP-based applications. Advancements in software development tools, the use of high-level language, and the maturity of real-time operating systems have opened up the field to hundreds of thousands of people across a myriad of applications. Thanks to these software breakthroughs, DSP programming is no longer an art form practiced only by those versed in DSP theory.
So what does this software element look like? Well, the ideal software equation is made up of two halves: the host or integrated development environment (IDE) side, and the target or application content side. Each plays a vital role in developing DSP-based applications, and cutting-edge silicon vendors will offer a comprehensive package that includes both. In this way, DSP designers of all experience levels will have access to the many phases of code development.
The concept of an IDE did not exist in the DSP world 10 years ago. Now it's the first piece of the software puzzle for manufacturers to consider when designing with DSP.
Ideally, the IDE should fully integrate editing, debugging, project management, profiling, and emulation capabilities, to name a few. Together, these enable developers of real-time embedded applications to seamlessly manage projects of any complexity and make DSP code writing more efficient. They also bring dramatic improvements in code development, optimization, and debugging for single- or multiprocessor systems, ultimately empowering developers to fully use the power of the DSP.
As new DSP applications grow in scope and complexity, and as team sizes grow to realize these applications, there is an increased need to support a host of users working over a wide geographic area on systems with multiple, often heterogeneous, processors. Therefore a strong IDE offering will include multisite connectivity. This capability enables teams to be spread over multiple sites rather than shackling them to a single location, giving them an overall advantage in product development.
In addition to the overarching capabilities of the IDE, the host-side element should feature robust capability in foundational elements such as code generation, debug, and linker. It is also important to look for host capabilities to provide a visibility link to the target side of the software equation.
The target side consists of three elements from a vendor: access to a proven real-time operating system (RTOS) that includes a multithreading kernel; analysis tools and peripheral configuration libraries; and access to a vast array of third-party software. Together, these elements give OEMs flexibility in development approaches, allow make vs. buy options, and enable valuable engineering resources to concentrate on differentiating their end product rather than reinventing the wheel.
Licensing a proven, scalable RTOS allows OEMs to develop and deploy sophisticated products more quickly than with traditional DSP software methodologies. It eliminates the need for them to develop and maintain their own custom systems or control loops. It also makes DSP-based applications easier to maintain thanks to the clean partitioning made possible by the multithreading kernel. Through it, new functions can be added without disrupting real-time response.
Analysis tools are also vital to software design. These tools capture and display information on thread execution sequences, performance, and overall CPU loading without stopping the DSP. They also simplify the process of application debug and optimization, just as peripheral configuration libraries make device-driver development easy.
Add to this mix a network of third parties that delivers expertise across DSP and DSP-based applications, and manufacturers have a winning combination on the target side.
When choosing a vendor, it's important to look for a company that offers an extensive collection of digital signal processing support through third parties. Access to a third-party network can offer designers a broad range of application software, development hardware and software, and consulting services, all of which are built around DSPs.
Using third-party network support also allows customers the ability to access innovative products and services, including development boards, operating systems, software algorithms, function libraries, and system consulting services.
In a previous column, we stressed the importance of code compatibility, and frankly, it will be stressed in every column due to its fundamental importance. Customers should have access to next-generation DSP functionality without going through costly re-engineering. Existing software should be reusable immediately, protecting the huge software investments customers have made.
Leveraging a comprehensive and integrated software capability across host and target, you can focus on what is really important-differentiation of your product and nailing your time-to-market window.
Greg Delagi is vice president and manager of worldwide digital signal processing at Texas Instruments Inc. in Dallas. E-mail comments to firstname.lastname@example.org