While processors tend to grab the headlines, algorithms are the real essence of digital signal processing. Algorithms are what enable the incredible innovations we see in fields ranging from medical imaging to wireless communications to computer vision. And, algorithms are becoming even more important over time, as better processors enable more sophisticated algorithms to be implemented at acceptable price points. For system designers looking for ways to differentiate their products, better algorithms are often the most promising way to do so, considering that competitors usually have access to the same chips and tools.
But the path to better algorithms can be a circuitous one, with serious hazards along the way. For example, an algorithm that performs well under one set of conditions may fail under other (perhaps more realistic) conditions. Or, an algorithm that performs well when simulated using double-precision floating-point arithmetic may not perform at all when implemented with 32-bit fixed-point math.
A truly innovative, effective algorithm that addresses an important problem can be worth many millions of dollars. An algorithm that appears effective but is fundamentally flawed can torpedo a product development effort—and even an entire company.
Given the intense product development schedule pressures typical of the electronics industry, an independent evaluation can be a valuable way to quickly determine whether a candidate algorithm is right for the job—enabling decision-makers to proceed with confidence.
In one such situation, a start-up company that had developed a novel noise-reduction algorithm engaged BDTI to evaluate the algorithm. The company wanted validation of its technology from a respected third party in order to help it license the algorithm to chip and system designers.
BDTI performed a thorough examination of the algorithm using both theoretical and experimental techniques. Although the company’s simulation results looked promising, BDTI’s two-pronged analysis showed clearly that the algorithm didn’t work—nor could it, because it was based on flawed theory.
The start-up’s executive management team was—not surprisingly—disappointed with BDTI’s findings. Lacking in-depth technical expertise, the executives faced a tough choice: should they rely on BDTI’s evaluation, or trust the algorithm’s developers (who had impressive resumes and who believed the algorithm worked)?
The start-up chose to go ahead with its algorithm despite BDTI’s findings. Six months later, though, reality caught up, and the venture was abandoned. In the meantime, the start-up had consumed significant capital, effort, and time that could have been invested more fruitfully elsewhere.
In another algorithm evaluation project, BDTI was engaged by a medical equipment manufacturer to evaluate a conceptual design for an algorithm that would enable valuable new functionality to be added to an existing product. Through a combination of theoretical and simulation-based analysis, BDTI found that the proposed algorithm would not work under real-world conditions. Within a few weeks, however, BDTI was able to propose an alternative algorithm better suited to the system. The new algorithm enabled the manufacturer to proceed with its product development plans quickly and with confidence.
Are you developing a new algorithm, or considering licensing one for use in your product? If the efficacy of that algorithm is critical to your success, consider engaging BDTI’s team of digital signal processing experts to provide a technically knowledgeable, independent analysis merging theoretical and practical perspectives. The earlier you determine whether your candidate algorithm will really do the job, the more likely you are to meet your business objectives. To find out how BDTI can help with your specific requirements, contact Jeremy Giddings at +1 (925) 954 1411 or giddings@BDTI.com.
Add new comment