Articles by Casey Weltzin

Casey Weltzin is a Product Manager in the Real-Time and Embedded software group at National Instruments, with special focus on FPGA, multicore, and virtualization technologies. He holds a B.S in Electrical Engineering from the University of Wisconsin-Madison.

To Design Innovative Products, You Must Fail Quickly.

Friday, November 12th, 2010 by Casey Weltzin

While making incremental changes to existing embedded designs may be straightforward, engineers and scientists working on creating new, innovative designs live in a much different world. They are tasked with building complex electrical or electro-mechanical systems that require unique combinations of I/O and processing elements to build. Rather than starting by budgeting time and resources, these designers often need to begin the design process by asking “is this even possible?”

One example of this kind of innovative application is a system created by KCBioMedix, which teaches premature infants how to feed. With up to one-third of premature infants born in the United States suffering from feeding problems, the device called NTrainer, helps coordinate sucking, swallowing, and breathing movements to accelerate feeding without a tube. It is essentially a computerized pacifier that emits gentle pulses of air into an infant’s mouth.

Of course, this kind of innovation seldom takes place without skeptics. Innovative designs require investment that is often heavily competed for and scrutinized within organizations. Or, in the case of startup ventures, entrepreneurs require investment from venture capitalists that have many other places to put their funding. Ultimately, to make a commitment, management or third party sources of capital require the same things – proof that the concept will work and a sound business plan.

Let’s concentrate on the former. Complex devices and machines typically require tens or even hundreds of iterations during the design process; in short, failures. And these iterations can be time consuming and expensive. While making software modifications is relatively easy, changing I/O or processing hardware can take weeks to months. Meanwhile, business leaders and investors become increasingly impatient.

How can both large organizations and startups mitigate the risk of redesigns? One solution commonly employed is to carefully study design requirements and come up with an architecture that is unlikely to need modification. This is a poor solution for two reasons. First, even the most capable designers may fail to foresee the challenges associated with a new, innovative design – resulting in cut traces or a rat’s nest of soldered wires to modify a piece of hardware. Second, because engineers are likely to reuse the architectural patterns and design tools they are used to, innovative features are more likely to be traded-off to fit the constraints that those patterns impose.

A better solution is to use a COTS (commercial off-the-shelf) prototyping platform with a combination of modular I/O, reconfigurable hardware, such as FPGAs (field programmable gate arrays), and high-level development tools. Using this approach, extra I/O points can be “snapped-in” when needed rather than requiring an entire board or daughterboard redesign. Additionally, FPGAs enable designers to implement high-performance custom logic at several orders of magnitude less upfront cost than ASICs (application-specific integrated circuits). Finally, high-level design tools enable both experienced embedded designers and application experts to take advantage of FPGA, real-time operating system, and other technologies without prior expertise or a large team of experts in each technology. In other words, when equipped with the right tools, a small team can “fail quickly” and accelerate the innovation process.

There are a number of economic concerns that must be addressed when using COTS platforms for prototyping. First, since these platforms typically present a much higher up-front cost compared to the BOM (bill of material) components used in a final design, organizations must carefully consider the productivity savings they provide to determine the time to break-even on the investment. For many complex projects, COTS solutions have the potential to reduce the time to first prototype by weeks or months while also reducing the overall size of the development team required. And, it may be possible to reuse these tools between multiple projects in innovation centers (amortizing the upfront cost over a longer period of time).

Another economic consideration that must be made is how much the transition from prototype to final deployment will cost. For small or medium size deployments, it may be beneficial to use COTS hardware embedded in the final device (provided that it meets size and power constraints) – essentially a trade-off between higher BOM cost and reduced development time. On the other hand, for large deployments the benefits of a low BOM cost may warrant moving to a custom cost-optimized design after prototyping. In this case, organizations can save cost by choosing prototyping tools that provide a minimal-investment path to the likely deployment hardware.

Returning to the example of KCBioMedix, the company was able to reduce prototyping time of their premature infant training system from 4 months to 4 weeks using COTS tools – providing an estimated savings of $250,000. COTS hardware is also being used in the final NTrainer product to maximize reuse of IP from the prototyping stage.

The bottom line is that for both the aspiring entrepreneur and the large organization that wishes to maintain an entrepreneurial spirit, prototyping is an essential part of producing innovative designs in time to beat the competition. Organizations that encourage prototyping are more nimble at weeding out good ideas from bad, and ultimately producing differentiated products that command strong margins in the marketplace.