In last month's column, I profiled ultrawideband startup XtremeSpectrum. This month, I turn my attention to Anadigm. This startup is moving the field-programmable concept from the digital realm into the analog realm with its field-programmable analog array (FPAA).

Spun out of Motorola in January 2000, Anadigm has bet its 35 employees and $36 million (so far) on creating a product that will displace analog designs using discrete components. Anadigm's integrated circuits consist of input and output cells as well as configurable analog blocks that are interconnected with a dynamically reconfigurable switch fabric. The configurable analog blocks themselves contain banks of switched capacitors together with a few dedicated functions such as operational amplifiers and comparators. The banks of switched capacitors can be configured to mimic resistive and capacitive elements by adjusting the clock frequency. Though individual MOS capacitors tend to have widely varying tolerances, switched-capacitor designs get around this problem by instead relying on tightly controlled ratios of capacitors.

Programmable analog arrays have bounced around in the semiconductor market for years with less-than-stellar traction. Indeed, Anadigm faces semi-direct competition from Zetex's Trac family and Lattice Semiconductor's ispPAC family. Perhaps one can understand what has held programmable analog back by looking at its advantages and disadvantages relative to the use of discrete components. The most significant disadvantage of programmable analog arrays is a cost premium in the neighborhood of 50 percent.

Programmable analog arrays do have their advantages, the most significant of which is ease of design-particularly for more novice analog circuit designers. Ease of design of course translates into a time-to-market benefit. A second important benefit is that FPAAs can be reconfigured in the field. This attribute is important in environments where the condition of the signal changes-typically degrades-over time. An example would be a currency reader with an optical sensor that gets dirty over time and requires signal conditioning.

For any semiconductor technology, the key to success is finding a market segment that most benefits from the technology's attributes. For FPAAs, that means finding segments that are not highly cost sensitive, that have maximum frequency requirements of 500 kHz and that may require reconfiguration. The cash reader example above fits this bill as do a host of professional audio and industrial applications.

As the cost differential between FPAAs and discrete analog components narrows, the programmable technology will gain momentum. Anadigm must target the right segments today while pushing down costs to open new markets.

Jeremey Donovan (jeremey.donovan@gartner.com) is chief analyst at Gartner Dataquest.