SANFORD, Fla. -- Using fuzzy-logic constructs rather than conventional software, BasiConcepts claims to have created a processor that is "programmed in human language." The startup's FL1000 processor design uses the FL1000 Application Expert--a software-development environment that simplifies the creation of fuzzy membership functions and rules. A real-time hardware emulator is due out in the third quarter and working silicon devices are slated the fourth quarter of this year.

"I analyzed over 200 designs and discovered that from 40 percent to 90 percent of a microcontroller's code uses comparison functions--just the kind of operation that fuzzy logic is great at solving," said Paul Basehore, president of BasiConcepts.

BasiConcepts arose from the ashes of American NeuraLogix, where Basehore was vice president of research and development. The core engineering staff, which also moved to BasiConcepts from American NeuraLogix, has designed hundreds of fuzzy-logic systems as a consultancy.

The FL1000 off-loads the tedious comparison functions into an on-chip fuzzy processor, but also performs all the standard Boolean functions of a traditional microcontroller. In-house, the FL1000 is implemented as an expandable ASIC library from which several custom versions are being prepared for customers. Standard off-the-shelf devices are expected by year's end.

"We chose a fuzzy/Harvard hybrid architecture that, due to its high efficiency, reduces the number of microcontroller instructions required for a typical design by 80 percent, while requiring 50 percent less memory and running 20 percent faster," said Basehore.

The FL1000 has three major on-chip modules: a program store, a register file and the fuzzy processor itself. The fuzzy arithmetic-logic unit performs both fuzzy and standard Boolean operations. Internally, it consists of a fuzzifier, a rule processor and a defuzzifier. The fuzzifier takes crisp input data and applies the premade membership functions to create fuzzy data. The fuzzy data is input to the rule processor, which combines the output of all the applicable fuzzy rules to create a fuzzy conclusion. The defuzzifier then converts that to crisp data for output.

"With our development software, you can design a fuzzy system without knowing anything about the details of fuzzy logic," said Basehore.

Free Subscription to EE Times First Name Last Name Company Name Title Email address   Click here for your Free Subscription to EETimes Europe