AdapTrim* (Self-Adaptive Hardware)  Project Overview

Adaptive Self-Trimming Hardware & Software

• Background:  Many Analog circuits require precisely matched components that are impossible to guarantee in typical digital semiconductor processes.  Yet, customers demand mixed analog and digital circuits on one chip.   • So, Bill invented his AdapTrim system, in which components on a chip "learn to become matched."   A small, serial, on-chip bus interconnects all the components that must be matched.  A simple connectionist learning rule (MR3) is executed on an ordinary digital IC tester, using mainly the usual I/O pins/pads.  The result: Precision analog matching in an otherwise "sloppy digital process".  Match-learning can be done , even after packaging the chip, which can compensate for complex package-related capacitances and inductances.   • Keywords: Adaptive, Matching, Hardware, MR3, Backpropagation-through-perturbation, Balanced RF Mixer.          * "AdapTrim" is a contraction for "Adaptive Self-Trimming System"

Significance of this Accomplishment

• High Precision analog matching in an otherwise "sloppy digital process" (that is the most cost effective). • Precision analog technology is expensive and is highly inefficient for building ordinary digital circuitry. • Cell phones (the one product that drives the semiconductor industry) require precision mixed analog and efficient digital circuitry in one monolithic chip.

Skills Transferable to Your Projects:

Leadership / Management • Bill led the entire AdapTrim project. Theory / R&D • Bill's machine learning expertise and IC design experience enabled the AdapTrim innovation. • This includes Bill’s numerous insights in machine learning, neuroscience,  computer science, IC design, IC testing, RF components (LNAs, Oscillators, Mixers, etc.) and hardware design (CMOS digital and analog).  Software • Bill wrote C++ software to successfully demonstrate AdapTrim's use on an innovative doubly balanced RF mixer at 1 GHz. Hardware • Bill designed an innovative doubly balanced RF mixer, with AdapTrim.  The chip was fabricated and its adaptation was successfully tested.  Radu Secareanu, Bill's co-worker, was a major contributor, too.

Future Opportunities

By pushing beyond the current state-of-the-art, there are these opportunities: • This is one of the most critical challenges to the further miniaturized of semiconductors:  (The ability to automatically compensate for statistically probable defects in future semiconductor chips). • The experience gained by creating AdapTrim, gives  Bill a valuable advantage in meeting this future challenge.

Links to related material

• An Adaptive Circuits Concept to Address Mismatch in Analog Circuits IEEE ISCSAS (International Symposium on Circuits and Systems)  May 25, 2004