This thesis develops an engineering practice and design methodology to enable us to use CMOS analog VLSI chips to perform more accurate and precise computation. These techniques form the basis of an approach that permits us to build computer graphics and neural network applications using analog VLSI. The nature of the design methodology focuses on defining goals for circuit behavior to be met as part of the design process. To increase the accuracy of analog computation, we develop techniques for creating compensated circuit building blocks, where compensation implies the cancellation of device variations, offsets, and nonlinearities. These compensated building blocks can be used as components in larger and more complex circuits, which can then also be compensated. To this end, we develop techniques for automatically determining appropriate parameters for circuits, using constrained optimization. We also fabricate circuits that implement multi-dimensional gradient estimation for a gradient descent optimization technique. The parameter-setting and optimization tools allow us to automatically choose values for compensating our circuit building blocks, based on our goals for the circuit performance. We can also use the techniques to optimize parameters for larger systems, applying the goal-based techniques hierarchically. We also describe a set of thought experiments involving circuit techniques for increasing the precision of analog computation. Our engineering design methodology is a step toward easier use of analog VLSI to solve problems in computer graphics and neural networks. We provide data measured from compensated multipliers built using these design techniques. To demonstrate the feasibility of using analog VLSI for more quantitative computation, we develop small applications using the goal-based design approach and compensated components. Finally, we conclude by discussing the expected significance of this work for the wider use of analog VLSI for quantitative computation, as well as qualitative.
展开▼
