VLSI analog computation

Once the dominant tool for simulating dynamical systems, analog computers faded from use in the 1960s and 1970s. These physically large systems were time consuming to program and required a great deal of user expertise. Nonlinear partial differential equations, stiff systems of ordinary differential equations and stochastic differential equations are still potentially time consuming to simulate on a digital computer. Modern analog VLSI allows for the design of electronically programmable, tunable, small, and low-power analog computers, thereby eliminating many of the problems that plagued conventional analog computers.

This project showed that for certain applications, namely those in which speed is more important than the utmost of accuracy, analog simulation is a useful tool for dynamical system simulation. To this end, we have designed, fabricated and tested a 100 mm² programmable analog computer chip. A sample of this chip is shown in Figure 1. It features 80 integrators, and more than 400 linear and nonlinear functional blocks. The blocks are subdivided into a 4 x 4 grid of macroblocks. This structure is visible in Figure 1. Figure 2 shows a photograph of one of these macroblocks. Through a Simulink based user-interface, the device is automatically configured allowing the convenient simulation of a wide variety of dynamical systems.

G. Cowan, R. Melville, and Y. Tsividis, “A VLSI analog computer / digital computer accelerator,” Journal of Solid-State Circuits, vol. 41, no. 1, pp 42-53, Jan. 2006.

Y. Tsividis, G. Cowan, Y. W. Li, and K. Shepard, “Continuous-Time DSPs, Analog/Digital Computers and Other Mixed-Domain Circuits,” Proceedings of the 31st European Solid-State Circuits Conference (ESSCIRC), 2005, pp 113-116.

G. Cowan, R. Melville, and Y. Tsividis, “A VLSI analog computer/co-processor for a digital computer,” Proceedings of the International Solid-State Circuits Conference (ISSCC), February 2005, pp 82-83, 586.