Center for Circuits and System Solutions Overview

C2S2 is a multi-university research center chartered by the U.S. semiconductor industry and U.S. Dept. of Defense to develop long-range solutions for the design challenges that arise from semiconductor device scaling. Ten of America’s most elite research universities are participating in this center.

Moore’s law predicts that semiconductor devices double in speed and density every 18 months; it has held remarkably true for the last two decades. Unfortunately, ultimate physical limits are being reached as transistors continue to scale down. This causes enormous design challenges. C2S2 focuses on the critical middle of the semiconductor “food chain.” Our mission is to convert tomorrow’s transistors into useful performance—no matter how different, difficult, or downright hostile those transistors turn out to be. We invent new circuit, system, and software designs to prevent scaling from toppling this vitally important food chain.

Research Thrusts

Research in C2S2 is partitioned into one of six major technical thrusts:

• Models & Methods: Custom design flows and numerical modeling methods for analog, RF, MEMS, & DSP systems, emphasizing rapid design, and accurate prediction and tradeoff analysis.

• Robust Design: Error tolerance for circuits & systems. Circuits that work—from devices that may not. Isolation methods to integrate sensitive analog, RF or MEMS on hostile, scaled digital chips.

• Wireless Design: Circuits, architectures, and protocols that push the performance envelope for wireless transmission. Current design target is 1Gbit/sec self-organizing LAN, using a multiple-antenna scheme.

• Wireline Design: Circuits at the boundary between electronics & optics. Current design target is a 100Tbit/sec Internet IP router, using electronics for routing, and a micromirror switching fabric.

• Processors & System Software: Architectural mechanisms to handle soft errors in ultra-scaled technologies. Deep analysis techniques to massively customize system software—the code that is closest to the hardware and the circuits.

• Interfaces: Circuits that connect from chip to world. Focus on massively parallel analog systems, using local feedback for accurate element calibration, and to reduce manufacturing variations that will only increase with further device scaling.

Carnegie Mellon’s Focus

The objectives of Carnegie Mellon researchers are in Models and Methods, and include:

• Develop automated numerical model extraction, validation, and simulation techniques for nondigital circuits—analog, RF, microwave, MEMS, nanotech—that are to date too impractical, inefficient, or inaccurate to model with ad hoc methods.

• Develop radical new design flows to support rapid evolution from algorithms to architectures to silicon, for complex digital and nonlinear analog/RF designs.

• Develop silicon drivers that validate our models and flows, and push next-generation wireless, wireline, interface circuits, and heterogeneous system applications.