University of Michigan

Michael B Henry

David Theodore Blaauw

Dennis Sylvester

David Alan Fick

Laura E Fick

Yoonmyung Lee

Yejoong Kim

Kaiyuan Yang

A weighted sum is a key computation for many neural networks and other machine learning algorithms. Integrated circuit designs that perform a weighted sum are presented. Weights are stored as threshold voltages in an array of flash transistors. By putting the circuits into a well-defined voltage state, the transistors that hold one set of weights will pass current equal to the desired sum. The current flowing through a given transistor is unaffected by operation of remaining transistors in the circuit.

Floating-gate transistor array for performing weighted sum computation

A true random number generator comprises a ring oscillator which is triggered to start oscillating in a first mode of oscillation at an oscillation start time. The first mode of oscillation will eventually collapse to a second mode of oscillation dependent on thermal noise. A collapse time from the oscillation start time to the time at which the oscillator collapses to the second mode is measured, and this can be used to determine a random number. The TRNG can be synthesized entirely using standard digital techniques and is able to provide high randomness, good throughput and energy efficiency.

True random number generator

Signal value storage circuitryis provided which includes a first transistor stack, a second transistor stack and a third transistor stack. The signal value storage circuitry is controlled by a single clock signal. Keeper transistors and isolation transistors serve to permit static operation of the signal value storage circuitry (i.e. the clock signal may be stopped without losing state) and to prevent contention within the circuitry.

Static signal value storage circuitry using a single clock signal

Growing community of inventors

Ann Arbor, MI, United States of America

Michael B Henry