PCM MEMS Capacitor Memory for Low-Latency Multiply Operations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current processor architectures face challenges such as computational latency and energy/bit resolution issues due to high power consumption and electronic noise in electronic circuit elements, making them inadequate for brain-like computing and large machine learning applications.
Innovation Solution
The use of phase-change material (PCM) variable microelectromechanical systems (MEMS) capacitors, which include a substrate, electrodes, PCM, and a heater, allows for varying capacitance by controlling the PCM's temperature through voltage pulses, enabling efficient single multiply operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If electronic circuit elements are used for memory and computing, then the device can perform computational operations, but power consumption increases due to standby leakage currents
Solution Approach 1:
The patent replaces electronic circuit elements with MEMS capacitors that use mechanical displacement of movable electrodes to store and process data. The capacitive storage mechanism uses physical movement rather than electrical charge storage, eliminating standby leakage currents while maintaining computational functionality through mechanical-capacitive interactions.
Solution Approach 2:
The patent changes the operating parameter from electrical charge storage to capacitive coupling strength, controlled by the distance between movable and fixed electrodes. By varying the capacitance parameter through mechanical displacement rather than maintaining continuous electrical charge, the system achieves computational operations with minimal power consumption.
2Power
If electronic circuit elements are used for memory and computing, then the device can perform computational operations, but bit resolution is reduced due to inherent electronic noise
Solution Approach 1:
The patent substitutes electronic signal processing with capacitive coupling measurements between MEMS electrodes. The mechanical displacement of electrodes creates variable capacitance that can be measured with high precision, replacing noisy electronic circuit operations with more precise capacitive sensing that is less susceptible to electronic noise.
3Device complexity
If memory is separated from the processor core, then the architecture follows the traditional von Neumann design, but computational latency increases due to data transfer bottleneck
Solution Approach 1:
The patent merges memory and processing functions into a single integrated device where MEMS capacitors serve both as storage elements and computational units. The movable electrodes can be positioned to perform capacitive multiplication operations directly on stored data, eliminating the need for separate memory and processor components and the associated data transfer latency.
Solution Approach 2:
The MEMS capacitor structure serves multiple functions: it stores data through electrode positioning, performs computational operations through capacitive coupling, and can be read out through electrical measurement. This multi-functionality eliminates the need for separate memory and processing components, reducing computational latency while maintaining architectural simplicity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach reduces computational latency and improves energy efficiency and bit resolution by performing multiply-and-accumulate operations in constant time and minimizing leakage currents and noise, thus enhancing the performance of computational devices for AI and ML applications.
Implementation Method 1
The heater element is disposed within the PCM cavity and is coupled to receive a voltage pulse, whereby a temperature of the PCM varies
Implementation Method 2
The PCM is disposed within the PCM cavity. The heater element is disposed within the PCM cavity and is coupled to receive a voltage pulse, whereby a temperature of the PCM varies to thereby vary the capacitance gap
Data Source
AI summary
A computational device includes a phase-change material (PCM) variable microelectromechanical systems (MEMS) capacitor and a power source. The PCM variable MEMS capacitor includes a substrate, a first electrode, a second electrode, a PCM, and a heater. The first electrode is spaced apart from the substrate to define a PCM cavity. The second electrode is spaced apart from the first electrode to define a capacitance gap. The PCM is disposed within the PCM cavity. The heater element is coupled to receive a voltage pulse, whereby a temperature of the PCM varies to thereby vary the capacitance gap. The power source is coupled to the PCM variable MEMS capacitor and is operable to (i) supply the voltage pulse to the heater and (ii) a time-dependent voltage between the first electrode and the second electrode, to thereby implement a single multiply operation.