Serial Latching Array for In-Memory Arithmetic Speed

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The miniaturization of integrated circuits poses challenges in design and manufacturing, requiring stricter specifications and improved reliability, particularly in ensuring efficient and reliable storage and computational operations.

Innovation Solution

A storage device with an array of latching units connected through inter-unit transmission gates, operating in various modes to perform arithmetic operations as a finite state machine, utilizing pulse operations to change states efficiently and improve computational speed.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Weight of moving object

If miniaturization of integrated circuits is pursued, then device size is reduced and power consumption is lowered, but design and manufacturing specifications become stricter and reliability challenges increase

Engineering Contradiction:
Improvedevice sizeVSAvoidreliability
Core Design Contradiction:
Weight of moving objectVSReliability

Solution Approach 1:

The integrated circuit is divided into multiple storage units, each comprising serially connected latching units that can operate independently. This segmentation allows each unit to be miniaturized while maintaining reliable operation through modular design, where failures in one unit do not propagate to others.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Inter-unit transmission gates are introduced as intermediary components between adjacent latching units within storage units and between storage units. These transmission gates control signal flow and isolation, enabling reliable data transfer and state maintenance even as device dimensions are reduced, thereby addressing reliability challenges in miniaturized designs.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Length of moving object

If miniaturization of integrated circuits is pursued, then device size is reduced, but design and manufacturing specifications become stricter

Engineering Contradiction:
Improvedevice dimensionVSAvoidmanufacturing specification
Core Design Contradiction:
Length of moving objectVSManufacturing precision

Solution Approach 1:

The circuit is segmented into standardized storage units with identical latching unit structures. This modular approach allows design optimization at the unit level while simplifying manufacturing, as the same components can be replicated across the device, reducing the impact of miniaturization on manufacturing precision requirements.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Each latching unit is designed as a universal building block that can function in multiple contexts (within different storage units, in different positions). This universality reduces design complexity and manufacturing specification stringency, as the same standardized components are used throughout the miniaturized device.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Reliability

If conventional storage and computational operations are used, then device functionality is maintained, but computational speed and efficiency are limited

Engineering Contradiction:
Improveoperational reliabilityVSAvoidcomputational speed
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

Storage and computational functions are merged within the same latching units and storage units. The latching units can maintain stored data states while also participating in arithmetic operations through controlled state transitions. This integration eliminates the need for separate storage and computation operations, significantly improving computational speed and efficiency while maintaining operational reliability.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

Arithmetic operations are performed through periodic pulse signals that control state transitions in the latching units. By using synchronized periodic actions across multiple storage units, the system achieves concurrent arithmetic operations, enhancing computational throughput while maintaining reliable state management through rhythmic control signals.

Inventive Principle:
Principle #19Periodic action

4Ease of operation

If conventional sequential operations are used, then operational simplicity is maintained, but computational efficiency is reduced

Engineering Contradiction:
Improveoperational simplicityVSAvoidcomputational efficiency
Core Design Contradiction:
Ease of operationVSProductivity

Solution Approach 1:

The system uses periodic pulse operations to control state transitions in latching units. This periodic control mechanism maintains operational simplicity through standardized timing signals while enabling concurrent arithmetic operations across multiple storage units, thereby improving computational efficiency without complicating the control interface.

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

The latching units incorporate feedback mechanisms where the state of one latching unit influences the operation of adjacent units through inter-unit transmission gates. This feedback enables coordinated arithmetic operations across the storage array while maintaining ease of operation through automatic state propagation, eliminating the need for complex external control logic.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS20250357919A1Storage device having latching units serially connected for in-memory arithmetic operations
Publication Date: 2025.11.20 TAIWAN SEMICONDUCTOR MANUFACTURING CO LTD
  • US20250357919A1 patent drawing
  • US20250357919A1 patent drawing
  • US20250357919A1 patent drawing

AI summary

A method includes setting each latching unit in an array of latching units to a forward-connection mode for a duration of a first time period while each inter-unit transmission switch in a plurality of inter-unit transmission switches is at a connected state, and after the first time period, setting each latching unit in the array of latching units to a latch mode and setting each inter-unit transmission switch to a disconnected state. The method also includes setting each latching unit in the array of latching units to a backward-connection mode for a duration of a second time period while each inter-unit transmission switch is at a connected state, and after the second time period, setting each latching unit in the array of latching units to a latch mode and setting each inter-unit transmission switch to a disconnected state.