3D Memory Weight Storage Using Thermal Coupling for Analog Synapses
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory devices struggle to efficiently store and retain analog values, particularly in neuromorphic systems, due to limited sense windows and thermal disturbances that affect the reliability and precision of synaptic weight representation.
Innovation Solution
A memory device architecture utilizing a three-dimensional array of memory cells, including an aggressor memory cell and victim memory cells, leverages thermal disturbances during write operations to store analog values by altering the threshold voltages of victim cells based on thermal relationships, mimicking neuro-biological architectures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If thermal disturbances are used to store analog values in memory cells, then the precision and reliability of synaptic weight representation are improved, but the thermal stability of stored values deteriorates due to thermal noise and disturbances
Solution Approach 1:
The patent converts the harmful thermal disturbances into a beneficial mechanism for storing analog values. By applying controlled thermal disturbances during write operations, the invention enables analog weight storage in neuromorphic systems, transforming thermal noise from a detrimental factor into the core mechanism for achieving precise synaptic weight representation.
Solution Approach 2:
The invention changes the thermal parameters of memory cells by applying controlled thermal disturbances during write operations. This parameter change enables the memory cells to store analog values by temporarily altering their thermal state, which then translates into precise threshold voltage shifts for synaptic weight representation.
2Quantity of substance
If three-dimensional array architecture with aggressor and victim memory cells is used, then the density and integration of memory cells are improved, but the complexity of thermal management and cross-talk control deteriorates
Solution Approach 1:
The patent segments the memory array into distinct aggressor and victim memory cell groups organized in a three-dimensional architecture. This segmentation allows independent control and management of thermal disturbances, enabling high-density integration while simplifying thermal management by localizing thermal effects to specific victim cell groups rather than managing entire array-wide thermal patterns.
Solution Approach 2:
The invention transitions from two-dimensional to three-dimensional memory cell arrangement, stacking memory cells vertically to achieve higher density. This dimensional change allows better thermal isolation between layers and enables more efficient thermal management pathways, reducing cross-talk complexity despite increased integration density.
3Adaptability or versatility
If analog values are stored using threshold voltage alterations, then the representational capability for neural network weights is improved, but the susceptibility to thermal noise and read disturbances increases
Solution Approach 1:
The patent applies preliminary controlled thermal disturbances during the write operation to establish the desired analog threshold voltage state before reading. This preliminary action sets the memory cell in a stable analog state that is less susceptible to subsequent thermal noise during read operations, as the cell has already been conditioned to the target weight value.
Solution Approach 2:
The invention implements feedback mechanisms where read operations detect the analog threshold voltage and provide information back to control circuits, which then apply corrective thermal disturbances or voltage adjustments to maintain the stored analog value. This feedback loop compensates for thermal noise and drift, preserving the representational accuracy of neural network weights.
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
Enhances the ability to store and retain analog values with improved precision and reliability, enabling advanced computing tasks such as learning, vision, and auditory processing in neuromorphic systems.
Implementation Method 1
the thermal energy output by the aggressor memory cell during the write operation causes the state of the victim memory cells to alter based on a thermal relationship between the aggressor memory cell and at least one of, if not each of, the victim memory cells
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods, systems, and devices for mimicking neuro-biological architectures that may be present in a nervous system are described herein. A memory device may include a memory unit configured to store a value. A memory unit may include a first memory cell (e.g., an aggressor memory cell) and a plurality of other memory cells (e.g., victim memory cells). The memory unit may use thermal disturbances of the victim memory cells that may be based on an access operation to store the analog value. Thermal energy output by the aggressor memory cell during an access operation (e.g., a write operation) may cause the state of the victim memory cells to alter based on thermal relationship between the aggressor memory cell and at least some of the victim memory cells. The memory unit may be read by detecting and combining the weights of the victim memory cells during a read operation.