Data Sensing Device Current Adjuster Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional neural network computations using non-volatile memory require a large area and high power consumption due to the need for multiple reference currents to handle varying input signals, as memory cells cannot provide negative transconductances, leading to inefficient circuit design and power usage.
Innovation Solution
A data sensing device with a current adjuster and sensing amplifier that generates a shift current based on input signals to adjust the read-out current, allowing comparison with a reduced number of reference currents, thereby optimizing circuit area and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple reference currents are provided for different input signal amounts, then accurate data sensing is achieved, but circuit area and power consumption increase
Solution Approach 1:
The patent applies dynamics by making the read-out current adjustable through a current adjuster that dynamically shifts the current based on input signal amounts. This dynamic adjustment allows a single set of reference currents to effectively cover multiple input signal ranges, eliminating the need for multiple fixed reference current circuits and thereby reducing circuit area while maintaining sensing accuracy.
Solution Approach 2:
The patent changes the parameter of read-out current by introducing a shift current that adjusts the read-out current according to the input signal amount. This parameter change enables the system to use a predetermined number of reference currents for different input signal conditions, reducing the total number of reference currents needed while preserving measurement precision.
2Measurement precision
If multiple reference currents are provided for different input signal amounts, then accurate data sensing is achieved, but power consumption increases
Solution Approach 1:
The current adjuster dynamically adjusts the read-out current based on input signal amounts, allowing a single set of reference currents to serve multiple input signal ranges. This dynamic approach reduces the number of active reference current circuits needed, thereby lowering power consumption while maintaining sensing accuracy across varying input conditions.
Solution Approach 2:
By changing the read-out current parameter through shift current adjustment, the system can use fewer reference currents to cover a wider range of input signal amounts. This parameter modulation reduces the total power consumption of the reference current generating circuit while preserving measurement precision.
3Adaptability or versatility
If weights are added by a predetermined shift value to make all weights positive, then neural network computation is enabled, but the reference current generating circuit requires large area
Solution Approach 1:
The patent implements a dynamic current adjustment mechanism that shifts the read-out current based on input signal amounts. This dynamic shifting eliminates the need for multiple static reference current circuits that would be required to handle different weight ranges, thereby enabling neural network computation with positive weights while reducing the reference current generating circuit area.
Solution Approach 2:
The current adjuster serves multiple functions by adjusting the read-out current for different input signal amounts using a single set of reference currents. This multi-functionality allows the reference current generating circuit to handle various neural network computation scenarios without requiring separate reference current circuits for each case, reducing overall circuit area.
Data Source
AI summary
A data sensing device and a data sensing method are provided. The data sensing device includes a current adjuster and a sensing amplifier. The current adjuster corresponds to a memory string of a memory array, generates a shift current according to an amount of a plurality of input signals of the memory string, and generates an adjusted read-out current by adjusting a read-out current of the memory string according to the shift current. The sensing amplifier receives the adjusted read-out current and a plurality of reference currents, and generates a read-out data by comparing the adjusted read-out current and the plurality of reference currents.