Adaptive Reference Signal for MRAM Read Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As magnetic random access memory (MRAM) devices are scaled smaller, device-to-device variations increase, making it challenging for a single consistent reference signal to reliably differentiate between magnetic states, leading to overlapping resistance distributions and high error rates in reading operations.
Innovation Solution
An adaptive reference signal is generated from a plurality of reference cell pairs connected in parallel, with half programmed in high conductance (P) and half in low conductance (AP) states, allowing for individual programming to adjust to variations between magnetic states, and using averaging circuits to balance temperature effects and reduce neighborhood variability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single consistent reference signal is used for reading MRAM cells, then the reference circuit is simple and manufacturing is easy, but device-to-device variations cause overlapping resistance distributions and high error rates
Solution Approach 1:
The reference circuit is segmented into multiple reference cells (at least two) instead of using a single reference signal. Each reference cell can be independently programmed to different conductance states, allowing the system to segment the reference signal generation into multiple adjustable components that can collectively compensate for device variations.
Solution Approach 2:
The reference cells are dynamically programmable, allowing their conductance states to be adjusted based on observed device-to-device variations. This dynamic adjustment capability enables the reference circuit to adapt to different operating conditions and compensate for variations in memory cell characteristics.
2Reliability
If more reference devices are averaged to reduce variation, then reading reliability improves, but the area occupied by reference devices increases
Solution Approach 1:
Instead of increasing the number of reference devices, the invention changes the conductance state parameters of existing reference cells through programming. By adjusting the conductance values of at least two reference cells, the system achieves variation compensation without adding more physical devices, thus maintaining compact area while improving reliability.
3Reliability
If reference cells are programmed to adjust to variations in magnetic states, then covariance between magnetic states is minimized and yield improves, but the programming complexity and device complexity increase
Solution Approach 1:
The system employs feedback mechanisms where the conductance states of reference cells are programmed based on observed variations in memory cell magnetic states. This feedback loop allows the reference circuit to automatically adjust and compensate for deviations, minimizing covariance between magnetic states and improving yield without requiring complex manual calibration.
4Device complexity
If a single reference cell is used, then the circuit is simple, but it cannot accommodate device-to-device variations and temperature effects
Solution Approach 1:
The single reference cell is segmented into multiple reference cells (at least two), each capable of independent programming. This segmentation provides the adaptability needed to handle device-to-device variations and temperature effects while maintaining relatively simple circuit architecture.
Solution Approach 2:
The multiple reference cells serve multiple functions: they can individually compensate for device variations, adjust for temperature effects, and provide flexible programming options. This multi-functionality allows a single reference circuit to handle various operating conditions without requiring separate compensation circuits.
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
The adaptive reference signal effectively reduces error rates by centering the reference signal, accommodating outliers, and improving yield by minimizing covariance between magnetic states, even in small geometries, thereby enhancing the reliability of MRAM reading operations.
Implementation Method 1
magnetic memory cells to hold reference data... where 'P' signifies that the magnetic domains of the cell are oriented in the same direction and 'AP' signifies that the magnetic domains of the cell are oriented in opposite directions
Implementation Method 2
the magnetic domains of the cell are oriented in the same direction and 'AP' signifies that the magnetic domains of the cell are oriented in opposite directions
Implementation Method 3
a sense amplifier typically compares the current flowing through a memory cell under an applied voltage to a reference current
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A structure and method is described for an adaptive reference used in reading magnetic tunneling memory cells. A collection of magnetic tunneling memory cells are used to form a reference circuit and are coupled in parallel between circuit ground and a reference input to a sense amplifier. Each of the magnetic memory cells used to form the reference circuit are programmed to a magnetic parallel state or a magnetic anti- parallel state, wherein each different state produces a different resistance. By varying the number of parallel states in comparison to the anti-parallel states, where each of the two sates produce a different resistance, the value of the reference circuit resistance can be adjusted to adapt to the resistance characteristics of a magnetic memory data cell to produce a more reliable read of the data programmed into the magnetic memory data cell.