Die-Based Trimming for Non-Volatile Memory Yield
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current non-volatile memory fabrication methods face challenges in accurately programming memory cells due to fabrication tolerances, leading to inadequate yields and performance, particularly in battery-powered devices where low operating voltage is desirable to reduce power consumption and extend device lifetime.
Innovation Solution
Implementing an on-chip state machine for parallel programming of memory pages, allowing for die-by-die trimming of memory devices by determining and adjusting programming parameters such as voltage and pulse width to match individual die performance, thereby improving yield and reducing testing time and costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If wafer-based trimming is used to program memory cells, then the entire wafer can be adjusted to meet design specifications, but adequate yields and performance of individual memory dies cannot be achieved
Solution Approach 1:
The invention transitions from wafer-based trimming to die-based trimming, segmenting the trimming process into individual die level. Each die is independently tested and trimmed based on its specific performance characteristics, allowing precise programming adjustments for each die while maintaining overall yield improvement.
Solution Approach 2:
The invention implements local quality by tailoring trimming parameters to each individual die's performance characteristics. Instead of applying uniform trimming across the entire wafer, each die receives customized trimming based on its measured programming time and performance, ensuring optimal programming accuracy for each specific die.
2Manufacturing precision
If standard programming parameters are used for all dies, then the fabrication process is simple, but fabrication tolerances cause varying performance that prevents adequate programming accuracy
Solution Approach 1:
The invention performs preliminary testing of each die to measure its actual programming time and performance characteristics before applying trimming. This preliminary action enables data-driven trimming decisions, where the measured performance of each die is used to determine the appropriate trimming parameters, achieving high programming accuracy while managing complexity through systematic measurement and adjustment.
3Productivity
If die-by-die trimming is implemented to improve yield and performance, then individual die performance can be optimized, but testing time and fabrication complexity increase
Solution Approach 1:
The invention implements feedback by measuring the actual programming time of each die and using this information to determine the appropriate trimming parameters. The measured performance data feeds back into the trimming process, allowing iterative adjustment and optimization of programming parameters for each die, which improves yield while managing testing time through efficient feedback-driven adjustments.
Data Source
AI summary
Methods and structures are described to provide trims for die on a wafer. The trims are set on a die-by-die basis instead of a wafer basis. Accordingly, the individual die are more finely tuned and more die operate at the target specifications so that yield is increased. In an embodiment, the odd and even blocks of each non volatile memory die are erased and then programmed to test the program time. Statistical analysis of the tested program times is performed. Based on this analysis the trim values are determined and programmed into the die. Accordingly, each die on a wafer has its individual trim settings.