Distributed Driver Quilt Pattern for Memory Interconnect
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory device architectures are inefficient in terms of die size, lithographic requirements, and driver circuitry, particularly due to the peripheral placement of word and bit line drivers, which leads to increased IR drop and RC delay, and requires complex transistor and circuit designs.
Innovation Solution
The proposed solution involves distributing row and column driver regions across the footprint of the memory array in a quilt pattern, with drivers located under the memory cells and socket regions positioned centrally along the word and bit lines, allowing for mid-point driving of electrodes and reducing interconnection requirements, thereby simplifying the layout and reducing the number of interconnect metal layers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If word and bit line drivers are placed peripherally in existing memory architectures, then the layout is simplified, but IR drop and RC delay increase significantly
Solution Approach 1:
The driver circuitry is segmented and distributed across multiple locations within the memory array footprint rather than being concentrated at the periphery. This segmentation allows drivers to be positioned closer to the memory cells they serve, reducing the length of interconnect lines and thereby minimizing IR drop and RC delay while maintaining layout simplicity.
2Reliability
If drivers are distributed under memory cells in a quilt pattern, then IR drop and RC delay are reduced, but driver circuit complexity increases
Solution Approach 1:
The driver circuit design is replicated and copied to multiple locations throughout the memory array in a systematic quilt pattern. Rather than designing complex unique drivers for each location, the same standardized driver circuit is copied and distributed, which reduces individual driver complexity while achieving the benefit of distributed driving that minimizes IR drop and RC delay.
3Device complexity
If socket regions are positioned centrally along word and bit lines, then interconnection requirements are reduced, but lithographic precision requirements increase
Solution Approach 1:
The socket regions are merged with the centrally positioned driver regions to form integrated driver-socket units. This merging consolidates multiple functions into single locations, reducing the overall interconnection requirements by eliminating separate socket structures and simplifying the lithographic patterning process while maintaining the benefits of central positioning.
Data Source
AI summary
For multi-level interconnect metallization, each metal level maintains a parallel line arrangement within a region, and the lines of each adjacent metal level are orthogonal or otherwise cross with one another. Vertical shunting among levels for routing in different directions employs short paddles that stay within the parallel scheme, and multiple paddles within a region at the same metal level can be co-linear. Parallel lines in the same metal level can be rotated with respect to one another in adjacent regions, for example to better interface with driver circuitry with orthogonal orientations in the different regions.