Hierarchical Row Decoder for Phase-Change Memory Power Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing row decoders for phase-change non-volatile memory devices face challenges in optimizing electrical characteristics and power consumption, particularly in achieving a balance between static power and area occupation, especially during reading and programming operations.
Innovation Solution
A hierarchical row-decoder architecture with distinct current paths for reading and programming operations, utilizing low-voltage and high-voltage MOS transistors to separate and optimize the selection of wordlines, allowing for reduced static power and improved performance during reading while accommodating higher power requirements during programming.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If a traditional row decoder architecture is used for phase-change memory, then the memory array can be accessed, but the static power consumption during reading operations is excessive due to biasing all selector transistors connected to selected wordlines
Solution Approach 1:
The row decoder is segmented into multiple independent decoding stages (first row decoder and second row decoder), each handling a subset of wordlines. This segmentation allows selective activation of only the decoding stage corresponding to the desired wordline range, thereby reducing the number of simultaneously active selector transistors and lowering static power consumption during reading operations.
Solution Approach 2:
The patent implements dynamic control of decoder activation through enable signals (e.g., first enable signal and second enable signal) that selectively activate only the required decoding stage based on the address input. This dynamic switching ensures that only necessary selector transistors remain in active state during reading, minimizing static power dissipation while maintaining full access capability.
2Use of energy by moving object
If the number of storage elements connected to one row is reduced to lower static power, then power consumption decreases, but the row decoder requires larger area occupation
Solution Approach 1:
The row decoder area is divided into separate decoding stages (first row decoder and second row decoder), each with its own selector transistors and control logic. By segmenting the decoder, the active area for any single decoding operation is reduced since only one stage needs to be fully active at a time, while the total area accommodates multiple stages in a distributed manner throughout the memory array.
Solution Approach 2:
The patent employs a hierarchical decoder structure where first row decoders and second row decoders are distributed across different memory array regions. Each decoder stage is nested within its corresponding memory block, allowing the overall decoder area to be distributed and shared across multiple memory sections rather than concentrated in a single large block.
3Power
If high voltage is applied to wordlines during programming operations, then phase-change material switching is effective, but the electrical characteristics requirements differ from reading operations
Solution Approach 1:
The patent implements dynamic voltage switching capability where the same wordline infrastructure can operate at different voltage levels depending on the operation mode. During programming, high voltage is applied to achieve phase-change material switching, while during reading, low voltage is used to detect resistance states without causing unintended phase changes. This dynamic voltage adaptation is enabled by controlled activation of selector transistors and appropriate biasing schemes.
Solution Approach 2:
The electrical parameters (voltage and current) of the wordlines are dynamically changed based on the operation type. The patent employs different voltage levels and current magnitudes for programming versus reading operations, with the row decoder controlling these parameter changes through selective transistor activation and enable signals. This parameter adaptation allows the system to meet the distinct electrical requirements of each operation mode using the same hardware infrastructure.
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
This approach reduces static power consumption during reading and dynamic power consumption overall, while allowing for optimized transistor sizing and improved performance by separating the row-decoding paths, thus enhancing the efficiency and area utilization of the non-volatile memory device.
Implementation Method 1
this electrical current, by the Joule effect, generates the temperatures necessary for the phase change
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
Described herein is a row decoder (5) for a phase-change memory device (1) provided with an array (2) of memory cells (3) organized according to a plurality of array wordlines (WL) and array bitlines (BL); the row decoder (5) has a hierarchical architecture and has a global decoder (8; 8') that addresses a first (MWL_LV; MWL') and a second (MWL; MWL_WR) global wordline according to first address signals (GP, D); and at least one local decoder (9; 9'), which is operatively coupled to the global decoder (8; 8') and addresses a respective array wordline (WL) according to the value of the first global wordline and the second global wordline and of second address signals (WLSEL, WLSELN_LV; WLSEL_SW). The local decoder (9; 9') has a first circuit branch (25, 26, 31; 40, 41) generating, when the first global wordline (MWL_LV; MWL') is addressed, a first current path between the array wordline (WL) and a first biasing source (VDD) during a reading operation; and a second circuit branch (32; 42) generating, when the second global wordline (MWL_LV; MWL') is addressed, a second current path, distinct from the first current path, between the array wordline (WL) and a second biasing source (VCC) during a programming operation.