Multi-Stage Memory Clock Partitioning for Faster Low-Power Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory architectures face challenges in reducing access time and clock power consumption due to increased resistance and capacitance, particularly at lower technology nodes, leading to significant silicon area and power consumption issues.
Innovation Solution
A memory architecture that partitions the internal clock into multiple stages, specifically three segments, to distribute gate and metal capacitance load, optimizing access time and reducing power consumption without increasing silicon area.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If buffers are inserted into clock paths to reduce RC delay, then access time is improved, but silicon area and gate delay increase
Solution Approach 1:
The patent divides the memory into multiple banks and the clock distribution into multiple stages. Instead of using a single buffer for the entire clock path, the clock is partitioned and distributed through multiple smaller stages, reducing the RC delay per stage while minimizing the total silicon area required for buffering.
Solution Approach 2:
The patent introduces a temporal dimension to clock distribution by using multi-stage clock partitioning across different time phases. This allows overlapping operation of different bank stages, improving access time without proportionally increasing silicon area, as not all buffers operate simultaneously at full capacity.
2Speed
If buffers are inserted into clock paths to reduce RC delay, then access time is improved, but gate delay increases
Solution Approach 1:
The clock distribution is segmented into multiple stages with progressively finer granularity. Each stage uses smaller buffers that introduce less individual gate delay, and the multi-stage approach allows for optimized timing at each level, reducing the cumulative gate delay compared to a single large buffer approach.
Solution Approach 2:
The patent performs preliminary clock partitioning and buffer placement during the design phase to optimize the clock tree structure. By pre-calculating and positioning buffers strategically before fabrication, the design minimizes the propagation delay through the clock network while maintaining area efficiency.
3Speed
If more input gate capacitance and metal capacitance is introduced to reduce RC delay, then access time is improved, but toggle power increases
Solution Approach 1:
The patent segments the total capacitance load across multiple smaller buffer stages rather than concentrating it in a single buffer. This distribution reduces the toggle power at each stage because smaller capacitance transitions require less energy (E=1/2CV²), while the cumulative effect maintains or improves access time through reduced RC delay at each segment.
Solution Approach 2:
The patent changes the capacitance parameters by using progressively smaller buffer sizes in later stages of the clock distribution tree. This parameter optimization reduces the gate capacitance and metal capacitance in regions where large buffers would be detrimental, thereby reducing toggle power while maintaining the speed benefits through the multi-stage architecture.
Data Source
AI summary
A memory comprises a multi stage clock-partitioning circuit, and at least one upper bitcell memory array and at least one lower bitcell memory array. An input is configured to receive an external clock signal. A first stage of the multi stage clock-partitioning circuit is configured to receive the external clock signal and generate a first internal clock signal and provide the first internal clock signal to the bitcell memory arrays. A second stage of the multi stage clock-partitioning circuit is configured to receive the first internal clock signal and generate a second internal clock signal. A third stage of the multi stage clock-partitioning circuit is configured to receive the second internal clock signal and generate a third word line generated clock signal and provide the third internal clock signal to the at least upper and lower bitcell memory arrays.