Local Power Multiplexers for PDN Area Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity and size of power delivery networks in integrated circuits (ICs) are increased due to the need for multiple power rails to handle the worst-case current requirements of various cores, leading to a substantial area requirement for implementing these rails.
Innovation Solution
A system and method utilizing local power multiplexers, where each core has a primary power rail for lower demand levels and a secondary power rail for higher demand levels, with a dynamic clock voltage scaling (DCVS) controller to determine which rail to use based on the core's requested voltage level, allowing for reduced size and increased efficiency by routing the larger power rail to handle higher currents when needed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple power rails are provided to handle worst-case current requirements of various cores, then all cores can access required DCVS levels, but the overall size of the power delivery network becomes substantially large
Solution Approach 1:
Multiple power rails are merged into a single shared power rail through the use of power multiplexers. Each core has access to multiple DCVS levels via the multiplexer network, which dynamically connects the appropriate power rail to each core based on its current requirements. This combining approach reduces the total power delivery network size while maintaining full adaptability to all DCVS levels.
Solution Approach 2:
The power delivery network transitions from a static configuration where each core has dedicated power rails to a dynamic configuration using power multiplexers. The multiplexers dynamically reconfigure the power rail connections based on real-time DCVS level requirements, allowing a smaller set of power rails to serve multiple cores at different power levels as needed.
2Reliability
If each power rail is sized to handle the worst case highest current, then all cores can operate at maximum performance, but the area needed to implement the multiple power rails becomes substantially large
Solution Approach 1:
Instead of having multiple oversized power rails each capable of handling worst-case current, the system merges power delivery through multiplexers that dynamically route current from a smaller set of power rails. The multiplexer network allows any single power rail to be connected to any core that needs maximum current, eliminating the need for redundant high-capacity rails for each core.
Solution Approach 2:
Each power rail is designed to universally serve multiple cores at different DCVS levels through the multiplexer network. A single power rail can be dynamically connected to any core that requires its current capacity, making each rail multi-functional rather than dedicated to a single core or function.
3Adaptability or versatility
If multiple power rails are implemented to provide independent supply voltages to various cores, then voltage scaling can be achieved, but the complexity of the power delivery network increases
Solution Approach 1:
The system combines multiple power rail functions into a unified power delivery architecture using multiplexers. Instead of independently managing multiple power rails, the multiplexer network consolidates control, reducing the overall system complexity while preserving the ability to provide different voltage levels to different cores through dynamic switching.
Data Source
AI summary
A power delivery network (PDN) including a battery, a set of regulators for generating supply voltages, and an integrated circuit (IC) including power rails configured to receive the supply voltages. The IC further includes an IC chip having a set of cores. The power rails includes a larger rail configured to provide a full range of currents, and the other smaller power rails each configured to provide lower range of currents. The IC includes multiplexers having first inputs coupled respectively to the smaller rails, second inputs coupled to the larger rail, and outputs coupled to the cores. When the smaller rail is able to supply the current needed by a core, the multiplexer is configured to couple the smaller rail to the core. When the smaller rail cannot supply the current needed by the core, the multiplexer is configured to couple the larger rail to the core.