Power Delivery Network Worst-Case Voltage Drop Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current circuit design and simulation techniques, such as deterministic models, are inadequate for handling modern semiconductor technology requirements due to increased variation in semiconductor devices and interconnects, leading to complex delay and power consumption estimation, particularly with power supply voltage drops affecting path analysis.
Innovation Solution
A method for fast power simulation involving preselecting an input vector file, initializing supply voltage, performing event-driven simulation to extract time-varying power supply current waveforms, and using a linear network simulator to derive time-varying voltage waveforms, with iterative comparisons until convergence within a specified tolerance, while accounting for power delivery network models and dynamic voltage waveforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If deterministic models are used for power supply voltage drop estimation, then manufacturing precision is maintained, but simulation time increases significantly
Solution Approach 1:
The patent employs a statistical sampling approach where a limited number of input vectors are selected to represent the entire input space. Instead of exhaustively simulating all possible input combinations (which would be computationally expensive), the method uses a representative subset of test cases that capture the essential behavior of the power delivery network under varying loads and conditions.
Solution Approach 2:
The methodology transforms the problem from deterministic to statistical by changing the fundamental approach: rather than calculating exact voltage drops for every possible input condition, the system performs multiple simulations with different randomly selected input vectors and aggregates the results statistically. This parameter change from deterministic to probabilistic analysis dramatically reduces simulation time while maintaining acceptable accuracy for worst-case estimation.
2Loss of time
If statistical sampling with limited input vectors is used, then simulation time is reduced, but measurement precision of worst-case voltage drops decreases
Solution Approach 1:
The patent implements an iterative refinement process where the statistical sampling results are continuously improved. After initial simulations with a first set of input vectors establish a baseline worst-case voltage drop estimate, the method refines this estimate by performing additional simulations with a second set of input vectors. The feedback from the first simulation results guides the selection and weighting of subsequent test cases, progressively converging toward a more accurate worst-case estimate while controlling simulation time.
Solution Approach 2:
The methodology performs preliminary statistical analysis to identify which input vectors are most likely to produce worst-case voltage drops. By analyzing the results of initial simulations and using statistical techniques to identify critical test cases, the system prepares a refined set of input vectors for subsequent simulations. This preliminary action ensures that subsequent simulations focus computational resources on the most relevant test cases, improving measurement precision without proportionally increasing simulation time.
Data Source
AI summary
Systems and methods related to fast simulation of power supply networks and identification of a set of extrema (e.g., maxima or minima) waveforms associated with the power supply networks. In accordance with an embodiment, a method is provided for estimating the worst case voltage drop on the power delivery network of a circuit, comprising selecting a model of a power delivery network of a circuit, simulating the circuit over a predefined number of vectors, collecting dynamic voltage waveforms at each of a plurality of points on the power delivery network, calculating a dynamic worst case voltage waveform at each of the plurality of points, and reporting the dynamic worst case voltage waveform along with an associated confidence interval.