Charge Pump Modeling via Frequency-Domain Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for modeling low-dropout regulators and charge pumps are inefficient, producing unstable and complicated time-domain models that are dependent on load or power delivery networks, resulting in inaccurate simulation results and requiring frequent model updates with changes in load or PDN conditions.
Innovation Solution
The use of frequency-domain fitting methods, including vector-fitting algorithms, to approximate impedance functions and synthesize circuits for LDO regulators and charge pumps, allowing for the creation of stable, load-independent models that can simulate various loads and PDNs accurately without the need for frequent model rederivation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional time-domain modeling methods are used for LDO regulators and charge pumps, then the modeling process is straightforward, but the resulting models are unstable, complicated, and dependent on load or power delivery network conditions
Solution Approach 1:
The patent replaces conventional time-domain modeling methods with frequency-domain fitting methods. Specifically, it uses vector-fitting algorithms to approximate impedance functions in the frequency domain, then transforms these approximations into stable time-domain models. This substitution of modeling domain (from time to frequency) resolves the contradiction by producing models that are both stable and manageable in complexity.
Solution Approach 2:
The patent changes the fundamental parameters used for modeling by transitioning from time-domain parameters to frequency-domain parameters. By performing frequency-domain analysis at multiple frequencies and using vector-fitting to approximate impedance functions with rational expressions, the method generates models with improved stability properties while controlling complexity through the fitting process.
2Productivity
If conventional modeling methods are used, then the initial model creation is simpler, but frequent model updates are required when load or PDN conditions change
Solution Approach 1:
The patent creates universal models through frequency-domain fitting that are independent of specific load or power delivery network conditions. The vector-fitting approach approximates the impedance function in a way that captures the essential behavior of the LDO regulator or charge pump across different operating conditions, making the model universally applicable without requiring frequent updates when loads or PDNs change.
Solution Approach 2:
The method performs preliminary frequency-domain analysis and vector-fitting to create robust models that anticipate variations in load and PDN conditions. By establishing the model foundation in the frequency domain with appropriate fitting techniques, the approach pre-presents a model structure that is inherently more adaptable to condition changes, reducing the need for subsequent updates.
3Measurement precision
If conventional time-domain models are used, then the modeling approach is simpler, but simulation accuracy across different loads and PDNs deteriorates
Solution Approach 1:
The patent substitutes frequency-domain analysis for conventional time-domain analysis to achieve superior simulation accuracy. By performing frequency-domain analysis at multiple frequencies and using vector-fitting algorithms to create rational approximations of the impedance function, the method captures system behavior more accurately across different operating conditions, then transforms these accurate frequency-domain representations into time-domain models for simulation.
Solution Approach 2:
The patent moves the modeling process to another dimension by transitioning from time-domain to frequency-domain analysis. This dimensional change allows for more accurate characterization of the system's impedance behavior across different frequencies, which then translates to improved simulation accuracy when the model is used to predict performance under various load and PDN conditions.
Data Source
AI summary
Computer-implemented systems and methods are provided for modeling a charge pump. A relationship between an output voltage of the charge pump and a loading condition is determined. A frequency-domain analysis is performed at multiple frequencies to determine an impedance function representative of the charge pump's impedance at each of the multiple frequencies. A vector-fitting algorithm is applied to approximate the impedance function using a plurality of poles and residues. A circuit is synthesized based on the plurality of poles and residues. A model for the charge pump is generated, where the model includes the synthesized circuit and components that model the relationship between the output voltage and the loading condition.


