IIR Digital Filter for Transient Thermal Modeling

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for dynamic thermal management in integrated circuits require significant computational resources, making real-time temperature prediction and power management challenging due to the need for extensive numerical calculations and storage of previous data points, especially for components with varying time constants spanning multiple orders of magnitude.

Innovation Solution

The implementation of a Foster RC network converted into an infinite impulse response (IIR) digital filter, which allows for the calculation of temperature responses using a recursive relationship between current and previous power inputs, enabling rapid thermal response prediction by constructing an equivalent resistor-capacitor network that simulates the thermal response of the system.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If direct finite element modeling (FEM) is used for thermal simulation, then measurement precision of thermal response is improved, but computing time increases significantly

Engineering Contradiction:
Improvethermal response prediction accuracyVSAvoidcomputing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent creates a simplified thermal response model that copies the essential thermal behavior characteristics of the complex FEM model. Instead of using the full FEM model for real-time predictions, the invention develops a reduced-order model that replicates the thermal response curve key features, enabling fast computation while maintaining acceptable accuracy for thermal management decisions

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent transforms the thermal simulation problem by changing from solving the complete FEM partial differential equations to using a parameter-based reduced-order model. The thermal response is represented through simplified parameters such as thermal time constants and resistance values that capture the dominant thermal behavior, dramatically reducing computational complexity while preserving essential thermal characteristics

Inventive Principle:
Principle #35Parameter changes

2Power

If thermal equivalent circuits with discrete convolution integrals are used, then computational resources are reduced compared to FEM, but extensive numerical computation and storage of previous data points are still required

Engineering Contradiction:
Improvecomputational powerVSAvoidnumerical computation complexity
Core Design Contradiction:
PowerVSDevice complexity

Solution Approach 1:

The patent replaces the numerical computation-based discrete convolution integral approach with an analog electronic circuit implementation. The thermal equivalent circuit uses physical RC (resistor-capacitor) elements to naturally perform the convolution operation through electrical charge accumulation, substituting complex numerical calculations with straightforward electrical measurements that are inherently parallel and require minimal computational resources

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The thermal equivalent circuit performs self-computation through its physical structure. The RC network automatically integrates the power history through capacitor charging, eliminating the need for external computational systems to perform discrete convolution integrals. The circuit itself serves as the computational engine, requiring only simple voltage or current measurements to obtain thermal response

Inventive Principle:
Principle #25Self-service

3Measurement precision

If measurement of thermal step-response function is implemented, then thermal characterization is achieved, but response time for components with time constants spanning multiple orders of magnitude becomes excessively long

Engineering Contradiction:
Improvethermal characterization accuracyVSAvoidthermal response measurement speed
Core Design Contradiction:
Measurement precisionVSSpeed

Solution Approach 1:

The patent segments the broad thermal time constant range into multiple distinct frequency decades. Instead of attempting to measure the entire thermal response from microseconds to hours in a single step-response test, the invention divides the measurement into multiple frequency sweeps, each covering a specific decade range. This segmentation allows the system to capture both fast and slow thermal responses using appropriately scaled measurement intervals for each frequency range

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent employs periodic sinusoidal power excitation signals at different frequencies to probe the thermal response across multiple time scales. By applying periodic heating at various frequencies and measuring the temperature response amplitude and phase, the system can extract thermal characteristics for different time constants without waiting for complete step-response settling, dramatically reducing measurement time while maintaining accuracy

Inventive Principle:
Principle #19Periodic action

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 significantly reduces computational time for thermal response calculations, enabling real-time monitoring and management of thermal transients, allowing for efficient dynamic thermal management in integrated circuits.

Implementation Method 1

use the known thermal transfer function of an electronic system to generate an equivalent resistor-capacitor (RC) network having a dynamic response that is identical (or very similar) to a given power excitation as the actual electronic system would have to that power excitation

Methodology Applied
Scientific EffectThermal-electrical analogy:

Implementation Method 2

generate an equivalent resistor-capacitor (RC) network

Methodology Applied
Scientific EffectCapacitance: Capacitance

Implementation Method 3

resistor-capacitor (RC) network having a dynamic response that is identical (or very similar) to a given power excitation

Methodology Applied
Scientific EffectJoule heating: Joule Heating

Data Source

PatentUS8635044B2Transient thermal modeling of multisource power devices
Publication Date: 2014.01.21 ADVANCED MICRO DEVICES INC
  • US8635044B2 patent drawing
  • US8635044B2 patent drawing
  • US8635044B2 patent drawing

AI summary

Embodiments of systems and methods for improved measurement of transient thermal responses in electronic systems are described herein. Embodiments of the disclosure use the known thermal transfer function of an electronic system to generate an equivalent resistor-capacitor (RC) network having a dynamic response that is identical to a given power excitation as the actual electronic system would have to that power excitation. Using the analogy between thermal and electrical systems, a Foster RC network is constructed, comprising a plurality of RC stages in which resistors and capacitors are connected in parallel. Subsequently, the analog thermal RC network is converted into an infinite impulse response (IIR) digital filter, whose coefficients can be obtained the Z-transform of the analog thermal RC network. This IIR digital filter establishes the recursive relationship between temperature output at the current time step and measured power input at the previous time step. Using this IIR digital filter, temperature response subject to arbitrary time-dependent power can be calculated in very small amount of time compared with prior art methods.