MOSFET Gate Drive Waveform Tuning Under Environmental Variation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing gate driver technologies struggle to optimize switching waveforms for MOSFETs due to the inability to account for environmental variables like temperature and supply voltage, which are observable but not controllable, leading to suboptimal performance and increased switching losses.

Innovation Solution

The method employs Bayesian optimization to generate a model that maps design and environmental variables to output metrics such as current overshoot and switching energy loss, allowing for the selection of optimal gate drive waveforms that minimize these metrics by incorporating observable environmental variables into the optimization process.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If traditional single-step waveform is used to control switching device, then device complexity is reduced, but current overshoot increases and switching performance deteriorates

Engineering Contradiction:
Improvegate driver waveform complexityVSAvoidcurrent overshoot
Core Design Contradiction:
Device complexityVSObject-generated harmful factors

Solution Approach 1:

The gate drive waveform is segmented into multiple distinct phases: a first voltage level to turn on the MOSFET, a second voltage level to maintain conduction, and a third voltage level to turn off the MOSFET. This segmentation allows optimization of switching performance by controlling the transition characteristics at each phase, thereby reducing current overshoot while maintaining manageable device complexity.

Inventive Principle:
Principle #1Segmentation

2Device complexity

If environmental variables are omitted from the model, then model simplicity is maintained, but model accuracy and waveform optimization performance deteriorate

Engineering Contradiction:
Improvemodel complexityVSAvoidmodel accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The system utilizes readily available sensor data from the electronic device's existing environmental sensors (temperature, humidity, light, sound, acceleration, gyroscope, magnetometer, proximity, barometer) to inform the waveform optimization model. This self-service approach allows the model to incorporate environmental variables without requiring additional complex measurement infrastructure, thereby improving model accuracy while maintaining practical simplicity.

Inventive Principle:
Principle #25Self-service

3Object-generated harmful factors

If multi-phase gate drive waveform is implemented, then switching performance and current overshoot control are improved, but device complexity increases

Engineering Contradiction:
Improvecurrent overshootVSAvoidgate driver waveform complexity
Core Design Contradiction:
Object-generated harmful factorsVSDevice complexity

Solution Approach 1:

The gate driver circuit is designed to dynamically transition between multiple voltage phases based on real-time sensor feedback and optimization objectives. The circuit selectively applies different voltage levels (first, second, and third phases) to the MOSFET gate depending on the switching state and environmental conditions, enabling precise control of current overshoot while adapting to varying operational requirements without excessive complexity.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentUS11515872B1Active gate driver optimisation with environmental variables
Publication Date: 2022.11.29 KK TOSHIBA
  • US11515872B1 patent drawing
  • US11515872B1 patent drawing
  • US11515872B1 patent drawing

AI summary

A method for active gate driving a switching circuit, wherein: a characteristic of a waveform controlled by the switching circuit is represented by a function mapping an input variable to an output metric, and wherein: the input variable comprises: a design variable having a first set of possible values; and an environmental variable having a second set of possible values, wherein the environmental variable is observable but not controllable. The method comprising: performing Bayesian optimisation on the function to generate a model of the function, wherein a next value of the design variable for evaluating the function is selected based on values of an acquisition function associated with a predicted value of the environmental variable; determining a first value of the design variable that optimises the model of the function; and controlling the switching circuit according to the first value of the design variable.