GaN and SiC Power-Device Selection for High-Voltage EV Converters
A completed architecture-aware electronics study comparing GaN and SiC power-device evidence positions for 400 V and 800 V electric-vehicle converters.

Project definition
Problem statement
GaN and SiC devices are often compared through isolated datasheet figures even though voltage class, topology, current, switching frequency, packaging, cooling, gate drive, EMI, qualification, and cost determine the converter result.
The engineering problem is to reject voltage-incompatible options first and then compare credible device-architecture pairs without presenting normalized datasheet positions as measured supplier rankings.
Project objectives
- Define six traceable GaN and SiC device evidence positions.
- Define seven 400 V and 800 V onboard charger, DC/DC, and traction architecture positions.
- Calculate conduction, switching, reverse, gate-drive, electrothermal, density, and relative-cost indicators.
- Compare case-temperature and switching-frequency sensitivity.
- Quantify bounded uncertainty using 10,000 fixed-seed samples per compatible pair.
- Build architecture-specific decision and Pareto shortlists with explicit engineering limits.
Project structure
Project components
Evidence positions
Retains voltage, resistance, charge, switching, reverse, thermal, qualification, maturity, and cost positions with source anchors.
Compatibility gate
Blocks device positions that cannot satisfy effective architecture voltage stress and derating.
Electrothermal model
Calculates parallel count, component loss, indicative efficiency, junction temperature, and eligibility across 270 cases.
Uncertainty model
Produces P05, median, and P95 loss, efficiency, and temperature positions with eligibility probability.
Decision analysis
Keeps efficiency, thermal margin, density, reliability, cost, driver complexity, and EMI evidence visible beside the final score.
Evidence package
Retains CSV, JSON, figures, tests, references, editable documentation, and a staged validation plan.
Methodology
Project workflow
- 01Declare
Load the device and converter architecture evidence positions.
- 02Gate
Reject voltage-incompatible device-architecture pairs before loss comparison.
- 03Calculate
Evaluate loss and temperature at three case temperatures and three frequency factors.
- 04Test uncertainty
Run 10,000 bounded samples for each compatible central pair.
- 05Rank
Build architecture-specific scores and cost-loss Pareto sets from eligible cases.
- 06Review
Trace every report value to retained results, equations, source annotations, and limitations.
Demonstration scenario
Run the study once, inspect the voltage-compatibility matrix, compare the central loss components, increase case temperature and frequency, then trace one architecture winner through uncertainty, decision score, and Pareto status. Explain why 650 V GaN is blocked from a two-level 800 V traction link but becomes eligible in a three-level position.
Engineering
Tools and method
- Tools
- The project uses Python 3.12, NumPy, pandas, Matplotlib for subject analysis, simulation, and results.
- Device and architecture data
- Python dictionaries define the traceable evidence and design positions.
- Numerical analysis
- NumPy and pandas implement deterministic, uncertainty, decision, and Pareto calculations.
- Figures
- Matplotlib generates twelve labeled comparison and sensitivity figures.
- Documentation
- The build creates editable Word and fixed PDF report and guide files with contents, figure, and table lists.
- Reproducibility
- A fixed seed, automated tests, repository checks, and Docker repeat the retained model.
Testing
Evaluation
Evaluation measures
- Thirty voltage-compatible central device-architecture pairs
- Two hundred seventy deterministic temperature and frequency cases
- Conduction, switching, reverse-recovery, reverse-conduction, and gate loss separation
- Estimated junction temperature and minimum thermal-margin eligibility
- Ten-thousand-sample P05, median, and P95 uncertainty positions
- Architecture-specific multi-criteria scores and twelve cost-loss Pareto cases
- Seven automated tests and twelve reproducible figures
Project boundaries
- All device values are normalized screening evidence positions, not supplier rankings or production design inputs.
- Indicative efficiency excludes magnetic, capacitor, control, filter, and cooling-system losses.
- Thermal, power-density, EMI, reliability, and cost values are screening positions and not measured compliance, lifetime, volume, or quotations.
- The project contains no high-voltage hardware, safety case, functional-safety evidence, or certification.
- Any hardware extension requires competent supervision, rated equipment, and an approved high-voltage laboratory.
- No information collected.
Included
- 01Six normalized GaN and SiC device evidence positions
- 02Seven onboard charger, high-voltage DC/DC, and traction architecture positions
- 03Voltage compatibility before electrical comparison
- 04270 deterministic temperature and frequency cases
- 0510,000-sample uncertainty analysis for every compatible pair
- 06Architecture-specific decision scores and 12 Pareto cases
- 07Twelve generated result figures and one licensed literature figure
- 08Seven automated tests and clean-container reproduction
- 09Complete project files, calculations, results, and analysis in a private GitHub repository
- 1082-page project documentation in PDF and editable Word formats
- 1119-page setup and usage guide in PDF and editable Word formats
- 12Sixty annotated references with a complete source matrix
Project record
No information is collected on this page.
- Permanent project ID
- GP-EC-0H9T2WP
- Catalogued
- 21 Aug 2026
- Completed
- 30 Aug 2026
- Verified
- 30 Aug 2026
- Demonstration
- Included in repository
Handover
After purchase
- 01Payment is confirmed
The project is marked unavailable and cannot be purchased again.
- 02Repository access is granted
The buyer's submitted GitHub account receives access to the private repository.
- 03The purchase record is delivered
The certification sheet is prepared from the reviewed buyer details and sent privately by email.