Partial Discharge (PD) Mitigation in Power Module Design

Student: Shilpi Mukherjee

Degree: Ph.D., December 2020

Major Professor: Dr. Alan Mantooth

Research Area(s):

Modeling & Simulation

View Research Quadslide

Background/Relevance

  • Multi-chip Power Modules (MCPMs) are used in many power conversion utilities where electrical performance must be reliable and thermal/mechanical/EMI effects must be accounted for.

  • Partial discharge (PD) is a localized breakdown of insulation material. It leads to reduced life time of the insulation material.

Innovation

  • Manufacturing design rule check for partial discharge (and HV operation in general) has never been done before in a tool like PowerSynth where layouts are optimized for electrical and thermal performance.

Approach

  • Generate models for trace gap as a function of voltage and other parameters using FEA incorporating a derating factor to account for non-idealities. Fabricate and test to determine yield.
  • ∆x=f(ε_e, ε_c, d, ∆V) and ∆x ̂=D∗ ∆x

where

∆x is the trace gap,

ε_e is a characteristic of the encapsulating dielectric material,

ε_c is a characteristic of the ceramic dielectric material,

d is the copper pull back distance from the edge of the ceramic (ledge width),

∆V is the potential difference applied,

D is the derating factor based on non-idealities (such as metal edge terminations during wet etching), and

∆x ̂ is the adjusted trace gap based on non-idealities.

Key Results

  • E-field v. voltage v. tracegap plots are helping us determine appropriate design rules for power modules.

Conclusions

  • The new methodology helps determine design rules for power modules that operate at high voltages. The methodology can be used to determine yield of power modules that will pass a partial discharge test at a certain voltage.

Future Work

  •  Model spacing for various
    • encapsulation materials
    • With bubbles at various locations in the encapsulation material
  • Continue implementing the model in PowerSynth.

  • Implementing improvements as models become more accurate.