Draft of Thermal Standards for Power Modules

Deriving one number for thermal resistance of a power module is difficult for the following reasons:

  1. Different dice (in a multiple die module) are at different temperatures. In larger modules, paralleling is used to increase current ratings. Even among the same die type, different locations of the dice cause differing temperature distributions.
  2. Module baseplates are so large that the bottom of the baseplate is typically not isothermal due to lateral heat flux
  3. Mounting the module to a heatsink can significantly change the pattern of heat flux within the module. E.g. If the module is more tightly secured to a heatsink on one side, that side will have a lower thermal resistance to the sink. Thermal grease application is another major factor is heat flux and temperature patterns. Further, the heatsink geometry itself strongly influences the flux and temperature pattern within the module.

Proposed solution:

  1. Use a heatsink that is nearly isothermal across the module surface. This will almost certainly require liquid cooling, but the only mandate is that heatsink temperature not vary by more than 1 C under the module.
  2. The measurement of thermal resistance shall be a junction to sink measurement. The highest temperature on any die will be used for the high temperature measurement, and any temperature on the heatsink under the module may be used for the low temperature measurement. There must be at least a 10 C temperature difference from higher to lower temperature. Measurements of power dissipation and temperature must be accurate to three significant digits.
  3. If there are two separate types of switch (e.g. diodes and IGBTs), the two types shall be characterized individually.
  4. The mounting method of module to heatsink must be fully disclosed (e.g. thermal grease, solder (!), etc.).
  5. The heatsink temperature for the test must be disclosed (to avoid someone improving module performance by cooling to cryogenic temperatures where thermal conductivities are high).

Other issues:

  1. How to measure transient thermal impedance? Recommend similar procedure.
  2. Temperature measurement on dice: an IR camera has emissivity and accuracy issues, wire bonds get very hot and may skew measurement of die temperature for thermocouples.
  3. Will the separate measurements for diodes and IGBTs (point 3 in proposed solution) cause problems in that the total temperature will be higher when both are running? The heat from one affects the other. Superposition may work here.