Effect on observed package thermal performance of system board metal content

In a growing number of customer applications, the ultimate success of a semiconductor package design is measured by its thermal performance. Unfortunately, thermal databases consisting of junction-to-ambient thermal resistance (θ_ja) alone fall short of allowing the customer to adequately predict the actual thermal performance of a package, since the actual θ_ja of a package is sensitive in varying degrees to the details of the system board on which the package is mounted. Though this fact is often acknowledged, the magnitude of the effect is generally not appreciated, and it is rare to find sufficient information in a product data sheet to allow such an evaluation. Variations in thickness and coverage of signal traces, power and ground planes, and insulating layers, can change the nominal θ_ja of a package by from 25 to 50%. To illustrate this effect, several non-thermally enhanced Quad Flat Pack (QFP) type packages were mounted on several types and sizes of test boards, which were then both measured and modeled. Changes in overall θ_ja are correlated with the metal content of the boards