last time we started talking about detailed modeling in its specifics, by looking at the appropriate approach for modeling a die. this month, let us consider modeling leadframes.
a leadframe is an essential constituent of all leaded packages. this includes qfp (quad flat pack), soic (small outline integrated circuit), plcc (plastic leaded chip carrier), etc.
a leadframe consists of a die mounting paddle (flag) and lead fingers. the die paddle primarily serves to mechanically support the die during package manufacture. the lead fingers connect the die to the circuitry external to the package.
one end of each lead finger is typically connected to a bond pad on the die by wire bonds or tape automated bonds. the other end of each lead finger is the lead, which is mechanically and electrically connected to a substrate or circuit board.
the leadframe is constructed from sheet metal by stamping or etching, often followed by a finish such as plating, downset and taping.
in the large majority of cases of a leaded package without a heatsink, up to 80% of the heat flows through the leadframe into the pcb. even in cases of such a package with an attached heatsink, up to half the heat could still flow though the leadframe. this makes modeling the leadframe correctly absolutely essential for creating an accurate thermal model of a leaded package.
structure of typical leadframes
most commonly, leadframes are two-sided or four sided. two-sided leadframes are seen in packages that have leads on two sides only, such as the soic and tsop. four-sided leadframes are present in packages with leads on all sides such as qfp and plcc.
four-sided leadframe in a qfp
tsop package with a two-sided leadframe
leadframe of a 20-lead soic
leadframes are also "single-sided", as in to type packages. such leadframes are asymmetric, and typically consist of only a few leads. the center lead in such leadframes is often a ground lead, and is fused to the die attach pad.
leadframe of a to-220 package
leadframes are commonly made of copper based alloys. however, alloy-42 leadframes are also seen. alloy-42 is a ferrous alloy whose conductivity is about 10-15 times smaller than a copper based alloy. hence, the choice of the material used for the leadframe has a critical bearing on the thermal performance of a leaded package.
for example, a lead-on-chip type tsop is almost always made up of alloy-42 because of the manufacturing challenges of using a copper alloy. this increases the thermal resistance of this package considerably.
the thermal resistance of the leadframe is also dependent on some other factors. a key one is the separation distance between the die flag and the leadframe.
since the intervening material is typically an epoxy-based encapsulant, the resistance across the gap is significant. minimizing the gap helps optimize the package thermal performance. a study has shown that the junction-to-ambient resistance grows by roughly 14 c/w for every millimeter of the gap in qfp packages.
the extent of the gap between the die flag and the leadframe affects the thermal resistance
to reduce the thermal resistance, occasionally some of the ground leads in a package are physically fused to the die flag. this technique works quite well in reducing the thermal resistance encountered by the heat as it flows from the die to the package.
modeling the leadframe
before deciding on how to model your leadframe in your thermal analysis tool, you must understand the critical thermal phenomena occurring as the heat flows through the leads. firstly, the die paddle/flag should be modeled discretely and separately from the set of lead fingers.
regarding the modeling onthe desirable solution would be to avoid modeling each lead discretely; this would save a huge amount of grid. however, this depends on the number and density of leads.
fortunately, if the leadframe has a large number of leads, you can usually get away by modeling it as a lumped cuboid with an orthotropic (i.e., directionally dependent) conductivity.
this means that the conductivity in the direction along the leads would be obtained by an assumption of parallel resistance, whereas the conductivity in the transverse direction would be obtained by an assumption of series resistance.
this would avoid spurious spreading of the lumped leadframe cuboid in the transverse direction in the model.
however, lumping the leads is a bad approach when few of the leads are fused to the die paddle. in this case, a more innovative strategy has to be utilized. modeling each lead finger discretely is ideal, but you could probably get away with modeling only the fused leads discretely and lumping the rest carefully in a manner such that they do not "short" (thermally speaking) with their fused counterparts.
another situation where modeling leads discretely may be called for is when the number of leads is small - say, less than 16. this is especially true if there are unusually large copper islands connected to some of these leads on the pcb. as in most modeling questions this is an issue of judgment, and is therefore a function of experience.
in some package families (e.g., powerpqfp from amkor), the leadframe is thermally enhanced by attaching it to the edges of the die paddle. naturally, a layer of insulation is required between the two but the overall effect is to provide a direct thermal path from the die to the leadframe.
this has a major impact on the thermal resistance. care must be taken to model this correctly.
|