Energy Fluxes from a
Typical Quad Flat Pack

Figure 1 - 208L PQFP in JEDEC Test Enclosure

One of the things that always used to worry me was that I didn't have a real feel for where the energy from component actually went. Did the majority get into the air through convection or did it pass into the board through conduction or to the surroundings through radiation?

So, in an idle moment, I started playing around with the FLOTHERM model of the JEDEC 208L SQFP. For those of you who aren't familiar with this model, it was created by Harvey Rosten (Flomerics' Technical Director) in parallel with experimental measurements being carried out by members of JEDEC JC15.1 and in support of work on the DELPHI Project. A comprehensive discussion of the modeling of the package including examples of its application to real situations and comparison with experimental data is contained in a paper which was presented at the 1995 ECTC Conference in Las Vegas.

The package itself is a specially manufactured 208 lead SQFP which is mounted on a test board. The test board is then mounted horizontally in still air and the package is set to dissipate 1W.

Figure 2 - Exploded View of the FLOTHERM Model of the 208L PQFP

When run, the FLOTHERM model gives results which are in good agreement with the experiments so we have a fair amount of confidence in it. If you want to have a look at it for yourselves, a slightly simplified version is delivered with every copy of FLOTHERM as a demo case called ZZQFLP. It has also been discussed in a number of technical papers at various conferences over the last year.

In playing around with the model, the only changes that I made were to "instrument" it by adding a series of Mean Flow Regions:

• on the top of the package;
• on the base of the package;
• and on the leads.

Using these, I could then see how much of the energy was passing in each direction.


Some Results

The results for the QFP in natural convection were as follows. Note that, since I am only using a half model of the component, the total power is 500mW.

The top of the QFP passes 97.3mW (19.5% of the total). Of this, 44.7mW (9.0%) is goes into the air and is convected away. The remaining 52.6mW (10.5%) is radiated to the environment.

Beneath the base of the component, there is a small air gap. Heat passes to the board through a combination of radiation and conduction. For the operating conditions we are looking at here, the radiative component is negligible and the conduction accounts for 79.8mW (16.0% of the total).

The sides of the QFP pass the majority of the heat lost from this component: 322.9mW (64.8% of the total).

Of the heat passing through the sides, a mere 9.0mW ( 1.8%) is lost through convection. This is hardly surprising given the little surface area available. However, a massive 313.9mW (63.0%) passes through the leads.

Of the heat passing through the leads, a small fraction - 19.0mW (3.8%) is radiated to the environment. But, by far and away the largest contribution to the cooling of this package is conduction through the leads to the boards: 294.9mW or 59.2% of the total.

Figure 3 - Energy Flux Budget

It would be dangerous to assume that these proportions are the same for every component and every mounting arrangement. A CPGA (for example) will have different flux characteristics and I suspect that we would get a very different set of results if the orientation of the board was vertical rather than horizontal or if we were dealing with forced convection. However, I would like to mention a couple of things:

The large amount of heat being passed to the board (>80% of the total) highlights the importance of having a good model for conduction within and through the board (and the socket if appropriate).

It also means that testing the thermal characteristics of a package and quoting them without saying something about the mounting arrangements is misleading.

With 80% of the heat going straight into the board,it would be easy to assume that we could use a conduction only solution with specified (correlation based?) heat transfer coefficients.

If anything, the opposite is true. Any heat that gets into the board must leave somehow. The only two ways for this heat to leave are conduction through the fabric of the board and convection and radiation from the board surface. We still need to look carefully at convective effects!

• if testing hardware and quoting experimental results for a component, be sure to specify the mounting arrangements (socket and board).
• if modeling, pay attention to the heat transfer from the board whether through conduction, convection or radiation.
  
  

If you want to carry out an analysis of this kind for yourself, you will find a simplified version of the Quad Flat Pack model included in your software delivery as ZZQFLP.PFM. There is some information about the case in the FLOTHERM User's Guide on page C-12.

Mean Flow Regions

The use of Mean Flow Regions is described in the FLOTHERM Reference Manual on pages 2­24 to 2-25 and 10-19.