Why are these SMD heatsinks designed to not touch the IC?

Question

I am using a 7106DG heatsink to dissipate the heat from an IC. It looks just like in its datasheet:

It leaves a gap between the IC and the heatsink, and the datasheet even mentions it, but I guess no one is surprised.

Other heatsinks (in the same datasheet) leave an even larger gap:

As our company sells products that need heatsinks for our ICs, I am trying to choose the best heatsink, and use it as intended, but I bend the heatsink to make it touch the IC, as this seems to be the best way to transfer the heat.

I chose the first one (7106 heatsink) which dissipates heat better, as its graph next to it indicates, but I also make changes to the heatsink to make it touch the IC better, like below:

1. Bend heatsink's legs/pads to make the top part of the IC touch the heatsink.

2. Cut the heatsink's legs to make the top and sides of the IC to touch the heatsink.

My questions and the possible answers I think are:

1. Why are these heatsinks designed and advertised as a means to transfer heat only from the bottom of the IC, while they could do it from the top as well? (Maybe the top part of the IC should NOT get hot, making my designs/above pics wrong?)

2. Should the gap be filled with thermal paste? Why do none of the pictures shown in the datasheet use thermal paste? (I guess gap filling paste for large gaps and liquid thermal paste for parts that touch the IC would be a good idea and they just don't show that on the pictures.)

It would be even cheaper for the factory to produce heatsinks with smaller legs, so there must be a reason they do not touch the top of the IC.

Answer 1

Thermal Pad

The bottom of the IC's TO-263 package has a thermal pad that directly solders to a copper pad on the PCB. Your heatsink then solders to that same copper pad on the PCB. Heat produced via the case (not via the pad) will be about one-tenth of that transferred to the PCB copper pad hence, the majority of that heat passes to the heatsink via the copper pad.

That's how it's meant to work. Note the words in the data sheet: -

It's a surface mount heatsink. Note the words on the page in the data sheet: Material: 0.63 (0.025) Thick Copper..... Finish: Tin Plated

In other words, it's not aluminium (like most heat sinks) and easily solders to the copper pad on the PCB (that is thermally connected to the IC via it's thermal pad). This picture (from the question) isn't the intention of the heatsink manufacturer: -

Recap

The IC's TO-263 footprint (as specified in the heatsink data sheet) has a thermal pad like so: -

That thermal pad should solder to copper on the PCB that extends wider than the IC so that the heatsink can also make a solder connection to it. Greater than 90% of the heat produced by the chip is transferred this way.

Answer 2

There are several things at work here, all giving the same result, the actual thermal path in the IC, mechanical tolerances, and the poor thermal conductivity of air.

In the IC, the active die is mounted directly on the lead frame. As it's designed as a high power package, care has been taken to make the lead frame thick and wide, and have a short path out to the solder points. If these are kept cool, the vast majority of the heat will leave the die by this route, little will make its way through the high thermal resistance plastic package above the die.

In order to make the heatsink touch the top of the package, there are two dimensional uncertainties that have to be kept under control, the package thickness and the heatsink leg length. Making the heatsink legs long so it's guaranteed to sit on the board solves this.

Merely making the heatsink touch the package is not enough to move any significant amount of heat. The air has to be replaced with some solid or liquid to make a better interface, which is messy, expensive, and an extra assembly operation.

Answer 3

The thermal conductivity of the metallic pads is much better than that of the case made of plastic. The amount of heat that may be removed via the case is negligible small.

Answer 4

From AAVID: Surface Mount Heat Sinks for D-PAK, D²PAK and D³PAK Packages

THEORY OF OPERATION: Heat from the D-PAK, D²PAK or D³PAK device is removed indirectly through conduction – no direct contact is made to the SMT device. Principally, the heat generated by the D-PAK/D²PAK/D³PAK device is conducted down into the PCB where it is picked up by a copper drain pad. The drain pad spreads the heat beyond the edge of the package to the point where the heat sink is mounted to the PCB. The heat is then conducted into the copper heat sink where it is dissipated into the surrounding environment.

The heat sink is not meant to make contact with the device. Rather the drain pad area is increased to allow heat to flow from the drain pad to the pcb to the heatsink.

From Intel - Physical Constants of IC Package Materials. Cu-W 90% (Tungsten-Copper) has a thermal conductivity of 180 - 200 W/mK (Watts per meter per degree Kelvin) and molding compound only 0.58 - 0.67 W/mK. Copper is 386 W/mK. Some heat will go through the package, but majority will go through the metal path ways. The heatsink is meant to facilitate this flow.

From Andys link, TO-263-3 drain pad is 10mm (8mm) by 7.55mm. Aavid recommends a drain pad of 16.3mm by 14.2mm (Red box) for TS-263 with a solder mask as shown. The drain pad recommended is significantly larger than the TO-263-3 drain pad. This allows heat to flow from package to copper to heatsink. Aavid calls this indirect conduction.

Ideally, as Aavid (in OP question) shows heatsink should sit on pcb and not rely on solder to transfer heat. I'd perform some test to see impact of gap on heatflow.

This has nothing to do with the question (and mainly why I did make an answer). In the area outlined in red, I see no change in coloring on pcb for copper drain pad. I fear because you did not understand how the heatsink is supposed to work, you used solder mask dimensions as copper dimensions. I see no vias on heatsink tabs either. The perspective is poor and hopefully, I am wrong about that.