日本語
SMT component clamping pressure guidelines?
Is there a good rule of thumb for the allowable clamping force applied to the top of an IC?
Say you have a board with some SMT power components (powerSO8 or DPAK transistors, ICs with thermal pads, etc), and this board will be mounted flat onto a heatsink. Naturally you want to ensure good contact between the PCB and the heatsink, so it makes sense to apply some clamping pressure to the packages to ensure that the area of the PCB with the dissipative components is in close contact with the heatsink. How do you establish a maximum design value for that clamping pressure?
A few things are immediately obvious:
- there will need to be some compliance to the clamping system to take up dimensional changes over temperature. This could be accomplished by designing elasticity into the clamping structure or by using a compliant thermal pad or similar between the clamp and the components.
- the minimum design pressure will be the amount of force required to overcome any out-of-flatness inherent in the PCB or imposed by its other mechanical constraints.
I see a few main factors that would be involved in a maximum clamping pressure:
- Changes in component/PCB electrical properties due to mechanical stress: since I have power circuitry in mind that is not particularly high precision, let's say this is negligible for now.
- Bending stress in the PCB: Since we're concerned with holding the PCB flat, this shouldn't be a serious concern for boards that are reasonably flat to begin with.
- Creep in the PCB substrate: I haven't seen specific numbers yet, but from what I have seen, it seems that the compressive stress for FR4 should be kept to something like 50MPa to avoid creep. However I'd like to find a good authority for this.
- Allowable stress in the device package/die: Although there is definitely literature out there, I haven't found any straightforward guidelines for this yet.
- Allowable stress in the solder joints: given that the parts in question have large under-package solder joints, this is a predominantly compressive stress, and since the joint area is relatively large, I don't expect that this is at all a limiting factor.
Has anyone analyzed this problem before, or have some good references for some quantitative guidelines?
- Comments(1)
A****min
May 13.2019, 14:45:26
May be hard to find or calculate, anyway; you'd have to do an FEA study, including device modulus and expansion rate, to figure it out anyway I think. Which is not at all infeasible even for small players today, but good luck getting part models of that level of detail, too...
Where larger clamping forces are expected, a bracket is often placed on the back side. Consider laptop CPU mounts, for example. FR-4 isn't very stiff, and a lot of the components placed on top, are (ceramic chip caps especially). Anything you can do to help distribute forces, without causing bending stress, is a big help, and allows that much more headroom to increase clamping forces.
I'm thinking of a lot of laptop hardware -- they need to do it right, and in very little space after all. Soft, high conductivity thermal pads and moderately stiff springs seem very much the way to go.
The soft, high-K thermal pads are particularly excellent. You get thermal conductivity from the face of the component, and from the copper pour around it. You don't have to worry about thermal conductivity vs. pressure, unlike with the prior generation's firm rubber pads. You do have to worry about how much distance it soaks up in the process (of compressing and spreading into the gaps), which is where the springs come into play.
An alternative modelling method might be to assume the board is rigid with an uneven surface (the components and the gaps between them), and see how much force is required for, say: 50%, 70%, 80%, 90%, 95%, whatever contact between the TIM and the board's outside surface. Assuming the TIM is homogeneous, isotropic, and has properties (modulus, strength, Poisson ratio..) consistent with the datasheet, this should be relatively straightforward to set up (I think even Blender could do such a thing -- if not necessarily quantitatively??).
Looking forward to any data others may have
