Designing reflowable modules

[asmi]

Hi guys,

I wonder how to design a module that can withstand a reflow after the module itself is assembled. Is it covered with some sort of potting, or all module's parts are glued using termosetting glue that can withstand repeated reflows?

[wraper]

If you pot it, most likely it will self destroy during reflow by solder squeezing out. If it's a simple PCB, then it's just like running PCB through second reflow which is quite normal process, especially for double sided assembly.

[wraper]

No wizardry.

[asmi]

The module I'm thinking of is going to have a tiny 0.4 mm pitch chip-scale BGA and I'm kind of afraid it will go out of alignment during subsequent reflows. Is this fear is unfounded?

[wraper]

The module I'm thinking of is going to have a tiny 0.4 mm pitch chip-scale BGA and I'm kind of afraid it will go out of alignment during subsequent reflows. Is this fear is unfounded?

Why would it? Solder surface tension tries to self align, not opposite.

[SiliconWizard]

Indeed. Components tend to self align. Depending on the initial quantity of solder paste and other factors though, you may occasionally experience tombstoning on some parts that didn't happen during the first reflow cycle.

One thing you have to worry about is that FR4 substrates only tolerate a limited number of reflow cycles. You should probably select a high-Tg FR4 PCB to minimize your risks.
On the same line of thoughts, you should probably bake your modules before each new reflow cycle. They will inevitably absorb some moisture and getting them as is in a reflow oven could have bad consequences.
Of course, regarding all the components on the module, you also have to figure out how many cycles they are able to take.

[ANTALIFE]

Like others have said should not be an issue, also would not recommend potting/coating the board as it introduces another variable that will probably go wrong

One thing to be extra wary of is max number of reflow cycles for components, for example I have come across some low ESR tantalum caps that could only be reflown once

[wraper]

On the same line of thoughts, you should probably bake your modules before each new reflow cycle. They will inevitably absorb some moisture and getting them as is in a reflow oven could have bad consequences.
Of course, regarding all the components on the module, you also have to figure out how many cycles they are able to take.

It's the same as with bare PCBs or moisture sensitive components. If they are packaged properly, baking is not needed.

you may occasionally experience tombstoning on some parts that didn't happen during the first reflow cycle.

That's impossible unless there was soldering defect to begin with. It cannot simply detach from the solder, surface tension won't allow it.

https://youtu.be/50ZBlDQJnAY

[mrpackethead]

Modules just annoy me.  What is the justification for using them.

[wraper]

Modules just annoy me.  What is the justification for using them.

For example not everyone wants to design and test RF part, use appropriate PCB for it, tune/calibrate the thing.
Say Blueetooth is relatively easy but making own GSM transceiver is very prohibitive.

[asmi]

Modules just annoy me.  What is the justification for using them.

Are you annoyed by anything you don't understand?

One good reason for module in my case is price. 0.4 mm pitch BGAs require expensive HDI process (laser-drilled microvias + very fine trace/spacing), but if the rest of the board can make do with cheaper more conventional process, there are savings to be had if expensive part of the board is implemented as a small module. This is my case - I want to use a very specific chip that is only available in chip-scale BGA package and has no substitutes I know of, so I'm currently researching if it would make sense to make a module for this part. I know for fact that at least one other user of that ICs - manufacturer of very popular MCUs so not a poor Joe I am - did the very same thing I'm planning to do, in fact this is what prompted me to seriously consider going with this. The good thing is that this IC is quite cheap, so parts cost for a single module would be very low, and the price of module would be dominated by PCB price. Thus it would make sense to do a sizeable run to amortize tooling cost (this is always the largest portion of any prototype run's cost for just about anything sans very simple 2- or 4-layer boards without any special requirements).

[IconicPCB]

Tg is important but so is Td.. .

[SiliconWizard]

you may occasionally experience tombstoning on some parts that didn't happen during the first reflow cycle.

That's impossible unless there was soldering defect to begin with. It cannot simply detach from the solder, surface tension won't allow it.

I do not completely agree with this, and happen to have seen this in a past project. This was not a module but a board that had to go through 2 reflow passes due to the way it was designed (not getting into details but basically it had to be tested after the first pass with incomplete assembly for testability reasons and due to very expensive parts).

Surface tension won't allow "detaching", unless there is enough unbalance between the two sides.

Although tombstoning is much more likely to happen on the first reflow cycle, this is not impossible for that to happen during a second cycle. It's due to an unbalance between the two sides and that unbalance may come from two main sources: wettability issues and surface tension. Obviously the wettability will become a non-issue during a second cycle, but there may still be an unbalance in surface tension during reflow if both sides have significantly different thermal masses. For very small parts such as 0201 or even 0402, that can cause enough torque to rotate the component. And yes this is usually a mix of a design issue (too different thermal masses at the pads), unsufficient or uneven solder paste quantity, and the reflow thermal profile.

The fact this is a module here and that the reflow profile during the first cycle and the second may be different could also be a reason why this could happen during the second cycle only.

Admittedly this is a relatively rare event and is bound to happen only with pretty small/light components and design specifics. Just thought it would be interesting to point it out.

[wraper]

Although tombstoning is much more likely to happen on the first reflow cycle, this is not impossible for that to happen during a second cycle. It's due to an unbalance between the two sides and that unbalance may come from two main sources: wettability issues and surface tension. Obviously the wettability will become a non-issue during a second cycle, but there may still be an unbalance in surface tension during reflow if both sides have significantly different thermal masses. For very small parts such as 0201 or even 0402, that can cause enough torque to rotate the component. And yes this is usually a mix of a design issue (too different thermal masses at the pads), unsufficient or uneven solder paste quantity, and the reflow thermal profile.

If the board was already reflowed, then imbalance does not matter. If one side is heated slower, then solder is simply solid on one side when other is already melted. Solid solder on one side won't allow opposite side to lift it up. If something what looked fine after first reflow was lifted after second reflow, then those parts were not wet by solder on one side to begin with. And there was a defect like on the pictures below.

[mrpackethead]

Are you annoyed by anything you don't understand?

Understand them and use lots of them.  They still annoy me, they just come with a lot of compromises. 

[Clear as mud]

What do Tg and Td mean?

[SiliconWizard]

If the board was already reflowed, then imbalance does not matter.

Sorry I missed your reply.

If one side is heated slower, then solder is simply solid on one side when other is already melted. Solid solder on one side won't allow opposite side to lift it up.

I'm no physicist, but I don't think it goes instantly from solid to liquid. The phase transition can be a tricky thing to model. So I think saying that one side would be transitioning faster than the other seems more accurate than saying that one side would be solid while the other would be liquid. And I have no clue which of the two forces would be greater at any given point in time during reflow. Would be interesting to have a physicist's point of view.

If something what looked fine after first reflow was lifted after second reflow, then those parts were not wet by solder on one side to begin with. And there was a defect like on the pictures below.

This could have been the case, although it passed undetected. On very small components (like 0201's) it can be pretty hard to see if it's barely lifted, unless you inspect all the boards with a microscope and from an angle... and electrical testing may fail to see that especially if it's bypass capacitors. So I don't deny it could have been that. But at least it can happen and not be *that* obvious to detect.

We didn't have the time or interest to investigate further back then, we just took actions with the assembler in order to avoid the problem (actions which would be valid at any reflow cycle anyway).

Again I'm still not convinced that's it's impossible. Phase transition is a tricky business. I'd be glad to see some real study on this though, such as both an analytical study with a decent model and a practical study.

[SiliconWizard]

What do Tg and Td mean?

You may be interested in the following paper: https://www.cedm.be/system/files/public/library/publications/2010-09_Global-SMT_Extract_PCB-Delamination.pdf

(especially the part on "The number of solder cycles-to-delamination".)

[OwO]

I'm no physicist, but I don't think it goes instantly from solid to liquid. The phase transition can be a tricky thing to model. So I think saying that one side would be transitioning faster than the other seems more accurate than saying that one side would be solid while the other would be liquid. And I have no clue which of the two forces would be greater at any given point in time during reflow. Would be interesting to have a physicist's point of view.

Shitty solder transitions from solid to sludge to liquid. It's this sludge phase that surface tension can be low and parts can tombstone if one side is colder than the other. Solution is to use eutectic 63/37 solder.