Ultra HDI design, or UHDI, is an approach that those in the packaging and IC substrate world are very familiar with. This design approach is used to design printed circuits with very fine lines for fabrication in an additive process, followed by sequential lamination of the layers into a substrate. UHDI is now moving into PCB design, with more commercial products slowly adopting the approach for ultra-high-density fabrication and assembly. Currently, the most common place it is used is in consumer products with high compute, small form factor, and dense component requirements, but we can expect to see it elsewhere.
In a recent podcast interview with Chrys Shea of Shea Engineering, NWES founder Zachariah Peterson discussed the use of a new test board for identifying HDI PCB assembly defects in volume manufacturing processes. The SMTA test board is available for use by assemblers targeting the UHDI or IC substrates markets, and the large number of placements on these boards provides reliable statistical data that can be used to qualify a UHDI PCB assembly process.
In the clip below, one of the main points of discussion was the prominence of tombstones in assembly, and whether or not "tombstoning" may be over-reported. As Zachariah points out, many guides on PCB assembly and DFA likely over-report the problem and thus recommend placing thermal relief pads everywhere in a PCB layout. CAD tools do not help with this perception, as their default design rules may enforce the placement of thermal pads when they are not necessary.
Watch the clip below with Chrys Shea or watch the full Altium OnTrack podcast episode here.
The Takeaway
Tombstoning generally arises when there is a thermal differential across an SMD part with two terminals, where one has a mask defined and the other pad is metal defined. The mask-defined pad has more metal, such as on a copper pore, which could draw away heat and thus create the temperature difference across the SMD part. Chrys agrees that, in reflow soldering assembly, tombstoning is probably over-reported, and a statistical qualification of tombstoning requires the inspection of hundreds of thousands of joints.
Other factors may lead to shifting of components or uneven wetting, such as inconsistent solder paste deposition, asymmetric paste mask openings, and asymmetric pad definition due to solder mask opening size on the mask-defined pad. When in doubt, consult with your PCB assembly house to determine which features on your PCB layout should include thermal relief pads to help prevent tombstoning.
Transcript
Chrys: Well, it turned out on the original board we ended up nicknaming some of the sections. You know, like the O4 BGAs, we called it Tic Tac Toe and these types of things. So, from the outset here, I decided to name the sections. For example, this here we call Tombstone Alley. This is another thing that's a DFM-related item that just does not get communicated to designers. These are capacitors, and capacitors love to tombstone, and it's because of thermal differentials across the device.
What we typically see is a metal-defined pad on one side of the cap and a mask-defined pad on the other. Let me see if I can bring the mask up.
Zach: Just for everyone that's listening, what we have here are some big copper pore regions, and then we have some SMD components lined up around the edge of these copper pore regions. I see here we have some of those SMDs have a thermal attach and some of them don't. I guess this is kind of a side-by-side comparison of the number of defects you would expect to see.
Chrys: Exactly. Again, that gives us the quality metrics to have the conversation with the designers and the owners of the product. Do you want to risk the defects, or can you put the thermal relief in?
Zach: Now, here's one thing I've been dying to ask somebody about just exactly what you're showing here with tombstoning. If you go online and start reading about DFA, you start reading about DFM, you start reading about assembly defects, it's almost obligatory that somebody says tombstoning. They talk about it as if it's the most common thing, and you need thermals everywhere. No matter what you do, if you don't have a thermal, you're going to see tombstoning. How common is it really? I feel like it's talked about as if it's more common than it actually is.
Chrys: I think you're right. When we do tombstoning tests, to get good sample sizes, we literally do hundreds of thousands of joints. Tombstoning is largely due to this type of scenario where you've got an unequal thermal, and they will exactly tombstone. Another thing that we've also been finding lately is if you control the tombstone, you can end up with a nice solder joint on your metal-defined side and a cold joint on your mask-defined side.
Zach: I see. So even though it's not a tombstone, it's still requiring rework. Every time we rework a board, we reduce its reliability.
Chrys: Right. That makes sense. We try to avoid that at all costs. We built these Tombstone Alleys for our O201 caps, our 015 caps, and our 0804 caps. Caps are a lot more likely to tombstone than resistors because they have five-sided terminations. The mechanism by which tombstoning occurs is the melted solder wets to one end, and the surface tension just pulls it right up. Whichever side melts first is the base of the tombstone.
Resistors don't tombstone as much because they only have three sides of metalization, so there's not as much metalization for the solder paste to grab. But now we've got this new phenomenon that we just designed in down here, and we set it up similar to the Tombstone Alley. I've talked to three different assemblers in the last month who are dealing with bottom-terminated Zener diodes. These are O201, OR8, or 015 packages, and they have very small bottom terminations and are very light.
What happens is, again, the side that melts first, the solder pulls it. It doesn't pull it up because it's bottom-terminated; it pulls it laterally, and you will see all of these skew by the same degree in the same direction when they have the uneven thermals.
Zach: Interesting. They would all do the same thing?
Chrys: Yes, and we've done things where we'll switch orientation in the machine, or we'll switch orientation in the reflow oven, or we'll run it at 90 instead of zero, and they all still skew in the same direction. I'm convinced it's due to thermal differentials, and that's why I put these on this revision of the board.