John McMillan
Welcome to the Siemens 3D IC podcast series, where we dive into the exciting world of semiconductor chiplet integration and advanced technology platforms using 2.5 and 3D techniques, brought to you by the Siemens thought leadership team. I’m your host, John McMillan. In this podcast series, I talk with industry leaders and subject matter experts to discuss the latest on 3D IC chiplet ecosystems, industry trends and roadmaps. In today’s podcast,
We’re going to talk about creating test vehicles and daisy chain design tests for semiconductor designs. I’m excited to welcome my guest, Kendall Hiles, senior 3D IC product specialist at Siemens EDA. I’m also excited that we’re able to meet face-to-face as the first video recording here in this wonderful studio. So thanks for being here.
Kendall Hiles
Well, I’ve been in EDA for 40 years, and I’ve been married for 40 years. So, for 40 years, I’ve been in EDA, you know, starting off working for a defense contractor and stuff like that. And then, we used the Mentor Graphics tools at the time, and I ended up being an applications engineer for Mentor Graphics for about 27 years. And the amazing thing is, I’ve worked for the same person for 40 years.
27 years, Mike Walsh. So, yeah, yeah. Mike’s been in it. We got into, um, basically the 3D IC stuff probably about 20 years ago. So, you know, working with the customer, and, uh, they needed solutions to some things they were doing. And so, I’ve been working with the, the, you know, the IC packaging side of it for, for going on 20 years.
That’s great. So today, we’re going to dive a little bit into the test vehicles and the daisy chain design test. What is a test vehicle for starters? So, a test vehicle, unlike a product, where a product you’re making to sell or to do something, what a test vehicle is, is you’re making a test bed to test something in the manufacturing process. Be it the bumps and balls connections, maybe you want new technology like running organic at 9-micron widths. Something that typically you wouldn’t do on an organic material, like nobody signs off on it, but you’ve got something that you need to go down and you need to push technology.
Well, before you spend millions of dollars on something and make a product, you want to make sure that the product you make can be manufactured. So that’s where the test vehicles come in. And it can be a number of things; any type of new technology you want to make a test vehicle to test out for the reliability and manufacturability of that product.
John McMillan
Gotcha. So, at what points, when is a test vehicle needed?
Kendall Hiles
Anytime something changes inside the, you know, inside the manufacturing space. So, we like you have a customer, I had a customer, they make these fancy glasses that have like little, you know, cameras in the temples. Well, in that 3D IC that they have in the temple, they have an embedded chip inside the laminate itself. So, you’ve got this embedded chip in here.
So before you go make millions of those, let’s go make sure that we can make the cavity, put the chip in the cavity. We have connectivity and everything else, so they’re testing out the manufacturing process before you commit to the full product and say, Hey, we’re going to make these. Because once you’re in production and everybody’s expecting you to have a product, if you have a delay, that hurts you more than coming to market a little bit late with something not having the availability to ship.
John McMillan
You sound like you are describing maybe a digital twin of something to see it functionally digitally before you have it manufactured. Is that kind of the same thing?
Kendall Hiles
Well, so let’s talk about what a test vehicle actually does, you know, so we can understand it. So if you have a, let’s just say it’s a flip chip device. You’re just flipping a chip and putting it on a die. You might have, you know, the corner bumps as they get heated and everything else, you’re going to expect them to crack. Okay. Your manufacturer, whoever’s manufacturing it says they will say, you know, we can guarantee this up to like an 80,000 pin chip. You just got a hundred thousand pin chip that you’re putting out. How many of those corner bumps are going to crack? So the way we do it is on the substrate itself, like from pin one to pin two,
You have a connection on the substrate from pin two to pin three, the connections on the die three to four substrates, four to five die. So, it’s literally a chain going through each one of these bumps and balls. And at the end of the chain, you bring that out to the BGA, and that goes out to your tester. So you can chain up everything inside of here. And typically, you wouldn’t do one chain throughout the whole thing.
Because if something broke, then the whole thing’s like, it’s dead. Right?. So they’ll have different chains, and then you can pull taps off of it. Like at the middle of each row, let’s pull a tap off to a BGA ball. So we can, we can do a resistance check on that. So now if this whole big chain breaks, you do a resistance on those taps and see which part of the chain breaks. And what that gets you to is if you want to go take it into the lab and slice and dice it. To see what actually broke. It’s pinpointing you on the device where to do it. that’s what the daisy chain part of it is in the test vehicle. So the test vehicle itself is testing the whole process. The daisy chain is mainly checking the connectivity on the bumps and balls inside your subsystem.
John McMillan
Gotcha. Well, it sounds pretty complicated. So, who is responsible for creating these test vehicles?
Kendall Hiles
It’s twofold. So if you’re the manufacturer and you’re going to say, “I’m going to guarantee this,” the manufacturer would probably make the test vehicles. If they’re having it in a process and they’re saying, “We’re guaranteeing that we can run the 12 and 12 spacing with an 80,000 pin die or something like that,” they’re probably going to make their own test vehicles to test out that process.
But as you and I know, technology pushes everything, right? So we’re making a product. We’ve got a 120,000-pin die. The manufacturer might not guarantee that. Oh, that’s 120. I’m not so sure about that, right? So, then it might be left to the actual customer to go off and make a test vehicle. And that die—it’s just a dummy die, right? The die just has an RDL layer with the connections across it.
But I see both the OSATs making them, customers making them, and foundries making them, just to test out all those processes. The actual person responsible for doing it, though, is just your typical layout engineer. Whoever’s probably laying out the product, they’re the ones who are actually going to be doing the test vehicle.
John McMillan
Gotcha. So.
John McMillan
Next question kind of segues nicely like how do test vehicles factor into that relationship between the OSATs and the OEMs?
Kendall Hiles
Yes, so I see both. The manufacturers themselves, you know, when they’re going through it, and I don’t know whether I can mention names, but one of our large customers probably makes 10 to 12 test vehicles for every product they make.
So they’re going off and making their own test vehicles to make sure that they’ve worked out all the bugs in manufacturing across this, whether they’re testing the bumps and balls and things like that. And then the customers themselves, like I mentioned before, are pushing the technology if they have embedded parts or any kind of chiplet, so you have this die-to-die interface.
Well, a lot of times we don’t route that die-to-die interface on the substrate directly. We use a silicon bridge. That silicon bridge is flipped over with the pins up. And then as your ASIC and your HBM come down, they, they, basically index off that bridge. So now you have this, you know, in an organic material, you have a silicon bridge talking die to die. I don’t know how they do that.
You know, got this thing sticking up and you’ve got these two huge parts coming down and they’ve got to meet this bridge perfectly right to make this connection going across. You’re going to make sure you have a test vehicle in there to make sure that the manufacturing process for that is working. Well, if you had a live chip, how would you do it? You’d have to run it at speed, run the memory and everything else. Whereas if you had a test vehicle, that’s just checking the connectivity through the bridge back across. Now, really quickly, you can go off and say, you know, did this bridge make it or not? You know, are we doing it correctly? And it’s as manufacturing changes, you know, like they found a better way of doing something. They might go back to that test vehicle and say, Hey, you know, we’ve already got this dummy die with the pins on it already connected. Let’s go make one of those and make sure that this is still working out correctly for them. Gotcha. Yeah.
John McMillan
That’s kind of familiar. Having spent decades, as you know, on the PCB side myself, why don’t we see test vehicles in PCB design?
Kendall Hiles
Well, we actually do. I mean, they do test vehicles for PCBs. But PCBs, you know, have become more of a commodity. Right? And they’re not pushing technology as much as they like the 3D IC spaces. I mean, 3D IC space right now is the Wild West, right? It’s, you know, each company is doing their own thing there. But for PCBs, they absolutely do. I remember I was working for General Electric, and we did high power for locomotives. And we had these inductors that we were putting on the board, and I was the librarian, right?
So I made the part for, for, I made the holes, the pads for the part to put the inductor back in. I made them too small; we were getting cold solder joints. Where’d we find that? production? Is that the best place to find stuff? No. You know, so in theory, when you add a new part like that, you probably need to have some kind of a test vehicle to make sure that what you’re making is actually, you know, going to work, especially for high production. So companies like Motorola and Ericsson and Nokia and things have made, you know, for their phones, when they did wacky things, they would make a test vehicle to test things out, to make sure that what they were doing is what they actually wanted, you know, because they were pushing technology for the space and everything at the time and in the phones shrinking everything. It was similar to what we’re doing now. In the IC space, right? You know, we were, you had this brick phone, and now it became this flip phone, right? But now it’s the same thing. We have all these dies and HBMs or memories that are separate. We’re pulling them all together on a substrate. And you know, the technology for that is still, I’m not saying it’s brand new. I’m just saying companies are coming up with more ideas than what can be manufactured right now. Yeah.
John McMillan
So, do test vehicles contain more than simple daisy chain structures?
Kendall Hiles
Yes. The basic function is that chain, which is resistivity. It would be like they’d have a four-wire connection on them. So, you have a voltage and a current to get the resistance, right? So you have a four-wire connection at each end. You can measure the resistance across the chain.
Well, the problem with that is like, you’re, you know, like I said, when you’re running something, we’re expecting the corner balls to probably break. Why did they break? Because of the temperature coefficient difference between the die and the substrate. So there’s other things they add onto the substrate. It’s it, they call them heaters, which to me and you, it would look like tuning patterns, right? But they’re going to put a voltage across that. that pattern heats up.
So it heats up the whole substrate. now they’re, they’re, you know, they’re trying to put something in there to mimic, you know, this under load and things like that. So you’ve got these heaters, though, add them into the substrate itself to try to heat it up. And the other thing, the way I actually put in like, these comb structures where you know, they don’t touch, but it comes kind of like this, where the combs come in.
They’re looking for, like, capacitive discharge because, you know, they didn’t clean off all the flux or something off the device when they did their mounting process and things like that. What’s left residual on the design. And we can test that across this, this capacitive pattern. The other thing I’ve seen them do is stack vias. Did you do HDI design? Yeah.
So when you have stacked vias, you usually have rules on how many vias you’re allowed to stack, right? Because each layer you put it on has its own registration. So what I’ve seen them do is like on this chain, let’s stack one to five. So you’re actually not necessarily going through the die at this point, but you’re just doing like a substrate chain. And this chain, the one to five, is completely stacked normally, right? They’re coincident with each other.
The next chain, all the vias in the stack are five microns off. The next chain, all the vias in the stack are 10 microns off. The next chain, they’re all 15 microns off. So what they’re trying to do is basically implement something that’s going to tell them how the misregistration is going to be handled. How far can these vias be off from each other before they actually start losing connectivity or something like that? And there are other chains inside, you know, other devices, things like that, that they use in the design, because it’s all the test, something new that you’re doing, some kind of new technology or whatever. if I can stack vias, that’s huge. I mean, that saves so much routing space rather than having a dog bone on one of those layers to move over and go down. Right. So I’d love to be able to stack everything, but you can’t in the same way. A lot of the substrates have a core which is much thicker than the buildup layers on either side.
Can I stack a via on top of the core via? Now, the core via is getting filled. How much reliability do I have if I stack the via above and below the core? And a lot of times you’re putting multiple vias because the core via is pretty big, right? So you might have four vias at the top of that. But you’re basically making that chain because
Somebody came to you in the last design and said, hey, we could only stack three layers. You just doubled the size of my design. I have the same number of layers with twice as many signals. How are we going to fix this? Let’s stack some more vias. Well, if you want to go do that, you’d better go prove that you can actually make it when you go do it. So again, it comes back to let’s make a test vehicle to test it.
John McMillan
This is one of the deep dives into the technical characteristics of what we are talking about with 3DIC. It’s been great to get these visualizations and talk about those details of daisy chain and stuff like that. We don’t really get to hear as much about that in the podcast. So it sounds like, taking some notes here, like testing is essential for reliability and quality assurance, risk mitigation is something you care a lot about, and performance verification?
Kendall Hiles
Yep. Cost efficiency is a big deal, as is ensuring regulatory compliance. Is that any part of this? On the regulatory side, automotive has real stringent things, like creepage, and you know, I dealt with a lot of that on the locomotive because, you know, what are locomotives? They were the first electric cars.
Right. It’s a diesel engine driving a, you know, an electric vehicle or whatever. So we were dealing with battling battery voltages. And when you get DC voltages that high, creepages come into it. Yeah. So if you’re going to go do something, you’re going to put a B-HAST test on it, right? You’re going to do basically the stress test that you’re going to put it in the chamber and change the humidity, the heat, the shake and the vibration on it and everything else.
A test vehicle is an excellent way to see whether something passes or not without having a device that’s running. Right? How are you going to tell when something’s breaking? If you have like the actual device in there, when you’ve got a daisy chain and you’re just measuring a few resistance points on it, but everything’s connected in the chain. Boom, you’ve got it. You know, you can even measure resistance between one net and another net. You know, did we get creepage between these? Yes, you know, regulatory kicks in, you know, especially the automotive market. They’re stringent.
John McMillan
Yeah, this has been a great conversation. Before we close this podcast, do you have any final comments?
Kendall Hiles
As you know, it is still an emerging technology. I see customers doing daisy chains by hand, you know, because when you have a chain and a net that’s going between us, you can’t call that whole net thread. If you did, what’s the layout you want to do?
It wants to connect all those pins together. Right. So, in a chain, this net is going to be like Fred one, the next little piece is Fred two. The next little piece is Fred three. I see people doing that by hand, you know, making their own netlist and things for that, or using a spreadsheet to try to drive it. You know, there are better tools than that today. You know, if you’re doing it by hand, don’t. Stop. Ask for help. Somebody can help you. Awesome. That’s great.
John McMillan
Thank you for joining me today. And thanks for sharing your knowledge and insights about the necessity and importance of creating test vehicles and daisy chain design tests. That’s interesting stuff. That’s it for today’s episode of 3D IC Podcast to all of our listeners and viewers. Thanks for joining us today. And be sure to check out the show notes to learn more about today’s topic. And also, be sure to follow this podcast on YouTube or your favorite streaming service so you don’t miss the next episode of the 3D IC podcast. Thanks. Thank you.
