Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Solder Joint X-Ray Inspection
Transcript
- Lucas: You design a board, you send it to assembly, you get back ten prototypes — and they all power up fine. Then you ship fifty units to beta testers, and three of them die after two weeks. What do you do? Luna: That's the nightmare scenario. And the worst part is, the root cause was probably there all along, just invisible. Lucas: Exactly. That's where automated X-ray inspection — AXI — becomes the difference between catching a defect at the prototype stage or catching it in the field. Today we're going to look at how hardware startups are using X-ray to inspect solder joints, specifically for components like ball grid arrays where you literally cannot see the connections. Luna: And we'll talk about a specific startup — a Bay Area electric skateboard company — that used AXI to cut their field failure rate from eight percent to under one percent. Lucas: Right. So let's start with the problem. When you have a BGA — a ball grid array chip — all the solder balls are underneath the package. You cannot see them. You can't inspect them visually, even with a microscope. So how do you know they soldered correctly? Luna: You can do electrical testing — boundary scan, flying probe — but that only tells you if a connection is open or shorted. It won't catch a cold joint that passes continuity today but cracks after thermal cycling. Lucas: That's the key insight. A cold solder joint might have enough mechanical contact to pass a resistance check, but it's weak. After a few dozen heat-cool cycles, it fractures. X-ray can see that void or crack before it becomes a field failure. Luna: So how does AXI actually work? Give us the thirty-second physics lesson. Lucas: An X-ray source fires photons through the board. A detector on the other side captures the shadow. Dense materials — like solder, which is mostly tin and lead — absorb more X-rays, so they appear darker in the image. Less dense materials, like the PCB substrate or air gaps, let more through. Luna: So a void in a solder ball shows up as a lighter spot inside a dark circle. That's the classic defect signature. Lucas: Exactly. And with 2D X-ray, you get a single projection — like a medical X-ray. It's fast, but you lose depth information. A ball with a void on the top might look the same as a ball with a void on the bottom. For many inspections, that's good enough. But for high-reliability applications, you want 3D. Luna: Which is laminography, or computed tomography, right? The machine takes multiple images from different angles and reconstructs a 3D model. Lucas: Yes. 3D AXI systems rotate the board or the source-detector pair to get slice images at different heights. You can look at the top of a ball, the middle, the bottom. That lets you measure void volume, detect head-in-pillow defects, check for proper wetting. It's more data, but it's slower and more expensive. Luna: And that cost is the barrier for most startups. Entry-level 2D systems start around 30 to 40 thousand dollars. A good 3D system is 80 to 150 thousand. Lucas: Right. So the question is: do you need to own one, or can you outsource? Let's talk about that skateboard company. They were doing about 200 units a month, mostly hand-assembled in a small shop in San Mateo. They had a 2D AXI unit on loan from a distributor, but they only used it when they saw a problem. Luna: Which is like only going to the doctor once you're bleeding. The real value is catching things early, not diagnosing failures. Lucas: Exactly. They had a BGA for the motor controller that kept failing after about 50 miles. They'd replace the chip and it would work again. But they couldn't figure out why. So they finally sent a batch to a contract manufacturer that had 3D AXI. The X-ray showed that the solder balls on one corner of the package had incomplete wetting — they were barely attached. Luna: And the root cause? Was it the reflow profile? Lucas: Partly. The board had a thick copper plane on that corner that was sinking heat. The reflow oven couldn't get the solder hot enough to fully melt and wet the pad. The profile was fine for the rest of the board, but that one BGA was marginal. Luna: So they fixed the profile, added a longer soak zone, and then used the 2D AXI to inspect every board. Defect rate dropped from eight percent to below one percent within two weeks. Lucas: And the cost? They bought a used 2D AXI system for 18 thousand dollars. That's less than the cost of replacing 50 failed motor controllers, not to mention the reputation damage. Luna: That's a clear ROI story. But not every startup has that volume. If you're doing 20 prototypes a month, does it make sense to invest in an X-ray machine? Lucas: Probably not. You can send boards to a test lab. A typical price is 50 to 100 dollars per board for 2D X-ray, maybe 200 for 3D. For a few hundred dollars a month, you can inspect your prototypes. But the key is to do it systematically, not just when something fails. Luna: So you're saying every hardware startup should have an X-ray inspection step in their NPI process, even if they outsource it. Lucas: Absolutely. Because the alternative is that you ship a bad board, a customer has a bad experience, and you may never know why. X-ray gives you the visual evidence to improve your process. Luna: And that evidence can feed back into design, too. You see a void pattern, maybe you change the pad geometry or the solder paste volume. It's not just inspection — it's process control. Lucas: That's the higher-level point. AXI data can inform stencil design, reflow profile, even component placement. One startup I know used X-ray to discover that their pick and place machine was placing BGAs with a slight rotational offset. That caused tombstoning on the corner balls. They fixed the placement program and the defect rate went to zero. Luna: Alright, let's talk about practical implementation. If a startup decides to outsource, what should they look for in a test lab? Lucas: First, make sure the lab has experience with your type of board. High-density boards with 0.4 millimeter pitch BGAs require higher resolution. Ask what the minimum detectable void size is. Also ask about throughput — how many images per board? A good lab will take multiple images at different angles to catch all defects. Luna: And what about data? Do you get a report with images? Lucas: You should get a report with annotated images showing all defects. Some labs will also give you a pass-fail summary based on IPC-7095 criteria, which define acceptable void sizes for BGAs. That's the industry standard. Luna: IPC-7095. Got it. And if a startup decides to buy their own system, what are the gotchas? Lucas: The biggest one is operator training. You can't just put a board on the stage and press a button. You need someone who can interpret the images, distinguish real defects from noise, and set up inspection programs. A bad operator will miss defects or generate too many false positives. Luna: False positives lead to wasted time re inspecting good boards. Lucas: Right. So you need to balance sensitivity and specificity. Modern systems have automated defect recognition algorithms, but they still need tuning. Plan for at least a week of training and a month of ramp-up before you trust the system in production. Luna: Let's talk about the other side: what X-ray cannot catch. It won't find opens in plated through-holes, for example, unless the hole is completely empty. Lucas: That's true. X-ray is excellent for solder joints, but it's not a substitute for electrical test. You still need flying probe or ICT for connectivity. And X-ray won't catch component polarity or wrong component values. Luna: So it's one tool in a suite. But for BGAs, QFNs, and other hidden joints, it's the only non-destructive tool. Lucas: Exactly. And as hardware startups pack more functionality into smaller boards, the number of hidden joints is only increasing. A modern IoT device might have three or four BGAs. Without X-ray, you're flying blind. Luna: Before we go further, I want to acknowledge something. If today's episode was useful to you — if you're building hardware and thinking about quality — the reason these conversations stay free and ad-free is listener support. You can help keep it that way at buy me a coffee dot com slash fexingo. Lucas: Yeah, it's a small way to ensure we can keep diving into these niche but critical topics. No pressure, just if you found value. Luna: Exactly. So back to the skateboard startup. They eventually scaled to 2000 units a month and bought a 3D system. Their process engineer told me the X-ray data helped them optimize the reflow profile for every new board design in one iteration instead of five. Lucas: That's the kind of efficiency gain that pays for the machine in months. And it's not just about cost — it's about confidence. When you ship a product, you want to be sure it's going to work. Luna: So what's the one takeaway you want listeners to remember? Lucas: If your design includes components with hidden solder joints, you need an X-ray inspection step in your NPI process. Whether you buy a system or outsource, do it before you ship to customers. The data will make your product better and your process smarter. Luna: And if you're at the prototype stage, start with a 2D system or a lab. You don't need 3D until you have high-density boards or reliability requirements. Lucas: Good advice. For Hardware Startups with Fexingo, I'm Lucas. Luna: And I'm Luna. See you next time.