Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Acoustic Microscopy for Hidden Defects
Transcript
- Lucas: So there is a failure mode in hardware that most startups never even think about until their first field return comes back with a cracked housing or a dead sensor, and they have no way to see inside the sealed module without destroying it. Scanning acoustic microscopy — SAM — is the tool that lets you see inside without cutting it open. Luna: I have heard SAM mentioned in the context of semiconductor packaging, but I did not realize it was accessible for smaller hardware teams. How does it actually work? Lucas: It is basically an ultrasound for solid objects. You couple the part to a transducer using water — usually a deionized water bath or a water jet — and then you fire a high-frequency sound pulse, typically in the 15 to 100 megahertz range. The sound travels through the material, and whenever it hits a boundary where the acoustic impedance changes — say, from epoxy to air, or from silicon to a crack — it reflects back. Luna: So the same principle as a medical ultrasound, but looking for delamination, voids, cracks inside potting compound or encapsulation. Lucas: Exactly. And the resolution is good enough to spot features as small as ten microns, depending on the frequency. Higher frequency gives better resolution but less penetration. For a typical potted assembly, 50 megahertz gives you a nice balance — you can see through maybe five to ten millimeters of epoxy and still spot a five-micron gap. Luna: That is impressive. When would a hardware startup actually need this level of inspection? I mean, if you are just building fifty prototypes, you probably do not have the budget for a fifty-thousand-dollar tool. Lucas: You do not need to buy the tool. You can send samples to labs like Sonoscan or Nordson Dage. They charge maybe two hundred to five hundred dollars per part depending on the scan area and number of gates. But the ROI is huge if you catch a delamination before you ship a thousand units that are going to fail in the field. Luna: Give me a concrete example. Lucas: Sure. There was a startup I worked with that made a waterproof smartwatch. They potted the barometric sensor port with a uv cure epoxy. First production run of five hundred units, they field-tested twenty — all passed. Then one user reported altitude readings drifting after swimming in salt water. They sent the unit to a SAM lab, and the C-scan showed a crescent-shaped delamination exactly where the epoxy had not bonded to the stainless steel housing. Luna: And that was invisible under X-ray, right? Because it is a planar gap, not a density change. Lucas: Correct. X-ray would have missed it completely because the density difference between epoxy and air is small. SAM sees the acoustic mismatch. That one scan cost them four hundred dollars and saved them from a recall that would have easily run into the hundreds of thousands. They adjusted their dispensing profile and all subsequent units passed. Luna: So SAM is really for catching adhesion issues and voids in materials that are otherwise opaque to other NDT methods. Lucas: That is the sweet spot. It is also used for detecting cracks in solder joints under BGAs — ball grid arrays — where the crack is under the package and you cannot see it with optical inspection. In fact, the automotive industry really drove SAM adoption after a 2018 Toyota recall where cracked solder joints in engine control units caused intermittent stalling. Those cracks were invisible until they used SAM. Luna: Right. And that is a high-volume, high-reliability environment. But for a startup doing maybe a hundred units a month, when does it make sense to start SAM testing? Lucas: Honestly, if your product will be potted, encapsulated, or bonded — and especially if it goes into a harsh environment like water, vibration, or thermal cycling — you should get at least one SAM scan on your first production batch. It is a one-time validation that your process is actually creating the bond you think it is. And if your contract manufacturer is already using SAM for their own quality control, you can ask for a sample report. Luna: So you are not saying every unit, just a few from the first batch as a process check. Lucas: Exactly. Five to ten parts from the first run, chosen from different cavities or dispensing needles. That gives you a statistical confidence that the process is within control. If you see any delamination, you fix the process before you build the next thousand. Luna: And there are different scan modes too, right? I have heard of A-scan, B-scan, C-scan. Lucas: Yeah, let me break those down quickly. A-scan is a single-point depth measurement — the transducer stays at one location and gives you a waveform showing reflections versus depth. B-scan is a cross-sectional slice — you move the transducer in a line and build up a 2D image of the internal layers. C-scan is the most common for startups — it is a plan-view map at a specific depth gate, so you can see the entire area of a bond line or an interface. Luna: That C-scan image is the one that looks like a heat map of delamination, with red areas showing where the reflection is strong. Lucas: Right. And the lab will usually provide a color overlay that makes defects immediately obvious. One thing to watch out for: the water coupling can be a problem if your device is not sealed yet. You have to make sure the water does not ingress into openings — otherwise you are damaging the part you are trying to inspect. For open assemblies, they use a water jet or a bubble couplant instead of full immersion. Luna: So there are workarounds. Look, if these kinds of practical deep dives are useful for what you are building or running, honestly, if today was worth a coffee to you, that is the link — buy me a coffee dot com slash fexingo. No pressure, just helps keep the show ad-free. Lucas: Yeah, and it really does make a difference. We do not run ads, so listener support is what keeps these episodes coming. Back to SAM — one more point I want to make: the frequency selection matters a lot. For plastic-encapsulated microcontrollers, 30 megahertz is usually enough. For thin film layers or die attach, you might need 100 megahertz or even 230 megahertz, but then you are only penetrating a millimeter or two. Luna: So you need to tell the lab what you are looking for so they choose the right transducer. Lucas: Yes, and they will often recommend a frequency based on the material and the expected defect size. A good lab will also do a reference scan on a known good part so you have a baseline. Another thing: SAM can also be used for thickness measurement of coatings or platings, if you know the speed of sound in the material. Luna: That is a nice bonus. I imagine for a startup doing a conformal coating on a PCB, SAM could verify the coating thickness and uniformity across the board. Lucas: Absolutely. And it catches pinholes in the coating that would otherwise lead to corrosion. So the use cases extend beyond potting and encapsulation. I have even seen it used to check the bond line of adhesive in a consumer electronics device — like the display to frame bond on a tablet. Luna: Is there any standard or specification that startups should reference when setting up a SAM test? Lucas: The main one is IPC/JEDEC j std-035, which covers acoustic microscopy for non-hermetic packaged devices. It defines the scan parameters, the acceptance criteria, and the reporting format. It is written for semiconductors but the principles apply to any encapsulated assembly. If your contract manufacturer is ISO 9001, they probably already have a procedure based on that standard. Luna: So if a startup walks into a SAM lab and says 'I want to test my potted sensor module per j std-035,' the lab will know exactly what to do. Lucas: They will. And it gives you a defensible quality record if you ever have a warranty dispute with a supplier. One more thing: SAM can also be used in failure analysis after a field return. If a unit fails and you want to know why, you can SAM it first before cross-sectioning. That way you see the internal defect in its original state, not damaged by the cutting. Luna: So it is both a preventive tool and a diagnostic tool. That is pretty high value for something that costs a few hundred dollars per sample. Lucas: And the turnaround is usually two to three days. So it is not a bottleneck. If you are shipping a product that costs more than fifty dollars to manufacture, spending four hundred to validate the process for a whole product line is a no-brainer. Luna: Alright, so the takeaway for a hardware startup: if you are potting, encapsulating, or bonding anything critical, get a SAM scan on your first production batch. Use a service lab, reference j std-035, and fix your process before you scale. Lucas: Exactly. And if you are curious about the acoustics behind it — the speed of sound in your specific epoxy versus the speed in air — that contrast is what makes the defect visible. It is a beautiful physics principle applied to real-world manufacturing reliability.