Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Accelerated Life Testing to Predict Failure
Transcript
- Lucas: Luna, I want to talk about something that keeps hardware founders up at night — how do you know your product won't fail a year after someone buys it? You can't wait a year to find out. Luna: Right, especially when you're a startup with limited cash and one shot at a first impression. So you cheat time somehow. Lucas: Exactly. The engineering term is accelerated life testing — ALT. The basic idea is you stress the product — heat, cold, vibration, humidity — in a way that speeds up the failure mechanisms you'd see in normal use. Then you use physics models to project how long the product would last in the field. Luna: And you do this before you ship, ideally before you commit to a full production run. How long does a typical ALT cycle take? Lucas: Depends on what you're accelerating. A common rule — the ten-degree rule — is that for every ten degrees Celsius you raise the temperature, the reaction rate roughly doubles. So if you run a device at sixty degrees Celsius instead of room temperature, one week in the chamber might simulate eight weeks in the field. Luna: That's powerful. But you're not just heating things up. What other stresses do startups use? Lucas: Vibration is huge for anything with moving parts or connectors. Thermal cycling — going from hot to cold repeatedly — that catches solder joint failures and material expansion mismatches. Humidity can reveal corrosion risks. Some startups do combined environments: hot and humid while vibrating. That's closer to real-world wear. Luna: Let's talk about a specific example. I was reading about a smart thermostat startup that did ALT before their first production batch. They found that a capacitor was degrading after the equivalent of eighteen months. They switched the part, cost them maybe two dollars more per unit, but avoided what could have been a recall. Lucas: That's exactly the kind of story that makes the case. And the math behind it comes from chemistry — the Arrhenius equation. It's not as scary as it sounds. It basically models how the rate of a chemical reaction — like oxidation of a connector, or breakdown of an electrolyte — increases with temperature. Luna: So if you know the activation energy of the failure mode, you can calculate the acceleration factor. But most startups don't have a chemist on staff. How do they get started? Lucas: They start with the ten-degree rule as a rule of thumb. It's not perfect, but it's conservative enough for many failure modes. Then they run a small pilot — say ten units in a thermal chamber for two weeks at seventy degrees Celsius. If nothing fails, that's a good sign. If something fails, they dig into why. Luna: And if those ten units are all you have? That's a common startup constraint — you don't have a hundred samples to sacrifice. Lucas: That's the real tension. ALT is statistically more meaningful with larger sample sizes. But for a startup, even five units can give you directional data. You're not trying to calculate a precise mean time between failures to six sigma confidence. You're trying to find the 'oh no' failures before they reach customers. Luna: So it's more about finding failure modes than proving reliability. That's a different mindset from what a big company like automotive supplier would do. Lucas: Exactly. A Tier One auto supplier might run a full design validation with hundreds of samples and multiple stress profiles. A hardware startup needs to find the one or two critical failure modes that would kill their business. ALT is a filter, not a guarantee. Luna: And if these conversations help you think about what's worth testing in your own product, that's the point. We keep this show ad-free and independent because listeners like you support it. If you find this useful, you can throw a coffee our way at buy me a coffee dot com slash fexingo. Lucas: Yeah, that kind of support lets us keep digging into the practical side of hardware — the stuff that's not in the marketing brochure. So back to ALT — one thing I want to emphasise is that you don't need a million-dollar chamber to start. A cheap consumer oven with a temperature controller can get you started for under a thousand dollars. Luna: Really? A toaster oven? Lucas: Basically. You can buy a benchtop thermal chamber from a company like Thermotron or even a used one on eBay for a few thousand. Or you can build a simple one with a temperature controller, a heating element, and a fan. The key is that you need to be able to hold a steady temperature and have some way to monitor the devices under test. Luna: And you need to think about safety — don't burn down your garage. Lucas: Absolutely. Use a thermal fuse, don't leave it unattended. But the point is, the barrier to entry is lower than most founders think. The real investment is time — you need to run the test for long enough to see failures, and you need to be disciplined about logging data. Luna: Let's talk about what kind of data you log. Is it just pass/fail, or are you measuring something like current draw or signal quality over time? Lucas: The best tests measure performance parameters continuously. For example, if you're testing a drone's battery management system, you could log voltage and temperature during the thermal cycle. If the voltage drops more than a threshold, that's an early warning. Some startups use a simple Arduino with sensors inside the chamber to record data. Luna: So you're not just waiting for a catastrophic failure. You're looking for drift. Lucas: Right. Catastrophic failures are obvious — the unit stops working, or catches fire. But more often, you see degradation: a voltage that slowly creeps down, a signal that gets noisier, a motor that draws more current. Those are the signs that a component is wearing out faster than expected. Luna: And catching that early lets you choose a different part or adjust the design before you're in production. I know a startup that makes outdoor security cameras. They ran a thermal cycling test and found that the housing seal started leaking after fifty cycles. That would have caused moisture ingress in the field after about two years. They switched to a different gasket material and the problem went away. Lucas: That's a perfect ALT story. And the cost of that change — maybe ten cents per unit — versus the cost of replacing cameras under warranty — huge. So the return on investment for ALT is often massive, even if you only catch one issue. Luna: What about the acceleration models for non-thermal stresses? Vibration, for example. Lucas: Vibration testing uses a different model. The standard approach is the fatigue damage spectrum — you apply a random vibration profile that mimics what the product would see in transport or use, and you accelerate by increasing the amplitude. There's a rule of thumb that doubling the vibration level roughly quadruples the damage rate. But that's more complex to set up because you need a shaker table and a controller. Luna: So temperature is the easiest place to start for most hardware startups. Lucas: Yes, and it's also the most common cause of failure. Over sixty percent of field failures in electronics are temperature-related. So if you do nothing else, a simple hot soak and thermal cycling test will catch a lot. Luna: One thing I wonder about — how do you know you're not overstressing the product and causing a failure mode that would never happen in real life? That's a risk with ALT, right? Lucas: It's a real concern. If you push the temperature too high, you might melt the plastic or activate a failure mechanism that only occurs at extreme temperatures. The trick is to stay within the material limits — for most electronics, that means below the glass transition temperature of the circuit board, which is typically around one hundred thirty to one hundred forty degrees Celsius. You also want to avoid temperature shock — rapid changes — unless that's part of your test plan. Luna: So you need to understand your materials' limits before you start. That's another reason to talk to your component suppliers early. Lucas: Exactly. And the datasheet often gives you the operating temperature range and the storage temperature range. Use those as guardrails. For a typical consumer device, testing at sixty degrees Celsius with twenty percent relative humidity is a good starting point. Luna: Let's talk about statistical confidence briefly. If you only test ten units and none fail, what does that tell you? Lucas: With ten units and zero failures, you can say with about ninety percent confidence that the failure rate is less than about twenty percent. That's not great — it means one in five units could still fail. To get to a failure rate of less than one percent with high confidence, you'd need hundreds of units. But for a startup, the goal is different: you want to know if the product is catastrophically bad, not whether it's perfect. Luna: So you're using ALT to filter out the obvious problems, not to certify the product for a decade of life. Lucas: Exactly. And that's fine for an early stage. As you scale, you can invest in more rigorous testing. But the worst thing you can do is ship a product that hasn't seen any accelerated stress at all. That's how you get a recall on your first batch. Luna: Do you have a recommendation for a minimum ALT plan for a startup with a new hardware product? Lucas: Sure. Take five to ten units. Run them at sixty degrees Celsius for two weeks while monitoring key parameters like current draw and output signal. Then do a thermal cycling test: fifty cycles from zero to fifty degrees Celsius with a ramp rate of one degree per minute. Then do a vibration test — even just taping the units to a paint shaker for an hour. That'll catch the biggest issues. Luna: And if you find nothing, you have some confidence. But you're not done — you should also test for things like drop, ESD, and moisture. But those are different tests. Lucas: Right. ALT is specifically about long-term wear. The other tests are about survive-ability. But combined, they give you a much better picture before you commit to a thousand units. Luna: I think the key takeaway is that ALT doesn't have to be expensive or academic. It's a practical tool that any hardware team can start using with basic equipment and a little physics. Lucas: And the earlier you do it, the cheaper it is to fix what you find. Changing a component at the prototype stage costs a few dollars. Changing it after you've ordered ten thousand units costs months of delay and thousands in scrap. Luna: So for anyone building hardware today, consider this your sign to start a test plan. Even a simple one. Your future self — and your customers — will thank you.