Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Jigs and Fixtures for Repeatability
Transcript
- Lucas: You ever watch a hardware startup try to assemble fifty units of the same product and end up with fifteen slightly different versions of the same device? Luna: I've seen it happen. One board sits a millimeter off, a connector is torqued differently, and suddenly the firmware team is chasing phantom bugs. Lucas: Exactly. And the fix isn't always more expensive robots or vision systems. Often it's something far simpler: jigs and fixtures. We're talking about physical guides, clamps, and alignment tools that force every assembly step to happen the same way every time. Luna: So this is the episode where we get into the unsung heroes of the workbench. I love it. Lucas: Let's start with the distinction. A jig guides the tool — think of a drill jig that ensures every hole is exactly 20 millimeters from the edge. A fixture holds the workpiece in a precise position — like a cradle that locks a PCB at the perfect angle for soldering. Most early-stage startups use both, often without realising there's a name for it. Luna: And they often start with tape and cardboard shims. Which is fine for prototypes. But when you hit volume — say, a few hundred units a month — that improvisation becomes a quality bottleneck. Lucas: Right. So I want to anchor this in a concrete case. There's a startup I've been following — let's call them SensorGrid. They make a multi-gas sensor for indoor air quality monitoring. Two founders, contract manufacturing in Shenzhen, but they do final assembly and calibration in-house. About 500 units a month. They were seeing a 12 percent rework rate because the optical sensor module was shifting during the alignment step. Luna: Twelve percent is brutal. That's sixty units a month you have to disassemble and reassemble. Lucas: Worse — each rework took about 25 minutes, plus the risk of damaging the flex cable. So they designed a custom alignment fixture — cnc machined aluminum, with a spring-loaded plunger that indexes the sensor module into the exact same position every cycle. Total cost was about $1,200 for the fixture, including the prototype iteration. Luna: And the result? Lucas: Rework dropped from 12 percent to under 1 percent. Assembly time for that step went from 90 seconds to 55 seconds. The fixture paid for itself in just over two months — call it 500 units at maybe $2.50 of saved labor and scrap per unit. That's a classic ROI story. Luna: But I imagine not every fixture story ends that neatly. I've seen startups spend $5,000 on a complex fixture for a part that changed two weeks later. Lucas: That's the trap. You have to match the fixture's sophistication to the design's maturity. If your PCB layout still has a chance of moving the connector, you don't cut a $3,000 hard-tooled fixture. You use a modular system — something like 80-20 aluminum extrusion frames with adjustable stops. You can reconfigure it as the design evolves. Luna: So when do you lock in custom metal fixtures? Lucas: The rule of thumb I hear from manufacturing engineers is: once your BOM and mechanical drawings have been revision-frozen for at least two production runs. If you're still changing hole locations or adding standoffs, stay modular. The moment you're confident the geometry is stable, then you invest in hard tooling. SensorGrid was smart — they'd gone through three prototype iterations before they cut that aluminum fixture. Luna: Another angle I want to bring up: 3D-printed fixtures. They're cheap and fast. But they wear out. Lucas: They do. A PETG fixture might last a few hundred cycles before the locating pins wear and start introducing slop. PLA even less. For low-volume — say, under 200 units total — 3D-printed fixtures are often sufficient. But if you're running 500 units a month, you want aluminum or at least machined Delrin. The cost difference is maybe a few hundred dollars, and the durability gain is orders of magnitude. Luna: And you can get those machined parts from services like Xometry or Protolabs in a few days. Hardly a barrier anymore. Lucas: Exactly. I think the biggest mistake I see is not jigging the inspection step. A lot of startups build fixtures for assembly but then inspect with calipers and eyeballs. That's where repeatability breaks down. You want a go/no-go gauge — a fixture that the assembled product either fits into or doesn't. That catches dimensional drift before it becomes a field failure. Luna: So a pass-through fixture for final quality check. What does that look like in practice? Lucas: Simplest version: a block of aluminum with cutouts that match every critical external dimension. If the product slides in and seats flush against all reference surfaces, it passes. If it doesn't seat, you know something shifted. One startup I visited had a fixture that checked the height of four rubber feet simultaneously — took two seconds per unit and eliminated a whole category of wobble complaints. Luna: That's elegant. Cheap, fast, and catches the thing customers actually feel. Lucas: Right. And that's a theme: jigs and fixtures are the most cost-effective quality system in hardware manufacturing. They don't need software updates. They don't need calibration certificates. They just need to be rigid, accurate, and well-maintained. Luna: Maintenance is a good point. A fixture that gets knocked off the bench or dropped can lose its accuracy. Do startups overlook that? Lucas: All the time. I've seen a fixture with a bent locating pin that had been in use for six months. Nobody noticed because the assembly still fit — it just fit consistently wrong. So part of the discipline is periodically checking fixtures against a master reference. For most startups, that's a monthly 10-minute check with a pair of calipers. Luna: OK, so let's say a listener is building a hardware product right now, maybe at the 100-unit pilot stage. What's your one piece of advice on fixtures? Lucas: Start with the step that's causing the most variation or the most rework. Don't try to jig everything at once. Pick the one assembly operation where the operator has to guess alignment, or where the torque is critical. Build one simple fixture for that step. Measure the improvement. Then do the next one. Iterate your fixtures the same way you iterate your PCB. Luna: I like that. It's the same build-measure-learn cycle, but applied to the tooling instead of the product. Lucas: If these conversations are useful for what you're building or running, listener support is what keeps this show ad-free and independent. We don't run sponsorships, and that's because listeners who find value here choose to chip in. You can do that at buy me a coffee dot com slash fexingo. Just a small way to say this content matters. Luna: Yeah, and it genuinely helps us keep the focus on practical takeaways — no filler, no ads. Lucas: Back to fixtures — one more thing I think is worth touching on. The rise of open-source fixture designs. Communities like the Open Hardware Repository and certain GitHub repositories now share CAD files for common fixture types — alignment blocks, press-fit jigs, even PCB panel break-out fixtures. Luna: So you can download and modify rather than design from scratch? Lucas: Exactly. For a simple push-pin alignment jig, you might find a design that's 90 percent right. Adjust a few dimensions, export to STEP, send to your local machine shop. That can turn a two-week design cycle into a two-hour one. Luna: That's a huge advantage for a two-person team where neither person is a mechanical engineer. Lucas: Yeah. And I think that's the final takeaway: fixtures are not mysterious. They're just physical constraints that remove human guesswork. The best hardware startups treat them as a first-class part of their production system, not an afterthought tacked on when quality problems appear. Luna: Alright, so next time you're assembling ten boards and three of them are slightly different, consider that the problem isn't your soldering — it's your lack of a jig. Lucas: Exactly. Thanks for listening, and we'll be back next week with another episode.