Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Flexible Printed Circuits
Transcript
- Lucas: So you’re building a wearable — say a smart ring. You’ve got the sensor, the battery, the Bluetooth chip. You open your CAD, lay out a rigid PCB, and you realize the board takes up so much space that the ring has to be a chunky twelve millimeters thick. That’s not shipping. Luna: Right. And you can’t just shrink the battery, or the thing dies in four hours. Lucas: Exactly. So what do you do? You start looking at flexible printed circuits — FPCs. Instead of a rigid fiberglass board, you’re using a thin polyimide film with copper traces, sometimes as thin as twelve microns. It bends, it folds, it snakes around the inside of your device. And that lets you reclaim a huge amount of space. Luna: Is that the same as those ribbon cables you see inside laptops? Lucas: Related, but not exactly. Ribbon cables are usually just conductors laminated between layers of polyimide — they’re meant to connect two rigid boards. What we’re talking about today is when the entire circuit is flexible. The traces, the components, everything mounts on the flex substrate. You can actually solder parts directly onto the flex. Luna: How much space does that really save? Give me a number. Lucas: I talked to the founder of a smart ring startup last month. They were trying to fit a heart-rate sensor, an accelerometer, a Nordic nRF52840, and a 30 milliamp-hour battery into a ring form factor. Their first rigid prototype was eleven millimeters thick. They redesigned with a single four-layer flex that wrapped around the inside curve of the band. Final thickness: six point five millimeters. That’s a forty percent reduction. Luna: And the battery life? Lucas: Same battery, same capacity. The flex freed up volume that would have been empty space with a rigid board, so they actually had room for a slightly larger battery in the end. But the headline is the thickness. That’s the killer metric for wearables. Luna: What’s the trade-off? I assume flex circuits aren’t cheaper. Lucas: Not at low volume. A typical four-layer rigid PCB might cost you around two to three dollars per board at a thousand units. A comparable four-layer flex with stiffeners — which you often need where components mount — can run eight to twelve dollars per board at the same quantity. The tooling NRE is also higher. You’re looking at a thousand to three thousand dollars for flex tooling, versus maybe five hundred for a simple rigid board. Luna: So if you’re doing a hundred units for a Kickstarter, flex might not make sense. Lucas: Depends. If your product simply doesn’t work as a rigid — like a smart ring or a hearing aid — you have no choice. But if you’re building a desk gadget, a rigid board is almost always smarter until you hit volume. The crossover point where flex becomes cost-competitive is usually around ten thousand units, but it varies hugely based on layer count and complexity. Luna: Let’s talk about the different types. I’ve heard of static flex and dynamic flex. Lucas: That’s the key distinction. Static flex means you bend it once during assembly — like folding it into a housing — and then it never moves again. That’s the vast majority of consumer products. Dynamic flex means the circuit actually flexes repeatedly during use, like the cable in a laptop hinge or the flex in a robot joint. Dynamic flex requires different design rules: wider traces, larger bend radii, and often a different copper alloy to resist fatigue. Luna: What happens if you ignore that? Use a static flex design in a dynamic application? Lucas: Copper fatigue. The traces crack after a few thousand cycles. I’ve seen a robotics startup that had a flex connecting a sensor on a gripper jaw. They used a standard single-layer flex with twelve-micron copper. After about eight thousand cycles, the resistance on the power trace doubled, and then the sensor started dropping data. They had to redesign with thicker copper and a larger bend radius. Cost them three weeks and a respin. Luna: What’s the general rule for bend radius? Lucas: For static flex, the rule of thumb is ten times the thickness of the flex. So if your stack-up is two hundred microns, your bend radius should be at least two millimeters. For dynamic flex, you want twenty to thirty times the thickness. And you never, ever bend across a component — the solder joint will crack. Luna: And stiffeners — those are the rigid sections you add where the components sit? Lucas: Yes. You can add a polyimide or FR4 stiffener in specific areas so the board behaves like a rigid PCB where the chips mount, but flexes everywhere else. That’s a common hybrid approach called rigid-flex. It’s the best of both worlds, but it’s also the most expensive. A two-layer rigid-flex board with four layers total — two rigid, two flex — can cost three to five times a standard rigid board. Luna: Are there assembly gotchas? I imagine reflow soldering a flex board is different. Lucas: Very different. Flex boards are thin, so they don’t have the same thermal mass as rigid boards. They heat up faster and can warp in the reflow oven. You typically need a dedicated fixture — usually a metal carrier — to hold the flex flat during solder paste printing and reflow. That adds a fixture cost, maybe two to five hundred dollars per design. And you have to be careful about the solder paste volume; too much and you get bridging on fine-pitch parts. Luna: What about connectors? I’ve seen those ZIF connectors on flex cables. Lucas: Zero insertion force connectors are the standard way to terminate a flex circuit to a rigid board. They’re reliable, but you need to design the flex tail with the correct pitch and length. The connector itself is cheap — maybe thirty cents at volume — but the alignment during assembly has to be precise. If the flex is even a half millimeter off, the contacts won’t mate properly. Luna: One thing I’ve heard startups struggle with is impedance control on flex circuits. Is that real? Lucas: It’s a challenge. The dielectric constant of polyimide varies more than FR4 — it can shift by ten percent depending on humidity and manufacturing batch. So if you’re running high-speed signals like USB 3.0 or MIPI, you can’t just copy your rigid board stack-up and expect it to work. You need to work with your flex fabricator to model the impedance and specify controlled dielectric materials. Some fabricators offer pre-characterized stack-ups for common impedances like 50 ohms or 100 ohms differential. Luna: Let’s talk prototyping. How do you prototype a flex circuit without spending thousands on tooling? Lucas: There are a few options. One is to use a rigid PCB fabricated with a very thin board — 0.4 millimeter or less — and simulate the flexibility. It’s not the same, but it proves out the circuit. Another is to use a service like PCBWay or JLCPCB that offers quick-turn flex prototypes. You can get a small batch of two-layer flex boards for around fifty dollars with two-week turnaround. But the real trick I’ve seen some hardware founders use is to prototype the flex shape with a laser cutter. Luna: Tell me more. Lucas: You take a sheet of polyimide film — you can buy it from McMaster-Carr — and a sheet of copper tape. You cut the polyimide to your outline, stick the copper traces down by hand, and solder your components. It’s ugly, it’s not reliable, but it lets you verify the mechanical fit in your enclosure before you spend money on a real flex prototype. One founder I know did three iterations of that in a single afternoon. By the time he sent the design to the fab, he was confident the shape would fit. Luna: That’s scrappy. I like it. Lucas: It’s the kind of thing you do when you’re bootstrapped and you can’t afford a five-hundred-dollar mistake. And honestly, it works for simple circuits — a few passives, an IC, maybe a connector. You won’t get high-speed signal integrity, but you’ll get fit and function. Luna: So what’s the takeaway for a hardware founder listening? When should they seriously consider flex? Lucas: Three scenarios. One: your device has a curved or irregular interior space — like a wearable, a medical implant, or a handheld tool. Two: you need to reduce the number of connectors and cables — a rigid-flex can replace a rigid board plus several cables and connectors, saving assembly time and improving reliability. Three: you have extreme space constraints — every millimeter of thickness matters. In those cases, the added per-unit cost is worth it because the product wouldn’t exist without flex. Luna: And if you’re on the fence, start with a static flex design. Avoid dynamic flex unless you really need it. Lucas: Exactly. And if you’re prototyping, use the laser-cutter trick to check your shape. It’ll save you a spin. Luna: You know, it’s conversations like this that make me wish we had a way to keep these episodes ad-free and accessible to everyone who’s building hardware. If these deep dives help you avoid a costly mistake, that’s exactly why we do the show. Lucas: Yeah, totally. And listener support is what keeps it that way — no sponsors, no ads. If you find value in episodes like this one, you can toss a coffee our way at buy me a coffee dot com slash fexingo. It genuinely helps us keep the lights on and keep the content coming. Luna: Exactly. Now back to flex — one more thing I want to touch on. What about the environmental resilience of flex circuits? Can they handle moisture or vibration better than rigid? Lucas: Generally, yes — as long as you design for it. Flex circuits have lower mass, so they resonate at higher frequencies, which can actually reduce vibration fatigue compared to a heavy rigid board. But polyimide absorbs moisture — about 0.3 percent by weight — so in high-humidity environments, you need a conformal coating. Most flex fabricators offer acrylic or parylene coating as an option. It adds maybe fifty cents per board at volume. Luna: Good to know. Alright, I think we’ve given listeners a solid toolkit for evaluating flex. Thanks, Lucas. Lucas: Yeah, happy to. Next time we’ll look at another overlooked piece of the hardware stack — the humble connector. There’s more to it than you think.