Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Wire Bonding for Chip-on-Board Assembly
Transcript
- Lucas: When you picture a circuit board, you probably imagine a green slab with little black packages — those are the chips, each one already sealed in plastic or ceramic. But there's another way to assemble electronics, one that's more common in things like hearing aids, smartwatches, and disposable medical sensors. Luna: You're talking about taking the bare silicon die and attaching it directly to the board? Lucas: Exactly. It's called chip-on-board, or COB, and the key connection method is wire bonding — using extremely fine wires, typically aluminum or gold, to connect the bond pads on the die to the traces on the PCB. We're talking wire diameters around 20 to 50 micrometers — thinner than a human hair. Luna: That's tiny. So how does a startup decide whether to go this route instead of just using a packaged chip? Lucas: The biggest driver is size. A packaged IC might be ten or twenty times the footprint of the bare die itself. If you're building a product where every millimeter matters — like an implantable sensor or a compact gesture-recognition module — removing that bulky package can shrink your board by half or more. Lucas: There's also a thermal advantage. With a bare die, the heat can flow directly into the board or a heatsink, rather than being trapped inside a plastic package. And for high-frequency signals, you avoid the parasitic inductance of the package leads, which can be a big deal in RF applications. Luna: Okay, but what's the process? How do you actually take a fragile little square of silicon and reliably attach those hairs-thin wires? Lucas: There are two main techniques. The first is ball bonding, where a gold wire is threaded through a capillary tool. A small electrical discharge melts the tip into a ball, and then the tool presses that ball onto the bond pad — applying ultrasonic energy and heat to create a solid weld. Then the tool moves to the board pad and makes a second bond, often called a stitch bond. Luna: So the ball is the first bond, and the stitch is the second? Lucas: Right. The second method is wedge bonding, which uses aluminum wire and a different tool shape. Instead of a ball, you get a wedge-shaped bond at both ends. Wedge bonding typically works at room temperature with just ultrasonic energy, so it's more forgiving on temperature-sensitive substrates. It's also the go-to for fine-pitch applications — pads that are very close together. Luna: And which one do startups typically use? Lucas: It depends on volume and cost. Ball bonding with gold wire is very fast — you can do ten to fifteen bonds per second — but gold is expensive. Wedge bonding with aluminum is slower and requires more precise alignment, but the material cost is lower. For prototyping, many startups use a manual or semi-automatic wedge bonder. Luna: What about yield? If one bond fails, is the whole board scrap? Lucas: That's the scary part. Wire bonds are essentially the last step before encapsulation. If a bond lifts or breaks during testing, you often can't repair it easily — especially if the die is already glued down. Some assemblies allow a rework where you can re-bond, but it's risky and time-consuming. That's why process control is so important. Lucas: Startups should track two key metrics: pull strength and bond shear. Pull strength measures how much force it takes to detach the wire from the pad. Bond shear tests the adhesion of the ball or wedge itself. A good manufacturing partner will give you a statistical process control chart for these values on every production lot. Luna: That level of quality assurance sounds expensive. Is there a volume threshold where COB starts to make sense? Lucas: Generally, I'd say below ten thousand units per year, it's hard to justify the NRE — non-recurring engineering costs. You need a custom tooling setup, you need a qualified wire bonder operator, and you need to qualify the encapsulation material — typically epoxy or silicone glob top that covers the die and wires. That can run twenty to fifty thousand dollars upfront. Luna: So for a small batch of prototypes, you'd probably stick with packaged chips. Lucas: Yeah. But if you're scaling to tens of thousands or more, and size or weight is a differentiator, COB can actually reduce your total cost — fewer board layers, smaller enclosure, lower shipping weight. I've worked with a startup making a wearable ultrasound patch. They replaced a connector with a wire-bonded die array and cut the overall package volume by sixty percent. Luna: That's a huge reduction. And what about reliability? Do wire bonds hold up under vibration or temperature cycling? Lucas: They can, but it depends on the materials. Gold ball bonds with a good epoxy glob top have passed thousands of thermal cycles from minus forty to plus one hundred twenty-five Celsius. Aluminum wedge bonds are more susceptible to corrosion if the encapsulation isn't perfect, so they're often coated with a conformal coating or a secondary overmold. Lucas: There's also a failure mode called 'wire sweep' — during encapsulation, the flow of epoxy can physically push the wires out of place. If they touch each other, you get a short. So the design has to account for the wire loop height and the viscosity of the encapsulant. Luna: This is getting into the weeds, but I think that's exactly what our listeners need — the concrete gotchas. One more thing: where do you actually find a contract manufacturer that does wire bonding for low-volume runs? Lucas: Great question. Most large EMS providers like Jabil or Flex are set up for high volume. For startups, I'd look at specialized micro-electronics assembly houses — often the same ones that do hybrid circuits or MEMS packaging. There are a handful in the US, like Promex or SyChip, and others in Shenzhen. You want a partner that's comfortable with small lots, say a few hundred boards at a time. Luna: And I imagine you should visit the facility before committing, given how many variables are involved. Lucas: Absolutely. Walk the line, ask to see their pull test records, check the cleanliness of the bond area — wire bonding is extremely sensitive to contamination. A speck of dust can cause a non-stick bond. Luna: All this talk about yield and process control makes me think about how much we rely on the people who do this day in and day out. If you're a founder listening, and you're trying to figure out whether COB is right for your product, maybe this episode helped you save some time. If it was worth a coffee to you, that's the link — buy me a coffee dot com slash fexingo. Lucas: Yeah, honestly, listener support is what keeps this show independent and ad-free. We don't do sponsors, we don't do affiliates — just the conversations we think are useful for builders. So if that resonates, the link is buy me a coffee dot com slash fexingo. Luna: Alright, back to the nuts and bolts. Lucas, you mentioned encapsulation — what's the typical material used for glob top? Lucas: Most common is a two-part epoxy that's dispensed as a liquid and then cured at temperature — typically around one hundred twenty-five degrees Celsius for about an hour. It has to have a coefficient of thermal expansion closely matched to both the silicon die and the PCB to avoid cracking. There are also uv curable silicones for applications that can't handle heat. Luna: And if you mess up the glob top, can you remove it and redo it? Lucas: Rarely. Once it cures, it's chemically bonded. You could potentially grind it off, but you risk damaging the die and the wires. That's another reason why the first-pass yield is so critical in COB. Every rework attempt is basically a gamble. Luna: So for a startup, the learning curve is steep. But the payoff in miniaturization can be huge. Lucas: Exactly. If you're building something where size, weight, or thermal performance is a competitive advantage — and you have the volume to justify it — chip-on-board with wire bonding is one of those superpowers that bigger companies have been using for decades. It's accessible to startups now, as long as you understand the process and the risks. Luna: Great advice. That's all we have for today. Next time, we'll dive into something completely different — stay tuned.