Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Surface Mount Technology Prototyping for Faster Iteration
Transcript
- Lucas: If you're building a hardware startup and your PCB prototype cycle takes longer than a week, that's your bottleneck — not the funding, not the software, not the go to market plan. Luna: And yet most founders I talk to ship their first boards out to a full-turnkey assembler and wait four to six weeks. They treat it like a necessary evil. Lucas: Right. But there's a middle path that a lot of hardware people don't seriously consider until they're already behind schedule: bringing surface mount technology prototyping — SMT prototyping — in-house at a very small scale. Not a full production line, just a desktop pick and place and a reflow oven. Luna: I've seen that work. I know a medical-device startup — they make a wearable ECG patch — that was stuck on board rev three for months. They brought in a $15,000 desktop pick and place and cut their turnaround from six weeks to four days. Lucas: Four days. That's the kind of number that changes how you think about hardware iteration. And if today's conversation gives you a concrete reason to rethink your own prototyping approach, that's the whole point. The way these episodes stay ad-free — no sponsors, no mid-rolls — is listener support. You can find us at buy me a coffee dot com slash fexingo. It's a way to keep these conversations going, nothing more. Luna: Yeah, totally. And Lucas, walk me through the actual setup. What does a 'desktop SMT line' look like for a startup that's maybe doing five to fifty boards per prototype spin? Lucas: You're looking at three pieces of gear: a solder paste printer — or at minimum a stencil alignment fixture — a pick and place machine, and a reflow oven. The printer and oven can be surprisingly cheap. A decent entry-level reflow oven is maybe $2,000 to $4,000. The pick and place is the big ticket. A used or new desktop model like a CHM-T36 or a TVM802 runs between $10,000 and $20,000. That's the whole line, $15,000 to $25,000 all in. Luna: Compare that to what you'd spend on one or two full-turnkey prototype runs with a contract manufacturer. Those can easily be $5,000 to $10,000 each, plus shipping and lead time. Lucas: Exactly. So after three prototype spins, the equipment pays for itself. And that's before you factor in the speed. But there's a catch: the quality of your SMT prototyping depends heavily on your stencil and your reflow profile. If you get those wrong, you waste boards faster than you can place them. Luna: Let's dig into the stencil part. For low-volume prototyping, you're probably not buying a laser-cut stencil from a high-end supplier. What do you actually need? Lucas: You need a stencil that matches your board's pad pattern, obviously. For prototype volumes, a laser-cut stainless steel stencil from a quick-turn shop costs maybe $50 to $100 and arrives in a day or two. You don't need an electroformed stencil until you're dealing with 0.3-millimeter-pitch components. The critical thing is alignment. If your stencil is off by even 0.1 millimeters on a fine-pitch QFP, you get solder bridges and you're spending your time reworking instead of iterating. Luna: So a good manual alignment fixture — basically a frame with micrometer adjusters — is worth the investment. Those run maybe $200 to $500. Lucas: Worth every dollar. And then the reflow profile. This is where a lot of startups screw up. They buy a cheap toaster-oven conversion and they don't have proper profile control. You need a preheat zone, a soak zone, a reflow spike, and a cooling ramp. If you're using a converted oven, you can still get good results with a thermocouple attached to the board and a PID controller. But honestly, a purpose-built desktop reflow oven with a programmable profile is worth the extra couple hundred bucks. Luna: And the pick and place machine — is it really feasible to hand-place all the fine-pitch components and only use the machine for passives and larger ICs? I've seen some startups do a 'hybrid' approach. Lucas: Absolutely. In fact, that's the smartest way to start. Use the pick and place for all the 0402 and 0603 resistors and capacitors — the stuff that's tedious and error-prone by hand — and hand-place the fine-pitch QFPs, BGAs, and connectors. For a prototype run of twenty boards, you can place all the passives in maybe twenty minutes with the machine, then spend fifteen minutes hand-placing the tricky parts. That's way faster than hand-placing everything, and you avoid the alignment headache of teaching a machine every single fine-pitch part. Luna: So the machine handles the grunt work, and you handle the precision. That makes sense. But what about component sourcing? If you're doing small runs in-house, you can't rely on full reels from distributors. You're buying cut tape or loose parts. Lucas: Right. Cut tape is your friend. DigiKey and Mouser will give you cut tape in lengths of ten, fifty, a hundred. You load those into the pick and place feeders — most desktop machines have eight to twelve feeder slots — and you manually advance the tape as needed. It's a little more hands-on than production, but for prototypes it's totally fine. And you avoid the minimum order quantities for full reels. Luna: And the yield? If I'm a startup founder, I'm worried about wasting boards. How many boards do you typically lose to SMT defects on a first run with a new line? Lucas: If you've done your stencil alignment and reflow profile right, you should see maybe five to ten percent defect rate on your first prototype run. That means you lose maybe one or two boards out of twenty. That's acceptable. And most defects — solder bridges, tombstoning, insufficient solder — are fixable with a soldering iron and some wick. So your usable yield can be close to a hundred percent after rework. Luna: And the inspection step. You need a microscope. I've seen people skip that and then wonder why their board doesn't work. Lucas: You absolutely need at least a stereo microscope with ten to forty times magnification. And a good light source. Inspection time is the single biggest time-sink in low-volume SMT. You can place and reflow twenty boards in an hour, but inspecting and reworking them might take two or three hours. But that's still faster than waiting four weeks for a contract manufacturer. Luna: And there's a hidden benefit: the learning curve. Once your team has done a few in-house prototype runs, they understand the assembly process deeply. That translates to better design for manufacturing decisions down the line. Lucas: Hugely important. When you've personally struggled to place a 0.4-millimeter-pitch connector, you design pads that are a little more forgiving. You avoid castellated holes that are hard to solder. You think about component orientation. That kind of tacit knowledge is hard to get when you're just sending Gerbers out to a CM. Luna: So what's the threshold? At what point does in-house SMT prototyping stop making sense and you should transition to a contract manufacturer? Lucas: I'd say once you're consistently doing prototype runs of more than a hundred boards, or you need multiple spins per week, the economics shift. At that volume, a full-turnkey CM with panelized boards and automated optical inspection will give you better cost per board and higher consistency. But for the early stage — the first ten to fifty boards, the first five to ten design iterations — in-house SMT is a superpower. Luna: And one more thing: you need to account for the time to set up the line and maintain it. It's not zero overhead. Lucas: No, it isn't. But if you're a three-person hardware team, dedicating one person to SMT for two days a week is a lot cheaper than waiting. And the ability to design, assemble, and test a board in the same week is the kind of speed that big companies can't match. That's your edge. Luna: Alright, I'm sold. Let's say I want to start next week. What's the single piece of advice for someone who's never done SMT? Lucas: Start with a simple board. Two-layer, maybe fifty components, all 0805 and larger passives, one QFP. Get your process dialed on that before you attempt a six-layer board with BGAs. And use a solder paste that's designed for prototyping — a no-clean, type 4 or type 5 powder. That gives you a wider reflow window. Luna: And if you screw up a board, that's a lesson. The cost of a few prototype boards is nothing compared to the time you save. Lucas: Exactly. The hardware startups that iterate fastest are the ones that control their own prototyping. In-house SMT is one of the most underrated ways to do that.