In the autumn of 2018 I had just started university. A senior’s old laptop had a minor fault, and a few of us volunteered to fix it. The moment our soldering iron touched the flexible circuit board, the board burned. We had an idea: a laser is an excellent heat source, so why not solder with a laser?
With a sharp snap, a laptop with one small problem became a laptop with no future.
We never fixed that laptop, but it led me into laser soldering. Over the next four years my team turned that idea into a machine we could sell. I was a partner on the team and its product manager. Looking back, the most valuable part was how we got unstuck, three times.
Why it burned
Once we calmed down and read up on it, we learned that laser soldering was already mature in Japan and Germany, and that the hard part was exactly temperature control. A laser’s energy density is so high that the board heats up almost instantly. Without control, the whole board burns.
Reading further, we found a real gap. Parts such as phone camera modules, micro antennas and earphone coils have tiny solder joints and are sensitive to heat. At the time there were only two ways to solder them precisely: buy imported laser soldering equipment, at more than a million RMB per unit and out of reach for small contract manufacturers, or solder by hand with an iron, which is slow and inconsistent.
Our school had just opened an innovation platform for smart manufacturing that offered student teams space and equipment. A few of us talked it over: why not give laser soldering a try? We were young enough not to be afraid, and there was little to lose if it did not work.
Stuck, the first time: the algorithm
For three or four months we were stuck on temperature control.
The usual approach was to mount an infrared thermometer beside the laser’s focusing lens and correct the temperature with software. We tried fuzzy control and neural networks, and neither worked well. It took us a while to see why: we were measuring at a different spot from the joint. However good the algorithm, every correction arrived too late.
The turning point was an offhand remark from a professor: “Have you thought about changing the physical structure?”
Going through the literature, we came up with the idea of building the temperature sensing path into the laser’s optics, so the sensing beam and the laser travel on the same axis and meet at the same focal point. Wherever the laser hits, that is where the temperature is measured. Following that idea, we ended up with five light paths on one axis (the laser, a pointer, illumination, CCD vision and infrared sensing) combined with variable-structure PID control. It held the laser temperature within ±2°C and could solder joints as small as 80 microns, roughly the width of a single human hair.
Stuck, the second time: the team
The machine was built by mechanical, electrical, optics and software people together, and each had their own “best” design. Even something as basic as routing cables through the frame could start an argument: the electrical side wanted clean, tidy wiring, while the mechanical side, thinking about the structure, insisted that a hole could not go there. Engineers tend to believe their own design is the most reasonable one.
My fellow co-founder and I found a way through. From the product’s point of view, we wrote down a few options with their trade-offs and let the team choose. The side whose change was easier, or better for the product, usually gave way. When we still could not agree, we built both and tested them. How far to push a parameter or which process to use rarely has a textbook answer, so we let experiments decide.
Even after a decision, someone would grumble: “Another change?” “Why didn’t you say so earlier?” That was when we talked to them one on one. Disagreement was never the problem; the problem was letting it turn into resentment.
Stuck, the third time: the market
With the technology working, we went to competitions, pitch events and trade shows. Plenty of business owners liked the prototype, until they learned that the person in front of them was a third-year student. Interest turned into doubt on the spot.
At first we asked our professors and a PhD senior to come along and vouch for us. Once we had the first orders, we changed our pitch: let the technology speak. The same kind of machine cost over a million RMB imported and about half a million from us, and in a real processing test on a customer’s line it reached a yield of 98.7%.
The machine went on to be used for light- and heat-sensitive parts such as phone camera modules, micro antennas, earphone coils and optical communication modules. The company made about RMB 3 million in revenue in 2021, earned high-tech enterprise certification and Hubei startup funding for college students, and grew from five or six people to more than twenty.
Three things I learned
Find a real pain point before choosing a technical route. Our starting point was a failure, but what kept the project alive was the next step: working out who pays for the problem and how they are coping with it today.
When algorithms cannot get around a problem, try changing the structure. If no amount of tuning fixes something, ask first whether you have been measuring in the wrong place all along.
Settle disagreements with options and experiments. Turn arguments into things that can be compared. If they cannot be compared, run an experiment.
In 2022 I was recommended for a master’s place in energy engineering at Huazhong University of Science and Technology, and I began handing day-to-day work over to teammates. Around then I said something at a sharing session that I still believe:
You don’t have to wait until middle age to start something, or wait for the perfect moment. If you are creative and willing to work hard, you can start any time.
When I switched fields, people asked me why. My answer: we only get a few decades, so why not try more things and make the road a little richer?