Clay Tech

"clay-works make things real"

translated from clazytech.com

I trust no one when I build a prototype

I get this question sometimes, but the people who ask it are almost always people who don’t build prototypes themselves.

That’s because the people who do build prototypes don’t trust anything.

Whenever I had an assembled board made, I always arranged for one extra bare board (a board with no components mounted) as well. That way I could check the quality of the bare board myself, separately from the assembly process. One time, an entire inner layer of a multilayer board’s circuitry turned out to be missing. The cause was a data mix-up between PCB manufacturers.

It’s an embarrassing story, but before I caught it I wasted about two days: mounting components on the board, powering it on, checking various waveforms, finding the power supply wasn’t working, probing point after point to check waveforms. In the end I stripped every component off the board, returned it to a bare-board state, and checked continuity at each point — that’s when I finally spotted the problem described above.

“If only I’d had a separate bare board, I would have seen it in an instant,” I thought, my shoulders sinking.

<Sometimes I trust the assembly shop, sometimes I don’t>

Prototype boards sometimes warp. A board is a stack of multiple materials in a mille-feuille structure, so naturally the thermal expansion rate differs by material, and heating or cooling it carelessly can cause it to bend.

Some of you might say, “I’ve never seen that happen.” You’ve been very fortunate in the people around you.

Several factors determine whether a board warps — broadly, board size and temperature — but in practice you can’t be sure until you actually try it. So a competent assembly shop tests this in advance. They arrange extra bare boards, prepare spare components, and test the assembly conditions beforehand to check the results. If the board does warp, it can cause all kinds of problems that aren’t visible at a glance: solder contact failures, destruction of small components, and so on. Avoiding that kind of thing is thanks to the behind-the-scenes effort of assembly shops.

But this kind of care is pure risk mitigation on their part, and it’s almost never listed on the “menu” of services. So it’s really the developer’s own skill to lay these safeguards themselves.

Sometimes the bare board has no trouble at all, the assembly goes smoothly, and yet the thing doesn’t work. This happens. Actually it happens all the time. It turns out the design was wrong, but as a small mercy, it’s a spot that can be fixed by hand without too much trouble. So, let’s modify it. On my one precious board?

A good engineer keeps spares precisely to avoid a situation like that. Just in case, I’d made three boards, so even after breaking one during modification, I still succeeded in modifying the other two.

Once the basic functionality check was done, I moved on to firmware development, but the time lost to the modification work put the schedule in jeopardy. I brought in another team member to help.

With two boards on hand, the two of us split the work between us without issue.

Development is like this in every respect. Anyone who thinks they can get through with a single prototype unit isn’t someone with exceptional confidence in their skills — they can be confidently identified as simply an amateur with little experience.

The content of this post is an excerpt (original text) from the following book. If you’re interested, please get a copy.

The Shape of a Happy IoT Startup

The Shape of a Happy IoT Startup


Originally published in Japanese at https://clazytech.com/2022/09/1155/. Translated with LLM assistance and reviewed before publication.