Over the next two weeks, I'll be visiting the rest, as well as a few additional potential wood enclosure suppliers in Shenzhen. I'll also be visiting the Global Sourcing Fair in Hong Kong. You can think of the GSF as 'sort of like CES, except with very, very few brand names you've ever heard from'—it's a four day festival of all of the names behind the names. I'll be scouting additional potential suppliers for the Model 01.
Since we last wrote, we’ve overhauled our product documentation and "bid packet," engaged professional help with sourcing and manufacturing contract negotation and sent the bid packet out to around a dozen potential manufacturers.
We've also finally gotten a full build of the new Model 01 design we'd expected the ex-manufacturer to build for us. That includes new circuit boards, updated keycaps, an updated enclosure, a new interconnect hinge design and new feet.
Below, you can see a quick snapshot of the updated enclosure and keycap shapes. Because this prototype was primarily intended to test shape and fit, we didn't spend the extra cash on making the enclosure out of wood or painting and engraving the keycaps like we will for your keyboards.
About a week and a half ago, we got five sets of the new circuit boards back from a quick-turn PCBA (Printed Circuit Board Assembly) shop in San Jose California. Sitting in the parking lot after picking up the boards, I plugged a USB-C cable into one of them, connected it to my MacBook and was pleased to see the keyboard tell my Mac that it was ready for a firmware download.
I drove home, plugged the board into my firmware flasher and...nothing. Actually, it was a little worse than nothing. I saw a tiny little puff of magic smoke rise from the board.
48 hours later, I found myself back in the South Bay. Jacob Alexander from Input Club spent a day working with me to triage design and build issues with the new circuit boards.
The magic smoke issue turned out to be caused by the same mistake as the not-being-programmable issue. We'd inadvertently had a diode mounted backwards on the PCB. When the programming cable was plugged in correctly, it couldn't tell the microcontroller to reboot into programming mode. When the programming cable was plugged in upside down, the diode got hit with 5 volts of power coming in the wrong direction and popped. Happily, the diode was intended as a safety feature—once we pulled it off, the board happily took a new firmware.
After that, we discovered that the right-hand PCBs were drawing insane amounts of power, not ever really powering up properly and were getting really, really hot. A close reading of the schematics for the two boards turned up a tiny, tiny miswiring. One of the pins on the keyscanning controller was wired to ground when really, it was supposed to be wired to the 5 volt line. So it was doing something screwy inside the controller to try to get that power, but being thwarted by the external ground connection. Jacob, who has a much steadier hand than I do, cut the relevant pin, tacked on a patch wire and things started to behave better.
The next issue we ran into was that...most of the LEDs didn't work. A little bit of triage with a magnifier revealed the problem: about 60% of the LEDs looked cloudy inside. It turned out that the PCBA shop hadn't properly treated the reel of LEDs we'd consigned to them. Electrical components typically ship from the factory in air-tight packaging. They can absorb moisture when they're exposed to the open air for more than a few days. As it happened, there was a nick in the airtight foil packaging of our sample reel, so our LEDs had been slowly absorbing a bit of moisture over the course of a few months.
The process of building circuit boards is generally to have a machine lay down a coating of solder paste, have another machine drop the components in the right places on top of that solder paste and then put the whole thing in a special kind of heater called a reflow oven. The reflow oven gets very hot, very quickly. The solder melts. Components get stuck in place. You've got a working circuit board. Unless your components have been absorbing moisture for a few days, weeks or months.
When you boil water, it turns into steam. Turning water into steam in a confined space is kind of amazing. It can be amazingly useful when properly harnessed, say inside a steam locomotive that's designed to actually handle all that pressure. Even that can go wrong, though.
Our little LEDs never had a chance. Well, they would have had a chance if the PCBA shop had baked them for long enough. As it turned out, the PCBA shop had done...mostly the right thing. They'd put the LEDs in a low-temperature oven to bake out all of the moisture before dropping them onto the circuit boards. They just hadn't prebaked them quite long enough or at quite high enough a temperature. The only solution to something like this is to throw away the blown LEDs and put new ones on the boards.
So back we went to the PCBA shop on Monday afternoon to drop off the boards and a test jig to show off the problem (and help verify the fix). By the time we picked them up 24 hours later, they'd prebaked new LEDs and manually worked all of the boards.
The triage would have been a little bit easier if we'd had good test points on the boards for all of the LEDs' clock and data pins, so we'll likely be adding those. The test points don't really change the cost or complexity of the boards, but will make debugging easier.
There are a couple other tiny issues we've found with the board design that we'll improve for the next spin. So far, none of them are issues that would stop the production lines if they'd snuck onto the final board design, but they're things that we're happy to fix while we can.
On the mechanical side, we're pretty happy with the new enclosure shape and updated keycap shapes, though we'll likely do one more revision of the palm key shapes to make them even more comfortable. The feet and interconnect hinges are another story entirely. Neither turned out quite as we'd hoped. The hinges are fiddly to use, hard to assemble and not as sturdy as we'd hoped. We have a few new designs we're prototyping and we're getting some incredibly helpful suggestions from the manufacturers I've been meeting with. The feet have similar problems. While it's somewhat petty of me, I feel a need to point out that both designs were recommended by that manufacturer we're not working with any more.
We've been working with a company called HWTrek on sourcing. What that means is that HWTrek has been helping us find new manufacturers and suppliers. They do the initial scouting and vetting before introducing us and shepherding us through the process of getting a quotation. They've also helped take care of transportation on the ground in Taiwan and Shenzhen, as well as translation and haranguing folks on our behalf.
This week, Jack and Joe from HWTrek have taken me to visit two keyboard manufacturers in the Taipei area. In each case, the manufacturer had already signed an NDA and had a copy of our bid packet in advance.
In both cases, the manufacturer's salesperson brought an electrical engineer and a mechanical engineer to the meeting. We spent a lot more time talking about the mechanical design than the electrical design, with both manufacturers' EEs saying that they were pretty comfortable with our board design. The general manager of each company joined the meeting after a bit and weighed in on the design. Typically, if the GM's excited about a project, you're going to have a much easier time getting to a successful outcome.
Both manufacturers came into the meeting having identified that the feet and interconnect mechanism were going to need work. One of them walked in with 3D CAD models showing off a redesign of the internals of both the feet and the hinge. The redesigns are outwardly identical to what we handed them, but make them cheaper to make and more reliable to operate. The other manufacturer's mechanical engineer brought up the issues in the meeting, but didn't come in with designs. Instead, he started sketching potential solutions while we talked through the business side and the electrical design.
Both manufacturers are about the same size in terms of production volume, though one exclusively makes products for their customers and the other is split evenly between customer projects and their in-house brand. One manufactures their products in Taiwan and the other does their manufacturing in China—I'll be visiting their factory late next week.
Each has their strengths. The factory in China has more experience with injection molding. The factory in Taiwan has been making keyboard for much longer, has a lot more automation and we're more directly familiar with some of their customers' products. We're hoping to get quotes from both manufacturers over the next several weeks.
I also visited two wood factories scouted by HWTrek in Taichung. (These are options in addition to the wood factory in mainland China that made the cool samples for us a few months back.) One turned out to be more of a wood contract manufacturer and tourist attraction. Their in-house wood shop doesn't do CNC or wood finishing, just more traditional cutting, sanding and a bit of engraving.
As we pulled up, we were a bit surprised to see four tour busses full of elementary school students. Turns out that their other business is "come see a real live wood factory!"
The second wood factory was a very, very different experience. Nestled down a side street off a side street in an industrial district, we pulled up in front of a big open warehousey sort of space.
The woman who met us turned out to be a second generation owner who brought us into their workshop to show off their six-mill 3-axis CNC and eight-mill 4-axis CNC. After showing off a bit of their work product, she told me that they'd designed and built the machines from scratch.
From there, we visited the other side of their workshop, where there were several CNC machines in various states of assembly. They too have a second line of business. They design and build CNC machines.
We sat down to talk though the project and pretty much instantly were talking through what size of CNC router bits we'd need to do various bits of the design most efficiently. I asserted that the entire underside was designed to not need anything smaller than a 1/4" bit. She looked at the design for about 30 seconds and told me that I was wrong. Turns out she was right. There's one corner that's just a bit tighter than that.
We talked through the production process, spending a fair bit of time on surface coatings (which they also outsource), color-matching the left and right halves of the keyboard, and the brass screw inserts they'd need to place to make it possible to attach the wood enclosure to the rest of the keyboards.
They brought out some pretty impressively nicely milled samples of their work before asking if we wanted to check out their 3D scanner and desktop laser engraver.
We'll see what both quotes look like, but as you might imagine, one of these shops is....much more our speed and sort of people than the other ;)
Next week, I have one more meeting in Taipei before heading off to HK and Shenzhen for more meetings and factory visits. With a little luck, we should have a whole lot more information by our next backer update.
<3 j + k