Hardware is slow

It’s January 2015 and you still can’t buy a Model 01.

The conventional wisdom floating around the startup community is that “hardware is hard.” That’s true, but more fundamentally, hardware is slow.

The first keyboard Jesse built took him about 40 hours of work once he had all the parts on hand. After a few iterations and a whole lot of experimentation, he could go from design concept to working keyboard in less than a day. That keyboard would have been laser-cut from plywood at Danger!Awesome in Cambridge, MA; populated with Cherry MX mechanical keyswitches scrounged from anywhere he could find them, Signature Plastics keycaps bought retail, and a Teensy or Arduino Micro; and all of it was soldered together completely freehand. Those early keyboards were fairly brittle creatures, but they worked. Most importantly, we could learn what did and did not work inside a day.

Over the past year or so, our cycle time has lengthened dramatically. It’s been frustrating, but it’s really important. The first keyboards Jesse built were great proofs of concept, but they weren’t robust, pretty or particularly manufacturable. Many of our recent iterations have focused on designing and testing a keyboard that we can make in the (small) thousands. Unfortunately, the bits and pieces that we can make easily and reliably in a production setup are incredibly slow and expensive to prototype.

The only ‘stock’ components we’re still using are the keyswitches. Rather than point-to-point soldering the keyswitches, they now slot into PCBs we’ve designed. The Arduino is gone… well, not gone exactly, but replaced with equivalent components on our left-hand circuit board. After a flirtation with aluminum, we’re making the enclosure (body) out of wood again. It’s not lasercut anymore, though. We’re milling it out of solid maple. The enclosure wraps around a pair of custom designed aluminum plates. We’ve replaced the commodity keycaps with custom caps we’ve sculpted to better fit your fingers.

Lately, each major design iteration of the enclosure has been taking us anywhere from 2 weeks to a month. In an attempt to speed this up, Jesse learned how to CNC-mill the enclosure prototypes himself, but found that occasional access to the ShopBot CNC machines at TechShop wasn’t going to get us something useful fast enough. For the most part, we’ve been getting our prototype enclosures cut by StrongD in Shenzhen. We’ve tried out a few different wood options. So far, it looks like maple’s winning the race for its combination of easy machinability, durability and longevity, and beautiful tight grain. It helps that maple is a hardwood that is relatively easy to source sustainably (maple trees grow fast!) Recent changes to the enclosure design have all been focused on making it easier to mill and more comfortable to type on. In general, StrongD can turn around a design for us in about 10 days. We could cut that down to less than a week by using a US based shop, but it’d cost an order of magnitude more. (We’ve gotten the quotes.) StrongD do a fantastic job and have been a joy to work with.

We’ve designed the keyboard PCBs to fit snugly inside the wooden enclosure. Consequently, we’ve been doing PCB revisions in lockstep with enclosure revisions. Once designed, each new turn of PCBs takes about two weeks to come back from Seeed Studio. Early on, Seeed was building us bare PCBs and we were soldering all the components onto the boards ourselves. As the complexity of our electrical design has outpaced Jesse’s electrical engineering skills, we’ve started getting circuit design help from professional electrical engineers and have started to take advantage of Seeed’s PCB assembly services. It takes longer than if Jesse were doing it himself, but it’s also much more likely to result in working boards.

The keycaps are the bit of the keyboard you interact with most directly. We’ve made the unusual choice to design 64 unique keycaps so they’re just right. It’s not surprising that they’ve taken many, many hours of work to get right. With coaching from Taylor Stein at Autodesk, Jesse’s built up a solid model of our keycaps in Fusion 360 over the past two months or so. Taylor has been generously printing test keycap sets for us on Autodesk’s high-end 3D printers. They’re great for getting a feel for the shapes, but too brittle to type on for more than a very short time before they crack. So far, the best technique we’ve found for making prototype keycaps is to have StrongD mill them from polycarbonate, paint them and laser-engrave them.

Now that we’ve got prototyping technologies reasonably well figured out, we’re moving on to making sure we can make you keyboards by the hundreds or thousands. Over the past month, we’ve been getting volume quotes for PCBs, keycaps and the enclosure. While our final selection of manufacturing partners is going to depend on the size of the production run, we’ve got a good feel for the economics and upper and lower bounds of various manufacturing partners’ production capabilities.

We’re finally moving on to the figuring out the mechanics of the crowdfunding campaign. After wading through over 120 directors’ sample videos and demo reels, we’re going to start filming our crowdfunding video on Monday. From there, we’ll make sure we have all our ducks in a row and then… Well, then the real fun starts.

Edited on Jan 15, 2015 to correct broken links. Thanks to @smylers2 for catching them.

Older Post Newer Post

$ 0.78
$ 456.78
$ 123,456.78