The perfect keyboard

Wednesday, Dec 26, 2018 · 2300 words · approx 11 mins to read

I recently finished building a new keyboard and I’ve now spent enough time typing on it to realise that due to certain aspects it’s, in aggregate, the best board I’ve ever laid my hands on to date. Yet it’s far from perfect, and in many ways it’s actually defined by its imperfections.

While building the new keyboard, the board I used in its stead has aspects of its own materials, build quality and practical systems that are, in their particular area, the best I’ve ever experienced.

That got me thinking about what the performance keyboard for me would be like, how it’d mesh aspects of those two boards, and whether I could possibly manufacture the idea of my mythical board in the future. I think I can, so I’m going to start by writing the main stuff down.

Case

A keyboard case needs to do two things: keep the board on the desk as you type, and present the keys to your hands so you can press them. My ideal case is one made of metal, so that it has a real heft to it and won’t move around on the desk as I type. Aluminium is heavy enough — the KBDfans TADA68 that I’ve been typing on for the last month or so has an aluminium case, and it’s one of its best aspects — but I’d love something made out of a lightly burnished, nickel-plated brass, with the barest of rubber inlays to stop it slipping.

The rake of the case is also important, at least for my hands. Flat boards like you have on most laptops, or even on something like Apple’s Magic Keyboard, aren’t angled enough so that the key rows are presented to the hand such that your fingers aren’t reaching too hard for the further rows. So my ideal board would be far from flat.

I’d also like that case, and the corresponding PCB internally, to expose the USB connector centrally at the rear. The switch or buttons for choosing which controller or flash would be exposed in a recess on the case bottom, a bit like the TADA68’s easy-to-press button to enter flash mode.

Switch plate

The construction of most modern mechanical keyboards is such that there’s a separate plate that holds the switches in place, that’s not part of the main case but anchored to it somehow, usually with screws. I’d build a board with separate plate I think, and the only requirement I have for it, aside from making it from the same plated brass that I’d use for the case, is that the plate be able to hold the switches in an ISO-like layout.

Layout

I’m a very big fan of the 69-key “ISO” layout sported by the original revision Input Club WhiteFox. It’s also used by the ISO variant of the TADA68, and it’s no surprise that they’re the two keyboards I talked about in the intro. However it’s not perfect. I’d add another vertical row of keys on the right, giving me 5 more 1U keys to use.

Visual symmetry would see me put those keys on the left, and I believe there are boards out there that do that, but for the keys I’d like to add I think they’re better placed on the right, at least for muscle memory’s sake.

PCB

The PCB details follow on from the case and layout: central USB connector and support for ISO-like layouts. One of the great features of boards like the Input Club WhiteFox and the TADA68 is that the PCB can support both ISO-like and ANSI-like layouts, and while my perfect board would only need to support the ISO-like layout that I love, if I were to ever sell the design (or give it away to the public domain), supporting ANSI would be a good idea since it’s by far the more popular descendant layout in use.

Secondly, mimicking modern PC mainboards and their dual BIOS support, my PCB would support either dual controllers each with embedded flash (most likely), or switchable external flash and a single controller, so that you can easily switch back to a known working firmware if you happen to make a mistake or the tooling somehow badly programs the controller firmware.

Lastly, I’d have hot-swap switch sockets. As you’ll see below, choosing a switch type isn’t the easiest thing in the world, so being able to infrequently swap them out to suit my tastes and mood would be much cheaper than buying a second keyboard! I don’t know what type of hot-swap sockets I’d use, but there are a few available.

Controller and firmware

Programmable controllers are one of the best things to happen to keyboards in the modern mechanical era, driven by open source controller designs. These days a fully programmable firmware is a given, allowing you to uniquely define the layout and functionality as you see fit.

For my perfect board I’d need programmable key macros (both recordable and immutable, the latter embedded into the firmware at compile time), multi-layer support, and individual control for each key’s LED. All of that is supported by the popular QMK firmware, so I’d start with that as the basis for my own — open source of course — board’s firmware.

Software

Programming the firmware for my perfect board would come with visual software for defining the layout, setting up the layers, editing key macros, and setting up the LED functionality. While there’s a part of me that enjoys editing the firmware source, compiling it, and flashing it, all via the command line, most users would be comfortable in a more visual piece of software.

Something like the Input Club configurator for their keyboards, which runs in a browser, as the basic idea.

Switches

One of the biggest practical differences between my WhiteFox — which I call the WeirdFox by virtue of its use of a revision 1 ISO plate from a WhiteFox, combined with the USB-C PCB and black case from the modern revision NightFox! — and my TADA68, are the switches I use.

The WeirdFox has NovelKeys x Kailh BOX thick click switches, and I use the jade variant which requires less weight to actuate and bottom out the switch compared to the navy variant. The switch also uses a thicker click bar compared to other clicky mechanical switches, making the click itself much louder and deeper compared to what you get on something like Cherry’s MX blue switch.

I love the BOX jade switch dearly; typing at full speed on my WeirdFox creates a raucous symphony of noise that’s like heavy artillery fire in comparison to an MX blue’s raspy and much less voluminous click. It’s brash and obnoxious and absolutely perfect in every way, and I’d choose it in a heartbeat, except…

I’ve got a completely different type of switch on my TADA68: 67g purple Zealios from round 8 of their production. Purple Zealios are a tactile switch without a click when you press them. Tactility without a click means that there’s a bump in the stem that the switch contact rubs against when you press the key.

That bump is what you feel when you press the key, giving you noticeable feedback that you’ve gone far enough as you depress to actuate the switch, and therefore don’t need to bottom out, which is what’s missing from a completely linear switch.

What separates the purple Zealio from the canonical tactile bump switch from Cherry, the MX brown, or their most Zealio-like effort, the MX clear, is the feel of the bump and the sound it makes. The bump of a purple Zealio is more pronounced in feel compared to the Cherry MX brown, but isn’t as sharp as the bump you experience with other tactile switches. That pronounced-yet-smooth bump is what gives purple Zealios their very particular feel and sound.

Depending mostly on the internal switch components — mainly contacts, stem and spring, along with any lubricant — but also on the switch housing, and the plate and case you’ve mounted the switches to, the sounds that a keyboard makes are almost as intoxicating to a keyboard enthusiast as the feel when typing.

What would I choose for my perfect board? I don’t think I could choose between those two switches today, despite them being so different to each other. Click versus bump; loud and deep shotgun crack of the BOX jades versus the smoother sounds of the purple Zealio; different actuation points and actuation weights — 67g bottom out on the Zealios vs 60g on the BOX jades — despite identical 50g initial actuation.

And even that’s a completely myopic view of the world of mechanical keyboard switches. What about BOX whites, which are the same as jades just with a thinner click bar, or the 40g actuated tactile BOX royals? Maybe I’d love V2 purple Zealios? Who knows!

What I do know is that in 2018 I learned that there’s a very wide and wonderful world of MX compatible switches outside of what Cherry themselves make, and I’m certain that my own perfect board would feature something from that vest selection of non-Cherry hardware.

Currently I lean towards a tactile click-free switch, but that’s only because I’m typing on something like that to write this.

USB cable

You might not think that the USB cable matters, but to a keyboard enthusiast it definitely does. In fact I’m typing this on the TADA68 just now with a very plain, thin, black USB micro USB cable and it ruins the premium look and feel of the board.

There’s not much to do here, but I’d have a custom cable made that was colour matched to the final case colour with an accent highlight the same colour as the accent colour of the caps on the board.

Keycaps

Last but not least, the thing that drew me into the world of bespoke mechanical keyboards in the first place: keycaps. Even though they’re not part of the board design, strictly speaking, they complete the board idea in my head. If you’re only used to the world of OEM or Cherry profile caps, or weren’t even aware that other keycap profiles exist, the huge variation in caps can come as a bit of a shock in the custom keyboard world.

Spend any time immersed in the community and you’ll quickly discover the huge variety in plastics, profiles, colourways and legends available for keycap sets, and that’s just the “normal” caps. There’s also a vast world of usually hand-crafted caps, called “artisans” by the community, which are low-volume products that tend to only fit Cherry MX-compatible switches, and where the designs are nothing like traditional key caps at all.

Ignoring any notion of artisan caps on my perfect board — although there are companies that do some incredible work on them, like Dwarf Factory in Vietnam — my favourite keycap profile today is MT3. It’s only surfaced on a single key set so far, although two more sets will reach production in 2019.

Its high-profile deep-dished sculpt is unique, feels fantastic to me, and also helps my typing accuracy because the dish locates my fingers on each key with less chance of slipping off into the void or onto a neighbour. Plus it looks really good.

The only available MT3 colourway is called /dev/tty and was sold exclusively through Massdrop. Two rounds have been produced so far, but there isn’t a third one planned at the time of writing, meaning the only way to get /dev/tty today is on the second hand market, usually at a steep mark-up for a set that’ll actually cover a board.

Two other MT3 sets are on the way: Godspeed (at the time of writing the round 1 drop is still available to join, but only if you hurry) and Susuwatari (round 1 is already over). If I have to pick an existing keyset for my perfect keyboard, Susuwatari is it. Light on dark, tastefully colourful, and that perfect profile.

If it was possible for me to design a keyset to my own exacting specifications for colour and key sizes, and have it manufactured, it would be Susuwatari MT3-like.

Summary

In summary, my perfect keyboard would use nickel-plated brass for the case and switch plate hardware, sport an ISO-like layout mostly cribbed from the WhiteFox but with an extra column on the right to give me 5 more keys, have a dual-controller PCB with central USB port, provide a well-designed GUI configurator for layout and layer changes, connect to my system via colour-matched USB cable, and my fingers would dance across MT3-profile Susuwatari caps. Switches would be a dilemma but the hot swap sockets would take care of that.

If it sounds fanciful and expensive then you’d be right. Rama Works made something very similar to what I want not too long ago, and I was very close to spending the almost £800 to get the Tank variant that Rama Works made out of brass.

Sadly, no ISO-like layout made it impossible to pull the trigger on a board that’s more than twice the cost of anything else I’ve ever put together, but I still swung between relief and regret for a couple of weeks after the Koyu’s limited run availability close alternately because of and despite the price.

The Koyu Tank means I know it’s possible to make what I want, and I even contacted Rama Works to ask if they’d take a commission for an ISO version of that board during one of my regretful swings. They said they’re not taking commissions at the moment, so maybe I can try and make it myself in 2019.