hbar alpha, an ergonomic keyboard testbed

hbar · 24 Jun 2015, 21:52

I recently decided that all keyboards I have or can buy have something in them that I don't like. Therefore, the only thing worth living for is a new, custom design that has never been made before. My intention is, therefore, to create a keyboard that (maybe after a few iterations) can satisfy my needs (and hopefully those of others, too, hence I just opened this thread).

Learning a lot from reading lots of articles on the internet, I set my goals like this:

1. Exclusively use parts available new. This rules out using donor boards for switches or electronics.

2. Split, ergonomic layout, with both halves physically being mirror images of each other.

3. Staggered columns, determined by the shape of my hands.

4. Compact design with minimal number of switches, with layers instead of dedicted clusters for navigation, etc.

5. No multi-key thumb cluster. I only expect the thumb to move inward, not outward, so Alt and Compose may be thumb keys, apart from the space bars (which the thumbs shall rest upon).

6. Curved plate. In order to avoid excessive complexity, the curvature will be cylindrical. If required, this can be made up for by using sculptured key caps.

7. Laser-cut chassis. The plate and its entire support structure shall be made by others (since I don't have the time) so I only have to design and assemble it.

8. Throw-away chassis. If I deem the design to be less than optimal, I discard it and design a new one. All valuable parts, the electronics, the keys, and the caps, shall be recycled in that event.

9. Individual switches with widely used mount and stem. This directly implies having to use the de-facto standard, Cherry MX. I don't particularly like any of its flavours, but it would be unpractical, if not impossible, to use BS or Topre instead. Maybe some day someone comes up with a viable way of building custom curved keyboards with BS or Topre, but until then, MX should be fine for testing the layout. I think that the layout is at least as important as the switch type used.

10. Hand-wired electronics. For a one-off project, this is much quicker and cheaper than designing a PCB that becomes obsolete the moment I decide to redesign the chassis. Having a diode next to each switch should also pose no problem so we can wire things up without worrying about modifier keys etc.

11. Commercially available USB microcontroller board with tmk_keyboard firmware, and minimal, if any, additional circuitry.

I hope that, eventually, this thread will be useful to other keyboard designers. Maybe you can find an idea or two that can be re-used. It appears to me that some things I have in mind haven't been tried before, so watch this space.

There's no guarantee that I'll follow through with this project, let's see how it turns out. It's fun to get started, anyway.

ħ

hbar · 24 Jun 2015, 21:55

So this is the layout that I currently have in mind:

: hbar_alpha_layout1.png (73.81 KiB) Viewed 7434 times

As you can see, the bottom row is shifted to the right by one, which I find looks more symmetrical than all column-staggered keyboards I've seen so far (which put T, G, and B above each other, leaving much more letters for the left hand to press than the right).

Apart from the letters, I tried distributing the symbols in a sensible manner, but practice will show what layout really makes most sense. Some of it was done by putting missing symbols into empty places, so I expect some changes here.

If I counted right, this makes it 62 keys, just perfect for a 4x8, 5x7, or 6x6 scan matrix per side.

I ordered a few bits and pieces (keys and caps for testing). Next on my list is drawing and submitting a test design to the laser cutter (I haven't used such a service before).

ħ

Muirium · 24 Jun 2015, 21:59

Worth a read:

http://deskthority.net/workshop-f7/my-d ... t2534.html

hbar · 24 Jun 2015, 22:22

Blimey. Not bad.

Muirium · 24 Jun 2015, 22:26

Yeah, Suka's the man to talk to about custom split ergo boards. He's been through a few builds, to say the least!

hbar · 24 Jun 2015, 23:02

The trouble is that 3D printing is quite expensive (except if you work at the source, as Suka does) and requires 3D design capabilities (software and skills). I'm looking into a much simpler way of building the chassis, more about that later.

ħ

jacobolus · 25 Jun 2015, 02:12

hbar wrote: ↑4. Compact design with minimal number of switches, with layers instead of dedicted clusters for navigation, etc.

Minimal is a lot fewer keys than you have here. I’d say you can get by with about 40 keys without sacrificing function or needing to learn anything too weird, or easily with 50. You have 62. (Nothing inherently wrong with including this many keys, just saying...)

5. No multi-key thumb cluster. I only expect the thumb to move inward, not outward,

Any particular reason? For example, are your thumbs injured, or does your hand have unusual shape / lack of flexibility? Otherwise, I highly recommend adding more thumb keys.

6. Curved plate. In order to avoid excessive complexity, the curvature will be cylindrical. If required, this can be made up for by using sculptured key caps.

Cylindrical which direction?

Personally I’d recommend using flat halves with the most aggressively sculptured keycaps you can find.

A cylindrical plate puts the switch axes at angles which make pressing keys more awkward / less efficient. If you want something more sculptured, go all the way and mount the switches at different heights, e.g. in a 3d-printed or milled case/plate (like Maltron-style, but more carefully designed). In particular it’s nice to mount the middle finger switches lower, and the pinky switches higher, than the other fingers, but again with the axes of the switches close to aligned with each-other.

7. Laser-cut chassis. The plate and its entire support structure shall be made by others (since I don't have the time) so I only have to design and assemble it.

Laser-cut wood/acrylic plate as well, or metal?

9. Individual switches with widely used mount and stem.

Go for Matias switches rather than MX. MX is garbage.

* * *

Comments on your proposed design:

- Full 1u of stagger between pinky and ring finger is probably a bit more than you want. My personal preference is about 3/4u stagger there, but in user testing I found several people who didn’t like more than about 1/2u, so that’s what I’d recommend. If you want to improve reach for the pinky column, the best solution is to raise the keys vertically.

- Assuming 'A' is roughly your "home position" for the left pinky, those escape and backslash keys are going to be an annoying stretch.

- I’d recommend less keys in the inner two columns, and maybe bigger ones. hitting those comma, period, tilde, and equals keys is going to be a pain in the ass, and requires reaching your hand. Consider just dropping those and putting some of them on a layer.

- Consider swapping ctrl and shift, putting backspace on a thumb key

- Consider putting numbers and symbols on 1–2 layers near the home row instead of their own top row. They’re much easier to type that way.

- Move the fn keys somewhere easier to reach.

- Do some careful testing to figure out where your thumb naturally lands, and make sure the spacebar is placed there. Most important, consider moving thumb keys away from the finger keys/toward your body by about 1/2u, so you can relax your fingers naturally land on both spacebar and home positions. Place the spacebar(s) directly where the thumbs land, and then place other keys afterward. Consider making the 'alt' key 1u long to make 'comp' easier to reach.

Muirium · 25 Jun 2015, 02:28

I'm still a card carrying member of the MX skeptic club, but I've got to admit that Matias switches aren't much better. They are no rival to classic (complicated) Alps. Which in turn is a poor rival to Topre or buckling spring. And the extra hassle of having to source Alps mount caps for them rules Matias switches out, in my view. You've a lot of extra work for very slim rewards.

To be fair though, I haven't tried a board of damped clickies though. (Matias regular clicky switches with "soft click" stems.) Those could have something the plain soft clicks don't have. Maybe…

Findecanor · 25 Jun 2015, 02:32

hbar wrote: ↑The trouble is that 3D printing is quite expensive (except if you work at the source, as Suka does) ...

No "maker space" in your area?

jacobolus · 25 Jun 2015, 03:38

Muirium wrote: ↑I'm still a card carrying member of the MX skeptic club, but I've got to admit that Matias switches aren't much better. They are no rival to classic (complicated) Alps. Which in turn is a poor rival to Topre or buckling spring.

It’s all subjective. I find Matias switches noticeably more enjoyable than a Model M or most of the Alps switches made after about 1990, and I strongly dislike Topre switches. YMMV.

Anyway, to the OP: use whatever you like. I was half kidding about MX, hence the smiley face.

hbar · 25 Jun 2015, 08:56

Gosh, that's more response that I could have hoped for. I'll have to go through Suka's thread a few more times to figure out what layout is best to start out with. To be honest, I never considered anything with fewer keys than the HHKB, given that until a few weeks ago, I wouldn't have even thought it possible to have a useful keyboard without cursor and F1..F12 keys and having an Fn key that's used for more than adjusting screen brightness and turn the ThinkLight on and off.

I do expect a few iterations, but starting with something sensible could keep that from becoming too many iterations.

hbar · 25 Jun 2015, 21:05

So it looks like I'm going to revise that layout in any case. A much smaller, more suka-like layout quite appeals to me. Plus, there won't be any need for stabilizers, and one TKL donor (in case I decide to recycle parts) would be enough for two of those boards. Neat.

In any case, laser-cutting still looks promising (if it works, which I'm not even 50% sure of) and would be much less expensive than laser-sintered nylon.

Damn, why didn't I know of Suka's thread before I started this one?

ħ

Muirium · 25 Jun 2015, 21:31

The pick of DT's workshop threads, all in one place:

http://deskthority.net/workshop-f7/work ... t7192.html

Suka is in there, of course. That's the required reading section that you seek!

What's your doubt about laser cutting? We've made a fair few alu and stainless steel laser cut boards around here. Something about going beyond simple layered plane shapes, right?

hbar · 26 Jun 2015, 22:04

The first test file is off to the laser printer, fingers crossed. (Mu: yes, I'm thinking of something beyond simple layered plane shapes.)

I had a quick glance at laser sintering as well (even though I don't think I would currently be able to design anything that works), and it looks rather expensive. I don't think I would be able to make a prototype keyboard chassis for less than €50, and that's quite a lot of money to throw away if you find out you have to modify the design a bit. Or am I looking at it the wrong way?

ħ

hbar · 04 Jul 2015, 20:16

Just a quick update: I got the first test cuts back from the laser cutter, they seem pretty much as expected or even better.

It's a shame I'm away this week-end and can't try them right away. I'm itching to assemble it all and put switches in.

ħ

hbar · 07 Jul 2015, 00:39

I finally assembled the two test "boards". The fit was perfect, it was only my clumsy glueing that wasn't, so here's the result. I made two in order to find the geometry that I would be aiming for in the real thing.

I hope you can see the two different curvatures in these photos.

: t1.jpg (48.92 KiB) Viewed 7156 times

: t2.jpg (31.95 KiB) Viewed 7156 times

The cylinder that fits on the top of the keycaps (DSA ones, not these OEMs, but I haven't got them yet) should be 20cm and 10cm in radius, respectively.

After only a few minutes of testing, I immediately like the rather aggressive 10cm "bowl" much more.

Lessons learned so far: the material (made out of timber, apparently) is strong enough to hold the MX switches, yet flexible enough to be bent. My simple scaffolding needs improvement of course, I expected that from the start.

ħ

hbar · 20 Jul 2015, 22:46

Thanks to the load of information in this thread and others (Suka's DIY in particular), I decided to go with a truly "minimal" design, currently I'm considering a layout with 44 keys with AdNW layout (thanks, Suka!). It's a shame my DSA keycaps still haven't arrived, so I can't test my 2x5 assemblies with them, though I think the 10cm one will be better in any case.

It's amazing how much this thread has already helped me think outside the box and consider choices I wouldn't have thought would make much sense.

ħ

vvp · 21 Jul 2015, 09:51

Another contoured keyboard. Nice!
I do not believe I did not notice this for a whole month.
As for as the diodes on the switches. Here are two different approaches how to do it.

Most people do it like Muirium here.
I did a bit different way (which I believe is easier) here. Though I switched to an enameled wire for the rows. This way I do not need to use a separate insulated wire to connect rows with thumb cluster. The nice thing about my approach is that the you do not need to hold diode/wire in hand while soldering. You get one hand for iron and the other one for adding solder.

hbar · 31 Jul 2015, 00:33

Finally got those DSA keycaps in today, here they are.

: ha1.jpg (43.18 KiB) Viewed 7025 times

Since starting this thread, my plans for the keyboard have changed considerably. Today I learned too things, immediately obvious once I put the caps on those two 2x5 layout testers I made earlier:

: ha2.jpg (36.64 KiB) Viewed 7025 times

First, aiming for a 19.05mm pitch measured at the top of the keycaps was a bad idea, the keys can be moved closer together vertically; the goal is probably to maintain the 19.05mm at the base of the keycap, so that the gap between neighbours is uniform. Second, given that the alpha keys only have three rows now, it appears to me that any curvature in the board is a waste of effort. My first reaction (remember, it's quite late and tomorrow I might change my mind again) is to ditch the curved plate concept altogether. For a flat plate, sheet metal and acrylic are much better materials (stronger and they don't require the scaffold that is hard to design), so I'll probably switch to one of those.

ħ

Findecanor · 31 Jul 2015, 16:29

Still, that was a very interesting way to construct a curved keyboard. I have considered laser-cut wood before, but never for the plate.

BTW. If your keyboard is going to be only 3 column-staggered rows plus a row of thumb keys, and no more than 6 switches wide, then a PCB for such a keyboard could fit within 10×10 cm. Then construct it symmetrical, in the same way as the ErgoDox is and you could get PCBs for (approximately) five keyboards made from DirtyPCBs.com for only $25.
The outermost switches on the left and/or right side would hang on the edge, though.

hbar · 31 Jul 2015, 23:01

The material isn't wood, it's called "Kraftplex", which is made of wood without added glue or binder, apparently, and mainly consists of cellulose fibres. The type I used was the thinner of the two strengths available (0.8mm, the other being 1.5mm). Bear in mind that it isn't waterproof, so I tested a light clear coat on the 20cm assembly and it seems to work (as in repel water) very well. I expect that normal use would cause it to start looking dirty very quickly unless protected somehow.

What I like about Kraftplex (and what I didn't expect before I handled it myself) is that it's less elastic than wood, so it's very easy to bend and will even stay in place if you do it right. The two layers of the plate on both assemblies I made were pre-formed by hand, then glued together essentially with no tension at all. I used clamps mainly because I (falsely) expected it to constantly want to straighten out, but I think you could easily do without them with a little practice.

I've made PCBs before and I can only say that I avoid them when I can (mainly because every single piece PCB software I've ever seen is a pain to use, and I've seen quite a few -- 2D CAD with Librecad is so much nicer and bears less risk of getting it wrong). A sturdy plate and hand-wired matrix is the way I'm going now. Currently working on the plate design. I've got plans for the next keyboard, though (maybe you can guess its name), which will use a PCB without any doubt (because there's no way around it).

ħ

hbar · 17 Aug 2015, 19:54

Two parcels arrived today, containing this (1.5mm stainless, one side brushed):

: ha1.jpg (115.17 KiB) Viewed 6916 times

as well as this (8mm clear acrylic):

: ha2.jpg (139.92 KiB) Viewed 6916 times

Both shipments surprised me: the acrylic came out better than I thought, especially given that I drew holes for M3 bolts to pass through in acrylic 8mm thick. The stainless top plates, on the other hand, have remains of a plastic foil (top left in photo) on the top side, which leaves a very nasty residue (top right in photo) when removed. Currently, I haven't got a clue how to remove it -- any hints anyone can give me?

ħ

hbar · 21 Aug 2015, 22:42

OK, a good alcohol bath and a bit of sobering up with a brush cleared that gunk. Some thread cutting and light sanding later, I'm now glueing the acrylic to the top plates.

ħ

hbar · 21 Aug 2015, 23:37

Just couldn't wait to see the whole thing assembled:

: h1.jpg (75.33 KiB) Viewed 6877 times

: h2.jpg (88.59 KiB) Viewed 6877 times

I realize that I need smaller washers, otherwise there were no surprises.

It's such a shame that the matrix and controller will have to wait until next week at least.

ħ

hbar · 09 Sep 2015, 21:13

Finally got around to start on the wiring. Pictures of the right-hand unit:

: w1.jpg (90.08 KiB) Viewed 6817 times

: w2.jpg (119.01 KiB) Viewed 6817 times

pcaro · 09 Sep 2015, 21:40

It looks like so great hbar! I am not fan of DSA keycaps , but I can imagine it in SA profile ...

Ray · 10 Sep 2015, 10:57

First question: is none of that wire of the matrix insulated?

and the second: what's your planned layout on that tiny board? I am asking because you don't get both umlauts and basic punctuation on the alphalayer.
Is it even made for general purpose computing? With 6 Thumbkeys left after backspace and space, 6 mods for ctrl, alt, shift and windows (you could skip that mod actually) there's only two left for additional layers, which get quite tight on that small board.

Edit: forgot to mention, it's looking very cute

. And quite comfy

Muirium · 10 Sep 2015, 12:08

The matrix won't work if there are short circuits between the rows and columns. Hopefully you've got them insulated in some fashion.

hbar · 10 Sep 2015, 13:07

Mu: I'm aware of that, that's why there is lots of insulation (about 2mm of air) between crossing wires. I'm planning to route the (insulated) wires from here to the controller in such a way as to maintain that separation, rather than pushing the row and column layers closer together. The rule-of-the-thumb arcing threshold in dry air is 2kV for 2mm of separation, USB only delivers 5V, so that shouldn't be an issue.

I did play with a few methods of insulation but found all of them more trouble than they're worth (remember this is a prototype, nothing I build for somebody else). That said, since my acrylic is only 8mm thick, this causes the switch pins to get very close to the (conductive) bottom plate, therefore I'm planning to add an insulating sheet to that.

BTW, the row wires are made of the snipped-off cathode ends of the diodes.

Ray: My intention is to use a layout similar to what Suka uses on his "RedTilt", which combines some common keys (Return, Backspace, etc.) with modifiers. I have no idea what that's going to be like, ask me again in a few months' time. And, on your list of keys, replace Win (which I have no use for) with Compose Character (which I do use a lot, being used to real computers).

ħ

Ray · 10 Sep 2015, 14:59

Ah okay, you're not sure about the layout yet. That's fine, just be weary you might spend some time changing it again and again. E.G. on my board ( http://deskthority.net/workshop-f7/my-c ... 10410.html ) I am back to pinky shift, after having it on my left thumb. I really liked it on my left thumb, but I couldn't handle it plus a regularly used layer-shift on the same thumb. And I kind of don't get used to left thumb backspace, which is strange, since even mainstream ergos have that…
And if you have backspace combined with a mod, I find you need nice access to ctrl to kill whole words, since you can't hold the key until it kills it.
It is kind of fun to play around with