Curved Split Keyboard Prototype

[Zekromtor]

The thing that could have been pretty, had I wired it nicely.
Yep, that's felt covering the key caps. Don't knock it til you try it.


The layout. All numbers on left hand = good for coding.


I tried a lot of curves/steps.


Ultimately I'm not incredibly pleased by this design. The angles of the bottom two rows didn't work for me and I'm back on my split kinesis. It's a decent way to make a curved board if that's what you're after though. I'm going to try something a bit different for the next one.

[Findecanor]

Have you tried a keymap with the home row one step up?

BTW. I have a theory that a non-uniform curve would work the best, with the lower rows being more curved inwards than the higher rows.

[Zekromtor]

That would be more like the Kinesis layout, and for whatever reason my fingers don't like making use of the bottom row much in that case. I mixup the bottom two rows way too often and it's a mess. I specifically moved the home row down to avoid that issue, with only one step to reach the bottom row. Reaching the top row was still no issue.

I'm just a little disheartened with curvature as a means of shrinking the Y axis in general, and I'm going to do a more traditional layout with no curve, a small step, and keys that are much shorter on the Y. Not as exciting, but should work.

[jacobolus]

Most of what I have to say on this topic I said in several posts over here: http://geekhack.org/index.php?topic=60394

Long story short: you need to think about which joints in your finger are being used to reach the key and which joints are being used to press the key down. Then (1) locate each switch in 3-d space so that moving to the top of the key is as easy as possible, but also importantly (2) orient the switch in 3-d such that pressing the key is as easy as possible. Your design above tries to optimize for (1) but completely ignores (2), and the predictable result is that it’s not all that easy to type on.

The keyboard currently available that does this best (though still far from perfectly) is the Maltron. Even better was the prototype DataStealth keyboard from the mid-1990s, which never made it into production.

Improvements you can make:
* Add some stagger between columns to account for different-length fingers
* Add some height variability between columns, again to account for different length fingers
* Orient most of the switches to face approximately the same direction, because when you reach for different rows with the same finger you still end up pressing the key using the same main joint at the base of the finger
* If you want more compression in the "Y" direction, use smaller or rectangle-shaped keycaps (since you’re using Alps switches this is a bit easier, or you can go even further with Cherry ML, etc.)
* Remove at least one row of switches: right now you have way more keys than necessary and some require substantial reaching.

[Zekromtor]

What do you currently type on? And are you a hand floater or do you type with your palm planted? What works best depends greatly on that, and I'd agree that a classic typist that floats their hands will prefer a keyboard with keys that are more uniformly orientated unquestionably.

I'm not convinced yet that extreme Y curvature is or isn't the way to go for someone with a planted palm, but even in the case of this keyboard that was aimed to test that, the only keys that were difficult to press were the bottom-most rows, likely due to a bad palm rest angle. To know for sure, my next board will still be a 3D design, this time with an aggressive step between rows and more uniform switch orientations.

I see no harm in having extra rows on top that are closer to reach than a typical F key row and not necessary to hit during normal typing. Plus, due to the curvature, the only key that required shifting of the hand and maybe should have been omitted was in the F1/F6 position. I didn't find it to be more difficult to press a key with a lancing motion with the finger as opposed to the more typical downward press, and using different muscles/motions for different rows may help with stress related injuries.

[jacobolus]

Currently I split my time typing on a laptop keyboard, a standard Model M, and a couple of standard Alps keyboards. I spent some time typing on an Ergodox but I’m not a huge fan of the thumb section. I’ve also spent a tiny bit of time typing on standardish split keyboards, a Maltron, a Kinesis Advantage, a μTRON, etc. Hopefully sometime soon I can switch to mostly typing on my own custom keyboard. I’m still working on building it though. I am mostly a hand floater, though I often rest my elbows or forearms on something, and don’t mind leaving my palms on a palmrest if it’s sufficiently high (most of them are much too low).

Don’t get me wrong: please experiment and make whatever layout suits your own preferences. I can’t argue with what works / is comfortable for you.

I just meant to emphasize that it’s worth *thinking about* what finger joints / movements you use to press the keys, and try to optimize switch orientation for comfort and strength. If a “lancing”/stabbing motion is comfortable, then go for it.

One other thing you might for the very nearest row is to tilt it up to a full 70–90° angle from the next row over, and press those keys with a “squeezing” type motion. I suspect this motion would work better for modifier keys than standard letters, but it might be worth a shot.

[Zekromtor]

I agree with all you've said. About the bottom row aggressive tilting as well, I'd like to try that. It will require a substantially higher palm rest, and it already needs some significant changes in that department. Do you know anyone currently using one of those datastealth type keyboards where you're squeezing keys tucked almost beneath your palm?

[jacobolus]

I don’t. I want to build at least one keyboard like that though, to try it.

[Zekromtor]

Well I'm giving this keyboard another try. The bottom rows now can be pressed properly now that I have an extra 3/4" pad on the palm rest, but still going to require the usual tweaking to get the palm rest, tenting, etc dialed in to where my hands like it.

I just shouldn't have made such a snap judgement without giving it time to test. I'm testing these ducky green clicky alps switches as much as the keyboard shape, and the high activation point takes some getting used to but I think I'm starting to like it. Very crisp, very tactile, and I no longer need the audible cue that I had play for each keystroke like I had with the Kinesis.

I'll say that the "lancing" motion when activating the top row keys is not the greatest for rapid strokes, but for me it beats having to shift layers. Still planning to build another design to see which I prefer.

[Zekromtor]

I'm starting to really love these switches. There's never any mental pause where you're wondering, "did I hit that sequence in the right order, better look." and it results in faster typing.

[jacobolus]

The Ducky green XM switches? Have you tried other Alps switches before? [Yes, the strong tactile feedback and high actuation point of Alps switches are very nice in general.]

[Zekromtor]

Yes they are the XMs off a 1087, hard to tell in the overexposed pic, but they are green. I could go on a rant about that keyboard... how they kinda screwed up making it so cheap when they could have put some more money into solid components and marketed the alps clone keyboard more properly as a high end alternative to cherry... oh well.

I have a bag of Matias clickies which I'll use for my next project, but I haven't used any other alps. I know there are some cool hybrids you can make that people like.

I'm not 100% sold on the high activation point yet. I might prefer a legit buckling spring (I cringe thinking about the keyboards from my youth that were thrown out). I still think I'll be choosing this style switch over any cherries, though.

[Zekromtor]

Welp, I'm learning why the 1087 was rightly considered a cheaply built keyboard. These switches are not very reliable. Individual switches like to start to report multiple hits, and it's not from the common mechanical bounce that can be solved via firmware. I don't know if this is a problem common to all of these alps style switches, ie if Matias switches suffer from this as well. I'd replace all of them with Matias, but they're like 3-4 thou too large for the holes that were purpose made for these damn duckies. Time for a new build

[Laser]

I have some white ALPS switches looking for a compatible PCB, perhaps I could put a Ducky 1087 XM PCB to good use
(if PCB holes are compatible, and if you are willing to part with it for some cash)

[Zekromtor]

Sorry it was pretty disgusting (used) and got thrown in the trash.

[Laser]

Pity,
Thanks for answering!

[Zekromtor]

Above is the layout I moved to after months of tweaks. I ended up much preferring having to shift my hand ever so slightly up to use the arrow keys than to use secondary layer. The screen lock is true to its name, locking my mouse (tablet) cursor inside the current computer's screen to avoid automatically hopping to any of its many neighbors.

I replaced all switches with Matias with good results. I'm not overly infatuated with alps switches anymore. In spite of their incredible tactile feel, they leave something to be desired for gaming, where my new theory is that precision timing is more important than the .5ms (estimate) speed increase you get from having a higher actuation point.

I can see now why cherry reds are the preferred choice for gamers, but honestly I think a bottom actuated switch would be preferable so that there is no question in the player's mind of when the key is actuated, and to make double taps faster. This applies to timing games like Osu! as well as FPS games - you're far more likely to be timing shots based on you and your enemy's strafe motion than to be reactionarily pressing a key when rounding a corner. When I have more time I'll be frankensteining my own bottom actuating mechanicals with the help of some microswitches.

[vvp]

How do you implement screen lock? I do not know how to do it without special application running on the machine (as opposed to something running in the keyboard).

If you want quick response time in games it may be useful to change the debouncing algorithm. Detect press at the first signal that a key is pressed. Then do not detect release at least for a given amount of time (during which no press can be detected). This way you can speed up key detection by at least 5 ms (debounce time for cherry mx). I may do this sometimes but it is one of least important things for me.

[Zekromtor]

The screen lock is via custom software running in the background of windows that forks input to the appropriate network computer.

The keyboard's firmware does already send an immediate signal when a key is pressed and checks for that signal more than once per millisecond, regardless of whether another key is being pressed. I was pretty disgusted when I saw the methods employed by popular firmware for handling debouncing. Using sleep as a means of debounce control is sloppy coding. Each key should have its own debounce timer that does not interfere with the response time of other keys when it is employed.

[vvp]

Katy is based on chrisandreae's firmware. It does not try to detect key press as soon as possible. I'll probably change that in some far future. Good you already have it better

[Zekromtor]

6 months later, and I'm liking these Matias clickies more. A blast of air fixes the issue when they start registering double clicks, and that hasn't been needed in a while.

I _love_ having the numbers all on the left hand. It is just awesome for coding. Right now I couldn't be any happier with this keyboard...except perhaps if I actually had some games worth playing and some time to play them. I'll settle for reading fantasy books, and doing typing tests in the mean time

[vvp]

Can you post a picture of the latest version you use?
I do not mean the layout. I mean the keyboard shape.

[Zekromtor]

It's still the same exact shape as the first picture up there, it just uses Matias switches now instead of those green duckies.