A keyboard with no name

[Nlight]

On the frist part of the journey,
I was looking at all the parts,
there was caps, switches and controllers
there was arcs, clusters and staggers.
The first thing I met, was a switch with a click
And a plate with no keys.

I've been through the build of a board with no name,
It felt good to be out of the rain.
In the matrix, you can't remember pins name,
Cause there are so much to give you much pain.
La, la, la la la la, la la la, la, la
La, la, la la la la, la la la, la, la

Ok, more seriously,

2018-08-02 :
The project has changed : new bean bag chair -> new ergonomics -> new requirements -> project evolution

Primary usage : programing, typing
Secondary usage : gaming ? If I can, this will be a bonus.

here are the characteristics of the project :

• full split keyboard, linked through HDMI cable, for sake  simplicity.
• 44 keys
• arc of 4 thumb keys plus 6*3 keys on each side
• fully 3d printed parts. As I own a 3d printer, it would be stupid not to use it
• mostly (if not all) switches are MX Browns
• custom printed keycaps, modified DSS profile (or DSA, don't remember)
• column staggered, adjusted to my hand
• wrist rest inspired from Razer OrbWeaver (yet to design)
• firmware : QMK

current status : case modeling and designing



2018-06-19 :

More

I make this post for my project, that has no name, well maybe it does now... Anyway, as I am new to this, i would to resume all I think I know on the process, and expose my goal (at least rougly at first)

Goal :
- Build a "split" keyboard, not a true split mind you cause it has to sit confortably on my lap while I am in a bean bag chair, but a split design nonetheless, plus It will be easier to do than a full split.
- the ergonomics will look a lot like the Diverge 3.0 with less keys, so : column staggered, thumb arc of 3 keys top, no F key row, no numpad, probably no arrows, and a wristrest.

- the keycaps : right now I hope to 3D print them, especially as they wont be standard size (16.5x16mm), if prototyping is not conclusive, well, revert to standard keycaps.
- the switches : Cherry MX Brown, seems pretty standard (edit : ok, I also orderd some reds, and a few silvers)
- Plate mounted on aluminum sheet
- hand wired : PCB is too expansive for the functional gain it offers.
- Diodes : 1N4148
- Controler : Teensy 2.0

Now the process, and please correct me if I'm wrong :
- make a matrix of keys : each diodes connect each switch to its row and are NOT mounted in series, in order to avoid successive voltage drop
- on the controller, each column needs a pin, each row needs one too
- make a Firmware, for ease of use, use EasyAVR, or Keyboard Firmware Builder as both are graphical and allow to set custom matrices
- flash the firmware onto the controler, well I need more info on how to do that (on windows), I guess through Teensy Loader, but I have not yet gathered info on this aspect.

There will more likely be more questions once I'll have done all of this, which is not for now (some furnitures to build and well, daily life)

[MatchstickMan]

I'm interested. Do you have a KLE setup of what you're looking to do?

[Nlight]

Unfortunately no, as KLE, as good as it is, is a hassle to deal with curves. But here is a first draw, some keys I think are very badly placed, like Ctrl, more work has to be done but the rough idea is here (in an ugly way, I insist)

[T14]

Cool ideas. I especially like how you stack the far left cluster of keys to accommodate the angled alpha group. It is very elegant 

[Nlight]

Thanks, I think I will keep them where they are, however, due to custom shape keycap, building stabilizer for large keycap will undoubtedly turn to be a hell to make, So I will revize the form factor of these keys, unfortunately...

At first I dissmissed inward columns next to G and H, but now, I think this not too much of a change to be so destabilizing that I can't place in there ENTER, Backspace and DEL. Also, I think I will get rid of PgUp and PgDwn and include them into a layer. the layer in question will be momentarily turned on with one left hand thumb key, it will host vim like arrows (H:Left, J:Down, K:Up, L:right), Y will be HOME, U:PgDn, I:PgUp, O:End, that way, shift can still be pressed with the left pinky.

Here is the work in progress :


As you can see thumb key are now four, they have been lowerd and while it may seems weird, they should theoretically fall under my thumbs joint (only the lower half of the key)

On the far right side, I think I will add keys to turn on some layers.

I also orderd some MX reds, and silvers for some keys, for example for the left hand thumb arc, from left to right : Silver, Brown, Brown, Red. If the Ctrl key stays where it is, a silver would be nice

[Nlight]

ok, here are some lil'update :

I printed a test support to get staggering right to my fingers :

 and also printed some caps, with crappy resolution cause it just is for test purposes, only one has been printed with .19mm thickness.


Once I had the column stagger as I wanted (or seemd to, as the test plate did not allow vertical stabily), I made another test plate, with better fit and fixed slots :


And the stagger placement and cap dimensions feel right. However when printing the caps I tried the DSS profile - as this is what I had on the M$ Natural 4000 and loved it - and well, the inclined top of the caps came out very crappy due to a too strongly attached support (caps were printed with their top as bottom of the pint for the stem to come out good enough)


So, my next steps will be to test the thumb arc, and probably devise a better way to print the inclined top caps in two parts that will have to be assembled, if this is a mess, a flat top with slight inner vertical groove is fine.

Only the finely printed cap does not have its stem interfering with the switch housing on the course to the bottom. Most other will feel ok, but needs to be pressed in a straight manner, but this will be fixed in the final stage. Also, as the caps are printed, the stem can not fully enclose the switch shaft without being subject to some dimensional errors which are hard to fix afterwards, I opted for  the shown stem design : easy and allow good precision while still having enough strength to endure my normal usage for quite a time before eventually breaking. But this design choice lead me to abandon the custom shapes for the TAB, Shift and Control key, they would definitely not endure as well the stress provoked by not being pressed in their center, just above their stem.

[Nlight]

I made yet some other modifications.

For years, as I do programming, I have used a custom layout that heavily relies on the AltGr modifier to output common characters right under my fingertips like ("[]<>{}.+/*-%_|&) to name some. But as I lacked other modifiers that would unlikely be used as shortcut by programs, I made the numerical row to respond as a QWERTY keyboard (meaning you don't have to press shift in order to ouput a numerical value).

Well, now that I have access to layers, why keep this row ? I removed it.

After a simple test, I saw that an inverted Z row DSS profile (-16° instead space row DSS 8°) allowed simpler placement of thumb keys while remaining easy to press and comfortable positions.

DSS profile is as such :


current physical layout looks like this :


I have printed two parts test keys, I still have to acetone glue it to see if it is good enough joint, but the look and feel is nice enough.

I am currently printing a test case with inclination, and thumb keys. I will upload a photo when it's done.

[Nlight]

Ok, my thumb arc need to be tuned a bit, it needs to be less accentuated than it is right now


here are the immensley difficult two parts keycap design )

These parts are jut fused together with "ABS Glue" (just ABS scrap dissolved in acetone till the texture is white glue like), works pretty well.
Here is what they look like now, you can easily spot them

Surprisingly enough, while these new caps looks a lot better, I kinda prefer their lesser precise versions, especially the Q row, they have a spot that spouse your fingertips just right. With higher thickness print, each row has a very specific feeling to them that clearly identifies them without looking, higher precision ones feel more bland and are less ditinguishable...
As you can see I reversed the Q row caps, as it feels a lot better like this now that there is no numeric row. Here is what the current rofile roughly looks :