I've been using an Ergodox for about a month now and I love the idea of a split keyboard/spacebar and considered picking up one to use at work. But I took a vacation for a week and came back to the ol' 100% and had nearly forgotten how to use it. At that moment I decided that I needed a standard layout in at least one of the areas where I type or I'll forget how to use any other keyboard and make myself look like a tard.

If I were to make a keyboard for work, it would need the function keys and it would need the number pad (I don't really want to deal with layers at work). This is what I've come up with.

http://www.keyboard-layout-editor.com/#/gists/52c8a7a87553cbd648e59f30058ccbb2

I'd like to hear what you guys think of the idea and possibly if anyone else would be interested in it.

Some assorted semi-random notes:

• Not sure if I want to design a PCB for this or if I want to hand-wire it.
• Spacebars are both 2.75u, but that was more of a guess and I'm not sure how readily available that length cap is.
• Both halves are the same length at 11.75 and I think I want to keep both halves at the same width, especially for tenting reasons.

Either way, the easiest way is to use a different controller on each half. It's a poor design approach, but surely saves you some time with the firmware.

That's what I'd planned on doing. I'm not sure I have many other options since I'll have so many keys on each half. Unless I do a wire for each row or something, which could look pretty nice...

GMK makes 2u convex spacebars, other than that, you can use two 2.75u right shifts as spacebars.

This is good to know. I might add a function key to the side of each spacebar to allow the use of the 2u convex bars.

Thanks for the input! Work has kinda slowed down for a bit, so I may look into it a little deeper today.

You can also get convex XDA spacebars in 1.75u, 2u, 2.25u, and 2.75u.

Excellent! I love having options!

Well, I took last night and most of today using Ruiqimao's KiCad guide. This probably wasn't the great "tutorial" project, but it seems to have turned out well enough.

I'm not entirely sure how the USB passthrough is supposed to work. I may look at the ergodox PCBs to see how that's implemented and try to guess how that works (unless someone has a guide for me about how they work). But that's a problem for later me.

It is recommend to avoid vias on the data line (USB) traces and to keep them as close as possible as the USB is a differential signal.

Yeah, this was a major learning experience on how to route lines on a PCB. I'm probably going to redo all the tracking and pin-out completely at some time because of how garbage it is.

By close as possible, are you talking about "to the controller" or "to each other"?

Also, your traces seem extremely thin, how many mils are they?

Honestly, it's whatever the default is for KiCad. Looking at the program, it looks like 0.25mm (9.84mil)?

The pin out was just really messy to route and I could have done it a lot better, but I was going into it blind. Now I know what to expect with regards to location of physical pins on the chip and I think that can help my routing immensely.

I will look at routing the data lines closer to each other next time. That's something that I didn't really know about (like much of this haha).

The guide I was using was pretty basic, but was for keyboard PCBs. I certainly welcome the opinion of those more experienced than me though. Maybe the prototype phase will expose an issue with using that thin of a trace.

Between your advice and some discussion on the r/MK Discord, I've got a much better idea of how this is going to work next run I go through and I think that will help.

should I be looking into trying to use a pre-assembled microcontroller setup like a ProMicro?

You should go for whatever method you feel more comfortable with but, if you abide by the general design recommendations, designing your own board from scratch won't be hard.
Read through The Living PCB Design Thread, it won't take much, and don't be afraid to ask questions for the gentlemen there are friendly and helpful.

I have no experience in Surface Mount Soldering, I feel like there are a lot of opportunities for me to mess this up on a prototype run.

With proper tools, SMT can actually be quite easy.
Prototypes are made to mess around and with them. Unfortunately they cost money, but learning from your own mistakes is unvaluable. I talk from experience.

- if you can help it, don't use 3u space bars, they're v difficult to find

Yeah, that's why I'm trying to stick with 2.75 for the time being. 2.25 looks to also be a relatively easy option as well.

- I went with an off-shelf teensy controller for my first pass - it saved me a lot of effort and time and limited the amount of things that could go wrong, though actually caused the biggest headache in debugging (pin D6 is connected to ground!). I've only every soldered one smd controller and it was a pain for me. Having said both of those two points, I kinda wish I had gone the full smd route - it's neater and you're less constrained by the premade controller geometry, though I may still be debugging issues on the first prototype.

Yeah, I'll probably stick with the smd route for now because my board is already a little constrained. Go big or go home, right? haha

- I'm digging the split and the left hand numpad - I would go one step further and ortho-linearise the alphas. I did that with my build and after two weeks of use I feel how bad the staggered arrangement of standard keyboards is.

I considered ortho, but I'm worried that if I vary too much from a standard layout then it's going to mess with my ability to use a standard keyboard. Having taken a week break from work (using my ergodox at home), I went back and it took me way too long to readjust to using a standard layout. If I go full hard on an ortho layout at work and at home, I'm concerned with permanently destroying my ability to type on another keyboard.

- I'm definitely glad I went the PCB route - it's a lot of effort in the design, and certainly more expensive, but once the PCB is there you can so easily make multiples. Indeed with minimum order quantities you'll have a few lying around to make a keyboard for your family and friends!

I'm really excited about the idea of making a pcb of my own. I design mechanical things for a living, but I've kinda stayed away from electrical because it's basically magic to me.

I still really need to figure out connection between the two boards or if I need to connect them at all. If I press Alt on my ergodox and 3 on my separate numpad, I get a ♥. So a Shift, Ctrl, etc. should work just the same, right? If I don't have to connect the two together to get Shift functionality then I don't think I need them to talk to each other. If I want to be able to add layers, that's probably a different story.

Some thoughts.

First, you can use software-side modifiers even when the two halves don't know about each other, that's OS' job. What you wouldn't be to do is keyboard-side stuff like layers.

Second, you don't have to use TRRS to connect the halves, there are other kinds of interconnection. I've seen SATA, HDMI, and obviously networking cables.

Third, is there any reason why you have a right-handed numpad on the left side? The standard design is such that the strongest fingers press the most common digits (0, 1) and there are vertical keys for the pinkie. Using it with the left hand breaks this, hence I recommend a mirrored physical layout.

First, you can use software-side modifiers even when the two halves don't know about each other, that's OS' job. What you wouldn't be to do is keyboard-side stuff like layers.

That's what I was hoping for. For a board of this size, I don't think there is much of a need for layers at the moment. That may be an upgrade if there is positive reception to an IC.

Second, you don't have to use TRRS to connect the halves, there are other kinds of interconnection. I've seen SATA, HDMI, and obviously networking cables.

Yeah, I understand that. I just need to get a better grasp of the I2C stuff and the limitations of each type of connector through that protocol. Main thing it that is should be an easy cable to find. HDMI and TRRS are that, but SATA a bit less so (outside of the enthusiast realm).

Third, is there any reason why you have a right-handed numpad on the left side? The standard design is such that the strongest fingers press the most common digits (0, 1) and there are vertical keys for the pinkie. Using it with the left hand breaks this, hence I recommend a mirrored physical layout.

Honestly, it was just taking the group of keys and moving it to the side. I didn't put much thought past that. But you bring up a really good point. I'm trying to think of "left hand" number pads in the real world though and it seems like every time I come across one, it throws me way off. It's the whole brain mapping that MNLOP brought up. I'll have to think on that one a little.


Thank you for the feedback! I appreciate it very much!

You could easily make the numpad ambidextrous by making it out of 1u keys and add 2u keys on both sides. Like so.

I wonder how difficult it would be to path the pcb to allow for either rh or lh layout. It would probably force plate mounted stabs, right?

Careful with the placements if you do go for a multi-option layout - I messed up but putting resistors/diodes too close to the stab system needing to grind out a bit of the solder which is a terrible thing to be doing. That's even without considering overlapping drillings for switches and stabs.

With hindsight I'd have done this. I make no promises that it's a better layout though!



TalkingTree has just posted an impressive custom board with a bunch of placement options, def speak to him if you want to have a flexible PCB - this complication is way over my head. Don't forget to project the placements to the switchplate!

All traces are 12mil, except VCC at 20mil

Feels familiar.

Are the components on the bottom side of the PCB?

Quote from: TalkingTree on Fri, 15 December 2017, 15:22:31

Are the components on the bottom side of the PCB?

Yeah, it seems people usually flip the switches to the other side of the board in KiCad? I did the opposite so I didn't have to mirror the switch layout.

No idea, I don't use KiCad, anyway having the electronic components on the bottomside is always better, in my opinion, because you have access to them even with a plate. Learnt that the hard way.

Yeah, they're on the opposite side of where the switches are mounted, so I ought to be good?

That's perfect.

This looks really interesting! What is the current status of this project?

Currently on hold. Things yet to do:
- Redesign the left hand PCB now that I have better knowledge of routing and whatnot
- Sanity check the design
- Build a prototype

For a while, I was considering using ProMicros as the controllers for each half, but I'm getting a build that has a bunch of SMD components on it as some sort of practice. Then I can make a decision from there.

I'm still really excited about this project and I'm a little upset with myself that I've put it on hold for so long, but it's definitely on my "Please, just do this" list.

Have you looked into keeb.io's BFO-9000? I just ordered the pcbs and am going to run this for a layout initially,will probably change it up over time though.

Bump!

I would really like to know if you got around to building this keyboard.

I am making a really similar split 110% keyboard. Using two Teensy 2.0++ connected via I2C. I am left-handed so I need the number pad to be on the right. I've finished the hardware and I'm just doing tweaks on the QMK firmware at the moment.

http://www.keyboard-layout-editor.com/#/gists/efdc60f745d231f7c2c590ed5ef5ad48

I haven't, yet! I decided to work on something a little smaller first (the numpad in my signature) before I come back to this. Though, I would be extremely interested in seeing how you set up your QMK and how your did your pcb tracing. I've had a really hard time wrapping my head around the I2C bits and how that works with a board as large as this.

I built it as a prototype using a unique method of drilling holes in stripboard (veroboard) and installing hot-swap sockets. Due to the number of switches I needed to use two Teensy 2.0++ clones.

I set up QMK in exactly the same manner as I did with my other split keyboards that use Pro Micros, but upon compiling, I found out that standard QMK does not support the Teensy 2.0++ for serial communication. It only supports ATmega32U4 controllers...

After modifying the hardware to use I2C, I still can't get the two sides to communicate to each other. Both sides work when plugged in individually, but there is no communication to the slave side. Anyway, didn't mean to hijack this thread. I'm just trying to find people who have successfully made full-sized split keyboards.