If your making a keyboard, your gonna need plenty of keys right? This is an easy enough task, you just need to plop a load of buttons onto the pcb and your done. I didnt want to waste i/o pins reading the status of all those buttons so I have made use of a chip that was made for the job. 

The TCA8418 chip can monitor up to 80 buttons, using minimal power and storing the key events in a nice little buffer so that you can process them whenever you are ready.

Using proper keyboard switches would be the ideal for the most usability, but they are not really pocket device friendly. there are smaller tactile switches, but once you start adding many multiples of buttons, the cost starts to shoot up.

If you have ever taken any small mobile device apart, you might have come across buttons that are stuck to the pcb under a plastic sheet. These are 'snap dome' switches, essentially taking a standard tactile switch and stripping it down to almost nothing. The pcb has 2 contacts on it, and there is a metal dome sat on top. pressing the dome deforms it, making it bridge the 2 contacts. The domes are arranged on an adhesive sheet, so they can be installed easily during manufacture. These custom sheets are actually really cheap, but the minimum order quantity on them is normally in the thousands, so prototypes and changes can get expensive.

It is of course possible to buy a bag of the domes, and make your own sheets with some tape and some steady hands... and the help of a 3d printed guide

lay some tape over the top, and then peel it off and transfer it to the pcb

Its not perfect sure, but its close enough to test :)

The other part to the puzzle is the strength of the domes. The ones I have used in the past were pretty 'strong' so when they were pressed and deformed, they had a loud and tactile click (hence the name clicker)

I wanted to get away from the noisy clicking (can you imagine a classroom full of clicky buttons being pressed) so I went for the lowest strength domes I could find. at a mere 55g activation force, you can almost trigger them by blowing on them!! I think there is some more experimentation needed to find the sweet spot of tactile feeling and minimal noise.

The quest continues....

I wanted to stick to the retro asthetic, so the screen being up the top left corner feels right. I considered a few options, but kept circling back round to e-ink.

Colour LCD

There are lots of really great little IPS screens with really impressive pixel densities, but they are generally designed for smart watches and the like, so they are super tiny (see TinyPi Pro ;) )

Monochrome LCD

I did try and hunt for some dot matrix lcd screens to really nail the retro aesthetic, but they seemed to be low resolution and pretty expensive. Also they can be a bit tricky with viewing angles and hard to see in low light

E-ink

So e-ink was the next on the hit list. I really love the way it looks, literally like printed paper, but the refresh rates are shocking. For displays with any colour on them, it can be up to 15 seconds to refresh the screen with lots of flashing, not really something that works well with a keyboard in a code editing situation. After some research, I found that the plain black and white screens could fully refresh much faster, and there was the option of fast or partial update which might make it usable...

The Curveball..

I had in my box of bits a Sharp Memory display, its a low power lcd that has a bit more of a reflective nature to it, and so has a bit more readability when the light isnt great. I hacked one onto my test boards so that i could compare and contrast

At first, things look nice!! (disregard that blob of damage on the screen) there are more pixels, and the refresh is instant, but it soon started to loose its shine. Firstly the extra pixels means that you have to use larger font to keep the same level of readability. While this is not the end of the world, it does kill off the advantage gained by the higher resolution. Also as i was moving the boards about, i noticed shocking contrast when your not looking directly at the device

Compared to the crisp e-ink, this was pretty bad. Coupled with the fact that i need 5v to get the screen to work, and the poor old Sharp Memory display went back into the parts bin

To make a project like this useful, you either need to spend a fortune adding all the sensors and 'radios' that someone might want/need, or give the mechanism to add them as needed.

Most systems use Qwiic or STEMMA connections to add additional sensors, but that really limits you to 2 pins, so it is relatively limited unless you use multiple plugs together (if you wanted something like SPI for example). This then opens up issues with getting plugs the wrong way round and things not working as planned.

I was hoping to push this towards the education sector, so using the Raspberry Pi style exposed header is open to damage by a bored student.

I therefore decided to switch it round, so that the more fragile header pins are on the cheap add on board, and the more robust socket is on the more expensive device. To make assembly simpler (cheaper) the header boards are mounted on the back of the board, using specially designed 'Bottom Entry' headers

The pattern of the sockets helps to make sure the add on board cant be inserted the wrong way, and limits any possible confusion. The case has a recessed corner with just a few holes as the only clue that there is something there

The add on boards only need simple headers to enable connection, giving access to power and ground, and 16 i/o pins

using thru hole headers means its more robust, but also could result in the weak point still being the more expensive device rather than the cheaper add on board