First One Done, 8 To Go

A Single-Channel MIDI Instrument

IMG_0250 2.jpg

IMG_0252 2.jpg


The many steps before that…

Prototyping and fabricating week!

The goal here is to make it as thin as possible in all dimensions, width, height, depth.

I started off by taking apart one of the slides to see what it’s made of and how I can make it thinner.


Turns out there is no way to make it thinner except if I make the slide itself a circuit board…

But, that’s not happening! Too risky and no parts.


I then tried different ways to stack all the components together to find the best solution.


What’s the best way to package them all?


After trying a few variations, I made a sketch of the ideal scenario.


Once I had the idea, I tried out with the cardboard prototype to get a sense of scale (the PCB here is used to represent the board + wires).

IMG_0172 2.jpg

I was quite happy with how thin I got it to be.

IMG_0173 2.jpg

I then tried milling with leftover materials I had to get a sense of the depth of the pocket for the slides and boards.


Once I had the pockets I then worked on the wiring.


And this is the compilation of everything I tried before finalising the design and parts.

IMG_0267 2.jpg

Finally I put all the best pieces together …

IMG_0260 2.jpg

Once put together, I was satisfied with the outcome.

Only 7 mm (around 5/16″) gap between top and bottom pieces! And the width x height is 10 x 2 cm.

IMG_0250 2.jpg

IMG_0252 2.jpg


Yay, this worked:

Sadly, the pictures are not what they seem…

This is the reality


Once I decided on the optimal measurements and fixed the files, I started shopping on McMaster-Carr. SO MANY OPTIONS! SO OVERWHELMING! I must have spent around two hours on this website.

Screen Shot 2563-04-21 at 17.58.25.png

Waiting for parts to arrive. These are the expected dates:

  • Soldering paste: April 24 // I have this problem where the very tip of my soldering iron doesn’t heat up (but it’s new and not oxidised, what’s wrong?)
  • Boards: April 24
  • Screws and threaded inserts: April 21
  • Black rubber feet: April 28

To do next:

  • Assemble parts this weekend
  • Screen Shot 2563-04-21 at 23.31.01.png
  • Program the ‘mass tuning-in’ software
  • IMG_9588.JPG
  • Playtest with people who live around Downtown Brooklyn

Random questions (note to self):

  • Milling pocket, CAD? Illustrator? Is there a reason they move? Re-home?
  • Best practices?
  • Add channels on top?

More stuff…


Eagle SCH and BRD


^ (Project using Beetle board instead of custom PCB)

Custom PCM Eagle files and PDF:


  • Do we need to connect GND and VDD on each side of the chip together?
  • How much warning is allowed regarding the wire touching (the pins on the chip already has warning)
  • What are the 5×2 pads on the sides of the micro USB? Are they to lock the cable? Or should they be connected to something?

NOTE: the 3 headers are to connect to the slide // I tried to model the slide but doing it this way saves space because the USB connector is at the same spot as the pins on the slide

Screen Shot 2563-04-18 at 18.49.56.png

(Maybe I want to make cut-corner edges to make them fitted to the wood/acrylic underneath)

This is the sketch of each layer of the device. The PCB in this case will replace the Beetle board (at the bottom).


Once assembled, the device will look like this







OSC MIDI + Waiting for Parts + Experiment


MIDI –> OSC –> Digital Ocean Server –> MIDI –> Ableton


Even though this worked, to not have to do the wifi set up for my case, it may be better to use webmidi directly i.e. slides act as standalone MIDI instrument like a keyboard.

2. Waiting for parts

Beetle boards x 8. The board is $3.95 each!

Beetle - The Smallest Arduino Board with bluetooth 4.0

3. Sidetrack experiment: slides control movement

Slides –> MIDI

This week I focused on getting the slides to behave as I want on screen i.e. move smoothly, no glitches or jumpes, no delays.


  1. The linear slides were acting like logarithmic slides (this took me almost the whole day to solve, the problem was the breadboard)

    How it should behave…
    Screen Shot 2563-04-03 at 10.00.15

    How it was behaving…
    Screen Shot 2563-04-03 at 10.01.53

    Once I realized the issue, I had a funeral for the breadboard

  2. It was so glitchy! All the values kept jumping

    EXPERIMENT 1: To solve this I tried the Kalman filter. However, the side effect of the filter is a ‘lag’ as the value jumps from one to the next because the filter tries to smoothen the values.
    (Change 100 to 127 to fit MIDI scale)Screen Shot 2563-04-07 at 18.18.55.png

    EXPERIMENT 2: Track the magnitude of change in the sensor value. If the change is greater than 2, count it as change, otherwise don’t. Since the steps of the digital slides were quite few (less than 127), this worked fine and the results of the movements were smooth.

    Screen Shot 2563-04-07 at 18.19.29.png

    EXPERIMENT 3: Although the movement transition between physical slide to digital slide is now smooth, sometimes when still, the slides are still glitching. This had to be a wiring effect so I built fitted box prototype of the slides to sit on.




So I got the desired movements in the end:

Next week: focus on the programming

Links to get started on:


Homemade Hardware Update

Prototyping board

Beetle - The Smallest Arduino Board with bluetooth 4.0

Chip ATMega32U4 (big pins)

ATMEGA32U4RC-AUR Microchip Technology | ATMEGA32U4RC-AURCT-ND DigiKey Electronics


WMYCONGCONG 10 PCS 10K Ohm Slide Potentiometer Single Linear 10K Electronic Potentiometer, 88mm

I had a problem with the slides because they turned out to be logarithmic instead of linear. I am still trying to figure out the algorithm to fix this.

What it looks like …

Screen Shot 2563-04-03 at 10.01.53

What it’s supposed to look like …

Screen Shot 2563-04-03 at 10.00.15

Other than fabricating the board. I’ll also be working on the synchronous web instrument next week.

Notes for parts

ATSAMD11C14 vs ATMega32U4?

Both native USB

ATSAMD11C14 (32 bit): less pins, low memory, faster, easier to solder, need J-link to program

ATMega32U4: more pins, higher memory, slower, difficult to solder, directly program
Bigger pins ATMega

Other parts and lead free solder @digikey

Tactile Fantasy, Unlearning the Keyboard






I started off this project with a vague objective to explore the relationship between the sense of touch on an HID controller and its corresponding visuals and sounds on a computer. The questions I had starting off the project were: must keyboards be so flat and modular? Must the interactions be so 1:1? Must each keystroke always be a command?

What if all these expectations were more loose?

Specifically, the questions I was trying to answer were:

  • How do I make the experience more of an exploration of the control as opposed to a control which I have to figure out how to use?
  • How do I take the sensations on the controller and create different modes of interactions on the screen / the speakers? e.g. unison, call and response, random etc.


PHASE I: Ideation – a keyboard with an internal life



PHASE II: Prototyping – tactilefantasy.js



Inspiration (reading)

Software testing

During this step I play-tested with 6-8 people. For me personally, the prototype was already working as an ‘explorative experience’. However, only 1 out of the 6-8 people agreed. The others were mostly confused. The key takeaways for me at this stage were:

  • Positions of the buttons matter more than their shapes/heights
  • There is a very strong need for any kind of immediate feedback after pressing a button

Version 1 ‘visual, sound, text’ (fixed keys)

Use the following keys to control

[‘a’, ‘s’, ‘d’, ‘f’]; // words
[‘z’, ‘x’, ‘c’, ‘v’]; // sounds
[‘q’, ‘w’, ‘e’, ‘r’]; // animation
[‘1’, ‘2’, ‘3’, ‘4’]; // brightness

Version 2 ‘visual, sound, text’ (random keys)

Use the following keys to control

[‘1’, ‘2’, ‘3’, ‘4’, ‘q’, ‘w’, ‘e’, ‘r’, ‘a’, ‘s’, ‘d’, ‘f’, ‘z’, ‘x’, ‘c’, ‘v’];

Version 3 ‘text, sound’

Use the following keys to control

[‘a’, ‘s’, ‘d’, ‘f’]; // words
[‘z’, ‘x’, ‘c’, ‘v’]; // sounds

PHASE III: Final fabrication + experiments


Applying the learning from prototyping

  • Increase size of buttons – the smaller ones were almost like pins poking into fingers.
  • Make positions less rigid and more freeform – make it say to users “be more playful! stop using the prefrontal cortex! stop trying to figure out how it works!”
  • Hide everything underneath – users were distracted by wires

Knowing how the wiring works already, I entered this phase by working on the enclosure. I wanted to learn how to powder coat so I got the steel conduit boxes to try. (more on, thank you Ben!)

The design, layer by layer


Prepare all the tools: powder coating gun + machine, powder, clamps + pliers, wire for hanging


Started off by heating the boxes in order to get rid of the oils. After heating, use scotch brite to scrub the surfaces.

IMG_0895 2

Hang the clean boxes on the rod


Charge the rod (and the boxes) + spray! … did not get a photo of spraying.


Put them in the oven again to let the powder sit. And viola!!! (Ben and I did this a few times because the new machine did not work and the powder did not stick to the surface, we ended up using the old gun).


Now time to make the top part. I started measuring the box and tried cutting cardboard pieces on the laser cutter to determine the actual size.


After finding out the size, I then designed the acrylic cover and the PCB underneath.

I ended up cutting both of these on the mill because the laser cutter was closed due to COVID-19. This was my first time milling acrylic! (Also, first time making an Eagle part)

Screen Shot 2563-03-16 at 18.09.36.png

Screen Shot 2563-03-24 at 14.34.19.png

Screen Shot 2563-03-25 at 11.07.40.png

The shape makes the copper look more like rose gold 😀


I was happy with how precise and fitted these buttons were on the surface. This saved me a lot of time – I didn’t have to wire much after this.


Of course, given they’re the same file, the acrylic was also perfectly fit (for most parts! I didn’t put the acrylic sheet on the machine properly and the bottom was cut off! luckily no bits were broken).


After that I wired it to my old perf board which I used for the prototype. Had to add the legos inside there to support the panel. I didn’t think through well enough about how to attach everything. Also, the holes on the pcb I made to attach to the box were too small and I didn’t know how to increase the size on Eagle, fix next time!)


After wiring everything, I finally plugged it on to the computer and tested the buttons with the serial monitor. Of course this took a few rounds of debugging and finding the right button / pins.

Here are the pin layouts and mapped keys! I color-coded them according to their heights in order to program easily after this.

Screen Shot 2563-03-24 at 10.30.17.png


Experiments with softwares!

I got sidetracked into texts, fancy visuals and rhythms during prototyping stage. Looking back at my original objective, I started exploring with the initial idea again.

First off, creating the same visual on the screen (using Bang object on maxmsp)

I then tried to map the visuals according to the height of the buttons. The lower, the darker.

Other combinations:

Screen Shot 2563-03-24 at 00.55.25.png

I then took the same circle pattern and built a mosaic-like visual. This was very satisfying even without sound. To visualize the tactile effect on a grand scale.

Inspired by the grand-ness of the previous experiment. I tried to make a whole wall of push-buttons. At this point I believed this was much more of a playful / exploratory experience and I had already lost the need to figure out what’s going on. The sounds on here are sound I pre-recorded with other buttons and switches: rotary encoder, missile launch switch, arcade button, toggle switch and push buttons.

If I could show this in class, I would have mounted the controller on a tripod, similar to a microphone. People can go up to it and play however they like.

IMG_9342 2.jpg

Imagine this:


I then tried to experiment more with direct sound mapping. With high quality sounds, the experience felt very surreal; expecting one sound but getting another. I think this could work very well if I had noise cancelling headphones (can’t hear the actual buttons at all but can only feel them).

Lastly, I wanted to take into account that people are very positional with their button placements. I modified Luke DuBois’s modular synth code to mimic the aesthetics of the buttons. Currently this is being controlled like arrows (left / right). I think if these were just holes on the synthesizers, it could be fun. (this is dismissing the heights of the buttons completely).

Code for modular synth:
Modified from:

Eagle board, Illustrator files, Arduino code, Max patches, all in here:

All in all I was happy to try a lot of new things on the fabrication side. I wish I had explored more on the modes of interaction through key combinations, delay in response – develop a more complex relationship between the pushbuttons and the results.

Idea: Final Project

Presentation slide


About the MIDI controller:
It’ll actually be 12-14 separate MIDI controllers for 12-14 people. Each one is just one slide potentiometer. Each one has its own channel. This will then be played through a synchronous web platform.

The idea is to have 12-14 people controlling different slides from different physical spaces but same digital space.

About variable clock: