Prior to the automaton class, me and Shloka discusses about the possibility of creating our own pen plotter due to 3 main reasons: (1) no pen plotter at ITP, (2) we got a crab pincer from Fire Island, and (3) just for fun. Combined with my recent realization of my obsession with my own brand name (@ekezia — I have included this name as an easter egg in a few of my prior projects), I’d like to create a name writer or signature forging automaton. Of course, this idea is made confident by related historical works such as The Draughtsman and The Writer by Jaquet-Droz.

These main inspirations impressively create a full drawing and indeed it includes a writing as well “Mon Toutou”, most of them are programmed through a series of cam, and the selection of the drawing is programmed with a large wheel mechanism that shift the cams up and down (this aligns a completely different set of pre-carved brass disks with the three cam followers (x,y,z) — similar to loading a new image from the mechanical memory). While it will be wonderful to program a series of cams for my own name, I’m more interested in designing a system that could scale and possibly be used by other people. The idea then transforms into a couple modular devices that create an arc and a line, both shapes which are minimal enough to create a letter. A remark from Josh Corn also suggests that this kind of idea can be an educational tool to visualize how mechanical device creates different shape.

Background Research

For this purpose, I’m limiting my study to just lowercase.

Fonts strictly composed of only straight lines and circular arcs are known as constructed geometric fonts or single-line (stroke) fonts. The origin of the single-line fonts is in fact coming from pen plotter, CAD (programmable drawing tool) or hand-drawn objects that maximize speed and legibility on blueprints (https://designincubation.com/publications/abstracts/the-machine-hand/).

ISO 3098 Font

Bauhaus typography also comes to mind. Bauhaus has strict design grid where letters are made of identical circular arcs attached to vertical lines. Also Bauhaus eliminates uppercase! They only have a unicase system (single-alphabet) to simplify typography (https://letterformarchive.org/news/bauhaus-typefaces-part-one/). The reasons of these uniformication of shapes and sizes are influenced by the rise of industrial mass production. Simple circles, squares, and straightl ines could be processed easier by heavy machinery. And we’d like to demonstrate this with our automaton machine (although not to discount the fact that the automaton is still capable of drawing more complex given a program). I’d also highlight that there is an intention to move away from the artistic part and wants to view the text from logical, scientific lens. It is something that is measureable, transferrable to another person.

In Biederman’s 1987 paper, Recognition-by-Components (https://pubmed.ncbi.nlm.nih.gov/3575582/) : A Theory of Human Image Understanding, complex visual objects can be broken down into basic geometry known as geons. Commencing from that idea, I believe ones my personal observation that most roman letters, at least the lowercase letters can be constructed by 2or less shapes: arc and line. This is fortunately reinforced by Gibson - The Feature-Analysis Theory of Letter Recognition. • In Principles of Perceptual Learning and Development, Eleanor Gibson established a matrix of unique features for the graphemes of the Latin alphabet. She explicitly categorizes the lowercase Roman alphabet by whether a letter possesses a "Straight Horizontal / Vertical Line" or a "Curve/Arc" (e.g., b, d, p, q are a line + an arc; o, c are arcs; i, l are lines). This paper will help me to make the recipe cook book for the automaton.

… insert Gibson’s recipe here

A sample creative exploration “Ukrug” found on IG from @atype_movement: https://www.instagram.com/p/DQO-sh4DdJX/ Strict forms such as circle and square, where either of them being scaled or repositioned can create alphabetic structure.

https://www.atypography.com/product-page/ukrug

Poem

lighght

— my own copy of the “lighght” poem by Aram Saroyan

https://thereader.mitpress.mit.edu/art-of-one-word-poem/

this poem denarrativized and decontextualized the language, removing much of the meaning from the word (since there is no context around it) and it placed the word in stark relief. Just like what Saroyan said, “…The one word isn’t “mine” but the one word in the center of the page is.”. This is similar to Joseph Kosuth’s 3 Chairs (https://www.moma.org/collection/works/81435), I feel like, where you have to choose the way that word “chair” is being denoted. I, assisted through the medium of keyboard and computer have created my own copy of Saroyan’s poem through the means of pixels. Now do I say that these lighght are a collection of pixels in a programmed positions? With this idea in mind, I’d like to transpose it to the automaton device that dissect each letter into just 2 shapes, somehow almost binary if we imagine the arc as 0 and line as 1. This way, we can perceive the “lighght” poem as a collection of arc and line shapes.

Breakdown of Mechanisms

While the inspirations are largely programmed with cams, I think I want to abstract it to the simplest shape that can be made just with gears, just like the first spirograph-y exercise that I did with the cardboard. For the arc line, we can use the gears combo:

And for the straight line, we can use the mechanism that either translate rotation into line or just use a simple linear transformation (need to confirm)…

Materials

The materials not sure yet but it may be nice to give a nod to bauhaus choices of materials.

Recipe Cookbook

My own simplification from Gibson’s analysis:

• Line Only: i (half line), l (1 line)
• Arc Only: c (1 arc), o (2 arc)
• 1 Line + 1 Arc (flippable): b, d, h, p, q, u, k
• Multiple Lines/Arcs:m, n, w, r, t (still purely composed of lines and arcs)
• Few Exceptions (include non 180deg rotation, diagonal crossings): j, t, v, x, y, z

Demonstrated on p5js for visualization:

A few bonus note for myself:

How to easily convert a text / shape into cams?

1. Vectorization: Convert the text / drawing into a time-based parameter curve, P(t) = (x(t), y(t), z(t)), where t is the rotation angle of the master driveshaft from 0° to 360°.
2. IK: Map the x,y,z to the physical length of the automaton structural arm levers to find how much the cam followers must displace at angle t.

There’s also a compiler pipeline caleld PDA-0 (https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8594443)