Embroidered Computer

Rina Chen’s living notebook on digital craft and design.

Embroidered Computer

Context

https://ireneposch.net/stitching-worlds/

with the advent of electronically conductive fibres, it might be possible to adapt the use of textile machinery to translate patterns into electronic functions. Altered processes and materials can be used to produce electronic components such as resistors, capacitors, inductors, and eventually complete electronic appliances.

What if electronics emerged from knitting, weaving and embroidery? How would technology be different if craftspeople were the catalyst to the electronics industry, via textiles manufacturing?

Reconstructing computation using materials and craft techniques allowed an embodied learning and walking through of the history of computation, revealing hidden histories and contingent choices.

  • visible to invisible,
  • analog to digital,
  • maths to symbols
  • hardware to software (separation)

The choice of techniques and materials are based on the most revealing of the technology.

  • flip dot replaces silicon
  • punchcard replaces software
  • maths is wired

Plan

  1. Introduction
  • 1a. Background:

The historical intersection of textiles and logic (Binary origin in weaving).

  • 1b. Research Question:

Can a functional, programmable 8-bit architecture be sustained using purely mechanical textile-based switching?

  • 1c. Project Plan:

Phase-gate approach: Discrete logic -> Arithmetic units -> Programmed Inference.

  1. Theoretical Framework: From Silicon to Silk
  • 2a. Taxonomy of Elements: Conductive thread as traces; metal snaps as contact points.
  • 2b. The Semiconductor vs. The Mechanical Switch: Contrast electron-flow control (Silicon) with physical gap-closing (Textile Relay).
  • 2c. The Embroidery Relay: Detailed design of a hand-coiled electromagnet pulling a ball to close a circuit.

  • 2d-e. Boolean Logic via Embroidery: * AND/OR: Physical series and parallel paths.

  • NAND (The Universal Gate): Explain that with NAND, you can build any other gate.

  • 2f. Arithmetic Foundations: Building Half-Adders (HA) and Full-Adders (FA).
  • 2g. Algorithmic Multiplication: Explaining “Shift and Add” logic, as mechanical computers cannot do complex math in one step.
  1. Hardware Architecture: The Textile CPU
  • 3a. The Clock: A “Slow-Clock” design. Suggestion: Use a large physical pendulum or a slow 555-timer circuit to show the “heartbeat” visually.
  • 3b. Registers: Using “Flip-Flops” made of two cross-coupled embroidered NOT gates.
  • 3c. The Textile ALU: Implementation of the Adder/Subtractor.

  • 3d. Volatile Memory (RAM):

  • The Embroidery Challenge: Instead of ICs, use Magnetic Core Memory. Threading conductive wire through ferrite beads. This is much more “textile-friendly” than building 128 individual relay latches.

  • 3e. The Control Logic (The “Loom” ROM): use a Diode Matrix (using small LEDs as one-way valves) or a Punch Card Reader made of conductive fabric.

  • 3f-g. The I/O Bridge: Converting light (Analog) to logic (Digital).

  • Finding: You may need to use one “hybrid” board here. Converting an LDR’s resistance to an 8-bit signal usually requires an ADC chip, but for a thesis, you could build a 1-bit Comparator (Is it Light or Dark?) using a single safety-pin relay.
  1. Software & Simulation
  • 4a. Circuit Simulator
  • 4b. The Development Chain: Logisim for logic verification-> Custom Assembler -> Hex editing the “Punch Card” patterns.
  • 4c. Build each steps
  • 4d. Logic analyzer
  • 4e. Subroutine Library: Writing the code for multiplication on a machine with very limited memory.
  1. Implementation of Artificial Intelligence
  • 5a. The Single-Perceptron: ”Training” the fabric. The “Weights” are determined by the number of thread wraps in a coil or the placement of a metal snap.
  • 5b. Linear Regression: $y = mx + b$. The computer takes the light sensor input and “predicts” a value.
  • 5c. Use Case (Weather/Light): Using historical data to create a “Threshold” where the computer triggers a “Wear Sunglasses” LED.
  1. Methodology: Fabrication & Assembly
  • 6a. Magnetic Mechanisms: Solving the friction and “sticking” issues of handmade relays.
  • 6b. Scaling: Moving from a single NAND gate to a complex 8-bit bus.
  • 6c. Integration: Managing “Line Noise” and resistance in conductive thread.
  1. Discussion: The Fusion of Analog and Digital
  • Exploring “Tactile Computing.”
  • The durability of e-textiles vs. silicon.
  • The philosophy of “Visible Logic”—making the “invisible” work of a CPU physical.
  1. Conclusion
  • Summary of findings.
  • Future implications for wearable, repairable, and sustainable electronics.

Experiment with 1 tiny bit (1/0)

![[WhatsApp Video 2026-01-26 at 9.02.21 AM.mp4]]

Logic gate

See [[A Thinking Textile, or a Textile that Made Me Think]]

Reference

How transistors run code? ![[Pasted image 20260215213954.png]] ![[Pasted image 20260215213724.png]] ![[Pasted image 20260215213900.png]] ![[Pasted image 20260215213932.png]] ![[Pasted image 20260215214047.png]] ![[Pasted image 20260215214216.png]] ![[Pasted image 20260215214226.png]] ![[Pasted image 20260215214319.png]] ![[Pasted image 20260215214435.png]] Full adder ![[Pasted image 20260215214600.png]] ![[Pasted image 20260215214642.png]] ![[Pasted image 20260215214807.png]] ![[Pasted image 20260215215348.png]] ![[Pasted image 20260215215254.png]] ![[Pasted image 20260215215644.png]] Binary decoders ![[Pasted image 20260215215619.png]] ![[Pasted image 20260215215848.png]] ![[Pasted image 20260215215857.png]] ![[Pasted image 20260215215929.png]] ![[Pasted image 20260215220010.png]] ![[Pasted image 20260215220235.png]] ![[Pasted image 20260215220327.png]]

Conventional 8-bit Computer

https://eater.net/8bit/parts

https://eater.net/8bit/kits https://www.youtube.com/watch?v=HyznrdDSSGM https://www.instructables.com/Making-an-8-Bit-Computer/ https://learn.sparkfun.com/tutorials/how-to-read-a-schematic/all

Clock

Why you need a clock

  1. load input
  2. multiply
  3. add
  4. threshold
  5. output Each step(eg. perceptron) needs a timing signal. Industrially, 555 timer, microcontroller are used

Memory

Machine Learning with 8-bit

The basic formula: y=sign(∑wi​xi​+b)

On an 8-bit system, that means:

  • integer or fixed-point math
  • very small input vector (2–8 inputs)
  • inference first, training optional

Inference, constant weight, and ROM

What is meant by inference? In a classical perceptron, there are two distinct phases:

  1. Training
    • Weights are updated
    • Learning rule applied
  2. Inference
    • Weights are fixed
    • System only computes output

The fabric 8-bit computer is designed to perform inference, not training, since the program flow assumes weight₁, weight₂, … are constants. So architecturally weights are read-only parameters, not variables.

This suggests ROM, not RAM. with ROM weights:

  • MUL k simply:
    • selects weight line k
    • feeds it into multiplier
    • no state change k is based on material.

Why inference rather than training?

The biggest reason is the feasibility, but there’s also a philosophical choice that I consciously made.

In normal computers, execution = interpreting symbolic instructions, where program and data live in the same abstract space.

learning-as-execution hides:

  • effort
  • material cost
  • irreversibility

In fabric computer, execution = physical signal propagation. There is not material structural change, execution is time passing through a fixed textile structure.

One idea that can be explored is: Learning = changing the textile, not the execution

  • The textile is the model
  • Updating the model means rebuilding part of the machine

This design makes explicit what modern ML hides:

  • learning consumes resources
  • learning changes infrastructure
  • learning is not free
  • learning leaves traces

In a fabric computer:

  • every learned change has:
    • a tactile cost
    • a visible history
    • a labor footprint

“Visible history” refers to the fact that learning in the system produces irreversible material traces—stitches, patches, deformations, and repairs—that make the system’s past states and interventions legible in its present form.

In conventional machine learning:

  • Weight updates are numbers in memory
  • Old values are overwritten
  • There is no trace of how a model arrived at its current state The system has no memory of its past, only its current parameters. But in a fabric-based ML system,
  • Weights are not overwritten
  • They are physically altered
  • Previous states are not fully erased In summary:
Digital ML Fabric ML
Weight update Material alteration
Overwrite Accumulation
Invisible training Legible intervention
Clean state Patched state
Forgetting by design Memory by residue

This aligns with:

  • craft traditions (repairs are legible)
  • feminist and care-centered technologies
  • Illich’s critique of opaque tools
  • post-digital material ethics

It is building a system where intelligence leaves marks. And where those marks:

  • slow future change
  • constrain behavior
  • demand care

Decision emerges from accumulated material interactions = The output bit is not computed symbolically, but arises only after currents, delays, resistances, and physical propagation across the textile have collectively stabilized into a state that crosses a comparator boundary. Philosophy = circuit diagram

(Generated with ChatGPT)

Different ways of doing:

  1. Material computation
    • The program is the textile topology
    • The weights are literally stitched
    • Reprogramming = re-stitching or patching - It’s not
    • Training (unless manual)
    • Precise repeatability - Only need comparator IC (LM393 or similar) to output “neuron”
  2. Digital fabric computer + perceptron in microcode
    • Hand-written assembly/microcode
Opcode Meaning
LOAD Load input into register
MUL Multiply input × weight
ADD Add to accumulator
CMP Compare with threshold
JMP Jump to next step 
Step 1: LOAD input_1
Step 2: MUL weight_1
Step 3: ADD accumulator
Step 4: LOAD input_2
...

Other readings:

https://petewarden.com/2015/05/23/why-are-eight-bits-enough-for-deep-neural-networks/ https://blog.lessaworld.com/2025/03/09/running-a-perceptron-on-an-8-bit-computer/ https://medium.com/data-science/ultra-tinyml-machine-learning-for-8-bit-microcontroller-9ec8f7c8dd12

The Challenge: Teaching an Old Computer New Tricks

I decided to implement Perceptrons, the most fundamental type of artificial neural network. Perceptrons are a supervised learning algorithm used in machine learning to classify data with simple binary decisions – perfect for a machine with limited power!

Now, let’s be real: running a Perceptron on an 8-bit computer is a serious limitation. But at its core, AI is just math, and I wanted to prove that you could implement the basic logic of artificial neural networks on even the most primitive hardware.

The goal? To use Perceptrons to model simple logic gates (AND, OR, NOT) – the building blocks of computation.

Irene’s Textile 8-bit Computer

![[レコーディング 2025-10-11 210533.mp4]] ![[assets/img/embroidered_computer.png]] ![[assets/img/embroidered_computer_signal_diagram.png]]

![[2851581.2891101.pdf]]

![[StitchingWorlds_Book_Kurbak_Ed__1_May2019_PDF-A.pdf]]

Process reference

1 Bit

  • https://www.kobakant.at/DIY/?p=5915

  • textile relays (prototype)

  • https://www.kobakant.at/DIY/?p=8059

    • what is pnp and npn
    • what is h-bridge circuit
    • what are transistors
  • ==enameled copper wire (0.1mm diameter)==
  • ![[Pasted image 20260115080614.png]]
  • used a pen which is about 1cm diameter and wind 80 times each. Use tape to hold them together when finished winding.
  • small pieces of copper fabric with fusible interface on the back
  • fix it to the base fabric by ironing it on. This will serve as a connection point to the power source.
  • ![[Pasted image 20260115080846.png]]
  • The magnet wire has enamel coating in order to isolate. We need to take out this isolation layer on the ends to make an electrical connection to the conductive fabric we added. For this, use lighter to burn the coating and scrape it off with ==fine gauge sand paper==. Check with multimeter if the isolation layer is taken out properly.
  • Heat up the fabric with soldering iron short time and add the solder to it. The solder will stick to the fabric like a paint. Then place the end of the coil wire on the applied solder and melt them together.
  • ![[Pasted image 20260115081605.png]]
  • polymorph + neodymium magnet (5mm diameter, 2mm high, axially magnetized) +thread or
  • ==magnetic bead (magnetite beads, hematite beads, polarised exactly around the through hole axis) if there is==
  • Place the finished “dot” on the coil, and thread the two end of the thread coming out of the dot to the back of the fabric. Make sure to give some extra length on the thread so there are a space to flip for the dot.
  • ![[Pasted image 20260115082107.png]]
  •  2-3V from bench power supply. You can also use 2xAA battery or lipo battery.
  • For multiple, The magnetic beads need sufficient distance to each other in order to be controlled individually
  • ![[Pasted image 20260115082737.png]]
  •  automatic pullback in putting a small magnet underneath the bead. The magnets force is overwritten by the active electromagnet, but has enough strength to pull the magnetic bead back in its original position ((S-015-01-N2-Ni magnet, 0,01g underneath the flipping pearl)
  • Relays are electrically actuated switches. […] For a typical mechanical relay, a current sent through a coil magnet acts to pull a flexible, spring-loaded conductive plate from one switch contact to another. (Scherz 2007)

Array

Community

Important questions/areas

  • Goldwork, craft, history/narrative of the object: uselessness or richness? Flattened world
  • Digital materiality, history/narrative of the digital: hidden, clean or disorderly [[Material Culture of Computation]]
  • Walk through the history of computation by making

Other ideas

Things that got me interested in:

  • lifting![[Pasted image 20260115083245.png]]
  • flipping like eyes

Other mechanisms

  • Mending with embroidery
    • Create a model house with switches and other electronics rethought
    • light switch, a clock, a computer, a timer