Learning AYAB and making garment samples with the knitting machine
I worked with Yafira and Defne Kelekci to understand the machine and create an inventory of all the parts we currently have. When the project started I had no experience with knitting machines, only regular sewing and therefore a big part of this was understanding those patterns and what are the challenges of a knitting machine user, from the perspective of a newbie.
What is AYAB?
The goal of the AYAB project is to provide an alternative way to control the famous Brother KH-9xx range of knitting machines using a computer. We have the Brother KH-910 (marketed in North America as the KnitKing CompuKnit) in the lab for this project, though we’re still waiting on parts to actually work with it.
AYAB connects directly to the electronics internally, which means it can support older machines that predate USB connectivity. There’s a shield that goes on top of an Arduino, which then interfaces with the machine’s existing electromechanical systems. Essentially, it hijacks the machine’s internal logic: it produces what looks like a clock line and turns the rotary encoder internally inside the machine, sending signals to the solenoid buffer of 16 bits. One clock cycle equals one turn on the machine.
Instead of being limited to Mylar pattern cards (those translucent plastic sheets with punched dots that act as physical binary code), you could send any digital pattern directly from your computer to the needles.
Evil Mad Scientist Labs used to distribute AYAB kits but stopped. Finding the hardware is difficult. So the question becomes: if it’s so hard to find, why focus on it? Are we trying to replicate it? Document it? Understand it as a proof of concept?
This matters especially if our goal is helping the disabled community have their own tools. Wouldn’t we also need to think about the accessibility of the process itself? The design files are on their Github for making our own Arduino Shield, which suggests one path forward that was spoken with in office hours with Daniel: manufacturing our own shields and documenting the process for the Ability Project’s YouTube channel. That feels more aligned with the project’s values than hunting down discontinued Etsy listings.
Key differences between electronic and analog knitting machines
The jump from purely mechanical to electromechanical knitting machines in the early 1980s represents a fascinating moment in craft-computing convergence. Before I get into AYAB’s intervention, it’s worth understanding what these early electronic machines actually did.
A purely mechanical machine like the Silver Reed LK750 (which Yafira and I actually worked with) relies entirely on manual needle selection and carriage movement. Every stitch is determined by the user’s physical setup: which needles you push forward, how you move the carriage, the tension settings you’ve dialed in. It’s tactile, immediate, and unforgiving.
The Brother KH-910 and its KnitKing equivalent introduced something different: electromechanical needle selection via Mylar pattern cards. Light sensors read the pattern of black dots on translucent plastic sheets, translating each dot into an on/off signal that triggers solenoids to select which needles knit or rest. The user still sweeps the carriage manually left and right, but the machine “remembers” the pattern through this optical reading system synchronized to each pass. It’s proto-programming. Physical data controlling textile behavior, row by row. Patterns weren’t code yet, but they behaved like code. The Mylar cards are essentially punch cards for fabric, a direct lineage from Jacquard looms through IBM data cards to knitting machines sitting in suburban basements in 1983.
What AYAB does is replace the Mylar reader entirely. Instead of optical sensors reading physical cards, the Arduino sends digital signals directly to the solenoid buffer that controls needle selection. You keep the manual carriage sweep (the human rhythm that makes these machines surprisingly meditative to use), but you gain the flexibility of digital pattern design. No more cutting custom Mylar cards or being limited to pre-printed patterns.
Task analysis: notes on the machine system
Working with Yafira and Defne Kelekci, we created an inventory of all the parts we currently have and tried to map out how the Brother KH-910 would actually function if we could get it running.
The machine is organized into several subsystems.
Structural foundation: A 4.5mm standard-gauge bed holding around 200 needles. Behind it, gate pegs define the needle path and create the geometry where loops form. This precision-milled surface is where everything else happens.
Knitting mechanism: Each needle has a hook and latch for catching yarn, plus a small butt that interfaces with the carriage cams. The carriage (K carriage) is what you sweep back and forth, and inside it, cams and levers choreograph whether each needle knits, slips, or tucks. Below the needle line, the sinker plate holds fabric in place so loops don’t rise too soon. Adjustment knobs along the front let you switch between plain, pattern, slip, or tuck modes through purely mechanical logic.
Patterning & selection: This is where it gets electronic. An electromechanical needle selector determines which needles knit or rest, guided by the Mylar cards. As the card passes through the reader, light sensors hit the pattern and trigger needle selection. Every pass of the carriage advances the Mylar one row, a synchronization between hand movement and stored instruction.
Electronic control: A simple logic board handles power distribution, signal reading from the Mylar sensor, and synchronization with the needle selection unit. There’s a control panel that counts rows and lets you start, stop, or mirror a design. It’s minimal but effective, a bridge between mechanical timing and digital pattern logic.
The modularity is interesting too. Users could add a ribber attachment for double-bed fabrics, garter carriage for reversible textures, intarsia or color changer for multicolor work, transfer and tuck tools for shaping. Each accessory extended the machine’s “computational vocabulary” without needing a full computer.
Creating first samples
Using the Silver Reed LK750 knitting machine, Yafira and I worked together on figuring out how it worked. The LK750 is fully mechanical, no electronics, which made it a gentler entry point than jumping straight into the Brother’s electromechanical complexity.
We still had some trouble and took a few sessions to get the hang of it. Some lubrication would have helped as well.
Videos not included for space.
Reflections
Machine knitting is incredibly sensitive to pacing, material choice, and setup. Thread tension, carriage speed, needle selection, yarn weight. Every variable affects whether you get clean fabric or a tangled mess of dropped stitches. We focused on producing swatches rather than trying to make finished pieces, treating each sample as a small-scale experiment to isolate variables. Whilst I would have loved to complete my initial plan of gathering these swatches into a cubic shaped lamp, i’m also confident that it will be done moving into this next semester.
This mode of working was refreshing in a way. Process over efficiency, iteration over output. It felt aligned with the Ability Project’s broader methodology and the hands on nature of these older machines. They’re cheaper precisely because they’re old and there’s a huge community present and interested in repairing them. Part of me kept wanting to figure out the hardware issues faster, but there were also some valuable lessons in the process.
Accessibility isn’t just about the final garment; it’s about who has access to tools, how knowledge is transmitted, whether learning environments support experimentation and failure. Our swatch work became a form of research in itself, whether we intended to or not.
It would be helpful to create documentation on how to clean and maintain the knitting machine. Essential, especially for tools that will be used by multiple people over time. But the bigger question is about expansion and innovation. Where can we actually contribute something new?
Manufacturing our own Arduino Shield feels fruitful regardless of whether we can eventually source an official AYAB kit from Etsy. The design files are open source on Github. Documenting the build process for the Ability Project’s YouTube channel would make this knowledge accessible to others trying to bridge vintage knitting machines and contemporary digital fabrication tools. There’s also a philosophical tension here worth naming: we’re focusing on technology that isn’t manufactured anymore. If the goal is helping the disabled community have their own tools, shouldn’t we also consider the accessibility of the process itself? AYAB is brilliant as a concept, but if the barrier to entry is soldering custom Arduino shields and reverse-engineering 40-year-old Japanese electronics, we’re not exactly lowering barriers for those we want to help
Maybe the value isn’t in the AYAB module itself, but in understanding how these machines think - their physical logic. Studying the KnitKing CompuKnit KH-910, even without running it, reveals an early conversation between craft and computation.
Thank you to Daniel, Yafira and every VIP member. This was almost entirely new for me and the support was invaluable.
Resource List
- Knitting Machines Repair
Yarn it Forward: An NYU Club Gives Back to NYC - Yarn it Forward members create warm garments and blankets that local organizations distribute to people in need around the city
KH-910 Machine Components - Inventory
This process isn’t finished as we decided at the time to focus on the existing complete machine at the lab. To be continued into the semester.
