Recitations

Table of Contents

Apart from the weekly lecture, there sometimes were additional lectures as introductions to practicalities needed for the respective assignment.

Programming & Debugging

  • fabacademy microcontroller families
    • attiny: small, very capable for the size they have if you use them properly
    • samd: are more for a fool. they can do 30bit operation, do usb etc.
    • rp2040: raspberry microcontrollers. much more advanced. small computer
    • esp32: if you need wireless
  • attiny
    • low-cost, low-power
    • suitable for simple embedded applications with limited processing and io requirements
    • programming over updi 8 bits - 1MHz - 20Mhz -> 32 MHz overclock
    • versatile io: ianalog, pwm, i2c, uart, spi
    • single pin programming
    • sandbox board: adrianino
  • samd
    • programming: swd/jtag
    • put small program in it and that will let your program it through usb
    • so only putting bootloader on it is the difficult operation
  • rp2040
    • difference raspi / xiao: raspi has more pins
    • good for projects requiring multitasking and connectivity
    • programming using uf2
    • bootloader is already on it. you cannot erase it. you cannot really break it
    • 32bits - 133MHz -> 250MHz overclocking
    • much more capable than attiny
    • uses more power
  • esp32-c3/s3
    • wifi and bluetooth on it
  • others
    • stm32, pic, msp (not used in fabacademy)
  • programming languages
    • here, we focus on rp2040
    • micropython:
      • justified since we have pio (whatever that is)
      • simple
    • c/c++
    • circuitpython
      • use of it is different than micropython but less efficient. rather go with micropython
    • javascript
      • only use that can justify this: if you want js everywhere for your project, you might use it.
    • rust
      • it works, but it is difficult to compile. thats not really an option that is recommended. the fun of it is that since he tested it: now he uses it for his everyday computer programming
    • go
      • rather go with cpp
      • go was made to ease the use of seriel processors
      • rp2040 has 2 cores
    • asm
  • debugging hardware
    • dont panic
    • pin down problem in pieces:
      • power
      • soldering
      • correct components?
      • placed correctly?
    • stay calm and think of a step by step plan
    • document when debugging
    • the pragmatic programmer: good book
    • rubberducking
    • dont panic, you gonna cause more problems
    • narrowing down
      • reproducing the problem. try to do it a secnod time
      • alsways first do visual check. inspect the circuit:
        • traces, soldering, components
      • problem might not be visible.
      • use microscope
      • use multimeter
      • logic analyzer
      • use good microhooks
      • osci
      • sdr

Artificial Intelligence

  • about embedded AI/ML

TinyML Concepts & Applications

  • idea is to do inference of AI on small and cheap hardware to make it more accessible
  • edge AI
  • tinyML (part of Edge ai)
  • use tinyML for
  • def
    • machine learning appl for mircrocnt.
    • ai inference on low-power devices
    • implementation method for edge intelligence
    • … (check slide)
  • toolchain and Platform Introduction
    • tensorfolow lite
    • pytorch mobile
    • edge impulse
    • stm32 cube.ai
    • sensecraft AI

questions

  • What does “Processing” refer to in the previous slide in the edge AI block?
  • Is it correct that Edge AI is an approach that aims to speed up inference by running AI algorithms on or near the device where the data is generated, and TinyML is an approach that focuses on enabling inference using ML models within very constrained machine resources?

LLM Intro

Text to 3D Models

Sustainable Materials

https://fabacademy.org/2025/recitation/materials/index.html

quantum

optics

  • openuc2
  • speaker talking about making optics more accessable, similar to what arduino did for electronics.
  • idea: put everything in a cube.
  • O3Q.de
  • openuc
  • going open source
    • because it benefits the
  • holography setup: https://openuc2.github.io/docs/Toolboxes/DiscoveryInterferometer/InlineHolography
  • open source business models
    • prusa as example
    • you can buy finished machine
    • or download design
    • not based on the product, but on a value.
  • openuc2.com Or in our wiki openuc2.github.io

materiom

Machine Building

  • different machines fab.cba.mit.edu/classes/865.24/people/rahul
  • Jani (oulu): different machines inspired by rahul gitlab.com/fab-lab-oulu/sampo
    • used some funny and simple gcode
    • used mods
  • leo mcelroy
  • jens dyvik
    • 2 years travelling through fab labs
    • 8 years building non profit organization
    • github.com/fellesverkstedet/fabricatable-machines/wiki/modules
    • osb used for machine.
    • github.com/Fabricatable-Machines/Timeline notes on fabricational machines
    • reason for open hardware:
      • greater good, contribute to the pool you drink from.
      • social
  • InMachines
  • jake read
  • quentin bolsee
    • gitlab.cba.mit.edu/neilg/urumbu
    • gitlab.cba.mit.edu/quentinbolsee/urumbubotxy-2.0
  • nadya peek
    • neils student, prof at uni of washington
    • not enoigh to be open source
    • other principles
    • be able to undestrasd the technology
    • be able to extend it
    • we dont build machines, but workflows
    • one machine that they work a lot with is tool changing machinne: jubilee. multitool fabrication.
    • making it easy for people to control machines. extending so that it is possiblle to use them for usages beyond fabrication
    • levitating tool for styropor