Latest:

Jan 25: Mac CI

• It’s crazy how fast the M1 chips are.
• Github provides 3x M1 CPUs in CI (for free). It compiles the project in just 3 minutes (as opposed to the 4x CPU ubuntu machines taking >30 minutes).
• I don’t wanna spend crazy time on this because it isn’t the target hardware, but still cool and useful for development I guess.

Jan 15: Fuzzer

• I’m really happy with this fuzz test.
• Basically it generates a random sequence of inputs (like KEY_W, KEY_X, etc.) and sends them to the program.
• The goal is to run random commands for 30 seconds without segfaulting
  • On linux it’s crashing pretty much every time because of the audio server stuff… This is like super important actually… I’m working on it…

Jan 4: Dockerfile

• I’m still learning docker but… I saw in openpilot how they use a base image and another image on top of that
• Definitely on the roadmap, I think it can make CI even faster. Also standardize the environment and stuff

Jan 2: clang-format

• Thanks to @powei-lin for adding clang-format to the repo

Dec 31: cache CI

• I started using cachev4 + ccache within github actions and it’s really good.
• The average ci time is ~3 minutes down from 30
• Right now the main bottlenecks are the CMake generation step (which can’t be cached?) and installing apt dependencies
  • But I’m happy enough with the speed for now.

Ever since I heard the tame impala live album, I wanted the Juno 106. The only thing was, it was insanely expensive, and I knew my parents would never let me get it, so I decided to make my own.

Original Synth

To start off, I chose the teensy 4.1 for this project because of the high price-performance ratio, and because of the teensy audio library, which makes the programming a lot easier

First, I wrote the software using the Teensy Audio Library. I used PureData to create a virtual midi instrument, which made testing a lot easier

Architecture

First, there are 4 oscillators, the volume of which are controlled by a mixer. They then go through the high and low pass filters, and are then modulated by the envelope. The signal is then duplicated and sent into the flangers, and ultimately ends is output as a stereo signal that goes to the SGTL5000

There’s more complexity than this, as most of these blocks are modulated in turn by the LFO and envelope.

Hardware

Once I had the software done, I started working on the hardware. In this case, it was pretty simple, I had midi in and out, and then a bunch of sliders buttons, and leds, which you use to change parameters.

Midi Input Circuit