Using fixed-point arithmetic in a modern FPGA to produce cool sounds by modeling a 1970s-era Moog-like synthesizer.
Python Verilog
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
instrument More thinking and coding. Jan 5, 2014
README.rst Here is everything so far. May 31, 2013 Here is everything so far. May 31, 2013
fpga.v Here is everything so far. May 31, 2013 Here is everything so far. May 31, 2013 Here is everything so far. May 31, 2013 Here is everything so far. May 31, 2013


An FPGA-based synthesizer

Back in the 1970s and 80s, I was interested in electronics and music synthesizers, and they were still a pretty big novelty. By that time, Bob Moog had already identified a collection of analog modules that could be set up like this to produce a very wide range of interesting, obviously synthetic musical noises.

So a friend and I used to play with this stuff because his father had gotten him started with it, and then he'd gotten me interested (actually I guess my first impetus was the vinyl album "Switched On Bach"), and I knew some electronics so that was a welcome addition to his pool of knowledge on the subject. We had a good time and I built a few electronic musical instruments in college and immediately after, but then got too busy with my career to pursue it more.

Fast-forward thirty years, and we have cheap 8- and 32-bit microcontrollers and powerful FPGAs. In my career as an electrical engineer I'd done some FPGA work, and it has occurred to me that the kinds of fixed-point arithmetic computations that FPGAs do well are the ones that would do the jobs of all those analog modules. So my hope/plan is to use a Papilio One or Papilio Pro FPGA board to make a synthesizer with very few analog parts.

Getting back into this stuff after 15 years as a software engineer, there has been a bit of a learning curve. I don't have the resources of a large corporation supporting the FPGA development effort, I just have what I find online and can purchase mail-order. Sometimes the simulation acts one way and the chip acts another, so there's a good bit of floundering involved. Some of the early commits in this repo will be of stuff that just barely works or doesn't work at all, but that's why they call it a hobby.

Hacking MyHDL

Installation on Ubuntu 10.04 (Lucid)

sudo add-apt-repository ppa:balau82/ppa
sudo apt-get update
sudo apt-get install myhdl gtkwave verilog

Making sure it works:

(cd /usr/share/doc/myhdl/examples/rs232; python

Online manual:


Installation on OSX 10.8.3 (Mountain Lion)

Download MyHDL from this website:


tar xfz Downloads/myhdl-0.7.tar.gz
cd myhdl-0.7
python build
sudo python install

Unpack the GTKWave zip in the /Applications directory:

Add this to your .bash_profile:

export PATH=$PATH:/Applications/

Add this line to your MyHDL source file:


This will produce a file called testBench.vcd, and now you just type:

gtkwave testBench.vcd

Open up the device under test in the upper left, select signals you want to look at, and drag them over to the waveform display. Right-click on the signal name to choose a data format, including analog formats that show waveforms.

The MyBlaze core

MyBlaze is a MyHDL implementation (LGPL) of the GCC-targetable MicroBlaze soft processor core which runs on Xilinx FPGAs. Some people have run MMU-less Linux on it, and it can also run FreeRTOS.