Quick tickle-the-itch project on taking an input audio file and generating some commands for the NES audio proccessing (APU) to try and best reconstruct the audio.
NES audio is composed of 4 signal components: 2 square waves, 1 triangle wave, and a noise channel. I remember when writing a NES emulator (back in University) that I was surprised at how the Super Mario Brothers song had used the noise channel as percussion.
The square and triangle waveforms have other features ( like frequency sweeping, and envelope decay) that I didn't try to use here.
Fourier analysis of the sequence is done in chunks (e.g., block sizes of 1024, 2048, 4096 samples), and the best parameters (frequency, duty-cycle type, volume) are selected for each of the 4 waveforms.
There is a simple c-code driver (rom/audio_main.c) that is used to feed the data to the NES APU. It has a simple equalizier.
Interface:
Restricting to square channels: square 2, square 1&2. In the interface, you can use left/right to highlight channel and down/up to mute/unmute.
Requirements:
- Octave (for doing the analysis)
- c65 (a c-compiler to generate NES files)
- A 44.1khz .wav file to sample from (use FFmpeg to generate a wav file if you don't have one)
- A NES emulator to play your ROMs
Some video examples:
Run the following command to generate a playable NES ROM from your input wav:
analyze_wav_file('wav/my_input.wav', 'my_output');
The output will be in the rom folder:
- my_output_tri0_2048.rom: Give square wave's first priority, using block size of 2048
- my_output_tri0_4096.rom: Give square wave's first priority, using block size of 4096
- my_output_tri1_2048.rom: Give triangle wave's first priority, using block size of 2048
- my_output_tri1_4096.rom: Give triangle wave's first priority, using block size of 4096
Using a smaller block size increases representation ability, but also leads to more audio popping due to changing the parameters more frequently.
The low and high frequencies caps can be specified. Cappign the upper frequency can get rid of some of the vocals (which sometimes cause confusion in the frequency selection).
attenuate_sigma: controls the range of frequencies attenuated after choosing a frequency
for a single channel (making this small will cause square and triangle waves to choose
similar frequencies)
clear_all_freq: When 1, similar frequencies chosen in square or triangle waves are not
allowed in the other wave types (triangle or wave).
There are a few examples (.c, and .rom) in rom directory:
- radiohead_tri1_4096.rom: Good example of noise used as percussion.
- californication_tri0_4096.rom: Giving square waves higher priority.
- hey_jude_tri0_2048.rom: Using a smaller block size.