https://github.com/mathaou/jdsp
Digital Signal Processing Utilities for Jai
https://github.com/mathaou/jdsp
Last synced: 9 months ago
JSON representation
Digital Signal Processing Utilities for Jai
- Host: GitHub
- URL: https://github.com/mathaou/jdsp
- Owner: mathaou
- Created: 2023-11-27T19:07:29.000Z (over 2 years ago)
- Default Branch: master
- Last Pushed: 2024-05-31T23:10:19.000Z (about 2 years ago)
- Last Synced: 2024-06-01T00:25:02.802Z (about 2 years ago)
- Language: C
- Size: 3.21 MB
- Stars: 2
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# jdsp - Jai Digital Signal Processing Utilities
| Feature | Description|
| --- | --- |
| `jfft` | Fast Fourier Transforms with various windowing functions |
| `jdsp_filter` | **WIP** Various filters including lowpass, highpass, bandstop, and bandpass |
---
#### `jfft`
`jfft` only processes real data sets at this time. No complex numbers. This means that instead of real/imaginary components we have sine/cosine.
`jfft_create_transformer :: (signal_length: int, allocator: Allocator = temp) -> *jfft_transformer #must`
> Sets up the memory needed to perform FFTs on inputs of a specific length. `signal_length` should be a power of two for best performance, but the only real requirement is that `signal_length` is even. Internal memory can be reused by `jfft_forward` and `jfft_backward`.
`jfft_destroy_transformer :: (jfft: *jfft_transformer)`
> If you used something other than `temp` as an allocator or you don't call `reset_temporary_storage` too often you can call this.
`jfft_forward :: (jfft: *jfft_transformer, input: []float32)`
> Performs an FFT on a given input buffer and puts sine/cosine components . `input` should match the `signal_length` of the transformer. DC is in bin `0`. Bins `1..N/2-1` contain frequency bins. Nyquist starts at bin `N/2`. Results can be fed into `jfft_magnitude_transform` or `jfft_magnitude_dB_transform` for useful analytics.
`jfft_forward_transform :: (input: []float32, allocator: Allocator = temp)`
> Creates `*jfft_transformer` internally. Same as `jfft_forward` otherwise. Good as a one-off.
`jfft_backward :: (jfft: *jfft_transformer, input: []float32)`
> Performs an inverse FFT on data returned from `jfft_forward`. Floating point errors mean that the original signal may not be able to be recreated exactly depending on original signal.
`jfft_backward_transform :: (input: []float32, allocator: Allocator = temp)`
> Creates `*jfft_transformer` internally. Same as `jfft_backward` otherwise. Good as a one-off.
`jfft_magnitude_transform :: (input: []float32)`
> Modifies input in-place. Converts results of `jfft_forward` to be magnitude data up to `N/2` due to Nyquist-Shannon sampling theorem. This means that an FFT can only provide useful info for frequency bins of `sample_rate / 2`. Anything beyond `N/2` is imaginary or something (useful for reconstruction).
`jfft_magnitude :: (input: []float32, allocator: Allocator = temp) -> []float32 #must`
> Same as `jfft_magnitude_transform` but it returns a new array without touching original data.
`jfft_magnitude_db_transform :: (input: []float32)`
> Similar to `jfft_magnitude_transform` except converts the first `N/2` frequency bins to decibels.
`jfft_magnitude_db :: (input: []float32, allocator: Allocator = temp) -> []float32 #must`
> Same as `jfft_magnitude_db_transform` but returns a new array without touching original data.
---
Modifications to the frequency-domain data from `jfft_forward` would require overlap-add for successful `jfft_backward`. The successful execution of Fourier synthesis from modified frequency-domain data is on the programmer and outside the scope of the API at this time.
---
#### `jdsp_filter`
Working on it!
---
#### Example
Go into `example` and run `jai example.jai -release`. You should have an `sdl2` library accessible somewhere in case the one in `modules` doesn't work for you for whatever reason.
Run it like `./example`.
`UP/DOWN` will increase/decrease your chunk for FFT.
`SPACE` will pause the graphs at the current chunk.
`LEFT/RIGHT` will move to next/previous frame when paused.
`K/L` will decrease/increase the time betweeen chunks when unpaused.
Click the `Window` combo box at the top to see how different windows affect your input signal and FFT results.