Expand description
synfx-dsp DSP real time audio synthesis, effect algorithms and utilities for Rust
Most of the algorithms and implementations in this library have been implemented for HexoDSP and used in HexoSynth. I factored them out, because they seem useful in other contexts too, for instance the synfx-dsp-jit crate.
I collected most of the algorithms in this crate from various GPLv3 compatible sources. They also were partially translated from multiple different C++ projects. I tried to document the source and source license diligently in the comments of this crate. I apologize if any attribution is missing and would welcome patches or reports.
Feel free to use these algorithms and utilities. Help, patches and additions are appreciated if they comply with the GPL-3.0-or-later license and don’t break the test suite in HexoDSP.
Attention: HexoDSP comes with a large test suite that also covers these algorithms. And that is the one that also has to pass if these algorithms are touched. The flip side is, that these implementations are actually covered by a test suite.
Requires Nightly as of 2022-10-02 due to std::simd!
Copyright, Licenses, Attribution, Contributions
Here is a list of sources parts of this library copied or translated code from:
- crate::quicker_tanh64 / crate::quicker_tanh
quickerTanh / quickerTanh64 credits to mopo synthesis library: Under GPLv3 or any later. Little IO <littleioaudio@gmail.com> Matt Tytel <matthewtytel@gmail.com> - crate::quick_tanh64 / crate::quick_tanh
quickTanh / quickTanh64 credits to mopo synthesis library: Under GPLv3 or any later. Little IO <littleioaudio@gmail.com> Matt Tytel <matthewtytel@gmail.com> - crate::tanh_approx_drive
Taken from ValleyAudio Copyright Dale Johnson https://github.dev/ValleyAudio/ValleyRackFree/tree/v2.0 Under GPLv3 license - crate::AtomicFloat
Implementation from vst-rs https://github.com/RustAudio/vst-rs/blob/master/src/util/atomic_float.rs Under MIT License Copyright (c) 2015 Marko Mijalkovic - crate::Biquad
The implementation of this Biquad Filter has been adapted from SamiPerttu, Copyright (c) 2020, under the MIT License. See also: https://github.com/SamiPerttu/fundsp/blob/master/src/filter.rs - crate::DattorroReverb
This file contains a reverb implementation that is based on Jon Dattorro's 1997 reverb algorithm. It's also largely based on the C++ implementation from ValleyAudio / ValleyRackFree ValleyRackFree Copyright (C) 2020, Valley Audio Soft, Dale Johnson Adapted under the GPL-3.0-or-later License. See also: https://github.com/ValleyAudio/ValleyRackFree/blob/v1.0/src/Plateau/Dattorro.cpp and: https://github.com/ValleyAudio/ValleyRackFree/blob/v1.0/src/Plateau/Dattorro.hpp And: https://ccrma.stanford.edu/~dattorro/music.html And: https://ccrma.stanford.edu/~dattorro/EffectDesignPart1.pdf - crate::process_1pole_lowpass / crate::process_1pole_highpass
one pole lp from valley rack free: https://github.com/ValleyAudio/ValleyRackFree/blob/v1.0/src/Common/DSP/OnePoleFilters.cpp - crate::process_1pole_tpt_lowpass / crate::process_1pole_tpt_highpass
one pole from: http://www.willpirkle.com/Downloads/AN-4VirtualAnalogFilters.pdf (page 5) - crate::FixedOnePole
Fixed one pole with setable pole and gain. Implementation taken from tubonitaub / alec-deason from https://github.com/alec-deason/virtual_modular/blob/4025f1ef343c2eb9cd74eac07b5350c1e7ec9c09/src/simd_graph.rs#L4292 under MIT License - crate::process_hal_chamberlin_svf
Hal Chamberlin's State Variable (12dB/oct) filter https://www.earlevel.com/main/2003/03/02/the-digital-state-variable-filter/ Inspired by SynthV1 by Rui Nuno Capela, under the terms of GPLv2 or any later: - crate::process_simper_svf
Simper SVF implemented from https://cytomic.com/files/dsp/SvfLinearTrapezoidalSin.pdf Big thanks go to Andrew Simper @ Cytomic for developing and publishing the paper. - crate::process_stilson_moog
Stilson/Moog implementation partly translated from here: https://github.com/ddiakopoulos/MoogLadders/blob/master/src/MusicDSPModel.h without any copyright as found on musicdsp.org (http://www.musicdsp.org/showone.php?id=24). It's also found on MusicDSP and has probably no proper license anyways. See also: https://github.com/ddiakopoulos/MoogLadders and https://github.com/rncbc/synthv1/blob/master/src/synthv1_filter.h#L103 and https://github.com/ddiakopoulos/MoogLadders/blob/master/src/MusicDSPModel.h - crate::DCBlockFilter
translated from Odin 2 Synthesizer Plugin Copyright (C) 2020 TheWaveWarden under GPLv3 or any later - crate::cubic_interpolate
Hermite interpolation, take from https://github.com/eric-wood/delay/blob/main/src/delay.rs#L52 Thanks go to Eric Wood! For the interpolation code: MIT License, Copyright (c) 2021 Eric Wood - crate::TriSawLFO
Adapted from https://github.com/ValleyAudio/ValleyRackFree/blob/v1.0/src/Common/DSP/LFO.hpp ValleyRackFree Copyright (C) 2020, Valley Audio Soft, Dale Johnson Adapted under the GPL-3.0-or-later License. - crate::PolyBlepOscillator
PolyBLEP by Tale (slightly modified) http://www.kvraudio.com/forum/viewtopic.php?t=375517 from http://www.martin-finke.de/blog/articles/audio-plugins-018-polyblep-oscillator/ - crate::VPSOscillator
This oscillator is based on the work "VECTOR PHASESHAPING SYNTHESIS" by: Jari Kleimola*, Victor Lazzarini†, Joseph Timoney†, Vesa Välimäki* *Aalto University School of Electrical Engineering Espoo, Finland; †National University of Ireland, Maynooth Ireland See also this PDF: http://recherche.ircam.fr/pub/dafx11/Papers/55_e.pdf - crate::Oversampling
Loosely adapted from https://github.com/VCVRack/Befaco/blob/v1/src/ChowDSP.hpp Copyright (c) 2019-2020 Andrew Belt and Befaco contributors Under GPLv-3.0-or-later Which was originally taken from https://github.com/jatinchowdhury18/ChowDSP-VCV/blob/master/src/shared/AAFilter.hpp Copyright (c) 2020 jatinchowdhury18 - crate::next_xoroshiro128
Taken from xoroshiro128 crate under MIT License Implemented by Matthew Scharley (Copyright 2016) https://github.com/mscharley/rust-xoroshiro128 - crate::u64_to_open01
Taken from rand::distributions Licensed under the Apache License, Version 2.0 Copyright 2018 Developers of the Rand project. - crate::SplitMix64
Copyright 2018 Developers of the Rand project. Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or https://www.apache.org/licenses/LICENSE-2.0> or the MIT license <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your option. This file may not be copied, modified, or distributed except according to those terms. - splitmix64 (http://xoroshiro.di.unimi.it/splitmix64.c) - crate::f_distort / crate::f_fold_distort
Ported from LMMS under GPLv2 * DspEffectLibrary.h - library with template-based inline-effects * Copyright (c) 2006-2014 Tobias Doerffel <tobydox/at/users.sourceforge.net> Original source seems to be musicdsp.org, Author: Bram de Jong see also: https://www.musicdsp.org/en/latest/Effects/41-waveshaper.html - crate::PolyIIRHalfbandFilter
Taken from va-filter by Fredemus aka Frederik Halkjær aka RocketPhysician https://github.com/Fredemus/va-filter Under License GPL-3.0-or-later originally translated from the freely available source code at https://www.musicdsp.org/en/latest/Filters/39-polyphase-filters.html - crate::DCFilterX4
Basic 4 channel SIMD DC-filter from Understanding Digital Signal Processing by Richard Lyons. Taken from va-filter by Fredemus aka Frederik Halkjær aka RocketPhysician https://github.com/Fredemus/va-filter Under License GPL-3.0-or-later - crate::fh_va::LadderFilter / crate::fh_va::Svf / crate::fh_va::SallenKey
Taken from va-filter by Fredemus aka Frederik Halkjær aka RocketPhysician https://github.com/Fredemus/va-filter Under License GPL-3.0-or-later
Modules
This module contains the VA filter code of Fredemus’ aka Frederik Halkjær aka RocketPhysician.
Macros
This macro allows you to float compare two vectors to a precision of
0.0001,
only every $decimate element will be looked at though. Useful for keeping the fixed
value tables in your DSP code small.Calculates the (linear) slope between consequtive values in
$vec and compares the slopes
with $cmp_vec with a precision of 0.0001. This macro only looks at every $decimate slope.Calculates the (linear) slope between consequtive values in
$vec and compares the slopes
with $cmp_vec with a precision of 0.0000001. This macro only looks at every $decimate slope.Calculates the (linear) slope between consequtive values in
$vec and compares the slopes
with $cmp_vec with a precision of 0.000000001. This macro only looks at every $decimate slope.Calculates the (linear) slope between consequtive values in
$vec and compares the slopes
with $cmp_vec with a precision of 0.0001.This macro allows you to float compare two vectors to a precision of
0.0001.Holds the previous
state.current value for $time_ms.Increases/Decreases
state.current value until you are at $value within $time_ms.Increases/Decreases
state.current value until you are at $value within an adjusted $time_ms.
Depending on how close state.start is to $value, $time_ms is linearily shortened.
For this to work, you need to supply the supposed starting value of the envelope.Returns the name of the distortion selected by the
dist_type parameter of the apply_distortion
function.Structs
An all-pass filter based on a delay line.
An atomic float that can be asynchronously read/written. Best combined with an
Arc<...>.The AtomicFloatPair can store two
f32 numbers atomically.2nd order IIR filter implemented in normalized Direct Form I.
Signal change detector that emits a trigger when the input signal changed.
Trigger signal detector with custom range.
Basic 4 channel SIMD DC-filter from Understanding Digital Signal Processing by Richard Lyons.
Dattorro plate reverb implementation.
This is a delay buffer/line with linear and cubic interpolation.
A retriggerable AD (Attack & Decay) envelope with modifyable shapes for the attack and decay.
Envelope state structure for the macros crate::env_hold_stage,
crate::env_target_stage and crate::env_sustain_stage.
Gate signal generator for HexoDSP nodes.
Implements oversampling with a ratio of N and a 4 times cascade
of Butterworth lowpass filters (~48dB?).
This is a band-limited oscillator based on the PolyBlep technique.
This is a polyphase iir halfband filter (cutoff at fs/4) consisting of cascades of allpasses.
translated from the freely available source code at https://www.musicdsp.org/en/latest/Filters/39-polyphase-filters.html
A ramped value changer, with a configurable time to reach the target value.
Random number generator based on xoroshiro128.
Requires two internal state variables.
You may prefer SplitMix64 or Rng which only use one
u64 as state.Random number generator based on SplitMix64.
Requires two internal state variables. You may prefer SplitMix64 or Rng.
A slew rate limiter, with a configurable time per 1.0 increase.
A splitmix64 random number generator.
An LFO with a variable reverse point, which can go from reverse Saw, to Tri
and to Saw, depending on the reverse point.
Trigger signal generator for HexoDSP nodes.
Trigger signal detector for HexoDSP.
Generates a phase signal from a trigger/gate input signal.
Vector Phase Shaping Oscillator.
The parameters
d and v control the shape of the sinus
wave. This leads to interesting modulation properties of those
control values.Constants
The threshold, once reached, will cause a trigger event and signals
a logical ‘1’. Anything below this is a logical ‘0’.
The lower threshold for the schmidt trigger to reset.
! Contains various utilities for trigger signals in a modular synthesizer.
Traits
Functions
Converts a coefficient/factor to decibels
Linear crossfade.
Linear crossfade with clipping the
v2 result.Constant power crossfade.
Linear (f32) crossfade with driving the
v2 result through a tanh().Exponential crossfade.
Logarithmic crossfade.
Hermite / Cubic interpolation of a buffer full of samples at the given index.
len is the buffer length to consider and wrap the index into. And fract is the
fractional part of the index.
Signal distortion by Bram de Jong.
Foldback Signal distortion
A faster implementation of cosine. It’s not that much faster than
Rust’s built in cosine function. But YMMV.
A faster implementation of sine. It’s not that much faster than
Rust’s built in sine function. But YMMV.
Converts gain in decibels to a factor/coeffient
Initializes [WHITE_NOISE_TAB]
Apply linear interpolation between the value a and b.
Apply 64bit linear interpolation between the value a and b.
Given the mutable
state generates the next pseudo random number.Converts a midi note (0 to 128) to a frequency
Process a very simple one pole 6dB high pass filter.
Useful in various applications.
Process a very simple one pole 6dB low pass filter.
Useful in various applications, from usage in a synthesizer to
damping stuff in a reverb/delay.
Process a very simple one pole 6dB high pass filter in TPT form.
Useful in various applications.
Process a very simple one pole 6dB low pass filter in TPT form.
Useful in various applications, from usage in a synthesizer to
damping stuff in a reverb/delay.
Process a HAL Chamberlin filter with two delays/state variables that is 12dB.
The filter does internal oversampling with very simple decimation to
rise the stability for cutoff frequency up to 16kHz.
This function processes a Simper SVF with 12dB. It’s a much newer algorithm
for filtering and provides easy to calculate multiple outputs.
This function implements a simple Stilson/Moog low pass filter with 24dB.
It provides only a low pass output.
A
pow like shape function for exponential envelopes.
It’s a bit faster than calling the pow function of Rust.Cheap 4 channel tanh to make the filter faster.
Another tanh approximation. See also tanh_levien.
Maps any
u64 to a f64 in the open interval [0.0, 1.0).