27 releases (14 breaking)

0.15.1 Feb 14, 2024
0.14.1 Jan 15, 2024
0.13.0 Nov 1, 2023
0.10.0 Jul 20, 2023
0.3.0 Nov 2, 2021

#168 in Algorithms

Download history 30/week @ 2023-12-22 22/week @ 2023-12-29 66/week @ 2024-01-05 216/week @ 2024-01-12 119/week @ 2024-01-19 107/week @ 2024-01-26 99/week @ 2024-02-02 305/week @ 2024-02-09 150/week @ 2024-02-16 229/week @ 2024-02-23 238/week @ 2024-03-01 120/week @ 2024-03-08 119/week @ 2024-03-15 212/week @ 2024-03-22 296/week @ 2024-03-29 150/week @ 2024-04-05

789 downloads per month
Used in 4 crates

MIT/Apache

125KB
2.5K SLoC

Embedded DSP algorithms

GitHub release crates.io Documentation QUARTIQ Matrix Chat Continuous Integration

This crate contains some tuned DSP algorithms for general and especially embedded use. Many of the algorithms are implemented on integer (fixed point) datatypes.

One comprehensive user for these algorithms is Stabilizer.

Fixed point

Cosine/Sine

[cossin()] uses a small (128 element or 512 byte) LUT, smart octant (un)mapping, linear interpolation and comprehensive analysis of corner cases to achieve a very clean signal (4e-6 RMS error, 9e-6 max error, 108 dB SNR typ), low spurs, and no bias with about 40 cortex-m instruction per call. It computes both cosine and sine (i.e. the complex signal) at once given a phase input.

Two-argument arcus-tangens

[atan2()] returns a phase given a complex signal (a pair of in-phase/x/cosine and quadrature/y/sine). The RMS phase error is less than 5e-6 rad, max error is less than 1.2e-5 rad, i.e. 20.5 bit RMS, 19.1 bit max accuracy. The bias is minimal.

PLL, RPLL

[PLL], RPLL: High accuracy, zero-assumption, fully robust, forward and reciprocal PLLs with dynamically adjustable time constant and arbitrary (in the Nyquist sampling sense) capture range, and noise shaping.

Unwrapper, Accu, saturating_scale()

Unwrapper, Accu, [saturating_scale()]: Tools to handle, track, and unwrap phase signals or generate them.

Float and Fixed point

IIR/Biquad

iir::Biquad are fixed point (i8, i16, i32, i64) and floating point (f32, f64) biquad IIR filters. Robust and clean clipping and offset (anti-windup, no derivative kick, dynamically adjustable gains and gain limits) suitable for PID controller applications. Three kinds of filter actions: Direct Form 1, Direct Form 2 Transposed, and Direct Form 1 with noise shaping supported. Coefficient sharing for multiple channels.

Comparison

This is a rough feature comparison of several available biquad crates, with no claim for completeness, accuracy, or even fairness. TL;DR: idsp is slower but offers more features.

Feature\Crate biquad-rs fixed-filters idsp::iir
Floating point f32/f64
Fixed point i32
Parametric fixed point i32
Fixed point i8/i16/i64/i128
DF2T
Limiting/Clamping
Fixed point accumulator guard bits
Summing junction offset
Fixed point noise shaping
Configuration/state decoupling/multi-channel
f32 parameter audio filter builder
f64 parameter audio filter builder
Additional filters (I/HO)
f32 PI builder
f32/f64 PI²D² builder
PI²D² builder limits
Support for fixed point a1=-2

Three crates have been compared when processing 4x1M samples (4 channels) with a biquad lowpass. Hardware was thumbv7em-none-eabihf, cortex-m7, code in ITCM, data in DTCM, caches enabled.

Crate Type, features Cycles per sample
biquad-rs f32 11.4
idsp::iir f32, limits, offset 15.5
fixed-filters i32, limits 20.3
idsp::iir i32, limits, offset 23.5
idsp::iir i32, limits, offset, noise shaping 30.0

State variable filter

[svf] is a simple IIR state variable filter simultaneously providing highpass, lowpass, bandpass, and notch filtering of a signal.

Lowpass, Lockin

Lowpass, Lockin are fast, infinitely cascadable, first- and second-order lowpass and the corresponding integration into a lockin amplifier algorithm.

FIR filters

hbf::HbfDec, hbf::HbfInt, hbf::HbfDecCascade, hbf::HbfIntCascade: Fast f32 symmetric FIR filters, optimized half-band filters, half-band filter decimators and integators and cascades. These are used in stabilizer-stream for online PSD calculation on log frequency scale for arbitrarily large amounts of data.

Delta Sigma modulator/noise shaper

[Dsm] is a delta sigma modulator/noise shaper in MASH-(1)^K architecture.

Dependencies

~1–1.7MB
~38K SLoC