Crate vecfit 

Pure-Rust relaxed vector fitting for rational approximation of frequency-domain data. Fits scalar, vector, and matrix-valued responses.

API overview

• Model — fitted rational approximation (poles, residues, evaluation)
• Options — solver configuration (pole count, convergence, weighting)
• Csv and Touchstone — parse sampled data from common file formats
• hz, rad, real, complex — axis wrappers for unit conversion
• Shape and Layout — describe the structure of flattened response data

Examples

Fit sampled data from CSV:

use vecfit::{Csv, Options};

let csv = "freq_Hz,re_f1,im_f1\n1,0.95,-0.31\n5,0.35,-0.72\n10,0.10,-0.98\n50,0.02,-1.0\n100,0.01,-1.0\n";
let model = Csv::from_csv(csv)?.fit(Options::new().poles(2))?;

assert!(model.shape().is_scalar());

Fit an analytic transfer function via closure:

use num_complex::Complex64;
use vecfit::{Options, Model, complex};

let s: Vec<Complex64> = (0..80)
    .map(|k| Complex64::new(0.0, 1.0 + k as f64))
    .collect();

let model = Model::fit(
    complex(&s),
    |s| 0.05 + 1.2 / (s + 3.0) + 0.4 / (s + 15.0),
    Options::new().poles(4),
)?;

let fitted = model.eval_scalar(&s)?;
assert_eq!(fitted.len(), s.len());

Multi-channel vector fit using the hz axis wrapper:

use vecfit::{Options, Model, hz};

let freq: Vec<f64> = (1..60).map(|k| k as f64).collect();
let model = Model::fit(
    hz(&freq),
    |f| {
        let w = 2.0 * std::f64::consts::PI * f;
        [1.0 / (1.0 + w), 0.6 / (3.0 + w), 0.3 / (9.0 + w)]
    },
    Options::new().poles(4),
)?;

assert_eq!(model.channels(), 3);

JSON round-trip:

use num_complex::Complex64;
use vecfit::{Options, Model, complex};

let s: Vec<Complex64> = (0..48)
    .map(|k| Complex64::new(0.0, 2.0 + k as f64))
    .collect();
let model = Model::fit(
    complex(&s),
    |s| vec![1.0 / (s + 2.0), 0.5 / (s + 8.0)],
    Options::new().poles(3),
)?;

let json = model.to_json()?;
let loaded = Model::from_json(&json)?;
assert_eq!(loaded.channels(), model.channels());

Re-exports

pub use crate::axis::AsComplexAxis;

pub use crate::axis::Axis;

pub use crate::axis::ComplexAxis;

pub use crate::axis::Hz;

pub use crate::axis::IntoAxis;

pub use crate::axis::RadPerSec;

pub use crate::axis::RealAxis;

pub use crate::axis::c;

pub use crate::axis::complex;

pub use crate::axis::hz;

pub use crate::axis::rad;

pub use crate::axis::real;

pub use crate::emt::ChannelStateSpace;

pub use crate::emt::DiscreteChannelStateSpace;

pub use crate::emt::DiscreteStateSpaceModel;

pub use crate::emt::DiscretizationMethod;

pub use crate::emt::RealSection;

pub use crate::emt::RealSectionChannel;

pub use crate::emt::RealSectionModel;

pub use crate::emt::StateSpaceModel;

pub use crate::error::Result;

pub use crate::error::VecfitError;

pub use crate::fit::AutoPoles;

pub use crate::fit::Options;

pub use crate::fit::ProblemRef;

pub use crate::fit::Report;

pub use crate::fit::SampleMatrix;

pub use crate::fit::SampleMatrixRef;

pub use crate::fit::SolverPolicy;

pub use crate::fit::SolverUsed;

pub use crate::fit::WeightStrategy;

pub use crate::io::ComplexModelJson;

pub use crate::io::Csv;

pub use crate::io::ParsedSamples;

pub use crate::io::RealKernelJsonModel;

pub use crate::io::RealKernelPoleJson;

pub use crate::model::ChannelErrors;

pub use crate::model::Model;

pub use crate::model::ModelParts;

pub use crate::shape::FlatResponse;

pub use crate::shape::IntoResponse;

pub use crate::shape::Layout;

pub use crate::shape::ResponseSample;

pub use crate::shape::ResponseScalar;

pub use crate::shape::Shape;

pub use crate::touchstone::DataFormat;

pub use crate::touchstone::FrequencyUnit;

pub use crate::touchstone::ParameterType;

pub use crate::touchstone::Touchstone;

pub use crate::touchstone::TouchstoneOptions;

Modules

axis

Axis types for specifying how sample points map to the complex plane.

emt

error

fit

io

model

shape

touchstone

Touchstone v1 (.sNp) file parser.