pyplnoise 1.4

Arbitrarily long streams of power law noise using NumPy and SciPy.

Warning

This package is no longer maintained. Compatibility with recent versions of Python, NumPy or SciPy cannot be guaranteed.

pyplnoise

Python power law noise generates arbitrarily long streams of power law noise using NumPy and SciPy.

The noise is generated with shaping filters and the methodology closely follows Stephane Plaszczynski, Fluct. Noise Lett. 7: R–R13, 2007. You can also find the article on arXiv:astro-ph/0510081.

pyplnoise consists of a single module providing classes implementing the following noise sources:

• general 1/f^α power law noise with upper and lower frequency limits (class AlphaNoise),
• a fast red (Brownian) noise generator with a lower frequency limit (class RedNoise),
• a convenience alias for pink noise (aka 1/f noise; class PinkNoise),
• and of course white noise (class WhiteNoise).

Quick example

The interface is very simple: just instantiate one of the above classes and run get_sample() to retrieve a single sample or get_series(npts) to retrieve an array of npts samples. Detailed interface documentation is available in the code.

import pyplnoise
import numpy as np

fs = 10. # sampling frequency in Hz

# instantiate a noise source with lower frequency limit 1e-3 Hz,
# upper frequency limit 5 Hz and 1/f^1.5 power spectrum
noisegen = pyplnoise.AlphaNoise(fs, 1e-3, fs/2., alpha=1.5, seed=42)

one_sample = noisegen.get_sample()
many_samples = noisegen.get_series(100000)

Detailed examples

Jupyter notebooks are provided in the /examples directory:

1. Overview of the noise sources and their properties
2. Application example: modeling the random signal errors of a gyroscope (Allan variance of synthetic noise)

Installation

Dependencies

• NumPy ≥ 1.17 (see NEP 19)
• SciPy ≥ 1.3

Installing from PyPI

pip install pyplnoise

Installing directly from GitHub

Download the release tarball and run

python setup.py install

Because everything is contained in the module pyplnoise, you can alternatively just copy the module and the LICENSE file into your project.

You may find pyplnoise useful, if...

• ...you're looking to generate 1/f^α noise with very long correlation times (frequencies ≪ 10^-7 Hz); particularly if your machine has limited memory resources.
• ...you like to superimpose many colored noise sources, possibly sampled at different frequencies and possessing different bandwidths.

You may not find pyplnoise useful, if...

• ...you're looking for a pink noise source for your software synthesizer or other audio stuff. There are lots of interesting solutions for such applications, notably "A New Shade of Pink", the Voss-McCartney Algorithm, which is also available in Python and some highly specialized filters.
• ...you want to generate finite 1/f^α noise streams with relatively short correlation times (frequencies ≥ 10^-7 Hz). In such a case Fourier transform methods are tractable and in some cases these methods deliver higher quality results than the shaping filters used by pyplnoise.