LIFE Interpolation Interface

We have developed a C++ package to provide a fast implementation of the windowed sinc interpolation of the LIFE interferometer data. It is developed in a package lifeinterif

Installation

The lifeinterpif package can be installed on 64 bits Windows, Python 3.7 using a pre-built wheel located on the arg server:

pip install --upgrade lifeinterpif -f https://arg.usask.ca/wheels/

A setup.py script is provided so users can build the the wheel for other operating systems, eg python setup.py bdist_wheel

Note

The variables include_dirs and library_dirs inside setup.py script will need to be modified on different systems. include_dirs must specify the path to the C++ boost installation as wells as the numpy include files inside python. library_dirs must specify the path to the C++ boost libraries. You may aslo have to explicitly include the boost libraries in the libraries variable.

Usage

The lifeinterpif package is a shallow swig wrapper for the C++ code and has been developed for development convenience rather than user ergonomics. It is recommended that another layer of wrapper code be placed around the package if ergonomic interfaces are desired. An example of usage is as follows:

>>> import lifeinterpif.lifeinterpif
>>> ok,Ginterp = lifeinterpif.lifeinterpif.windowed_sinc_resample(  tnew,  G, Gt, Fs, num_sinc, len_Ginterp)
windowed_sinc_resample(tnew, G, Gt, Fs, num_sinc, len_Ginterp)

Performs the windowed sinc resampling algorithm to the LIFE interferometer data.

Parameters:
  • tnew (numpy.ndarray(float)) – An array of unevenly spaced times at which interpolated signals are required. The array must be in ascending order. For atmospheric measurements this array is typically around 7840 elements. Read-only.
  • G (numpy.ndarray(float)) – The array of evenly spaced samples for either one pixel or 8 pixels. These are the values interpolated by the code. If processing one pixel it will have a size around 177,000 and will be the same size as Gt. If processing 8 pixels it will be exactly 8 times the size of Gt. The 8 samples for the 8 pixels at one instant in time must be ordered consecutively in memory. Read-only.
  • Gt (numpy.ndarray(float)) – The array of times of the evenly spaced samples. It is typically around 177,000 elements in size. It muts be in ascending order. (read only).
  • Fs (float) – The frequency scaling (Hz) to be applied to time measurements. Typically around 78125.0
  • num_sinc (int) – The total number of points from the evenly sampled data to be included in the windowed sinc interpolation. Typically half the points are to the left of each unevenly spaced time point and half are to the right
  • len_Ginterp (int) – The length of the sampled output array returned to the user. This will either be the same as the size of array tnew or 8 times the size of array tnew if 8 pixels are processed.
Returns:

Returns a two element tuple. The first element is a boolean indicating success of failure. The second element is a 1-D numpy array of size len_Ginterp containing the data interpolated to the unevenly sampled time points in tnew. It will either be the same size as tnew or it will be 8 times the size of tnew if 8 pixels are simultaneously processed. The output array is undefined and may contain garbage if the first element of the returned tuple is not true.
Return type:

Tuple[ bool, nnumpy.ndarray(float)]