Skip to content

Latest commit

 

History

History
59 lines (50 loc) · 2.41 KB

CHANGELOG.rst

File metadata and controls

59 lines (50 loc) · 2.41 KB
Raw

Changelog

For latest release notes, see the GitHub releases page.

v2.0.0

A major overhaul of the code, making it object-oriented and modular.

Features

  • Python 3 compatibility.
  • Class-based code, with specific objects for Observatory, Observation, and class templates for different Sensitivity calculations, including default PowerSpectrum.
  • Clean attrs-based classes.
  • click-based CLI interface.
  • Ability to pickle intermediate classes for re-use.
  • Ability to specify observation frequency more freely (and self-consistently)
  • Removal of all usage of aipy as it is far too complex for a simple calculation such as this.
  • Ability to specify Observation using pyuvdata objects.
  • Useful docstrings throughout.
  • Explicit cosmological calculations based on astropy.
  • More flexible: extra parameters for foreground model and integration time, among others.
  • Agreement with previous versions (not exact agreement, because of the increase in accuracy from using astropy).
  • Tools for plotting output -- and default 1D PS plot from the CLI.
  • All quantities have appropriate units (from astropy).
  • Example documentation, and example configuration files.
  • Configuration files are no longer python files... they are YAML.
  • Added a parameter systematics_mask to PowerSpectrum sensitivity, which enables arbitrary k-modes to be masked out in the sensitivity calculation.
  • track option to Observation. This is an alias for obs_duration but has a closer resemblance to the original 21cmSense v1.
  • New calculate_sensitivity_2d_grid method that makes it easier to obtain a gridded cylindrical power spectrum sensitivity for arbitrary bins.
  • Ability to use builtin observatories that set certain parameters.
  • antpos.hera() can now return a split-core configuration, and also outriggers.

Fixed

  • Bug in dL_df (missing square).
  • Use yaml.SafeLoader instead of yaml.FullLoader.
  • Bug when averaging from 2D to 1D that got about half the sensitivity.

Changed

  • Many computations altered to use numpy vectorization over for loop implementations. Including but not limited to:

    • 2D sensitivity calculation.
    • 1D sensitivity calculation.
    • UVW calculation as a function of time.

  • _utils.find_nearest can solve for the index of an array of inputs.