It is entirely possible (and indeed recommended for scripting) to just use the WebbPSF interface without the GUI, but the GUI will provide a quicker method for simple interactive calculations.
Given that the JWPSF package has been available for several years now, one might ask why do we need a new PSF simulator? From a user’s perspective this new code provides the following enhancements:
Perhaps even more importantly, the underlying codebase has been entirely replaced and revamped. The most significant additions from a programmer’s perspective include:
Most users may skip this section; read on only if you are interested in details of how the computations are performed. Otherwise, jump to Quick Start
The problem at hand is to transform supplied, precomputed OPDs (derived from a detailed optomechanical model of the telescope) into observed PSFs as seen with one or more of JWST’s various detectors. This requires knowledge of the location and orientation of the detector planes, the properties of relevant optics such as bandpass filters and/or coronagraphic image and pupil plane masks, and a model of light propagation between them.
Instrumental properties are taken from project documentation and the published literature as appropriate; see the References for detailed provenance information. Optics may be described either numerically (for instance, a FITS file containing a mask image for a Lyot plane or a FITS bintable giving a spectral bandpass) or analytically (for instance, a coronagraph occulter described as a circle of a given radius or a band-limited mask function with given free parameters).
WebbPSF computes PSFs under the assumption that JWST’s instruments are well described by Fraunhofer diffraction, as implemented using the usual Fourier relationship between optical pupil and image planes (e.g. Goodman et al. 1996). Two specific types of 2D Fourier transform are implemented: a Fast Fourier Transform and a discrete Matrix Fourier Transform.
The familiar Fast Fourier Transform (FFT) algorithm achieves its speed at the cost of imposing a specific fixed relationship between pixel sampling in the pupil and image planes. As a result, obtaining finely sampled PSFs requires transforming very large arrays consisting mostly of zero-padding. A more computationally attractive method is to use a discrete matrix Fourier transform, which provides flexibility to compute PSFs on any desired output sampling without requiring any excess padding of the input arrays. While this algorithm’s computational cost grows as O(N^3) versus O(N log N) for the FFT, the FFT’s apparent advantage is immediately lost due to the need to resample the output onto the real pixel grid, which is an O(N^2) operation. By performing a matrix fourier transform directly to the desired output pixel scale, we can achieve arbitrarily fine sampling without the use of memory-intensive large padded arrays, and with lower overall computation time.
Further optimizations are available in coronagraphic mode using the semi-analytic coronagraphic propagation algorithm of Soummer et al. 2007. In this approach, rather than propagating the entire wavefront from pupil to image and back to pupil in order to account for the coronagraphic masks, we can propagate only the subset of the wavefront that is actually blocked by the image occulter and then subtract it from the rest of the wavefront at the Lyot plane. This relies on Babinet’s principle to achieve the same final PSF with more computational efficiency, particularly for the case of highly oversampled image planes (as is necessary to account for fine structure in image plane occulter masks). See Soummer et al. 2007 for a detailed description of this algorithm.
First, download and install the software (as described in the next page of this document). Then just start python and
>>> import webbpsf >>> webbpsf.gui()
and you should be able to test drive things using the GUI:
Most controls should be self-explanatory, so feel free to experiment. Detailed instructions on the GUI are available here.