Using PSFs with SimCADO

Point spread function (PSF) kernels are used by SimCADO to mimic the spread in the beam due to the specific optical configuration of the ELT+MAORY+MICADO optical system.

A PSF kernel is in essence a two dimensional array which describes how the light from a point source is spread out over the detector plane. While SimCADO can accept 2D numpy arrays as input directly from the user, more often than not the user will probably want to use a set of pre-calculated PSFs.

Default PSFs for MICADO

Warning

The default PSFs are approaching their expiry date

The PSF files described below are useful for quick simulations which don’t require a super accurate PSF. Furthermore they were generated in 2015 (MCAO, NOVA), and 2016 (SCAO, LESIA), and only cover a single set of atmospheric parameters.

For newer PSFs, see the section on SCAO PSFs (from AnisoCADO) or contact the SimCADO team about MCAO PSFs (from MAROY)

By default the MICADO instrument package includes PSF kernels (arrays) in FITS format for the MCAO and SCAO observing modes. These default kernel FITS files are found in the MICADO data folder under the names: PSF_MCAO.fits, PSF_SCAO.fits.

SimCADO can be told to use either one of these PSFs by passing the SCOPE_PSF_FILE keyword to either the simcado.run function, or to a UserCommands dictionary:

hdu = simcado.run(... , SCOPE_PSF_FILE="PSF_MCAO.fits")

or:

cmds = simcado.UserCommands()
cmds["SCOPE_PSF_FILE"] = "PSF_MCAO.fits"

Note

The default PSFs are field-constant.

This means that they do not vary over the field of view. For the MCAO option this is the optimal case - a constant AO correction everywhere in the field - however for the SCAO case, this is only valid for the central ~2x2 arcsec.

If we are interested in simulating the full MICADO field of view, it will be unrealistic to use the PSF_SCAO.fits file. More on this below.

Both the MCAO and SCAO files were generated for reasonably good conditions and should be understood as offering an optimistic view of how MICADO will perform.

Field varying SCAO PSF

As stated above the default PSFs contained in the MICADO package are for the case where the PSF doesn’t vary over the field. For the SCAO observation modes this is an unrealistic assumption unless one is only interested in the central ~2x2 arcsec or so. SimCADO (v0.6 and above) accepts a so-called FV-PSF file (field-varying PSF file) in order to simulate the degredation of the PSF over the field of view.

FV-PSF files are quite large (several 100 MB to GB) as they contain PSF kernels for many different positions over the field (and generally for several different wavelengths).

We have generated SCAO FV-PSF files for three observing conditions which roughly correspond to “good”, “median”, and “bad” atmospheric conditions.

Warning

These files are ~1 GB in size.

Profile

Seeing

Zenith Distance

Wind Speed

Turbulence Profile

File Link

Good

0.4

0

8.8

ESO Quartile 1

AnisoCADO_Q1

Median

0.67

30

10

ESO Median

AnisoCADO_Median

Bad

1.0

60

13

ESO Quartile 4

AnisoCADO_Q4

This FV-PSFs were generated by AnisoCADO. AnisoCADO is python package derived from Eric Gendron’s code for generating SCAO PSFs for the ELT+MICADO for any wavelength and off-axis guide star position. To generate custom FV-PSF kernels, see AnisoCADO’s documentation for how to make SimCADO-readable FV-PSF FITS files.

Images and profiles of the central PSF from each of these files can be found here:

To use these PSF files we simply pass the filename to SimCADO (as with the other PSF), and SimCADO (! >=v0.6) does the rest:

fvpsf_path = "path/to/AnisoCADO_SCAO_FVPSF_4mas_EsoQ1_20190328.fits"
simcado.run(... , SCOPE_PSF_FILE=fvpsf_path)

Warning

SimCADO will (probably) run slower when using a FV-PSF file.

This is because SimCADO convolves the FOV of each detector with each PSF kernel that is present inside the borders of that FOV. This means, e.g., if there are 9 different PSFs to be used in the region covered by the central detector chip, then this region will be convolved 9 times with different PSFs (and masked accordingly 9 times). Hence simulating the central chip will take 9 times as long as when using a field-constant PSF.