search

AstroHuntsman / huntsman-mocks

MIT License
0 stars 1 forks source link

investigate gunagala dark ring in PSF issue #22

Open lspitler opened 6 years ago

lspitler commented 6 years ago

Was apparent in gunagala make_noiseless_image output data of Lee's hacked solution for inputing simulation data as a "star".

bazkiaei commented 5 years ago

@AnthonyHorton @lspitler

The function mocks.create_mock_galaxy_noiseless_image uses gunagala package to create noiseless images. I subtract the output of mocks.create_mock_galaxy_noiseless_image by its median and limited the plot between 0 and 2 ADU (I know the unit seems odd, let's talk about that later). Then I plotted it and here is the plot:

noiseless

As you can see there are dark rings where we do not expect them. Here is a zoom in to the centre of the above image:

zoom-noiseless

lspitler commented 5 years ago

@AnthonyHorton any idea what might cause this?

AnthonyHorton commented 5 years ago

The example uses mocks.create_mock_galaxy_noiseless_image only? No other processing, e.g. convolution?

All the mock code seems to be doing is creating an Imager with the PSF set to match the simulated data, and generating a noiseless image of a single star. So aside from intensity scaling and unit fiddling the only thing that is being done to the simulation data is to resample it onto the instrument pixel grid. In the example shown what are the simulation data pixel scale, output pixel scale, and oversampling factor?

bazkiaei commented 5 years ago

The example uses mocks.create_mock_galaxy_noiseless_image only? No other processing, e.g. convolution?

All the mock code seems to be doing is creating an Imager with the PSF set to match the simulated data, and generating a noiseless image of a single star. So aside from intensity scaling and unit fiddling the only thing that is being done to the simulation data is to resample it onto the instrument pixel grid. In the example shown what are the simulation data pixel scale, output pixel scale, and oversampling factor?

Yes, it uses mocks.create_mock_galaxy_noiseless_image only. The simulation resolution is 170 pc / pixel, the pixel scale for 10 Mpc distance to the galaxy is equal to 3.52 arcsec / pixel and oversampling is 10.

AnthonyHorton commented 5 years ago

Can you show a side by side comparison of the simulation data and the noiseless image, for the region with dark rings?

bazkiaei commented 5 years ago

Can you show a side by side comparison of the simulation data and the noiseless image, for the region with dark rings?

This is one:

side

and this is another one in more zoom:

side-zoome

AnthonyHorton commented 5 years ago

Hmmm. Seems to be an artefact of the resampling process in gunagala then. Might need to look the algorithm it's using for that and find a way to suppress these spurious high frequency components.

Possible that it become a non-issue when used 'properly' though, i.e. providing extended source objects and instrument PSFs separately. The expectation for the pixelated PSFs in gunagala was that pixels in the PSF input data would be smaller than the image sensor pixels, ideally by factor of a few, but here it's the other way around.