Please add input and reference file names to subtraction headers

[Michael Wood-Vasey]

Would it be possible to add the input and reference filenames to the headers of the subtraction FITS files?

There's probably a whole list of things to request for the subtraction headers, such as the convolution kernel used, etc. but rather than wait for the all-encompassing master list I thought I just start with a couple of easy ones. I know that much of this information is stored in the database, but it would be easy enough to store it in the file header as well so that if someone is just looking at the subtraction file much of the information is already contained in there. Thanks.

[Paul Price]

The 'easy one' is fixed in CVS head:

HIERARCH PPSUB.REFERENCE = 'epoch1.fits' / Subtraction reference
HIERARCH PPSUB.INPUT = 'epoch2.fits' / Subtraction input

Send me your wish list and I'll work those in too.

[Paul Price]

Now also adding a brief description of the kernel parameters.

HIERARCH PPSUB.KERNEL = 'ISIS(20,(1.00,2)(2.00,2),2)' / Subtraction kernel

[Michael Wood-Vasey]

Is there some way to store the numeric values of the kernel parameters?

In addition, an important thing to store will be the kernel sum normalization, since this isn't constrained to be 1 for all kernels. There are some ridiculous subtleties about whether your photons come from your convolved, non-convolved or template or reference image. We don't have to get into the details yet, but we should provide the information about

(a) which way we convolved
(b) which way we subtracted
(c) which way we normalized the flux (this will generally be to the non-convolved image; but it doesn't actually have to be)

One of either (a) or (b) is already covered depending on how we settle on the meaning of INPUT and REFERENCE image.

I'm sure there will be more and we will appreciate your patience and flexibility as we come up with more things.

[Paul Price]

kernel.jpg

[Paul Price]

I have code that generates and writes out a representation of the kernel. The kernel is evaluated at the -1, -0.5, 0, +0.5, +1 positions (relative to the full size) in x and y, and a 5x5 grid is constructed. I expect that this image will eventually be part of the whole IPP system (with filename defined by the FILE.RULES in the camera configuration) and sucked into ippMonitor, but for now it is generated as "kernel.fits". I would propose to add to this FITS tables with the kernel parameters and their values.

[Michael Wood-Vasey]

Sorry, what are the axis of the 5x5 grid? What is each different subimage showing?

[Paul Price]

It's showing the spatial variation of the kernel. Each of the 5x5 subimages are a realisation of the kernel at a particular position. The top left subimage is the convolution kernel for the top left of the image; the centre subimage is the convolution kernel for the centre of the image; etc.

[Michael Wood-Vasey]

Thank you for the clarification. That's simple and straightforward.