27 October 2018 at 7:51 pm #574153
Used BASS in the past but not extensively. Using it more now to process spectra from home made spectroscopes and it seems easier than setting up ISIS for a different user config (from my ALPY). So I got to grips with Andy’s (excellent) BASS tutorial. A couple of questions if I may…
1. If I’m trying to replicate the ISIS/Demetra type process (using a ‘standard’ star to compensate for atmosphere and instrumental effects) is the attached process diagram correct?
2. Got stuck with the ‘Hot’ pixel removal in one particular spectrum. Thought I’d dealt with them but I got two spikes in a spectrum that I can’t account for. I’m certain they’re not genuine. The issue is they’re ‘hot’ but not hotter than the top-end counts for the spectrum itself. How do I deal?
Other than that I’m getting to grips with the freehand drawing of continuum curves and producing strange results – but I’m sure its just my hand/eye coordination.
Any help much appreciated.
Attachments:27 October 2018 at 8:54 pm #580103
Can you track the rogue pixels down in the raw images? If it is in every image then it should have been in the master dark and corrected but you could add it to the cosmetic file and zap it that way.
I dont know about BASS but the ISIS cosmics removal tool works on the individual sub exposures where a rogue pixel stands out more clearly than in the combined spectrum image. It compares each pixel with its neighbours, looking at the differential between each pixel and the surrounding ones. You can set the threshold so if a pixel is poking its head above its neighbours, ISIS will zap it. (If you set it too low though, particularly if the spectrum is close to undersampled, it can zap a whole row or spectrum line so you have to use it with caution)
Robin28 October 2018 at 10:53 am #580109Andy WilsonKeymaster
I am glad you are finding the tutorial useful. While ISIS is great for bulk reliable processing, BASS is very flexible and relatively easy to use.
Robin gives good advice on the hot pixels. Similarly I would suggest trying to locate the position(s) in one of the raw spectrum images, then compare that to your master dark frame. When creating a hot pixel map in BASS, you should use your master dark frame, that way you do not need to worry about clipping out any of the spectrum. Then you simply apply the same hot pixel map to all images.
If it turns out to be a frame specific problem, probably a cosmic ray hit, then the BASS “Cosmic, Hot & Cold Pixel Removal” tool allows you to select any size of sub-region to work on and set the threshold. If it is on a single image you could zoom in on the problem region, setting a very small rectangle around the hot pixel, and set the threshold to just pick out that pixel. How well this works will depend on whether the pixel(s) stand out above the spectrum intensity. There is also a tool for editing 1D profiles, but this really should be used as a last resort, and I would be tempted to ditch the sub-frame is there is a low intensity cosmic ray hit that is proving troublesome to remove.
That is a useful workflow you have shared. The one bit that jumps out to me is whether you are applying identical rotate, tilt and active binning regions across all your images. The first 2 of these are particularly important, though for best results you also want the same active binning region. When I use BASS I have all the images loaded, so I can apply the same identical corrections to all images in one go, rather than process the calibration, standard star and target star images separately. Noting this easiest performed on the stacked calibration, standard star and target star images rather than all the individual sub-exposures.
Also, just to check you are not really drawing the response/continuum curve freehand. With BASS you select points in the raw response profile that avoid absorption/emission lines, or any other nasty sharp features, so you get a nice smooth curve.
Andy28 October 2018 at 11:41 am #580111
I really appreciate the detailed response.
Yes, I’d generated a defect map and had applied that. But hadn’t thought about treating a subframe so that’s an easy fix for me.
I’d applied the same rotate, tilt and binning by default. That is, without specific intent but I’ll ensure that i do that now. I’m assuming you mean the same ‘size’ binning area (and not the same physical position on the frames).
With regards the ‘freehand’ curve I followed the same technique as in ISIS. I do have the ‘Free Draw’ box ticked in the Response Shaper tool (not the ‘Linearise’ box) but I select multiple points on the spectrum to produce a ‘curve’ as in ISIS. Are you saying I should I use Linearise?
Also, when using a reference spectrum from the Miles db do you routinely use ‘normal’ or ‘Dereddened’?
As for the workflow, let me know if you want it amended and I’ll share with the community (or send to JohnP) if you think it useful.
Regards T28 October 2018 at 11:55 am #580112
Definitely the normal (ie the original) spectrum from MILES for response correction. You want the original spectrum as it was measured, not what it would look like without interstellar exinction. (When you see MILES referenced in literature it is usually the dereddened spectra as it is normally used for modelling of stellar populations in galaxies but fortunately for us the raw uncorrected spectra were also made available)
Robin28 October 2018 at 12:03 pm #580113
I filter the spectra to match the resolution and shift them slightly if needed to line them up before division. With MILES spectra the match is generally very good then and little editing is needed (I edit out any artifacts rather than choosing points on the curve as that way I can smooth less and include more subtle effects in the instrument response. eg as here)
Robin28 October 2018 at 1:17 pm #580114Andy WilsonKeymaster
When creating the response curve in BASS do not tick “Free Draw” or “Linearise”. You should find that neither is needed unless you have a particular problem you need to overcome. The “Free Draw” means you can click anywhere, which is usually a bad thing as it means you can click outside the response curve. Leaving it unticked forces each point you click to sit on the curve. Of course the proof of whether it works is always to see if your standard star spectrum is adjusted to be a good match to the Miles or other reference spectrum.
Where possible, I try to make all the spectra I take fall on roughly the same vertical position on the CCD chip. This is a tip I picked up from Olivier Thizy at one of the workshops. It is not essential, but he recommended it as a technique to get the maximum accuracy from spectra. While, flat fields, rotation, tilt, wavelength calibration should all work to allow you to place the spectrum anywhere on the chip, by placing the spectrum in the same physical location on the chip it means these corrections need to do ‘less work’. That is always a good thing as it means the corrections tweak the result rather than have to be relied on for large corrections. In reality my spectra are never exactly on top of the same chip position, but I try to get within a few pixels.
It might be an idea to send the workflow to John for comment before publishing. With my tutorial I posted it to the BASS Yahoo Group, and I got useful feedback from the user community enabling me to improve the tutorial. My preference would be not to have the “Lamp”, “Standard” and “Target” as 3 separate top level streams. Instead using the top level sections of the tutorial as the top level streams. That way it enforces the same geometric corrections on each image or master image.
Andy28 October 2018 at 9:58 pm #580118
So much so, a plane has just tried to land in my garden.
- You must be logged in to reply to this topic.