ISIS cosmic rays filter problem with Alpy

Design, construction, tuning of spectroscopes
Information and discussion about softwares (telescope remote, autoguiding, acquisition, spectral processing ...)

ISIS cosmic rays filter problem with Alpy

Postby Peter Somogyi » Sat Apr 14, 2018 9:12 pm

Hello,

It seems I've hit either a severe bug or a misuse of "Cosmic rays filter" in ISIS with the Alpy (f/4 Newton, ATIK 414, bin1x1, 18 and maybe 23 micron slit).

In the past, always enabled this feature above 5 minute by default.
When the signal is high, enabling the checkbox "Cosmic rays filter" alters the result this way, like cutting the sharp peaks (red is the filtered):
tcrb_alpy600_cosme_vs_nocosme.png
tcrb_alpy600_cosme_vs_nocosme.png (43.87 KiB) Viewed 475 times

Not to mention, this filter is be rather harmful when signal is high.
FHWM Y: = 2.22 - 2.45
ISIS version: 5.7.0 (I know behind the latest, but later features in the whatsnew list are all out of my use case).

I did notice this feature is harmful for the reference stars already, but never thought real spectra - mostly emissions - are affected so badly.

The manual here is hard for me to interpret:
http://www.astrosurf.com/buil/isis/quic ... ced_us.htm
"... Do not filter the cosmic rays in case of need (when observing faint objects)."

Does it read as we should leave it switched off for all the non-faint objects? (for the UV, yet long exposures needed, cosmic rays possible)
Note that altering the cosmic filter "value" between 10-100-10000 doesn't change anything.

For now, just alerting the Database that many of my high-SNR Alpy results might be hit by this error, and requires a large review + reprocessing effort from my side in the future.

Peter

PS.: a fits header update containing the altered range would be welcome.
Peter Somogyi
 
Posts: 331
Joined: Sun Jul 13, 2014 8:56 am

Re: ISIS cosmic rays filter problem with Alpy

Postby Christian Buil » Sun Apr 15, 2018 10:29 pm

A tool like the "cosmic removal filter" works by using the pixel values information around the processed pixel. If the spectrum is not correctly sampled (more than 3 pixels per FWHM) there may be a modification of the useful signal, because the algorithm spatially filters some true details.

Any deviant pixel detection algorithm that works on a single image can pose this type of problem. The problem is not new (see discussion here about the eShel spectrograph in particular and bin 2x2 acquisition, for example).

Ideally, the optimal sampling of a spectrum is between 3 and 5 pixels per FWHM. Unfortunately, the design of the spectrographs is not always perfect on this point and the size of the camera pixels are often too big (I recommend working rather in binning 1x1, but for your setup (Atik414), the problem is present because the Alpy spectrum image is too "sharp"). In this case, do not use this type of filter. Sorry.

The situation of very faint objects is peculiar: we first try to detect, and in this situation, a slight alteration of the profile is less serious (it is often masked by the weak signal to noise to noise ratio).

Christian Buil
Christian Buil
 
Posts: 1148
Joined: Mon Sep 26, 2011 6:59 pm

Re: ISIS cosmic rays filter problem with Alpy

Postby Peter Somogyi » Mon Apr 16, 2018 6:13 am

Hi Christian,

Thank you Christian for the clarification.

I disagree it was a spectrograph design problem - in the aspect of case of cosmic rays filtering at least. (On the contrary, any spectrograph optimized for faint objects should exactly produce sharp spectrum.)

Such an algorithm should get activated just above a given ADU threshold, around a few pixel surrounding. That an isolation is trivial to implement, even if less effective.
It's also a requirement that we report any alteration of the binning zone to professionals in details like which wavelength regions was altered, in case it was the binning zone alteration. Without this awareness and what the algorithm exactly does - anything else than the standard processing -, the whole produced spectrcum may be deceptive!

Firstly, the help web page must be modified accordingly, removing this false sentence: "... without removing actual details of the spectrum".

As a workaround, I'll preprocess my Alpy images before processing with ISIS.

Peter
Peter Somogyi
 
Posts: 331
Joined: Sun Jul 13, 2014 8:56 am

Re: ISIS cosmic rays filter problem with Alpy

Postby Robin Leadbeater » Mon Apr 16, 2018 10:43 am

Hi Peter,

I have seen similar effects. It can be because the difference between the peak pixel and the adjacent pixels in a strong line are higher than the cosmic ray filter threshold. The algorithm then reduces the peak pixel value to that of the adjacent pixels. Try increasing the threshold ("settings", "cosmic ray filter")

(Any filter has the possibility to alter the signal so it is always a good idea to check with it on and off. I have seen this filter remove a complete row when the spectrum width was narrow in good seeing !)

Robin
LHIRES III #29 ATIK314 ALPY 600/200 ATIK428 Star Analyser 100/200 C11 EQ6
http://www.threehillsobservatory.co.uk
Robin Leadbeater
 
Posts: 1059
Joined: Mon Sep 26, 2011 4:41 pm

Re: ISIS cosmic rays filter problem with Alpy

Postby Peter Somogyi » Tue Apr 17, 2018 5:04 am

Hi Robin,

Thank you for the hints, I agree comparing with the non-filtered version will always be a must.
Changing the parameter in ISIS / last settings tab somehow producing similar results.
My impression was that it's only working in the background zone (= far the biggest zone in my case, something too big and error prone for letting controlled by human eye, definitely something to deal by software). Reading together the web page saying spectrum won't be altered, was encouraging to leaving it on.
I don't think this was the only cosmic ray algorithm, so I will continue experimentation by other tools (e.g. I'll examine soon IRAF/lacos, other suggestions appreciated), and yes will keep eye on their activity.

Peter
Peter Somogyi
 
Posts: 331
Joined: Sun Jul 13, 2014 8:56 am


Return to Spectroscopes and Softwares

Who is online

Users browsing this forum: No registered users and 1 guest

x