[Aavso-photometry] Cosmic Rays

[Michael Newberry]

>
> Hi Michael,
>
> I knew you would have to change sides on this one sooner of later.  That's 
> a relief!

Hi Ben,

If I gave the impression I was taking a side on this, I'm sorry---that was 
not my intention. I was just being scientifically critical based on my own 
experiences. I know for a fact that the Max operation works for this 
purpose---that's old hat. I never said Wolfgang's added steps would not show 
you all the cosmic rays. What I said was that I did not see how adding the 
extra step(s) would give more information than using the Max operator on 
normalized images. I used math arguments to show why I believed that the 
stars and hot pixels would still be there after creating the difference 
image and the ratio image.

    After going through Wolfgang's suggestions using a set of 50 star-field 
images, I found that the second step (the ratio image) does make the CR's 
stand out quite a bit better than I had expected. It is as I anticipated 
that all the same CR information in the Ratio image is also in the 
Difference image and the Max image. After all, it is the Max operator that 
is picking out the positive deviants in the first place. So additional 
processing is polishing up these detections. But the Ratio image diminishes 
the stars far better than I anticipated. Conversely the Difference image 
confrms my expectation that it does little more than remove a large scale 
gradient like a tilt or a hump. I ran the 50 images through the tests with 
and without dark subtraction. If the images are not dark subtracted, then 
the Difference image will also greatly diminish the hot pixels. It was the 
Ratio image that really impressed me in terms of its clarity in showing the 
total number of positive deviants (e.g., cosmic rays) in all 50 images.

    As far as cosmic ray *correction* goes, for a long time we've had a tool 
in Mira Pro that implements the same basic idea as Wolfgang's Ratio idea but 
it works within a single image using neighborhood statistics rather than 
using inter-image statistics. It detects and corrects on the fly but does 
not produce an image showing you where the CR's are located. Maybe that is 
an option to consider adding to that method.

    Certainly the concept of CR correction is an open work in progress. 
Every method has its strength and weakness. This thread now has me now off 
in another direction, revisiting the CR detection issue using some other 
ideas I've been toying with. I had one at 4AM this morning...

>
> I tried to send you those darks over the weekend but the email was kicked 
> back.  Not sure if it was my server or yours that thought the package was 
> too large.

    Email has limits. I didn't know your images would be that large. 
Typically, you are rolling the dice sending anyone attachments larger than 
5MB because many nodes of the internet just reject Email attachments larger 
than that---so whether it gets through depends on the route. I'll send you 
our anonymous ftp where you can dump them. We set a 100MB quota on 
that---will they fit?

>    But now that you are looking at this, I am still interested in the flux 
> that you are finding - you know, hits / sq mm / min sort of thing.  I am 
> getting emails off the list that indicate my counts are lower than what 
> they see.  But the counting is pretty hard to do manually, especially if 
> you don't use  Justins' mostdev.exe.  Maybe you can automate this?
>
    I looked at your mexdev image. Your super-dark frame is curious in that 
there are a lot of low angle CR's. Not every bright thing you see is a CR. 
The vast majority of those single points look to me like shot noise from 
under-corrected hot pixels. You will have to blink or animate the 9 images 
to tell for sure. Another option is to create a pixel mask and apply it to 
the dark frames to kill the known hot pixels, then run the maxdev procedure. 
You could do it afterwards on the combined image but that would cut into the 
profiles of many of the cosmic rays---split one into 2 by fixing a pixel 
that connects it and you erroneously add to the tally ot total cosmic rays.

    My experience has been about 1 "whopper" cosmic ray per square cm per 
minute on a mountain at 7000-9300 feet elevation. I define a "whopper" as a 
really obvious cosmic ray (which isn;t very scientific!). It means a bright 
artifact that looks like it just has to be a cosmic ray. Using that rule for 
combining 9 x 30 minute darks would then give 270 expected CR's per square 
cm. I am guessing your chip is around 4 cm^2, so my value would give ~1100 
CR's in total on that 9x30 minute combination. Just eyeballing your image 
and visually bbreaking it into into 32 horizontal and vertical cells (32 is 
about the square root of 1100), I would then expect about 1 CR per cell. 
Your image looks much higher than that. I would guess we're talking about 
twice my estimate. So you may be sitting near something radioactive---that's 
no joke. Are you in Colorado? Got any other ideas?

> On the source of radiation: a LBNL study analyzed (as only they can) the 
> types of radiation and found that Potassium90 was the most common local 
> source of the gamma radiation's Compton scattering electrons.  It is often 
> a contaminant in glass and sometimes in aluminum and concrete.
>
    Concrete yes---especially if you are in Colorado. That is a different 
beast, though they show up similarly in an image.

Michael Newberry


> Ben Davies
> _______________________________________________
>
> Aavso-photometry mailing list
> Aavso-photometry at mira.aavso.org
> http://www.aavso.org/mailman/listinfo/aavso-photometry
>
>