search

sot / proseco

Probabilistic star evaluation and catalog optimization
https://sot.github.io/proseco
BSD 3-Clause "New" or "Revised" License
1 stars 0 forks source link

What should we do with pixels with large negative dark current? #265

Closed jeanconn closed 5 years ago

jeanconn commented 5 years ago

For star selection and review how should a pixel with a reported value like the -8570 below be handled? Is it a bad/broken pixel that should just be completely avoided for guide stars and fids? Or should it just be set to 0 so it doesn't cause numeric issues?

ipdb> img
<ACAImage row0=-215 col0=339
array([[  142,   147,    12,    42,    14,    13,    21,   190],
       [   18,    34,    36,   524,    83,    12,    28,    26],
       [ -408, -8570,    26,    42,    29,   126,    26,    25],
       [  179,   183,    85,    68,   359,    21,    24,    17],
       [  187,     0,    38,   612,    30,    27,     8,     9],
       [   27,   -12,    17,    86,    35,   153,    19,   133],
       [   16,    57,    11,    16,    18,    27,    62,    27],
       [   27,   205,    27,    18,    69,    15,    23,     1]])>
taldcroft commented 5 years ago

Going OT, at some point in the not-too-distant future you should process a couple of SRDCs and see if this value persists. Most likely not since it arises from a large value in im1.fits in 2019:002 and is not in im2.fits. My recollection is that the 5-sec read generally occurs before the 10 sec, so it is hard to imagine this is real. Nevertheless...

It is fair to just clip at 0.0. I'm almost inclined to put that upstream in mica dark_cal as a default value for clipping (with the option to set to None for no clipping).

But in any case clipping to 0 in proseco is just fine. The real dark current cannot be negative.

jeanconn commented 5 years ago

"My recollection is that the 5-sec read generally occurs before the 10 sec, so it is hard to imagine this is real. Nevertheless..." I just couldn't wrap my mind around the pixel having a large value in the 5 second readout and a small value in 10 second readout for each replica.

jeanconn commented 5 years ago

And I had the SRDC thought as well when we were wondering about the star selection, but we'll get there eventually.

jeanconn commented 5 years ago

Closed by mica 3.16 .