search

spacetelescope / STScI-STIPS

STScI-STIPS
https://stips.readthedocs.io
13 stars 16 forks source link

Correlated noise between STIPS runs #118

Open benw1 opened 3 years ago

benw1 commented 3 years ago

I am finding that the noise in different STIPS-produced images is correlated. That is, if I run two different bands with 2 different SEED values, the noise peaks in each will be in the same pixels. This caused much trouble for source detection algorithms using the combined significance of counts across different bands.

benw1 commented 3 years ago

This may be a reappearance of issue #21 ... not sure, except it is not using the cpu_pool

benw1 commented 3 years ago

Updated STIPS, and am now retesting...

york-stsci commented 3 years ago

Any results from retesting?

benw1 commented 2 years ago

This correlated noise is still occurring with the latest dev16 version. I'm setting the seed in the ObservationModule() using "random_seed" but the noise is correlated between runs. The background noise doesn't show this. I can see it best if I don't include a background, so it is likely something to do with how the read noise is being added.

york-stsci commented 2 years ago

To confirm, what you are doing is creating two ObservationModule() objects, each with a different input value of random_seed, but getting the same noise values?

benw1 commented 2 years ago

I may have set the seed in the wrong line of the code. Please ignore for the moment while I continue testing.