Noise and light polution question.

[RemcoDutch]

Good evening everyone! 

After my first few posts I have been busy enjoying myself with some untracked photography. Absolutely love the photography part and  a bit frustrated by the editing part 😂. All part of the fun!
Have been getting better at focusing and actually framing/finding the targets (Like Andromeda) with a DIY laser assembly.
On my vacation to Tenerife this year I am planning on picking up a Star Adventurer (or something similiar) to be able to have longer exposures. This will help alot to eliminate noise right?

Yesterday I picked up a Canon 75-300 EF f5.6. I know this is not the best and a (very) cheap lens but alot of people seem to be getting decent results with it.
Yesterday I stacked and today I edited/played around with the following:

292 lights @ 1sec, 300mm, F7.1, ISO1600 (I read that stopping down will help eliminate some CA)
30 flats
50 darks
40 bias

When strecthing the image in GIMP and/or PS CS6 the first few stretches are fine, but then the 5th or so stretch I get MASSIVE red noise / light polution?
I am not sure what it is. The position of the red / orange creep does correspond to the location of the nearest (15miles or so) big city.
Below my stacking and editing results. (REMOVED and REPLACED wrong stacked file, sorry for everyone who downloaded!)

Any people have some good feedback for me?

Autosave.tif

 

 

 

Edited January 5, 2022 by RemcoDutch

[vlaiv]

12 minutes ago, RemcoDutch said:

292 lights @ 1sec, 300mm, F7.1, ISO1600 (I read that stopping down will help eliminate some CA)
30 flats
50 darks
40 bias

Try stacking without darks.

@alacant has been pointing out for some time (and I did not listen), and I've recently read some documentation on Canon DSLRs - they internally modify subs so that darks are useless (they perform a sort of dark removal on long exposure subs).

Calibrate with bias as darks and do dither.

1 second is just simply too short. You have total of less than 5 minutes of exposure.

Good astrophotographs require much more exposure than that - think hours instead of minutes (even if it is half to an hour of total exposure). If you can - do longer single exposures as well. If you want to beat read noise you need longer single exposures.

Longer exposures cause issues with tracking, but if you can - go for at least 30s per exposure.

I'm going to download your linear fits to see what I can do with it ... will post results back.

[alacant]

9 minutes ago, vlaiv said:

read some documentation on Canon DSLR

Please share the documentation. We always recommend turning off in-camera dark removal.

 

Edited January 5, 2022 by alacant

[vlaiv]

1 minute ago, alacant said:

Please share the documentation. 

 

To be honest, it's few bits and bobs around the net.

Let me see if I can find those links again.

In the mean time, here is what I think is happening with Canon subs. There is part of sensor that is covered / shielded from light - like couple of dozen of pixels on one side of sensor (and probably on the bottom of it as well). After exposure is taken - for each row average value of those first 20-30 pixels is taken and subtracted from the rest of the row.

This is like "row by row" dark subtraction, so there is no single dark offset that is removed from whole image, but each row has calculated dark offset that is removed.

Problem is - I haven't managed to extract actual full raw data by any software to try to do dark optimization algorithm and to check bias. Anything I tried produces similar "calibrated" result. DCRaw does it, DNG converter from Adobe - btw. DNG file format supports this by row dark offset as well as one of its options, so check it there as well.

Only thing I have not tried is APT and recording raws directly on computer. I don't know if Canon DSLR SDK reads out true raw or this "calibrated" raw as well.

For example - some of information as well comes from this post:

https://openphotographyforums.com/forums/threads/offset-black-level-in-canon-raw-files.12280/

(pixels are referred to as sensels - like "sensing pixels" or something like that)

@RemcoDutch

Attached tifs seem to be processed images - but have huge resolution - second one is 800mb and 12000x9000 or something like that.

This suggests that you either used drizzle for some reason, or you resized image in PS? Either is unnecessary and former hurts SNR a bit.

Could you post linear (not processed) fits / tiff 32bit stack straight from DSS?

[RemcoDutch]

@vlaiv New file is about 300mb so ALOT smaller .

I added it to the first post.
Curious to see what other people can make from my TIFF.

Looking forward to get to trying longer exposures with a tracker soon.

Thanks for the replay and will read into the black level stuff from canon!

Edited January 5, 2022 by RemcoDutch