Confusing noise

[D4N]

It was nice, cold (-5) and clear out last night so I took a some images and tried out some different settings. However the image noise isn't really behaving as I would have expected.

I seem to get lots of hot green pixels on my 5 min ISO 800 subs (I have been fighting these a lot in my processing), on the 10 min sub there are far less.

The 5 min ISO 3200 sub has far less noise than the ISO 800 subs, this may be partly due to the fact that the R pixels are turned off above ISO 1000 but it's still odd.

I attach three consecutive images of the Soul nebula taken with my Fuji finepix S5 pro. These are single uncalibrated subs.

I have also attached identical crops from the centre of the images.

All images have been debayered and had an identical single stretch in PI, no other processing done and in camera noise reduction is off.

If anyone has any thoughts on this they would be appreciated.

5min ISO 800

10min ISO 800

5min ISO 3200

Crops in the same order:

TSED70Q, iOptron Smart EQ pro, ASI-120MM, Finepix S5 pro.

[andrewluck]

Can't really tell what's going on from these images. Can you make the original frames available please.

Some master calibration frames would also be very useful if you have them.

Andrew

[D4N]

Ok here is my master bias for ISO 800.

I stretched it to make the noise visible as otherwise it's black.

These frames are as original as I can post here, the non stretched frames are kinda dark so won't show anything as a jpg if I don't give them a stretch.

I can post my master dark and flat when I get on my laptop.

What I want to get rid of is the green hot pixels and I don't understand why they go away in the longer exposure.

TSED70Q, iOptron Smart EQ pro, ASI-120MM, Finepix S5 pro.

[andrewluck]

You need to upload to a service like Dropbox. The trouble is, you can only assess noise on an un-stretched, original size image.

Many cameras mess with the image even if it's saved as a supposed RAW and the manufacturers won't reveal what processing is being done.

Andrew

[D4N]

Ok I'll try to do that but I'm travelling at the moment.

What's confusing me is that the noise isn't behaving as expected. The longer sub should have a better SNR but there should be more noise not less.

I would also have expected the noise to go up when changing from ISO 800 to ISO 3200, instead it went down a lot.

I would like to be able to understand what's happening with the noise so I can avoid it.

TSED70Q, iOptron Smart EQ pro, ASI-120MM, Finepix S5 pro.

[andrewluck]

As I said in my last post, the camera is processing the image before you get to download it.

One comment you made I didn't understand: "the R pixels are turned off above ISO 1000". Explain please?

Andrew

[D4N]

The fuji super CCD has large S pixels and smaller R pixels arranged in a Hexagonal pattern. This is designed to give a higher dynamic range.

Above ISO 1000 only the larger S pixels are used.

TSED70Q, iOptron Smart EQ pro, ASI-120MM, Finepix S5 pro.

[andrewluck]

Ah, now I understand. I thought you meant Red  .  I've also been doing some reading on this camera model which has helped.

As I suspected, turning NR off only 'sort of' turns it off. This article is worth a read, especially the section on noise: http://www.bythom.com/s5review.htm

Andrew

[D4N]

Ah so the noise reduction scales with exposure length and gain settings. I hadn't seen that review before.

This begins to make more sense, I realised that as with most DSLRs the camera doesn't output a truly raw file but I didn't realise that the noise reduction was so harsh when it is turned off.

It doesn't seem to be eating the stars though so not all bad.

TSED70Q, iOptron Smart EQ pro, ASI-120MM, Finepix S5 pro.

[andrewluck]

The article also explains why your images are rotated 45 degrees. Not sure I could process those. Guess you need the Fuji software first?

Andrew

[D4N]

I had worked out the rotation thing, it did throw me at first though.

Processing is fine, just go as normal then dynamic crop before exporting.

It's actually kinda nice having the black corners as they are truly empty of data so they give a good black point comparison

Sent from my iPad using Tapatalk