Author Topic: CMOS/ADTG/Digic register investigation on ISO  (Read 568660 times)

Danne

  • Contributor
  • Hero Member
  • *****
  • Posts: 6039
Re: CMOS/ADTG/Digic register investigation on ISO
« Reply #1200 on: December 11, 2019, 10:32:33 AM »
Ok, I think I found your code now.

Danne

  • Contributor
  • Hero Member
  • *****
  • Posts: 6039
Re: CMOS/ADTG/Digic register investigation on ISO
« Reply #1201 on: December 11, 2019, 11:09:01 AM »
I tested iso 200 with the new tweaks and then with a build at iso 200 without the tweaks. Note. Only very briefly and the test scene wasn´t even static but settings were exactly the same:

Iso 200 no modification(comes out brighter)


Iso 200 modified(comes out darker)


iso 200 unmodified


iso 200 posterization in modified(more testing needs to be performed)?


Ok, so I have no interest in proofing anything bad here. On the contrary I really want this to work. I would suggest someone to do following test.
1 - Record at iso 200 with the exact same settings with modified and unmodified version. Even if brightness will vary.
2 - Match brightness in post and compare shadows for refinements or not.
3 - Maybe try other isos as well and do above test

Recording scene should be chosen with shadows in mind.
If we could take the next step by doing these tests we will immidieately understand if the registry changed are the correct ones etc.

timbytheriver

  • Senior
  • ****
  • Posts: 371
Re: CMOS/ADTG/Digic register investigation on ISO
« Reply #1202 on: December 11, 2019, 11:49:35 AM »
Let's remind ourselves of @a1ex's recent apropos post here first! https://www.magiclantern.fm/forum/index.php?topic=24601.msg221994#msg221994  :)

Here are my comparisons. Black Strength lowered to 0 to bring up shadows in all. Strongly suggest looking at full-res (Load Full Resolution) on the upload site.

1) Stock ISO 200


2) Mod ISO 200 – Brightness matched in MLVApp


3) Mod ISO 200 – Exp raised in camera to compensate for reduced gain ceiling (spotmeter now maxes-out at 95% on the hole of the lampshade)


I know which combination of settings I'd prefer grading from. :)

PS @Danne I'm not seeing posterization [yet]!
5D3 1.1.3
5D2 2.1.2

Danne

  • Contributor
  • Hero Member
  • *****
  • Posts: 6039
Re: CMOS/ADTG/Digic register investigation on ISO
« Reply #1203 on: December 11, 2019, 11:54:47 AM »
Not at my computer atm but how are the first two files when compared? Identical? Will take a look in a while.

If you increase brightness like in third example we again are not comparing in a good way imo.

timbytheriver

  • Senior
  • ****
  • Posts: 371
Re: CMOS/ADTG/Digic register investigation on ISO
« Reply #1204 on: December 11, 2019, 12:00:30 PM »
I'd say that the mod changes the quality of the noise for the better in 2) (matched brightness)

If the exposure were simply brightened without the mod, the highlights would blow. So with the mod + recorrected exp ceiling we retain the complete dynamic range of stock ISO  – but with marked improvement in shadow noise.

Maybe we should be thinking of this as a way of exposing rather than just settings. ?
5D3 1.1.3
5D2 2.1.2

Danne

  • Contributor
  • Hero Member
  • *****
  • Posts: 6039
Re: CMOS/ADTG/Digic register investigation on ISO
« Reply #1205 on: December 11, 2019, 12:11:01 PM »
That's the other aspect. Modifying exposure behaviour as a user. Wouldn't it only act as in cam programmed ettr compensation? I mean. It's of course positive to be able and expose "normally" but let's keep the information real.

timbytheriver

  • Senior
  • ****
  • Posts: 371
Re: CMOS/ADTG/Digic register investigation on ISO
« Reply #1206 on: December 11, 2019, 12:43:56 PM »
I hear you, but I'm not necessarily suggesting modifying ML ISO behaviour across the board at a deep system level. I'd be happy to see this as a "let's turn on the 'clean-iso preset' and ETTR to the new exposure ceiling because this shot would really benefit from cleaner shadows…"

That's why I'd prefer to see it working as Q-sub-menu choice as opposed to a hard-coded preset like I have at present. Then at least those who wanted to use it could do so, and others could happily ignore it. :)
5D3 1.1.3
5D2 2.1.2

Danne

  • Contributor
  • Hero Member
  • *****
  • Posts: 6039
Re: CMOS/ADTG/Digic register investigation on ISO
« Reply #1207 on: December 11, 2019, 12:48:20 PM »
Sure. Just go ahead. You came much longer than I ever did in this subject.
Looking forward to the continuation.

timbytheriver

  • Senior
  • ****
  • Posts: 371
Re: CMOS/ADTG/Digic register investigation on ISO
« Reply #1208 on: December 11, 2019, 01:34:30 PM »
Now might be a good time to split my objective out into a new thread:

Cleaner ISO Preset https://www.magiclantern.fm/forum/index.php?topic=24702.msg223374;topicseen#msg223374 :)
5D3 1.1.3
5D2 2.1.2