Author Topic: iso experiments demo  (Read 702 times)

70MM13

  • Member
  • ***
  • Posts: 232
iso experiments demo
« on: April 04, 2019, 06:57:26 AM »

setting up the camera, and comparing the outputs of iso 100, 111, and 200 under identical light exposure

5d3 1.13 using the december 2017 iso experiments and crop rec

timbytheriver

  • Member
  • ***
  • Posts: 243
Re: iso experiments demo
« Reply #1 on: April 04, 2019, 09:12:11 AM »
@70MM13 Great stuff - thanks so much! Have to dash now but will get the popcorn ready and watch in detail tonight! :P
5D3 1.1.3
5D2 2.1.2

timbytheriver

  • Member
  • ***
  • Posts: 243
Re: iso experiments demo
« Reply #2 on: April 04, 2019, 05:52:46 PM »
@70MM13 Results look great!  :P I have a couple of questions:

Seems to me the reference still image you take initially is very important, as I assume it sends crucial exposure data to the ISO Registers module.

1) It's not quite clear from the video how this still image should be exposed. Is it exposed hard to the right with 99% on the highlights – or to capture cleanest shadows.…? (I tend to use the RAW histogram with 'hints' turned on, so I'm not sure how this corresponds to the spotmeter  % readout.)

2) When matching the two ISO exposures (200 and 111 here) how do you calculate that the difference in ISO in terms of the spotmeter % is from 99% to 95%? Is this clever maths or just eyeballed? ;)

Thank you again for making the info video.


5D3 1.1.3
5D2 2.1.2

lucoctober

  • New to the forum
  • *
  • Posts: 7
Re: iso experiments demo
« Reply #3 on: April 04, 2019, 09:07:53 PM »
How much highlights is sacrificed to gain clean shadows? Did You tried to compare overexposing by 1ev on iso200 and correcting exposure by -1ev in mlvapp?

70MM13

  • Member
  • ***
  • Posts: 232
Re: iso experiments demo
« Reply #4 on: April 05, 2019, 01:30:04 AM »
@timbytheriver
the reference image was taken with the lens cap on.  i find that under "normal" conditions, the default values match the reference image every time.
regarding the maximum exposure, i'm using the histogram and the spot exposure sensor reading.  the results bear out with what gets captured as seen in mlvapp.  the more you reduce the gain, the lower the % value gets for clipping.  when you see that the % value stops going up when you are increasing exposure, you know you have the limit.  reduce exposure until the value starts to drop again, and then you have it...
also, as i said in the video, i'm literally copying and pasting the identical settings in each clip, and they are the same, because they were exposed the same.
so the three comparison images are indeed exposed identically.  the reduced noise in the iso 111 shot is real.

70MM13

  • Member
  • ***
  • Posts: 232
Re: iso experiments demo
« Reply #5 on: April 05, 2019, 01:35:00 AM »
@lucoctober
since i'm exposing for highlights in every setting, no highlights are lost at all.  since there is no clipping, and the shadows are all being pushed by the same very strong amount, there is no loss in dynamic range in this real world example.
try it for yourself!

timbytheriver

  • Member
  • ***
  • Posts: 243
Re: iso experiments demo
« Reply #6 on: April 05, 2019, 09:51:43 AM »
@70MM13 Ah! I understand. Thanks! Will share my results. :)
5D3 1.1.3
5D2 2.1.2

timbytheriver

  • Member
  • ***
  • Posts: 243
Re: iso experiments demo
« Reply #7 on: April 05, 2019, 11:15:36 AM »
My results following @70MM13's method do indeed yield noticeably better shadow noise. I took my reference (base ISO) still image at ETTR exposure / highlight at 99% spot.

:D Pretty happy!

(base ISO 200 > 92)






5D3 1.1.3
5D2 2.1.2

70MM13

  • Member
  • ***
  • Posts: 232
Re: iso experiments demo
« Reply #8 on: April 05, 2019, 12:53:15 PM »
seeing kubrick's name makes me happy too ;)

seriously though, i'm glad you are happy with it.  it serves me well because i'm always exposing for highlights and who doesn't want cleaner shadows?

more people need to try this, and improve on it!

timbytheriver

  • Member
  • ***
  • Posts: 243
Re: iso experiments demo
« Reply #9 on: April 05, 2019, 01:53:38 PM »
more people need to try this, and improve on it!

+1

Thanks again for your help!  :)
5D3 1.1.3
5D2 2.1.2

70MM13

  • Member
  • ***
  • Posts: 232
Re: iso experiments demo
« Reply #10 on: April 05, 2019, 02:05:16 PM »
you're welcome!

maybe you should make a better video showing what you are doing...

PS: i don't think i made it clear in the video, but i do the test image in stills mode.  i don't know if it makes a noticeable difference, but that's the only way i get reliable consistent values for the "copy canon settings" action, getting the right default values, and getting the same 111 iso for my lazy method...

something to be investigated!