ISO & Dynamic range - what is going on?

AlanSh
June 27, 2025
chat_bubble_outline 11
  • link
    Forum Admin
    June 27, 2025, 5:14 p.m.
    link

    A member of my club posted "I found out the other day that if I increase my ISO from 100 to 160 the dynamic range goes up and the noise goes down." and referenced the site photonstophotos.net/. When I looked at his camera, (Canon 5D Mk3), I saw this:

    image.png

    Which, as you can see has better dynamic range at 160 and 320 ISO. But why does it do that?

    My Fuji X-T5 doesn't (native ISO starts at 125).
    Screenshot 2025-06-27 175811.png

    A Canon 90D doesn't.
    Screenshot 2025-06-27 180406.png

    Here's all 3 overlaid (with scales left & bottom):

    image.png

    Can someone explain this in layman's terms?

    Alan

    image.png

    PNG, 103.3 KB, uploaded by AlanSh on June 27, 2025.

    Screenshot 2025-06-27 180406.png

    PNG, 30.8 KB, uploaded by AlanSh on June 27, 2025.

    Screenshot 2025-06-27 175811.png

    PNG, 29.2 KB, uploaded by AlanSh on June 27, 2025.

    image.png

    PNG, 43.3 KB, uploaded by AlanSh on June 27, 2025.

  • link
    Members 905 posts
    June 27, 2025, 10:14 p.m.
    link

    In a camera, the ISO setting normally affects the gain between the sensor output and the raw written data. In some cameras, that gain is fixed and the ISO gain is applied in the conversion software. In the green curve, the gain is proportional to the ISO value; in the other curves, it is not. The blue curve (your Fuji) looks to me like it has "extended" low ISO and high ISO; and the black curve has some trickery in addition to that.

    "Extended" means that, for example, that the camera aperture and shutter values are not changed by the ISO setting below "native", say 100 and some high value..

    The graphing site owner Bill Claff does respond to questions and can give a much more specific response.

  • link
    Members 2466 posts
    June 28, 2025, 6:58 a.m.
    link

    I dont take any of these testing sites (linked to manufacturers 🫣) for real any more, i borrowed a sony a7r5 the other week and put it through some simple tests. it doesnt compare to the a7iv for DR and hi iso performance , just as i had thought, and even worst case is the a6600 compared to the a6700. the a6700 is so far ahead its nearly FF performance based on images i looked at from past cameras with the same sensor, baseing DR on noise is just plain wrong, as a guy said on FM forums how can increasing noise reduction(reducing noise) increases DR 🤔its all a joke. just do your own testing thats the only way you will get the answers.

  • link
    Forum Admin
    June 28, 2025, 8:17 a.m.
    link

    Yes, it's that 'trickery' I don't understand. What is the camera doing and why?

    Alan

  • link
    Members 2466 posts
    June 28, 2025, 8:42 a.m.
    link

    quote from Apples new 20 stop sensor " Patent quote : "advanced on-chip noise suppression "

  • link
    Forum Admin
    June 28, 2025, 9:59 a.m.
    link

    Donald, that doesn't help at all. Sorry.

    Alan

  • link
    Team
    June 28, 2025, 11:05 a.m.
    link

    What is Apple‘s new 20 stop sensor?

  • link
    Members 2466 posts
    June 28, 2025, 11:50 a.m.
    link
  • link
    Members 2466 posts
    June 28, 2025, 11:59 a.m.
    link

    www.dpreview.com/forums/post/68332406

  • link
    Members 1468 posts
    June 28, 2025, 12:07 p.m.
    link

    Basing DR on noise is just correct.
    quote from AI:
    Relationship between Dynamic Range and Noise Floor:
    -A low noise floor is crucial for achieving a wide dynamic range.
    -A high dynamic range requires a sensor that can capture both very bright and very faint signals without being overwhelmed by noise.
    -Noise reduction techniques can improve both dynamic range and perceived image quality by lowering the noise floor.

  • link
    Members 905 posts
    June 28, 2025, 1:28 p.m.
    link

    The graphing site owner Bill Claff does respond to questions and can give a much more specific response.

  • link
    Members 905 posts
    June 28, 2025, 1:37 p.m.
    link

    The ISO definition of DR is based on the noise floor specified as the point when the signal-to-noise ratio is 0 EV.

    Discussed by a human here:

    dougkerr.net/Pumpkin/articles/ISO_Dynamic_range.pdf