ISO independent of Exposure Value now?

Maybe it's been like this for awhile, but I just happened to notice in 4.20. I thought shutter, aperture, and ISO setting were all just different ways to adjust exposure value, but ultimately exposure value is what changes the overall exposure. But now, setting EV to 13 and ISO to 200 is twice as bright as EV at 13 and ISO at 100.  I thought EV 13 was EV 13 regardless of how you get there with shutter, aperture, and ISO.

I feel like maybe this is a bug, because when you adjust ISO, it still adjusts EV the same number of stops, but if ISO is supposed to be independent now it shouldn't do that. If you increase an Ev=13/ISO=100 setting to ISO=200, the EV goes one stop brighter to 12, but the exposure is actually increasing by two stops unless you manually adjust EV to 13 again.

Comments

  • PadonePadone Posts: 2,727
    edited February 20

    So 4.20 also breaks tone mapping other than materials, thank you for reporting this.

    Post edited by Padone on
  • I never adjust EV directly, partly bcause it has in the past seemed inconsistent in how it reacts to the different "real" controls (ISO, fStop, Shutter speed). EV is actually a proxy for Shutter Speed - I thought previously it was a proxy for ISO. if that has changed it would explain why you are seeing unexpected results, but I don't think it's wrong per se.

  • cgidesigncgidesign Posts: 264
    edited February 20

    To my experience none of the fstop, iso, shutter speed have any effect like in real world.

    FSTOP does not do anything because depth of field is controlled in a different way

    ISO does not do anything because there is no film or sensor grain in DS

    SHUTTER SPEED does not do anything because there is no motion blurr in DS

    The only relevant value is sensitivity to light. Whether it is changed by EXPOSURE, ISO, FSTOP or SHUTTER doesn't matter (of course the "real" tonemapper variables like crush blacks, burn highlights are a different story).

    But let me know if I miss something.

    Post edited by cgidesign on
  • thoreandanthoreandan Posts: 122
    edited February 20

    I think EV's usefulness was as an aggregator of the three settings below it. No matter how you set shutter, aperture, and ISO speed, you could see the effect on the tone mapped exposure by checking EV. This is especially true because none of the three individual settings do anything other than change exposure, you don't get more grain with higher ISO settings, you don't get different DOF results with different f/stops, you don't get motion blur with shutter speed. So if all these three settings do is change exposure, the Exposure Value setting should reflect the effect of all three. Otherwise, what makes ISO speed special that it should have an effect on top of EV?

    Also, it's not behaving the way you'd expect even if ISO is considered independent. I just did some experiments, and a one-stop change in ISO does not change tone-mapped light levels the same amount as a one-stop change in EV. This gets even more confused because changing ISO by one stop changes the EV setting by one-stop. So, if you start with EV 12 ISO 100, and change ISO to 200, it also changes EV to 11. That would only make sense if ISO was part of EV, and not separate. But even worse, I noticed that changing EV by one stop doesn't have the same effect as changing ISO by one stop. See the attached example, with a grid showing results for EV 12 and 13, and ISO 100 and 200.

    In the old system, the top row and bottom row should have been the same, since EV just aggregated the other three settings. But putting that aside, even if we accept that the effect of ISO is separate from EV now, then EV 12 ISO 100 should be the same as EV 13 ISO 200, since that's a one stop change in EV and ISO in opposite directions. But all four results on this grid are different, so changing ISO by one stop must be changing exposure by more than one stop of change to EV.

    (I should note the attachment is made from screenshots of viewport preview renders, I have not checked if the effect is different in full render.)

    EV vs ISO.jpg
    1200 x 1200 - 340K
    Post edited by thoreandan on
  • thoreandanthoreandan Posts: 122

    As another follow up, I compared EV 14 ISO 200 with EV 12 ISO 100, and found the EV 14 ISO 200 slightly higher exposure in the result than EV 12 ISO 100. That means changing ISO by one stop is changing the exposure by slightly more than two full stops of EV.

  • The default values are 13/128/8/100. Changing ISO to 200 gives me 14/128/8/200 as expected. Changing EV to 14 directly gives 14/256/8/100, which seems wrong to me (and is why I don't use EV - and certainly don't mix using EV and the other three values).

  • FishtalesFishtales Posts: 5,697

    Might be worth a read.

    https://en.wikipedia.org/wiki/Exposure_value

    When setting up Tone Mapping I don't even look at EV only using the other values to get what I want using real world Photographic settings that I use with my camera. I realise all the arguments about noise, DOF and motion blur aren't affected but I just find it easier and quicker :) For DOF I do that in the camera settings.

  • thoreandanthoreandan Posts: 122
    edited February 20

    I'm trying to make two points. One is that the other three values used to be largely irrelevant, all that mattered was the resulting EV. Regardless of what combination of shutter, aperture, and ISO settings you dialed in, the same lighting and the same EV value would generate the same exposure in the resulting render. This makes sense not only for consistency, but from a photographer's perspective as well: from the perspective of exposure (ignoring motion blur/grain/etc), if you used 200 film instead of 100, your exposure would increase by one stop all else being equal, and changing another setting by one stop would bring you back to the same exposure. But with the way this is working now, two scenes with exactly the same lighting and same EV value can have different exposures depending on how ISO is combined with the other values.

    The other point is that one stop of shutter speed is not the same as one stop of ISO in these settings now, even though it used to be and absolutely should be. Adjusting ISO by one stop and adjusting shutter speed one stop the other way does not yield the same results. Similar to your example, if you were to compare 256/8/200 with 128/8/100, the scene should have the same exposure, but right now they don't. Not sure what the EV resulting from those combos would be because I'm not at my computer, but they should be the same.

    (edited for mistake in shutter/iso)

    Post edited by thoreandan on
  • Are you sure? I just did 256/200 and 128/100, layered them set the blend to Difference, and everything went black.

  • thoreandanthoreandan Posts: 122

    Richard Haseltine said:

    Are you sure? I just did 256/200 and 128/100, layered them set the blend to Difference, and everything went black.

    Maybe it's only in the viewport, then, or I'm overlooking something obvious (always a possibility). I'll test it again when I get back to my computer. Thanks for looking into it.

  • thoreandan said:

    Richard Haseltine said:

    Are you sure? I just did 256/200 and 128/100, layered them set the blend to Difference, and everything went black.

    Maybe it's only in the viewport, then, or I'm overlooking something obvious (always a possibility). I'll test it again when I get back to my computer. Thanks for looking into it.

    Actually I had just eyeballed it - when I added a  levels adjustment and pushed the gamma way up there were differences in the shadows, though not, I think, anywhere near a full stop's worth.

  • thoreandanthoreandan Posts: 122

    You were right, 128-8-100 and 256-8-200 produce the same result. So from that standpoint, if you ignore EV then one stop of shutter speed counters one stop of ISO, so that's okay. I was focused on EV and ISO and lost track of what was happening with shutter speed in my first experiment.

    But 128-8-100 results in EV=13, while 256-8-200 results in EV=15, despite producing the same overall tonemapped exposure in the result. That part doesn't make sense. But now I understand what is happening: when increasing ISO by one stop (increasing exposure), it's increasing the EV by one instead of decreasing by one, which I'm pretty certain is a bug, and I think it started with 4.20. It's still calculating exposure based on the three values correctly behind the scenes, but it's reporting the wrong EV number in the interface. If EV went down by one when increasing ISO to 200 in the example we've been using, for example, EV would land on 13 for both 128-8-100 and 256-8-200, which would make sense.

    I'll report it now that I understand better what's going on, and at least it's not as big a deal as I thought. It should only be a problem if somebody has a scene where they set both EV and ISO directly, which I guess is unlikely. I was only doing that because I noticed something weird happening to EV when I changed ISO. I usually only use EV as a first approximation for how many stops to reduce exposure on EXR canvases in Photoshop.

  • I think the issue with EV going the wrong way for one input is long-standing, probably since it was first added.

  • barbultbarbult Posts: 19,359
    edited February 23

    Richard Haseltine said:

    I think the issue with EV going the wrong way for one input is long-standing, probably since it was first added.

    I submitted a help request about that issue years ago and was told it was not a bug. I guess they finally changed their mind. Or are you saying that EV still "goes the wrong way"?

    Where is the release note to explain this change? You are good with coming up with change log links, Richard. wink

    Post edited by barbult on
  • barbultbarbult Posts: 19,359
    edited February 23

    Is this also going to break our existing scenes, or store products that load entire scenes or provide Render Settings presets? Or does this change not effect existing scenes and presets? Or is there no change in 4.20 and people are just now noticing the strange behavior? I'm so confused!

    Post edited by barbult on
  • barbult said:

    Richard Haseltine said:

    I think the issue with EV going the wrong way for one input is long-standing, probably since it was first added.

    I submitted a help request about that issue years ago and was told it was not a bug. I guess they finally changed their mind. Or are you saying that EV still "goes the wrong way"?

    Where is the release note to explain this change? You are good with coming up with change log links, Richard. wink

    I don't think it has changed, and I submitted a bug report too (2016).

  • barbultbarbult Posts: 19,359

    Richard Haseltine said:

    barbult said:

    Richard Haseltine said:

    I think the issue with EV going the wrong way for one input is long-standing, probably since it was first added.

    I submitted a help request about that issue years ago and was told it was not a bug. I guess they finally changed their mind. Or are you saying that EV still "goes the wrong way"?

    Where is the release note to explain this change? You are good with coming up with change log links, Richard. wink

    I don't think it has changed, and I submitted a bug report too (2016).

    Thank you for the clarification, Richard. I think we are all a little gun shy about 4.20.

  • cgidesigncgidesign Posts: 264
    edited February 24

    That's the reason I dropped using the various sliders and just change EV. The other ones anyway do either not have any relevance with respect to their real world counterparts or the relation between them and EV changes over time following the developers actual mood.

    Post edited by cgidesign on
  • cgidesign said:

    That's the reason I dropped using the various sliders and just change EV. The other ones anyway do either not have any relevance with respect to their real world counterparts or the relation between them and EV changes over time following the developers actual mood.

    EV just drives Shutter Speed, if set directly. You can use the separate sldiers, purely to control tone, or use the EV - but mixing EV and the camera settings can be confusing due to the interrelationship.

  • thoreandanthoreandan Posts: 122

    But if they'd just make EV go the right way when ISO is changed, there wouldn't be any contradiction or confusion, because the reported EV would always match the real tonemapped change in exposure, regardless of the three settings, and changing any one of the four sliders would yield consistent results. 

  • Yes, and as I said I have bug reported it.

Sign In or Register to comment.