Color shift with ETTR developing?

Discussion in 'Image Processing' started by palombasso, Sep 15, 2015.

  1. palombasso

    palombasso Mu-43 Regular

    61
    Jan 31, 2014
    Hello,

    I was trying my hand at exposing to the right (ETTR) and pushing back the exposure in Lightroom, but had a strange result (see below).
    Do any of you more experienced on the subject explain this results? Maybe only some color channels clipped?

    Camera is a Panasonic GX-7 and the lens was the Panaleica 25 1.4, shot at f 7.1, ss 1/60, iso 125.

    Here's the raw file as imported into LR, all default settings. Notice that there are no highlight clipping warnings:
    ettr1.

    As in-camera I had overexposed by 2 stops, I only dialed back down 2 stops in the exposure slider, but then the colors in the sky became really distorted:
    ettr2.

    Thanks for any info.
     
  2. fredlong

    fredlong Just this guy...

    Apr 18, 2011
    Massachusetts USA
    Fred
    If one or more channels are blown you'll see color shifts when you adjust the exposure.

    ETTR only works if you don't actually overexpose any of the channels.

    Fred
     
    • Agree Agree x 1
  3. b_rubenstein

    b_rubenstein Mu-43 All-Pro

    Mar 20, 2012
    Melbourne, FL
    You have demonstrated the Achilles's of ETTR: push it too far and one of the color channels will clip before the others and part of that channel's color will be lost. In this case you clipped the blue channel and things turned pink. 4/3 sensor cameras aren't that great beyond 1 stop of recovery. 2 stops may work ok for FF sensors, not these. A little shadow noise looks much better than pink skys. Try that ETTR stuff on people sometime and see how great it works for blowing out the red channel.
     
    • Like Like x 1
  4. palombasso

    palombasso Mu-43 Regular

    61
    Jan 31, 2014
    thanks for the replies.

    yeah, I read a lot about this method but am finding it's not worth it, at least for my taste.

    For me It makes it harder to visualize the potential pictures and I was already finding that the camera evf nor the histogram was accurate enough to show clipped highlights in the raw file. This channel clipping issue just makes it worse...
     
  5. David A

    David A Mu-43 All-Pro

    Sep 30, 2011
    Brisbane, Australia
    ETTR can be tricky. I don't find the histogram particularly useful because it can show clipping but it doesn't tell you what is clipping. There are some things, like a light source in the frame or reflections from metal or water (often called "specular reflections" which are going to clip no matter what, and if you try to reduce your exposure to avoid clipping that sort of highlight you're going to end up underexposing the things you're interested in severely.

    I don't know if the GX-7 has a highlight and shadows clipping display option like the red and blue "blinkies" that my Olympus bodies have, try using that instead. I'm prepared to ETTR until an area I want to retain detail in starts blinking red in the display and then I either go with that exposure or reduce the exposure compensation by a third of a stop depending on what is clipping and how much is clipping. I've never run into a colour shift like that in your example working that way.

    Whether or not you have a highlight clipping indication, I'd recommend making some experimental test shots. For a target I used a piece of white paper in direct sunlight and started with the recommended exposure and went with third stop increases and decreases from that for the full +/- 3 stops range that I can dial in exposure compensation, and I took a note of which amount of exposure compensation triggered the highlight clipping indication. If you don't have a highlight clipping indication you can use, note when the histogram hits up against the right edge to indicate clipping. Then process all of those images in Lightroom, or whatever application you use if you use a different one, and see which exposure is the maximum you can give without running into problems you can't recover. Note how that exposure compares to the exposure at which the highlight clipping display or histogram started to indicate clipping, and now you know where your limits are relative to your camera's clipping indication.

    Now for some additional comments and points:

    - you probably should repeat this test for each metering method you use, especially if you use spot metering and spot meter a highlight area. You may find that the amount of leeway you have varies slightly between spot metering and the averaging meter modes.

    - make sure you go out and take a few photos under normal conditions after getting your results, just to check out how valid your test result is in the field, before shooting something where you absolutely have to get good results. If you don't have the chance to take a few shots in normal conditions first, then make sure you do a bit of exposure bracketing on critical shots (and bracketing may be a good idea for critical shots anyway).

    Doing this sort of test doesn't take long, it only takes a couple of minutes to take all the test shots and all you have to do in processing is do the basic exposure/contrast/highlights/white point adjustment and white balance adjustments just to work out for what level of ETTR you can recover your highlights. You can even start processing with the first file to show clipping indication when you open it and process that. If everything can be recovered, move on to the shot with a bit more ETTR next, if you can't recover everything move to the shot with a bit less ETTR. Keep going like that until you know where the limit is for what you can recover and how what the shots with up to a stop or so less exposure than that look with basic processing. You will probably have to process less than half the shots you take but take the lot anyway from maximum underexposure to maximum overexposure because that gives you samples across the full range of exposure compensation available to you and you then don't have to worry about finding when you process that you didn't cover a wide enough range. If you discover you didn't cover a wide enough range to give you a good indication of where your real clipping point is, you're going to have to go out and start from the beginning again in order to get the range of shots you need.

    - you only have to do it for the file types you shoot, so just shoot RAW if that's the way you shoot normally, shoot JPEGs if that's what you normally do, and shoot both if that's what you normally do. Once you've done the test and got an answer you don't have to repeat it unless you get a new camera (different sensors can have different usable ranges) or you change your metering practice or what sort of file type you save your images in.

    - Finally, the histogram/highlight clipping displays are based on an in camera conversion to JPEG, not on actual RAW data so your JPEG and display settings for things like picture modes (vivid/natural/muted/dramatic etc), contrast and so on can affect at what point your clipping indicator/histogram shows clipping so watch any changes you make to them. Try to keep the JPEG settings constant so your results are reliable.

    Run those tests and base your exposure adjustments on what they tell you about your camera's range and you should rarely find yourself getting problems like the one you showed.
     
    • Like Like x 1
  6. mattia

    mattia Mu-43 Hall of Famer

    May 3, 2012
    The Netherlands
    With increasingly ISO-less sensors I'm moving a bit further away from ETTR - making sure you don't blow channels anywhere that matters is more important.
     
  7. b_rubenstein

    b_rubenstein Mu-43 All-Pro

    Mar 20, 2012
    Melbourne, FL
    ETTR was a way of coping with the crummy S/N ratio of 1999 sensors in 1999.
     
  8. You need a camera that you can rely on for accurate exposure preview to be confident with getting ETTR right first time, every time.
     
  9. palombasso

    palombasso Mu-43 Regular

    61
    Jan 31, 2014
    Yeah, I think I read about the GX7 sensor being ISO-less somewhere on dpreview forums. Could you elaborate a bit on this and the relation to ETTR? It means over/under-exposing is unnecessary? Could I just shoot at base iso and push in post?
     
  10. palombasso

    palombasso Mu-43 Regular

    61
    Jan 31, 2014
    Made another quick test including skies and thought to put here as reference in case anyone else is interested:
    4 pictures taken in sequence, GX7, ISO 125 and 200, as metered and +2 exposure in camera:

    First the 2 pictures as metered by the camera. As expected both are extremely similar with the only difference being the shutter speed:
    ISO200. ISO125.


    Next the pictures taken at +2 exposure in camera and recovered by -2 in LR.
    Even in the ISO 200 the sky already shows a change in the tone of blue. The ISO 125 again shows patches of purple on the left (although not as much as the first pic in the thread).
    I come to the conclusion that in ISO 125 some channels blown quicker. Maybe that's why it's kept as an "extended" ISO setting?
    Sorry if it was obvious to others but it wasn't intuitive enough for me.

    ISO200+2. ISO125+2.
     
  11. David A

    David A Mu-43 All-Pro

    Sep 30, 2011
    Brisbane, Australia
    Sensors have a base ISO. With my E-M1 that is ISO 200 but the camera offers a "Low ISO" setting which is effectively ISO 100. That is achieved by overexposure, effectively by exposing to the right. If your ISO 125 setting is lower than your GX7's base ISO, that means that the standard exposure at ISO 125 is already exposed about two thirds of a stop to the right so dialling in a further 2 stops exposure compensation is really exposing to the right by about 2 and two third stops.

    Read some reviews and test reports and find out whether your camera's base ISO is ISO 200 or ISO 125. I'd suggest not using ISO 125 if ISO 200 is the base ISO for your sensor. Just use ISO 200 and don't expose so far to the right as to cause problems. What the examples above suggest to me is that your camera's base ISO is ISO 200 and that your second ISO 125 image is basically the result you would get from exposing to the right by around 2.7 stops at ISO 200.

    ISO setting affects the dynamic range of sensors. Move away from base ISO and dynamic range can reduce, it will not increase. If the maximum you can expose to the right by without running into problems at base ISO is 2 stops, then the maximum you can expose to the right by at other ISO settings will be no greater than 2 stops and less than that at some ISO settings. You can't assume that because 2 stops to the right is fine at one ISO setting, it's going to be fine at all ISO settings. In fact it won't be fine. You will get your maximum "leeway" at base ISO and at ISO settings below base ISO you will lose whatever exposure advantage you gain as a result of the lower ISO setting from whatever your maximum leeway is at base ISO. If you're going to expose to the right you aren't going to get any advantage at ISO settings below base ISO. Lower than base ISO settings are really nothing more than a way of dialling in a certain amount of ETTR at base ISO.
     
    • Agree Agree x 1
  12. Clint

    Clint Mu-43 All-Pro

    Apr 22, 2013
    San Diego area, CA
    Clint
    You did have highlight clipping although it may not show in the main part of the screen. See that little arrow in the box at the upper right of your screen shot - that is the color that is clipped. So when you back off the exposure, that is the color that Lightroom is trying to replace.

    Most manufactures already have a sense of ETTR built into the ISO calculations - that is the reason for the differences in the manufactures ISO rating and the tested ISO results such as produced at DxO mark.
     
    • Like Like x 2