Luminous Landscape Forum
Equipment & Techniques => Digital Cameras & Shooting Techniques => Topic started by: pco98 on March 31, 2008, 10:05:49 am
-
Hello all,
I'm in Hong Kong airport sitting with a scanner, scanning my last batch of film (hey do you remember those days?). But as of tomorrow I'm hitting downtown HK to finally go fully digital with a 5D, now at their cheapest ever. I put all my money into some L lenses, so there's been a financial delay in getting a high-end digital body until now :-)
I've been reading up about exposing to the right as the preferred shooting technique and had a question about the practicalities of this. Am I right in assuming you don't know if the highlights have been blown until after the image is taken and you check the histogram? If so is there a lot of checking and deleting because a channel has been clipped or because it's not as close to the clipping threshold as it could be. Or, after a while do you develop a feel for how much to expose to the right, getting non-clipped shots most of the time?
Thanks in advance,
Ross
-
Hello. Exposing to the right allows more data in the shadows to be collected. The danger, as you highlight!!! is blowing the highlights. IMO, a badly blown highlight is generally worse than a dark area. Hence in some cases I actually have to expose, using the same terminology, left, and the push the shadows later.
The problem with the histogram is that you don't see it until after the image has been exposed, and then what you are seeing is based on the Jpeg, and also the WB set in the camera. Some people have produced "null" WB settings in order to combat this.
In most camera's there is the ability to pull back some highlight detail during RAW conversion, but this really becomes a Know your camera, convertor and scene style. 2/3 of a stop is about my working limit.
It does provide benefits, but on a "must have" shot only you can determine whether the slight improvements to shadows out-ways the risk of protecting any channel clipping.
There is a technique called "high-speed Bracketing" which is also worth exploring, where you set the camera to bracket either side of your exposure and then take the images in rapid fire mode.
Steven
-
Hello all,
I'm in Hong Kong airport sitting with a scanner, scanning my last batch of film (hey do you remember those days?). But as of tomorrow I'm hitting downtown HK to finally go fully digital with a 5D, now at their cheapest ever. I put all my money into some L lenses, so there's been a financial delay in getting a high-end digital body until now :-)
I've been reading up about exposing to the right as the preferred shooting technique and had a question about the practicalities of this. Am I right in assuming you don't know if the highlights have been blown until after the image is taken and you check the histogram? If so is there a lot of checking and deleting because a channel has been clipped or because it's not as close to the clipping threshold as it could be. Or, after a while do you develop a feel for how much to expose to the right, getting non-clipped shots most of the time?
Thanks in advance,
Ross
[a href=\"index.php?act=findpost&pid=185686\"][{POST_SNAPBACK}][/a]
It's true - you don't know until after you've made the shot what the histogram looks like. There is no live preview of the histogram. Typically, I find Canon DSLRs (the two that I own) have a tendancy of under-exposing relative to what one wants using the ETTR approach. I don't know exactly why - could be how they calculate middle grey, how they accommodate specular highlights - a number of possibilities; but the outcome is that I often find after making the initial exposure, it is necessry to add some exposure compensation and make another shot to get the ETTR correctly positioned.
One develops a certain "feel" (aided by careful observation and experimentation) for how much compensation is needed relative to the amount of empty space on the right which needs to be "filled" with tones.
But it is also important to remember that the histogram and the preview are based on an in-camera JPEG rendition of the file, not the raw file itself. This in-camera JPEG may or may not be a fairly accurate predictor of how the raw file will emerge. So this is a second variable you need to mentally adjust - by experimenting a bit. You make several exposures with what looks like best ETTR practice on the camera's histogram, then open them in your raw converter to see any gaps between what the camera predicts and what emerges from the file on your computer display. I've found there is a bit of leeway - when the camera begins to show clipping the raw file may still be fine, but you'd need to check the extent of any such leeway with your camera.
-
I've been reading up about exposing to the right as the preferred shooting technique and had a question about the practicalities of this. Am I right in assuming you don't know if the highlights have been blown until after the image is taken and you check the histogram? If so is there a lot of checking and deleting because a channel has been clipped or because it's not as close to the clipping threshold as it could be. Or, after a while do you develop a feel for how much to expose to the right, getting non-clipped shots most of the time?
Yep, you are right that you don't know until after the fact, and yes, I look at the (RGB, not luminance) histogram primarily to determine that. I find that in most situations, autoexposure gives me something "pretty good", and that I'll tend to "shoot, glance at histogram, move on", and when that doesn't work it's more "shoot, glance at histogram, quickly sketch in some exposure compensation, shoot again, glance at histogram, move on".
I don't delete in the field. My shooting style tries to keep my brain as much as possible aimed at "seeing", and less time (but not less than necessary) on "intellectual thinking".
How much? It depends a little, but if I think the highlights will be visible in the histogram (this can be tricky when you only have a couple tiny highlights in a darker scene, but that's a rarer case), I tend to aim to put them about, oh, I dunno, an tenth of the visual width of the histogram from the right hand side.
A lot of times I will accept a capture with wider spacing if I know that the exposure I eventually want in the print is darker or about the same as the exposure I captured, I'll worry about ETTR more if I feel like I'm going to pull up the shadows, or if my histogram has pegged or nearly-pegged shadows I'd like some level of detail in.
This all sounds like a lot of work, but it becomes, for me, very intuitive and quick after a while.
-
ERRT, the good, the bad and the ugly:
http://www.digitalphotopro.com/tech/exposing-for-raw.html (http://www.digitalphotopro.com/tech/exposing-for-raw.html)
-
Also note that with recent cameras you have the option of Live View which does show a live histogram.
Same limitation of the histogram being generated from the in-camera JPEG applies, however.
-
ERRT, the good, the bad and the ugly:
http://www.digitalphotopro.com/tech/exposing-for-raw.html (http://www.digitalphotopro.com/tech/exposing-for-raw.html)
[a href=\"index.php?act=findpost&pid=185720\"][{POST_SNAPBACK}][/a]
This is a good article on exposure to the right, but I would take exception with the use of Adobe Lightroom (or the equivalent Adobe Camera Raw) to evaluate clipping in the raw file. The tone curve can affect the results and it is best to look directly and avoid these complications.
I recently did some tests with the Nikon D3, taking bracketed shots of a Stouffer step wedge and splitting out the green channel of the raw file in Iris (a free ware astronomical program) and analyzing it in ImageJ (a free ware program from the US National Institutes of Health). I used 14 bit NEFs with lossless compression.
Here is a plot of an image with step one of the wedge just short of clipping.
(http://bjanes.smugmug.com/photos/272951447_y3hyV-O.png)
And here is a histogram of step one, confirming the absence of clipping.
(http://www.smugmug.com/photos/272869144_qxukM-O.png)
Here is the image in Adobe Camera Raw (which uses the same raw conversion engine as Lightroom). ACR indicates clipping through step 3 (each step is 0.3 EV or 1/3 stop).
(http://www.smugmug.com/photos/271460357_vDqYZ-O-1.png)
To elilminate the clipping, it is necessary to use negative exposure compensation of -0.6 EV.
(http://www.smugmug.com/photos/271460363_jMYHL-O-2.png)
The camera histogram with the camera set to the Standard Picture Control (normal contrast, normal saturation) also shows come clipping:
(http://www.smugmug.com/photos/271460374_mRjof-O.png)
Decreasing exposure by 0.3 EV removes the clipping in the camera histogram, confirming that the camera histogram is slightly conservative in its treatment of highlights, but is within 1/3 stop, which is a reasonable amount of headroom.
(http://www.smugmug.com/photos/271460388_WDMpW-O.png)
My conclusion is that the camera histogram is useful in ETTR but that the result of the ACR histogram can be misleading as ACR by default places the highlights too high with this particular camera.
Bill
-
Also note that with recent cameras you have the option of Live View which does show a live histogram.
Same limitation of the histogram being generated from the in-camera JPEG applies, however.
[a href=\"index.php?act=findpost&pid=185781\"][{POST_SNAPBACK}][/a]
Hi Eric,
As far as I'm aware the only recent cameras with live histograms are non-DSLRs. Even my 1DsMk3, which has live-view, does NOT (much to my chagrin) have a live histogram. I wish..........
Mark
-
This is a good article on exposure to the right, but I would take exception with the use of Adobe Lightroom (or the equivalent Adobe Camera Raw) to evaluate clipping in the raw file. The tone curve can affect the results and it is best to look directly and avoid these complications.
I recently did some tests with the Nikon D3, taking bracketed shots of a Stouffer step wedge and splitting out the green channel of the raw file in Iris (a free ware astronomical program) and analyzing it in ImageJ (a free ware program from the US National Institutes of Health). I used 14 bit NEFs with lossless compression.
Here is a plot of an image with step one of the wedge just short of clipping.
(http://bjanes.smugmug.com/photos/272951447_y3hyV-O.png)
And here is a histogram of step one, confirming the absence of clipping.
(http://www.smugmug.com/photos/272869144_qxukM-O.png)
Here is the image in Adobe Camera Raw (which uses the same raw conversion engine as Lightroom). ACR indicates clipping through step 3 (each step is 0.3 EV or 1/3 stop).
(http://www.smugmug.com/photos/271460357_vDqYZ-O-1.png)
To elilminate the clipping, it is necessary to use negative exposure compensation of -0.6 EV.
(http://www.smugmug.com/photos/271460363_jMYHL-O-2.png)
The camera histogram with the camera set to the Standard Picture Control (normal contrast, normal saturation) also shows come clipping:
(http://www.smugmug.com/photos/271460374_mRjof-O.png)
Decreasing exposure by 0.3 EV removes the clipping in the camera histogram, confirming that the camera histogram is slightly conservative in its treatment of highlights, but is within 1/3 stop, which is a reasonable amount of headroom.
(http://www.smugmug.com/photos/271460388_WDMpW-O.png)
My conclusion is that the camera histogram is useful in ETTR but that the result of the ACR histogram can be misleading as ACR by default places the highlights too high with this particular camera.
Bill
[a href=\"index.php?act=findpost&pid=185786\"][{POST_SNAPBACK}][/a]
Bill, a couple of points: Firstly, I believe, based on my experience with the Canon 1Ds and 1DsMk3, plus discussion with a Nuikon D3 owner, that the Canons have a tendancy to under-expose and the Nikon D3 tends to expose more toward the right. The OP is in the Canon "world". Secondly, when I examine an image in ACR, I start out with totally neutral conditions" W/B "As Shot", all the rest of the front panel are zeros and both tone curves are linear. This way you see what's coming from the camera without any tinkering. Non-linearity in either of the tone curves or any positive saturation values, or a warming of the white balance, depending on the image, could generate some highlight clipping which isn't a product of what emerged from the camera.
-
Lightroom was doing this to my A700 RAW files, but I've since set the default import settings of Contrast, Brightness, and Black level to zero, as well as the linear tone curve, and this seems to be more indicative of the actual RAW.
-
This is a good article on exposure to the right, but I would take exception with the use of Adobe Lightroom (or the equivalent Adobe Camera Raw) to evaluate clipping in the raw file.
[...]
Here is the image in Adobe Camera Raw (which uses the same raw conversion engine as Lightroom). ACR indicates clipping through step 3 (each step is 0.3 EV or 1/3 stop).
(http://www.smugmug.com/photos/271460357_vDqYZ-O-1.png)
[a href=\"index.php?act=findpost&pid=185786\"][{POST_SNAPBACK}][/a]
Mmmmmm, it seems that your default preset for that shot has a Contrast of +25... I'd say, better to set it at 0 for that particular purpose?
Ditto for the Blacks : +5 should mean clipping, shouldn't it?
-
Bill, a couple of points: Firstly, I believe, based on my experience with the Canon 1Ds and 1DsMk3, plus discussion with a Nuikon D3 owner, that the Canons have a tendancy to under-expose and the Nikon D3 tends to expose more toward the right. The OP is in the Canon "world". Secondly, when I examine an image in ACR, I start out with totally neutral conditions" W/B "As Shot", all the rest of the front panel are zeros and both tone curves are linear. This way you see what's coming from the camera without any tinkering. Non-linearity in either of the tone curves or any positive saturation values, or a warming of the white balance, depending on the image, could generate some highlight clipping which isn't a product of what emerged from the camera.
[a href=\"index.php?act=findpost&pid=185825\"][{POST_SNAPBACK}][/a]
Mark,
Thanks for your comments. A contrast curve (S-curve) affects the three quarter tones much more than the highlights themselves. Here are my results with the settings you suggest. The point curve was also set to linear.
(http://bjanes.smugmug.com/photos/273026804_ESG9i-O.png)
(http://bjanes.smugmug.com/photos/273026841_24iWv-O.png)
Negative exposure compensation is still necessary to bring step 1 below clipping.
Bill
-
Mmmmmm, it seems that your default preset for that shot has a Contrast of +25... I'd say, better to set it at 0 for that particular purpose?
Ditto for the Blacks : +5 should mean clipping, shouldn't it?
[a href=\"index.php?act=findpost&pid=185851\"][{POST_SNAPBACK}][/a]
That is the default ACR for the D3 and other Nikon cameras that I have. See my reply to MarkDS for a linear tone curve. It does not affect the highlights very much, but does affect the three quarter tones. The Black setting does not affect the highlights.
Bill
-
Mark,
Thanks for your comments. A contrast curve (S-curve) affects the three quarter tones much more than the highlights themselves. Here are my results with the settings you suggest. The point curve was also set to linear.
Negative exposure compensation is still necessary to bring step 1 below clipping.
Bill
[a href=\"index.php?act=findpost&pid=185857\"][{POST_SNAPBACK}][/a]
Bill,
Thanks for taking the trouble to test that and illustrate it. Most interesting. Yes of course - what you say about the S-Curve is mainly correct, what it is supposed to do - and does; but depending on the image it (especially combined with boosted brightness and contrast in the Basic Tab) can have the unwanted side-effect of clipping some of the highlights. That is why it's safer to start in ACR with zeroed Basics and linear curves, as your own results show. The extent of clipping with my suggested (flat) settings is far less than it was with the previous settings - only about half a stop. In the example you show here, the preferred correction for the slight highlight loss would be to increase Recovery till those highlights just recover. This will work as long as it is not all three channels that are clipped.
Mark
-
IMO, any test about ETTR to analyse if a RAW file was clipped, needs neutral (1.0) white balance multipliers. Since ACR and any other commercial developer don't allow them, I would not take too much seriously precise results obtained using them.
DCRAW achieves neutral WB by -r 1 1 1 1
Regards.
-
Also note that with recent cameras you have the option of Live View which does show a live histogram.
Same limitation of the histogram being generated from the in-camera JPEG applies, however.
[a href=\"index.php?act=findpost&pid=185781\"][{POST_SNAPBACK}][/a]
Hmmm .... not on my D300 it would appear. Live histogram could be useful.
-
Hi Ross. With the size of cf cards now, I don't bother deleting until the images are in the raw converter. Seeing how much leeway I have or haven't got has taught me a lot. And yes, you quickly get used to checking the back of the camera and getting a feel for how to set it up.
I set the exposure to manual most of the time, take a shot and check the histogram, and then only check it every half hour or so. On the 40D, if the clipping warnings are just starting, then it's usually about right when opened in the raw converter. If I'm photographing birds, I take a reading off the sky and add about 1 to 1 1/2 stops. You'll have a lot of fun finding out what works for you. I hope you really enjoy your new gear. Cheers, David
-
It does not affect the highlights very much, but does affect the three quarter tones.
[a href=\"index.php?act=findpost&pid=185858\"][{POST_SNAPBACK}][/a]
Surprinsing, but well proven! Thanks to have taken the time for it.
Like Mark, I had the experience of the Contrast setting leading to clipping with borderline exposures (Canon 300d for me).
Side note : as my camera shows only a luminance jpeg-based histogram, I took the (bad) habit to consider that ETTR was OK as long as I had :
- on one hand, at least a bit of information in the upper stop,
- on the other hand, no jpeg clipping, and the more saturated colors in the hightlights the more margin I take with it (needs a well-tuned ouija board to be precise ).
Matter of tastes, but I prefer the Scylla of noise to the Charybdis of clipping-related hue shifts in the highlights.
-
In the example you show here, the preferred correction for the slight highlight loss would be to increase Recovery till those highlights just recover. This will work as long as it is not all three channels that are clipped.
[a href=\"index.php?act=findpost&pid=185870\"][{POST_SNAPBACK}][/a]
Mark,
Recovery is for the purpose you suggest, but in the case of global overexposure where all tones are lifted by say 0.5 EV, it may be better to use the exposure slider and bring down all tones by 0.5 EV. Recovery is useful in high dynamic range situations where you want to recover the highlights and not affect the three quarter tones and mid tones.
As an example, here are the characteristic curves for the stepchart images. On the left, negative exposure was used and highlight correction on the right, both in ACR using the previously posted example.
[attachment=5860:attachment]
In the recovery example, the slope of the curve is too flat, and the mid tones and highlight tones are not well separated
-
Mark,
Recovery is for the purpose you suggest, but in the case of global overexposure where all tones are lifted by say 0.5 EV, it may be better to use the exposure slider and bring down all tones by 0.5 EV. Recovery is useful in high dynamic range situations where you want to recover the highlights and not affect the three quarter tones and mid tones.
As an example, here are the characteristic curves for the stepchart images. On the left, negative exposure was used and highlight correction on the right, both in ACR using the previously posted example.
In the recovery example, the slope of the curve is too flat, and the mid tones and highlight tones are not well separated
[a href=\"index.php?act=findpost&pid=186076\"][{POST_SNAPBACK}][/a]
Bill, yes, if all three channels are clipped there is nothing to recover, why I suggested one needs some information to make use of it. Your second sentence is the point I was trying to make.
While the charts are interesting for confirming what one observes on the image itself, the nice thing about ACR and a well-calibrated and profiled display is that we can try any combination of adjustments and see how they play-out on the image - an embarassment of riches really. Pushed too far on certain images, I agree Recovery can flatten the highlights too much - as your charts show, in which case use of the tone curves becomes more appropriate. Fortunately, once we've finished doing all this stuff, ACR implements it all in optimal sequence for us.
-
Hi Eric,
As far as I'm aware the only recent cameras with live histograms are non-DSLRs. Even my 1DsMk3, which has live-view, does NOT (much to my chagrin) have a live histogram. I wish..........
Mark
[a href=\"index.php?act=findpost&pid=185822\"][{POST_SNAPBACK}][/a]
Are you absolutely sure about this, Mark?
I recently used a 1D Mark III in Live View mode and it had a live three-channel RGB histogram. I am fairly certain that if the 1D III has it, then the 1Ds III should as well.
The new Rebel XSi also has it.
Please check again.
-
IMO, any test about ETTR to analyse if a RAW file was clipped, needs neutral (1.0) white balance multipliers. Since ACR and any other commercial developer don't allow them, I would not take too much seriously precise results obtained using them.
DCRAW achieves neutral WB by -r 1 1 1 1
Regards.
[a href=\"index.php?act=findpost&pid=185954\"][{POST_SNAPBACK}][/a]
For shooting step wedges such as what Bill is doing, it doesn't really matter since the green channel will be the first to clip, and the green multiplier is nearly always 1 anyways.
-
Hi Eric,
As far as I'm aware the only recent cameras with live histograms are non-DSLRs. Even my 1DsMk3, which has live-view, does NOT (much to my chagrin) have a live histogram. I wish..........
Mark
[a href=\"index.php?act=findpost&pid=185822\"][{POST_SNAPBACK}][/a]
Mark,
Well, my 1Ds3 does display live histogram in live view mode. Read page 114 of your manual.
You need to set C.Fn IV - 16 to 1 (simulate exposure) to activate it. If it isn't displayed, then press the info button (once or twice).
-
Bill, as we discussed briefly on the Adobe User-to-User forums, the issue here is the baseline exposure differences between various cameras and across vendors. There is a fundamental tradeoff between (a) honoring the original distribution of raw values (between black point and white point) and ( b ) having a given exposure (e.g., f/8, 1/10th sec, ISO 400) be rendered the same way across different cameras. In order to have common controls such as Exposure compensation behave the same way across the different cameras, Camera Raw applies a baseline exposure compensation that varies from model to model in order to get them all to behave similarly when the exposure compensation control is set to its default value of zero.
In the case of the D3, the baseline exposure is set to +0.5. You can "undo" this by setting CR's exposure compensation to -0.5 (and save that as your camera default if that's what you prefer).
-
Mark,
Well, my 1Ds3 does display live histogram in live view mode. Read page 114 of your manual.
You need to set C.Fn IV - 16 to 1 (simulate exposure) to activate it. If it isn't displayed, then press the info button (once or twice).
[a href=\"index.php?act=findpost&pid=186095\"][{POST_SNAPBACK}][/a]
Eric, Francois - thanks - yes indeed it is there - we live and learn. Even the Canon rep at a trade show couldn't find it - neither of us had read page 114!
-
Eric, Francois - thanks - yes indeed it is there - we live and learn. Even the Canon rep at a trade show couldn't find it - neither of us had read page 114!
[a href=\"index.php?act=findpost&pid=186107\"][{POST_SNAPBACK}][/a]
Great! Intuitive is not always part of the dictionary of modern electronic equipment manufacturers.
-
Bill, as we discussed briefly on the Adobe User-to-User forums, the issue here is the baseline exposure differences between various cameras and across vendors. There is a fundamental tradeoff between (a) honoring the original distribution of raw values (between black point and white point) and ( b ) having a given exposure (e.g., f/8, 1/10th sec, ISO 400) be rendered the same way across different cameras. In order to have common controls such as Exposure compensation behave the same way across the different cameras, Camera Raw applies a baseline exposure compensation that varies from model to model in order to get them all to behave similarly when the exposure compensation control is set to its default value of zero.
In the case of the D3, the baseline exposure is set to +0.5. You can "undo" this by setting CR's exposure compensation to -0.5 (and save that as your camera default if that's what you prefer).
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=186098\")
Eric,
As long as we understand what is going on behind the scenes so to speak, we can deal with the issues as you suggest. My question is from where does the BaseLine exposure value come? From the DNG specification I see that it is one of the metadata values. Does this value come from the camera maker or is it filled in by the DNG converter from its own database? Also, how does one examine the DNG metadata fields to determine their value?
The point of my original post here was that without this knowledge, one can not really use ACR or Lightroom to determine if there is clipping in the raw file. IMHO it is best to examine the raw data directly with a program such as Iris. Another good way to look at the raw data is another freeware program [a href=\"http://www.cryptobola.com/photobola/rawnalyze.htm]Rawanalyze[/url].
DigitalDog does not go into these issues in his referenced article.
Bill
-
Eric,
As long as we understand what is going on behind the scenes so to speak, we can deal with the issues as you suggest. My question is from where does the BaseLine exposure value come? From the DNG specification I see that it is one of the metadata values. Does this value come from the camera maker or is it filled in by the DNG converter from its own database? Also, how does one examine the DNG metadata fields to determine their value?
The point of my original post here was that without this knowledge, one can not really use ACR or Lightroom to determine if there is clipping in the raw file. IMHO it is best to examine the raw data directly with a program such as Iris. Another good way to look at the raw data is another freeware program Rawanalyze (http://www.cryptobola.com/photobola/rawnalyze.htm).
DigitalDog does not go into these issues in his referenced article.
Bill
[a href=\"index.php?act=findpost&pid=186395\"][{POST_SNAPBACK}][/a]
Bill, at a purely operational level without the benefit of having those applications you recommend, would you agree that when you open a file in ACR or Lightroom with all the luminosity information in the Basic tab zeroed and the two tone curves set to linear, this should give a fairly reliable interpretation of whether or not the raw file has suffered from highlight clipping during exposure?
Mark
-
The point of my original post here was that without this knowledge, one can not really use ACR or Lightroom to determine if there is clipping in the raw file.
It's really not that hard to check for RAW clipping in ACR. Slide the exposure control to -4, and if the histogram data still is touching the right edge, or any of the channels exhibit spikes at their maximum values as shown below even if the spike isn't touching the right edge, then the RAW is clipped.
(http://www.visual-vacations.com/Photography/MiscImages/ClippedRAWHisto.gif)
Rawanalyze is also an excellent tool for evaluating exposure. It is particularly useful for determining the exposure offset between the camera histogram's clip indication and actual RAW clipping. It does so more precisely than ACR, and if you have underexposed, Rawanalyse will show you exactly how much.
-
Bill, at a purely operational level without the benefit of having those applications you recommend, would you agree that when you open a file in ACR or Lightroom with all the luminosity information in the Basic tab zeroed and the two tone curves set to linear, this should give a fairly reliable interpretation of whether or not the raw file has suffered from highlight clipping during exposure?
Mark
[a href=\"index.php?act=findpost&pid=186404\"][{POST_SNAPBACK}][/a]
Mark,
Not entirely, unless you use the BaseLine exposure correction. As shown below with a file that is not clipped when one examines the raw file directly, ACR with the neutral settings you suggest (and without the baseline correction) still indicates that steps 1 and 2 in the wedge are blown. That is 2/3 of an f/stop. This is not surprising since contrast affects the mid-tones and three quarter tones but should have no effect when the normalized pixel value is 1.0 (255/255, 16383/16383, etc).
As Jeff Schewe points out in p. 73 of his ACR book, increasing brightness to values over 100 can push the highlights to 255 in an 8 bit file. This can look like highlight clipping, but he says if you look at the converted 16 bit file, there usually is not clipping. The brightness control adjusts the mid-tones without clipping the highlights or shadows and determines the point about which the contrast slider operates.
(http://bjanes.smugmug.com/photos/273026804_ESG9i-O.png)
-
It's really not that hard to check for RAW clipping in ACR. Slide the exposure control to -4, and if the histogram data still is touching the right edge, or any of the channels exhibit spikes at their maximum values as shown below even if the spike isn't touching the right edge, then the RAW is clipped.
(http://www.visual-vacations.com/Photography/MiscImages/ClippedRAWHisto.gif)
Rawanalyze is also an excellent tool for evaluating exposure. It is particularly useful for determining the exposure offset between the camera histogram's clip indication and actual RAW clipping. It does so more precisely than ACR, and if you have underexposed, Rawanalyse will show you exactly how much.
[a href=\"index.php?act=findpost&pid=186405\"][{POST_SNAPBACK}][/a]
I would prefer to use a better documented method of checking for clipping. Here is a file that is clipped in steps 1, 2, and 3 as shown on the plot below and by visual inspection (green1 channel only is analyzed):
(http://bjanes.smugmug.com/photos/271460353_f2rBe-O.gif)
If I use your method in ACR with exposure of -4, the histogram does not touch the right and whether or not those spikes at maximum are significant is open to question.
(http://bjanes.smugmug.com/photos/273879576_3ya8r-O.png)
Rawanalyze shows clipping in the raw file step 3 at about 15970:
(http://bjanes.smugmug.com/photos/271834950_3FhNB-O.gif)
For those who are interested, here is the raw histogram:
(http://bjanes.smugmug.com/photos/273882918_WNkq2-O.png)
-
Thanks Bill,
By "use the BaseLine exposure correction" - what do you mean?
Not clear to me what pg 73 of Jeff's book has to do with this. What he says about the Brightness slider is of course correct, but how is it relevant to the problem we are discussing, insofar as we have the Brightness slider at zero? Something I'm not understanding here?
Mark
-
I would prefer to use a better documented method of checking for clipping. Here is a file that is clipped in steps 1, 2, and 3 as shown on the plot below and by visual inspection (green1 channel only is analyzed):
...
If I use your method in ACR with exposure of -4, the histogram does not touch the right and whether or not those spikes at maximum are significant is open to question.
I agree that Rawanalyze is much more precise; when the RAW is only slightly clipped, ACR's highlight recovery may disguise that. But in most such cases, the clipping doesn't detract from the image significantly after the highlight recovery process. But whenever you see the indications I described in ACR, the RAW is clipped.
-
Thanks Bill,
By "use the BaseLine exposure correction" - what do you mean?
Not clear to me what pg 73 of Jeff's book has to do with this. What he says about the Brightness slider is of course correct, but how is it relevant to the problem we are discussing, insofar as we have the Brightness slider at zero? Something I'm not understanding here?
Mark
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=186480\")
Mark,
For the baseline exposure correction, see Eric Chan's [a href=\"http://luminous-landscape.com/forum/index.php?showtopic=24354&view=findpost&p=186098]Post #27[/url]. Unfortunately, I am not that familiar with that correction and did not even know of its existence prior to this discussion. I posted a couple of questions for Eric and perhaps he will reply.
I think that some important concepts are involved here and it is strange that Schewe and his colleagues do not lend us their expertise. The last I heard, Schewe was ignoring me and DigiDog never admits error .
My point was that you see about the same highlight clipping information in ACR regardless of reasonable settings for contrast and brightness. It is only when the BaseLine exposure correction of -0.5 is applied in ACR that the ACR clipping results are accurate.
Bill
-
By "use the BaseLine exposure correction" - what do you mean?
BaselineExposure is a DNG specific TIFF tag indicating, that an "exposure adjustment" should be carried out by the raw converter. This tag is inserted in the DNG file by the Adobe DNG converter with the value of +0.5 EV for the Nikon D3 and D300, -0.5 EV for the D200, and D2X, -0.75 EV for the D70 and D100. The same "correction" values are applied by ACR automatically.
The automatic adjustment has legitimate usages. For example the Highlight Tone Protection feature of newer Canon cameras "requires" +1 EV adjustment in raw processing (and pulling back the highlights), because HTP is supposed to underexpose the shot in order to prevent clipping.
The problems with these adjustments are:
1. they are not always "legitimate". The argument
In order to have common controls such as Exposure compensation behave the same way across the different cameras, Camera Raw applies a baseline exposure compensation that varies from model to model in order to get them all to behave similarly when the exposure compensation control is set to its default value of zero
is plain bullshit. What we see here is a "correction" of the camera's ISO values. Some camera's gains are not corresponding to ISO 100, 200, etc. but to ISO 80 or ISO 135 or whatever. Camera manufacturers don't like to admit this, and call the settings "ISO 100" etc.
It is not the task of the raw processor to "correct" the camera's characteristics.
2. the adjustment can be destructive: if the raw data "encroaches" into the top 0.5 EV (with a Nikon D3 or D300), the adjustment causes clipping in the conversion. On the other hand, a negative adjustment can ruin the shadows.
In both cases the adjustment is without any reason. The numerical range of the pixel values for a camera (and in some cases for the actual ISO) is given, and the camera can occupy that full numerical range, depending on the exposure. There is no basis to remoive part of those values.
3. The worse on this habit of mutilation is, that ACR does not give any indication of the automatic adjustment: the "Exposure" slider is at 0, even though an adjustment occured. Therefor all those users, who don't know about the mutilation are led to believe, that their shot is over- or underexposed, even if the shot is perfect.
-
Pano,
Thanks for pitching in here. This is a very interesting perspective on the issue. If correct, I'm beginning to see the solution to this problem. There are a couple of layers to it, leaving out the DNG complication for the moment: (1) What the camera's JPEG is showing relative to what the raw file is really like; and (2) What the raw file is really like compared with how ACR portrays it, with all user-defined settings as neutral as they can be (i.e. zeros and straight lines).
It seems that the RawAnalyze program Jonathan recommends - according to Jonathan - is the one that "tells the truth" about whether the raw file actually has clipping or not.
Therefore the solution to this issue would seem to be to find a JPEG setting for the camera (by making test exposures ostensibly using ETTR) which comes as close as possible to a no-clipping histogram both on the camera LCD and in RawAnalyse. Then open that image in ACR and set ACR's Basic Tab luminosity controls to reference values which come as close as possible to replicating the "correct" histogram in ACR, and use those as one's defaults for that particular camera.
Perhaps further adjustments would be needed to deal with any baked-in compensation if one converts to DNG.
-
Mark,
For the baseline exposure correction, see Eric Chan's Post #27 (http://luminous-landscape.com/forum/index.php?showtopic=24354&view=findpost&p=186098). Unfortunately, I am not that familiar with that correction and did not even know of its existence prior to this discussion. I posted a couple of questions for Eric and perhaps he will reply.
I think that some important concepts are involved here and it is strange that Schewe and his colleagues do not lend us their expertise. The last I heard, Schewe was ignoring me and DigiDog never admits error .
My point was that you see about the same highlight clipping information in ACR regardless of reasonable settings for contrast and brightness. It is only when the BaseLine exposure correction of -0.5 is applied in ACR that the ACR clipping results are accurate.
Bill
[a href=\"index.php?act=findpost&pid=186492\"][{POST_SNAPBACK}][/a]
Bill,
I think it would be very helpful to get Jeff's and Andrew's technical input here; and to this objective it could be more useful to encourage them to participate rather than to presume upon their motives vis a vis yourself, the basis of which certainly isn't apparent to me.
So getting back to the substance of the issue, I agree, the Brightness and Contrast settings largely affect the midtones - though both can impact on highlights where there are tones near clipping before these adjustments are made. It is normal to expect that an Exposure change will have a greater impact on highlight clipping. All that said, in my just previous post, based on a contribution from Pano, I suggested an empirical procedure for being able to manage one's exposures between the histogram view, the raw file and ACR. Feedback on these suggestions would be most interesting.
Cheers,
Mark
-
Bill, I should have also have mentioned that Jeff and Andrew are probably both immersed or about to be immersed in PhotoshopWorld at Orlando, so that may explain why we're not hearing from them. I think if either of them sees this thread and they have the time we may hear from one or both as and when they can manage it. Let's just wait and see.
Mark
-
It is not the task of the raw processor to "correct" the camera's characteristics.
Disagree.
I think it's one of the raw converter's most important features, to compensate for differences among cameras including deficiencies. There are characteristics of cameras & lenses, such as hot pixels, chromatic aberration, noise, excessive infrared response, etc. which are all useful to treat in a raw processor. Most people consider these attributes "problems" or "deficiencies" and would like to be able to tame them in the raw converter.
There are also widely differing characteristics in sensor response which requires normalization (through the use of a color matrix or color lookup table).
If you want to examine raw data then I suggest ACR isn't the right place to do that even with the tone controls 'linearized' because ACR will still apply the color transform.
If you really want to start with the original data, run dcraw in document mode (-D) and enjoy.
Some (many, I would argue) people find it useful to be able to shoot a given job with two different camera models set at the same exposure and get the same results (in terms of tonality).
2. the adjustment can be destructive: if the raw data "encroaches" into the top 0.5 EV (with a Nikon D3 or D300), the adjustment causes clipping in the conversion. On the other hand, a negative adjustment can ruin the shadows.
I fail to see how a negative adjustment 'ruins' the shadows. It should not ruin the shadows any more than a conversion in which the baseline exposure compensation of +0.5 omitted.
Using negative EC in this case of -0.5 should not cause the shadows to clip. Perhaps you can clarify what you mean.
In both cases the adjustment is without any reason. The numerical range of the pixel values for a camera (and in some cases for the actual ISO) is given, and the camera can occupy that full numerical range, depending on the exposure. There is no basis to remoive part of those values.
As noted earlier, the data isn't really lost. Yes, those extreme highlights may be clipped in the default rendering, but that is easily adjusted (with no data loss).
3. The worse on this habit of mutilation is, that ACR does not give any indication of the automatic adjustment: the "Exposure" slider is at 0, even though an adjustment occured. Therefor all those users, who don't know about the mutilation are led to believe, that their shot is over- or underexposed, even if the shot is perfect.
Your point is well taken in the context of considering the "perfect shot" to be the one in which the raw data is not clipped and well exposed.
However, I don't consider this to be an issue, and I certainly think the use of the word 'mutilation' overstates the case (esp. since no real data is lost).
Eric
-
As long as we understand what is going on behind the scenes so to speak, we can deal with the issues as you suggest. My question is from where does the BaseLine exposure value? From the DNG specification I see that it is one of the metadata values. Does this value come from the camera maker or is it filled in by the DNG converter from its own database? Also, how does one examine the DNG metadata fields to determine their value?
Bill, the camera vendors who have their own raw formats (such as Canon and Nikon) do not provide this information. The metadata is filled in by the DNG Converter from its own database. I think the data comes from measurements that Thomas makes when he's adding new camera support (I don't know all the details here).
You should be able use any tool that can inspect TIFF tags to examine DNG metadata, since DNG is a TIFF extension.
The point of my original post here was that without this knowledge, one can not really use ACR or Lightroom to determine if there is clipping in the raw file. IMHO it is best to examine the raw data directly with a program such as Iris. Another good way to look at the raw data is another freeware program Rawanalyze (http://www.cryptobola.com/photobola/rawnalyze.htm).
I agree with you on this point. Given that ACR always performs some initial processing (besides baseline exposure compensation; things like white balance, etc.) I would suggest that if you're really interested in examining the raw data that you use Rawnalyze or dcraw for that purpose.
-
I think it would be very helpful to get Jeff's and Andrew's technical input here; and to this objective it could be more useful to encourage them to participate rather than to presume upon their motives vis a vis yourself, the basis of which certainly isn't apparent to me.
[a href=\"index.php?act=findpost&pid=186523\"][{POST_SNAPBACK}][/a]
[/quote
Probably so, but I have had some unpleasant experiences with these characters and have found that they are not open to ideas that differ from their preconceived notions.
So getting back to the substance of the issue, I agree, the Brightness and Contrast settings largely affect the midtones - though both can impact on highlights where there are tones near clipping before these adjustments are made. It is normal to expect that an Exposure change will have a greater impact on highlight clipping. All that said, in my just previous post, based on a contribution from Pano, I suggested an empirical procedure for being able to manage one's exposures between the histogram view, the raw file and ACR. Feedback on these suggestions would be most interesting.
[a href=\"index.php?act=findpost&pid=186523\"][{POST_SNAPBACK}][/a]
[/quote
If you look back to my original post in this thread that is what I did. I found that the D3 luminance histogram is within 0.3 EV of the clipping in the raw channel. That is about where I want it. It leaves just a bit of headroom.
Here is the raw histogram:
(http://bjanes.smugmug.com/photos/273996478_e24WK-O.png)
and the camera histogram:
(http://bjanes.smugmug.com/photos/271460374_mRjof-O.png)
and with clipping gone with another 0.3 EV exposure reduction:
(http://bjanes.smugmug.com/photos/271460388_WDMpW-O.png)
-
Therefore the solution to this issue would seem to be to find a JPEG setting for the camera (by making test exposures ostensibly using ETTR) which comes as close as possible to a no-clipping histogram both on the camera LCD and in RawAnalyse. Then open that image in ACR and set ACR's Basic Tab luminosity controls to reference values which come as close as possible to replicating the "correct" histogram in ACR, and use those as one's defaults for that particular camera.
I don't really understand, what problem should be solved this way. The solution for the problem at hand is simple: the auto-adjustment of ACR has to be reserved by the "Exposure" slider, so that one starts at "real zero". From there further adjustments can be done the usual way.
However, this advice does not help those users, who don't know about the auto-adjustment.
-
I think it's one of the raw converter's most important features, to compensate for differences among cameras including deficiencies
That's ok. However, we were not talking about a deficiency (except when making perfect shots is a deficiency of the camera).
As to the rest: I don't feel like playing in kindergarten today.
-
I don't really understand, what problem should be solved this way. The solution for the problem at hand is simple: the auto-adjustment of ACR has to be reserved by the "Exposure" slider, so that one starts at "real zero". From there further adjustments can be done the usual way.
However, this advice does not help those users, who don't know about the auto-adjustment.
[a href=\"index.php?act=findpost&pid=186586\"][{POST_SNAPBACK}][/a]
I explained what the problem is about in the previous paragraph and it has been running (or meandering) through the thread, so kindly just remain in kindergarten a bit longer to sharpen your understanding: inconsistency between the histogram reading from the in-camera JPEG and the structure of the raw file, and seeming inconsistency between the latter and the ACR histogram resulting from "zero and linear" default settings. That's how the issue has evolved from the OP. Jonathan and Eric recommended two pieces of software for helping to understand the raw file absent the issues of the JPEG and ACR rendition. Therefore a procedure developed around the use of that software can help one to determine a baseline of exposure practice and default ACR settings that will accomplish ETTR without clipping, as reliably and predictably as possible with present tools and techniques. Clear now?
-
Bill, as we discussed briefly on the Adobe User-to-User forums, the issue here is the baseline exposure differences between various cameras and across vendors. There is a fundamental tradeoff between (a) honoring the original distribution of raw values (between black point and white point) and ( b ) having a given exposure (e.g., f/8, 1/10th sec, ISO 400) be rendered the same way across different cameras. In order to have common controls such as Exposure compensation behave the same way across the different cameras, Camera Raw applies a baseline exposure compensation that varies from model to model in order to get them all to behave similarly when the exposure compensation control is set to its default value of zero.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=186098\")
I think the main problem is that the camera vendors are not using uniform calibration procedures for the ISO of the sensor and the rest of the camera system and thus the results are not consistent between different cameras of the same or different makes.
The main variables are:
1) Calibration of the light meter itself. According to Phil's tests on DPReview, Nikon has always been spot on for meter calibration and the recent Canon's are also. The calibration is done by comparison with the reading of a calibrated hand held meter. A source of confusion is that the ISO standard for light meters is for the equivalent of around 12% gray, not the commonly assumed value of 18%. See [a href=\"http://www.bythom.com/graycards.htm]Thom Hogan's Post[/url] for details. The upshot is that if you determine exposure by a reading from a Kodak 18% gray card, you have to add 0.5 EV to the exposure if you want the result to come out as middle gray in the photograph.
2) Variations in the shutter speed and aperture control. Electronic shutters in modern digital cameras are quite accurate and repeatable (precise). The aperture control is mechanical and subject to more variability, but the mechanism usually works well unless something is sticking.
3) Absorption of light by multiple lens elements. With complex lens designs, this can be a significant factor. It is taken into account by TTL metering, which will then give more exposure than indicted by the hand held meter.
4) Calibration of the ISO sensitivity of the sensor itself. This is where I think most of the variations arise. Norman Koren (http://www.normankoren.com/digital_cameras.html) explains the ISO 12232:1998 standard for determination of the ISO of digital cameras. The standard defines two methods: saturation based and noise based. the Saturation based standard is usually used. When exposed according to a standard illumination (determined by a meter reading or derived from luminance values directly), an 18% gray card should result in the sensor being 18/106 saturated resulting in a raw pixel value of 18/106 of full scale since the sensor is linear. The denominator of 106 rather than 100 is used to allow some headroom for highlights. If a gamma curve of 2.2 without any other adjustments is applied, this results in a pixel value of 114 in the resulting 8 bit file. Most cameras and raw converters apply additional tone corrections, so it is best to examine the percent saturation in the raw file directly. In ACR you can try a linear setting with all the sliders on the main panes set to zero (exposure, recovery, brightness, contrast, etc) and the point curve set to linear. If you then take card of the baseline exposure factor and take a picture of an 18% card using the exposure indicated the camera meter, you should get a pixel value of about 114 if everything is working correctly and the camera is calibrated to the 1996 standard.
The new ISO standard is 12232:2006 (http://en.wikipedia.org/wiki/Film_speed) as explained on Wikipedia and it is slightly different. It retains the saturation based method, and amplifies on noise based methods and adds a standard output sensitivity (SOS), which will lead to a pixel value of 118 in a gamma 2.2 space with 8 bit files.
With the saturation method, an exposure of an 18% gray card based on a standard light meter reading will result in an image with a grey level of 18%/√2 = 12.7% of saturation, which corresponds to a pixel value of 100 in a gamma 2.2 space. The square root of 2 is to allow 0.5 EV of headroom for specular highlights. Since the standard light meter is calibrated on the basis of the equivalence of 12% reflectance, you have to add 0.5 EV of exposure if you want the gray card image to have a pixel value of 118 in the gamma 2.2 file.
The Wikipedia article concludes, "Despite these detailed standard definitions, cameras typically do not clearly indicate whether the user "ISO" setting refers to the noise-based speed, saturation-based speed, or the specified output sensitivity, or even some made-up number for marketing purposes."
Bill Claff, a respected Nikon guru, has published ISO values necessary to attain 18% saturation of the sensor (essentially the SOS standard) on his Web Site (http://home.comcast.net/~nikond70/Investigations/Sensor_Characteristics.htm). For example the D3 is rated at ISO 200 but an ISO setting of 138 is needed to attain 18% saturation. By contrast, the D80 is rated at ISO 200 and Bill has rated the native ISO at 93. The high ISO performance of the D3 may not be quite so good as it is cracked up to be.
Bill
-
Bill,
Thanks very much for posting this clear and detailed explanation. I shall retain it for future reference re some tests on the 1DsMk3 I intend to undertake later this month in connection with another exercise. Meanwhile, I notice the standards and the math are based on 8-bit depth. Our current crop of high-end DSLRs are working in 14 bit depth. Does this affect the numbers much?
Mark
-
Bill,
Thanks very much for posting this clear and detailed explanation. I shall retain it for future reference re some tests on the 1DsMk3 I intend to undertake later this month in connection with another exercise. Meanwhile, I notice the standards and the math are based on 8-bit depth. Our current crop of high-end DSLRs are working in 14 bit depth. Does this affect the numbers much?
Mark
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=186719\")
Mark,
You are welcome and I look forward to seeing the results of your testing. I would also suggest that you participate in some of Bill Claff's collaborations as he is now extending his methods to Canon cameras. He does a very sophisticated analysis at multiple ISOs with automated programs that he has devised. Also it serves as a good check on your own results.
[a href=\"http://home.comcast.net/~nikond70/Collaborations/ISO_Collaboration.htm]Claff ISO Collabroations[/url]
Here are some results including the Canon 1Ds Mark II, which does quite well considering its pixel count: Results (http://www.nikonians.org/forums/dcboard.php?az=set_threaded_mode&forum=147&topic_id=53868&prev_page=show_topic&gid=53868#53891)
If you go to the trouble of testing, I think it makes sense to use the highest bit depth possible so as to get the best results. Thus far the extra two bits do not seem to make much difference. If you normalize the results to 1.0 (e.g. for 14 bit normalized pixel value = observed pixel value/16383) the results are the same for different bit depths.
Bill
-
Thanks Bill, I'll keep this on referral as it will be a bit of time before I can get down to this work. I'll report progress as and when........
Mark
-
inconsistency between the histogram reading from the in-camera JPEG and the structure of the raw file, and seeming inconsistency between the latter and the ACR histogram resulting from "zero and linear" default settings
These two issues are totally independent.
The former is an old hat, the solution has been discussed in a thread started by Guillermo for many weeks ago. Since then I have been using a special WB preset "template", which makes the previews ridiculously off-color, but I achieve ETTR always exactly (i.e. within 1/3 EV).
The latter issue is an ACR-specific problem, the solution of which lies with Adobe.
Your post was in response to the description of ACR's unwanted adjustment, thus I did not understand, what you want to achieve.
-
These two issues are totally independent.
The former is an old hat, the solution has been discussed in a thread started by Guillermo for many weeks ago. Since then I have been using a special WB preset "template", which makes the previews ridiculously off-color, but I achieve ETTR always exactly (i.e. within 1/3 EV).
The latter issue is an ACR-specific problem, the solution of which lies with Adobe.
Your post was in response to the description of ACR's unwanted adjustment, thus I did not understand, what you want to achieve.
[a href=\"index.php?act=findpost&pid=186892\"][{POST_SNAPBACK}][/a]
The two issues are separable but related in terms of the end objective. Thanks for your guidance about what is "old hat". Whatever Adobe does or doesn't do, to the extent any workaround is needed, I believe that is wha some of us have been discussing.
-
I really don't see a problem with ACR's exposure bias adjustment. It is nothing more than an offset between the actual exposure value and the displayed exposure value. Under no circumstances does it cause unrecoverable clipping. If the exposure bias is +.5 stops, then the actual exposure adjustment is +.5 stop when the displayed exposure adjustment is 0. If you have a perfect ETTR shot, sometimes you'll see clipping when the exposure slider shows and adjustment of zero, but in all cases, moving the slider to -.5 will eliminate the clipping. There is no case where the exposure bias adjustment will unrecoverably clip channels in a RAW image. All you need to do is move the exposure slider to the offsetting value, and the "issue" is completely eliminated.
-
Jonathan, yes, that's fine. But to remind about the main point of this discussion, starting from the OP, it's about the predictability of getting a good ETTR: <<do you develop a feel for how much to expose to the right, getting non-clipped shots most of the time?>> In this context, ideally one would want a set of defaults for both the in-camera JPEG and ACR which start by giving a reasonably reliable read of how the camera histogram portrays the raw data, and then in ACR, hoiw its histogram portrays the same data. The idea - and ideal - is to operate in both stages with as little custom adjustment as possible after say the second capture, once the first histogram is available - or even the first capture if we're talking live-view with live histogram. In this regard, the comments by Andrew on the camera's JPEG settings, and by Eric, Bill and yourself on raw file programs for evaluating the raw data apart from Camera Raw have been very helpful. I think the answer to the OP is that one needs to work with these tools and experiment till one finds the settings that work best for various kinds of imaging conditions.
-
I've just noticed this interesting old thread.
Let me add my opinion about Adobe's implementation of the baseline exposure adjustment. I think it s flawed because, as a matter of principle, these things should be DOCUMENTED and USER SELECTABLE. There are good arguments for not applying baseline exposure compensation and, IMO less good, arguments, for doing so. So, just document the damn thing and let the user decide whether he needs this function or not. Just my 2 cents worth.
-
I've just noticed this interesting old thread.
Let me add my opinion about Adobe's implementation of the baseline exposure adjustment. I think it s flawed because, as a mater of principle, these things should be DOCUMENTED and USER SELECTABLE. There are good arguments for not applying baseline exposure compensation and, IMO less good, arguments, for doing so. So, just document the damn thing and let the user decide whether he needs this function or not. Just my 2 cents worth.
I agree that better documentation, not only from Adobe but also the camera makers, would be helpful. After all, it is the camera makers decision to allow for varying amounts of headroom for highlights that has prompted Adobe to use the baseline exposure adjustment. If a maker allows extra headroom for the highlights, it will likely use a baseline exposure adjustment so the previews do not appear dark. Since a common complaint about ACR is the images do not match the camera JPEG previews, Adobe is more or less forced to use a similar offset. As Eric Chan has explained, the baseline adjustment is user adjustable, since the user can make an exposure adjustment a part of his/her ACR defaults.
Bill
-
As Eric Chan has explained, the baseline adjustment is user adjustable, since the user can make an exposure adjustment a part of his/her ACR defaults.
It is no good being user adjustable if one doesn't know it's there in the first place, is it? How many LR/ACR users have read this thread (or any other thread / forum where somebody might have mentioned it) to know about it? Suppose they've heard about it, how are they to know by how much Mr. Knoll has decided to adjust each camera's output?
I've had it with all sorts of undocumented processing and features by Adobe. Yes, other vendors might do similar things but it's the Adobe tools I use. Camera vendors not being documented enough is no excuse for third party sw vendors to be undocumented also. And no, based on the explanations given above, this is not done to match the out of camera jpg, it is done to match different cameras' output. I bet the users who have a need for that feature are in a minority and since, most probably, they are advanced pro users they will run their tests and know how to adjust the ACR defaults to provide them with a uniform exposure across cameras. Nobody needs Mr. Knoll deciding what's best for them without even letting them know about it.
-
I agree that better documentation, not only from Adobe but also the camera makers, would be helpful. After all, it is the camera makers decision to allow for varying amounts of headroom for highlights that has prompted Adobe to use the baseline exposure adjustment
The so-called "headroom" is a myth. It relates to a certain way of metering in a certain setting; as such it is a concept of very limited use. Tell about the headroom to all those, who sometimes overexpose when metering with the camera despite the headroom.
If a maker allows extra headroom for the highlights, it will likely use a baseline exposure adjustment so the previews do not appear dark. Since a common complaint about ACR is the images do not match the camera JPEG previews, Adobe is more or less forced to use a similar offset
I don't remember to have read anywhere, that Adobe had been asked to "correct" camera makers' mistakes.
As Eric Chan has explained, the baseline adjustment is user adjustable, since the user can make an exposure adjustment a part of his/her ACR defaults
1. The baseline exposure adjustment is paternalistic. Who the hell asked Adobe to make such decisions on part of the photographer? What is Adobe's aim? To shild the camera makers from the justified criticism?
2. The baseline exposure adjustment is clandestine. The users have no idea that this happens, thus they can not react by adjusting the ACR defaults. They are plainly misled by ACR.
3. The baseline exposure adjustment is counterproductive, for it hides the real exposure issues; in some cases it simply defeats the purpose. Example: Highlight Tone Protection with Canon cameras. Such a shot is underexposed by 1 EV (relative to the user's intention/camera metering). The camera's JPEG processing and Canon's raw processing software corrects this underexposure, in a non-linear fashion; this is the essence of HTP. What is ACR doing? It increases the intensity in a linear fashion without telling about it. The result is the worse of both worlds.
4. The baseline exposure adjustment is incorrect in several cases.
All in all: Adobe should get out of the business of fooling the photographers.
-
When talking about baseline exposure does this include the base contrast curve applied to the demosaiced sensor data that turns the dark linear preview AFTER demosaicing into a rendered image that somewhat resembles what the eye sees give or take varying contrast levels?
Have you ever examined this curve? It's not an easy curve to tweak when compensating for highlights or contrast because of its huge arc. One slight nudge or tweak to the curve and the highlights either go too dark or blow completely out. Same happens adjusting contrast. You can even see this artificial software driven simulation of light increase within the in-camera processor as it tries to deal with this crappy tone curve with each 1/3 to 1/2 step increase in exposure. One exposure shows the highlights located in the 1/4 tone section of the in-camera histogram and then the next step bunches it up on the right wall. WHAT? NO IN BETWEEN! LIGHT DOESN'T BEHAVE THAT WAY AND I DON'T THINK SENSORS DO EITHER!
So on top of this major base tone curve are the sliders and curve tools to apply finer tweaks. I think ACR does a remarkable job of allowing the user to smoothly and finely remap all tones without having to deal with this unruly base curve.
You could work directly off this dark linear data from scratch by editing this big arcing base curve or creating and applying a 1.0 gamma profile after the raw file has been rendered to 16 bit tiff and edit from there. Some raw converters provide this option. But from my experience fiddling with two raw converters that claim to deliver TRUE LINEAR data, I've gotten two different previews of the same raw file between them. So even at this level it's still interpretive and so it now becomes a question of not of who's right but what works.
No software be it the camera vendor's or a third party raw converter is going to get it right all the time because the data is interpretive right from the moment light hits the sensor.
-
I agree with Panopeeper, Adobe's implementation of baseline exposure just flat-out sucks. If their supposed goal is to match the tonality of in-camera JPEG then they have failed, at least in the case of my D300.
My approach to exposure for landscape work is to use manual exposure mode and UniWB custom white balance, exposing as far to the right as possible without blowing important highlights. When I first got the D300 I could not figure out why all of my NEF's were opening up in ACR signifcantly over-exposed, with blown highlights that weren't there in-camera. I'd still be in the dark if not for finding out about this baseline exposure adjustment. Not only do D300 NEF's look overexposed by default in ACR, but I think this behavior probably also accounted for some of the early criticisms about shadow noise at ISO 200 on the D300.
Now that I know about this baseline adjustment crap I've set my camera defaults to automatically apply -0.5 EV adjustment to offset it, and whenever I shoot with a new camera I'll be sure to use Rawnalyze to figure out what the baseline adjustment is so that I can 'undo' it. But I shouldn't have to do that, and the fact that Adobe was doing this behind my back because they think they know better than me what my exposures should look like is aggravating and insulting.
-
As Eric Chan has explained, the baseline adjustment is user adjustable, since the user can make an exposure adjustment a part of his/her ACR defaults.
How many people know how to determine the correct offset? You have to convert your file to DNG and open it in Rawnalyze to see what the baseline adjustment is, Adobe completely hides this information in their own UI and metadata displays. And to be thorough, you have to do this at every ISO you use, since the baseline adjustment can vary.
For that matter, how many people who don't hang out in technical photography forums even know this baseline adjustment even exists?
-
My approach to exposure for landscape work is to use manual exposure mode and UniWB custom white balance
Have you verified your Uni-WB? Rawnalyze from version 2.9.9.3 displays the WB coefficients on the histogram panel. They should not be farther apart than 5%.
Now that I know about this baseline adjustment crap I've set my camera defaults to automatically apply -0.5 EV adjustment to offset it
Are you aware of the fact, that the baseline exposure for the D300 is MINUS 0.5 EV @ ISO 100? Accordingly, your adjustment should be +0.5 EV.
This is one of the cases of incorrect adjustments I mentioned above; it should be about -0.15 (it should be zero, but compared to +0.5 EV @ ISO 200, the -0.5 EV @ ISO 100 is incorrect).
-
Have you verified your Uni-WB? Rawnalyze from version 2.9.9.3 displays the WB coefficients on the histogram panel. They should not be farther apart than 5%.
Shot WB coeffs RGGB: (1.0000, 1.0000, 1.0000, 1.0039). The WB setting file was courtesy of Iliah Borg.
Are you aware of the fact, that the baseline exposure for the D300 is MINUS 0.5 EV @ ISO 100? Accordingly, your adjustment should be +0.5 EV.
Yes, the baseline exposures for the LO/HI ISO settings are different. Before I found out about this baseline exposure stuff I shot at LOW -.33 (ISO 160) because the shadows seemed a bit cleaner, but that's because a smaller baseline exposure is used for that setting was lower. Nowadays I just shoot at ISO 200.