On the one hand, one sees advice to leave exposure compensation set on -1/3, in order to get better color. This is especially urged on small sensor cameras, which have a tendency to blow out highlights.
On the one hand, one sees advice to leave exposure compensation set on -1/3, in order to get better color. This is especially urged on small sensor cameras, which have a tendency to blow out highlights.If your sensor clips, you will have erroneous data in the affected pixels. Clever raw developers may do a good job of concealing that error if the clipping is not too severe, but it will still be a "guess".
I am confused about two pieces of advice one often sees, that seem to me to be contradictory. On the one hand, one sees advice to leave exposure compensation set on -1/3, in order to get better color.As already said, reducing EC to get better color is wrong.
This is especially urged on small sensor cameras, which have a tendency to blow out highlights.It's wrong here as well. Highlights are usually blown by the one-size-fits-all image processing of the camera. Also, so-called "intelligent" metering modes may increase exposure to avoid underexposure. These modes attempt to apply their own exposure compensation to the shot. They may over or underexpose the image if they guess wrong.
On the other hand, one reads articles urging photographers to expose to the right, which I gather means that in taking a photograph, one should examine the histogram and if necessary boost exposure compensation so that the curve shifts over toward (but does not touch) the right side of the histogram.
These pieces of advice seem to be at cross purposes. Can someone enlighten my understanding?
If there's any white in the scene, then you can't ETTR. If there's any bright red in the scene, then you can only ETTR by 1/3 stop or so before getting into trouble.
Care to explain these statements? They don't really jive with theory or my experience ...Sure.
I think the problem that faces most people is that of being able to set standard exposure reliably. In just about any scene that's static enough to implement ETTR, you should be able to use a gray card to set standard exposure. But you do need to have a card with a good quality matte surface...otherwise the reading will be all over the place.
There is no such thing as "standard exposure". Not only will each scene have its special exposure considerations, but different photographers will meter each scene slightly differently, depending on how they visualize the resulting print. As for an 18% gray card, they pretty much became useless with the advent of the histogram which, unlike an 18% gray card, actually provides you with useful/vital exposure data.
I'm not sure where you are getting your information, but much of it is simply not correct.
Actually, all my information is correct.
However, your notion of exposure is out of place in this thread because we're discussing ETTR.
And the rules for setting exposure via ETTR are short and sweet...maximize your exposure so that the brightest highlights you care about are near saturation.
I personally find the histogram to be useless. It's too small to indicate certain important clipped highlights, such as the whites of the eyes, and it doesn't tell you if those overexposed areas are specular in nature....I find the "blinking" function to be far more useful
So for every possible image there is only one exact place to put "the grey card" and one and only one exposure that meets "ISO standards"?No. For whatever you meter, you get standard exposure. If you meter a white wall, you'll get standard exposure and the wall will come out gray. If you meter a black wall, you'll get standard exposure and the wall will come out gray. That's how reflective metering works. The meter starts with a presumption that the world is gray...12.7% gray, to be exact. It then computes exposure so that the metered area comes out a gray of 12.7%.
Umm....you were the one who started talking about gray cards and standard exposure (neither of which was relevant to the OP)
I was only responding.
Actually, that is incorrect. When using ETTR, you should only increase exposure enough to ensure that there is relevant detail in the important shadow areas (minimize noise). That might mean increasing exposure only one stop. Pushing the histogram so far than you're near clipping is simply bad practice as you're possibly sacrificing tonal range in the highlights. ETTR is all about increasing the SNR in the shadows. That's all does.
Sure, the histogram doesn't plot every exposure level on the image, but it offers a helluva lot of information for the photographer. Most importantly, it displays placement and amount of the brightest and darkest areas of the image. Also, the slope of the end points, for instance, tells a lot about the tonal ranges of shadow and highlight regions. As the camera uses the jpeg preview to calculate the histogram, worrying about specular highlights is a waste of time as the resulting image, especially if RAW, will have a much greater overall tonal range than the camera histogram indicates. As for the blinking function, it uses the same data set as the histogram, so offers nothing new.
Exposure on digital sensors is linear.
http://forums.dpreview.com/forums/readflat.asp?forum=1041&message=42236844
"...However the question is of colour linearity, that is - if we apply a matrix transform to two shots of the same scene with exposure difference, we will have different L* obviously; but do we have same a/b? In other words, what is the deltaE2000 between those two shots after the linear exposure compensation in the raw converter? One even does not need to have an accurate colour profile to check it - just any matrix will give an idea being assigned to a binned raw data..."
so did you test yours ? provided that your raw converter indeed does linear EC
Furthermore, depending on the sensor you have, you don't always need to limit yourself to the right side of the histogram. Some sensors allow you to clip by upwards of 1 stop and still recover detail in the highlights.
Check your system carefully by shooting a scene with bright highlights (e.g. brightly lit clouds) using what I call "highlight bracketing": Shot 1-not clipped; Shot 2-clipped by 1/3 stop; Shot 3-clipped by 2/3s stop; Shot 4-clipped by 1 stop; etc. Then bring them into your editing app (LR, Aperture, CaptureOne) and process them to see what detail you can recover. That way, you can learn about the boundaries of your equipment, your visualization and shooting techniques. I have been amazed at how much can be recovered in my D800e files using LR 4.1 from areas that show as "clipped" on the histogram and with the highlight clipping indicators.
An ETTR discussion on an m4/3 forum?? Tell them to just buy bigger sensors! :P
The problem with the blinkies is that you don't know where the camera manufacturers have set the limit? I think it is wrong to assume it is 255 255 255 or 100%. I believe it is a lower value but they won't tell.
Are you serious when you say that you are concerned about the lightness values in a small area such as a person's eyes?
I must confess that I fail to see why "18%" or "12.7%" would ever have to be mentioned in a discussion about digital cameras. Film cameras might be another story, but I don't know much about those.
.that is, set your exposure by spot metering a 12.7% gray reference, then any objects that are white will be near saturation.you are assuming that amount of light falling on two does not influence the outcome... you might gave a grey object illuminated so that it will reflect more light in the end than white object... plus 12.7% is how spot meter assumes the reflectance of the target (and it varies - you need to check how your camera model meters, even around 12.7% is how most of them do), not actual reflectance of the gray reference.
you are assuming that amount of light falling on two does not influence the outcome... you might gave a grey object illuminated so that it will reflect more light in the end than white object... plus 12.7% is how spot meter assumes the reflectance of the target (and it varies - you need to check how your camera model meters, even around 12.7% is how most of them do), not actual reflectance of the gray reference.
...Yes, ETTR is a sound concept. Just have to realize that on scenes with white objects and bright highlights, standard exposure is pretty much the same as an exposure arrived at through ETTR...
The gray reference is still important to digital cameras because that's how the camera meter works.
I take it then you aren't a fan of the zone system of metering?
I take it then you aren't a fan of the zone system of metering?
"This topic deserves a discussion of its own, but for the purposes of this tutorial simply know that each camera has a default somewhere in the middle gray tones (~10-18% reflectance)."
I don't see why you differentiate between producing and capturing?
And the 12.7% doesn't vary. Otherwise, you couldn't use a handheld meter to set exposure.
Spot meter a 12.7% gray surface, and white objects are near saturation. That's how it works.
Really? It doesn't depend at all on the lighting of the scene, where in that scene the grey surface is or what the dynamic range is of the capture device?
This works the same for my RX100, my iPhone and my D700? Really?
actually it does - http://www.libraw.org/articles/Canon-5Dmk2-headroom.html
Really? It doesn't depend at all on the lighting of the scene, where in that scene the grey surface is
or what the dynamic range is of the capture device?
This works the same for my RX100, my iPhone and my D700? Really?
I'm not sure what I'm supposed to get from that. After presenting the same unsubstantiated information that everyone else does and making some quesitonable presumptions, it seems like he took a few pictures and then proceeded to draw conclusions that probably should have been experimentally verified with an additional set of captures demonstrating saturation at the predicted overexposure levels.
I must confess that I fail to see why "18%" or "12.7%" would ever have to be mentioned in a discussion about digital cameras. Film cameras might be another story, but I don't know much about those.
The gray reference is still important to digital cameras because that's how the camera meter works.
I don't need to meter anything to quickly obtain accurate exposure (which can either be ETTR or a perfect JPEG), basically because its electronic viewfinder allows me to use a real time pre-visualization method (you can see how the final image will result before clicking the shutter), versus the old fashioned metering + post-visualization scheme most DLSR's use (that often requires iteration to obtain accurate exposure).
well to his credentials he is one of two people behind rawdigger and libraw (along w/ Iliah Borg)... and yours ?
I don't have any. That's why any information I give can be easily verified by anyone reading.
to bring the bright parts as close to the clipping/significantly non-linear part of the sensor as possible.
From my point of view, positioning the middle tones at some pre-defined point along the sensor DR without regard for saturation is just backwards for digital sensors. The fact that it has worked for many years for film, and that many digital manufacturers/users are able to make good images using rules of thumb does in no way prove that it is the best way to do things.
but then you need to decide if those bright parts are more important colorwise than noise in shadows - may be you can allow yourself to place bright parts into non linear area and sacrifice there to gain better SNR for shadows (that is when bracketing, graduated filters, etc can't be used)... so it is case by case decision.Yes, if the scene DR is larger than the sensor effective DR, you might have to compromise both the top end and the bottom end (if you want to do it in one shot, no graduated ND etc). There are two ways of seeing this:
What appears totally blown out using the meter recommendation on the LCD is absolutely no indication the data (raw) really is blown out.I don't think that there is _no_ correlation between those two. The problem is that the correlation is far from perfect?
Pure guesswork:
- What it's new and interesting IMO is the correspondence between scene and output image. Camera metering/Zone System exposure based criteria said a 18% gray card should end in the middle point of the perceptual luminance scale. Does it make sense? or this was just an arbitrary rule that worked in the past but can be ignored? should every particular reflectivity in real life end at the same luminance point in the output? should a picture of a Kodak graycard be L*=50% in the output no matter if the picture was taken at 12h00 in a sunny day or at 21h00 with the sun set and the sky darkening?
;) ... an everybody w/ Canon 5D mkII can verify spot metering himself/herself based on the technique described
I can tell you that if you spot meter a neutral, evenly lit surface, then a white-balanced RAW conversion of the resulting capture, with no additional processing, will have sRGB values of 100, 100, 100 at the metered area.
What raw converter with what (default) settings?LightRoom with its neutral profile...I forget the name of it, but it doesn't apply any processing.
LightRoom with its neutral profile...I forget the name of it, but it doesn't apply any processing.
But what about tungsten light (as you've mentioned)metering will underexpose...
Neutral Camera profile? Neutral (default) rendering settings? Who says either are ideal?Not me. My words on the subject, located elsewhere in the thread, is that standard exposure is a tool, not a destination.
They are just a starting point.Exactly. Darn good starting point.
I can set the camera to ‘over expose’ 1.5 stops past what the meter recommends, normalize the image in LR using Exposure. So what makes my capture and your capture different, which is ideal?As I said, standard exposure is just a tool. It allows me to know what I'm getting. What I get may be just right, or may require an hour or more of processing. What's important is that I know I've captured a good starting point.
Again, I don’t see how we can separate the exposure from the processing. Setting a 1.5 stop increase will blow out the JPEG but the raw is just fine after normalizing in my raw converter.1.5 stop increase over what metering point? And to what JPEG are you referring to...an OOC or one with no processing? I'm trying to ascertain whether the blown highlights are from the initial RAW conversion, or from one-size-fits-all default processing. I'm no fan of OOC JPEGs, but I find that when using the Neutral picture control of my Nikon, that OOC JPEGs won't clip white unless RAW does (red is always another story.)
1.5 stop increase over what metering point? And to what JPEG are you referring to...an OOC or one with no processing?
metering will underexpose...When metering under tungsten light, if I increase exposure by 1/3rd stop over standard exposure, I will clip the green in bright white objects. So my Nikon, at least, doesn't underexpose in tungsten light.
Yes, 1.5 stops OVER what the incident meter suggested (ETTR) with raw normalization that shows zero blown highlights.And what did the camera meter suggested?
The JPEG I’m referring to is the camera generated JPEG that would be totally blown out using that meter suggestion. You can’t expose for raw and JPEG the same way. One major reason is the processing which I continue to ask: how can we separate that from the discussion and process. The JPEG processing is in-camera, the raw under my control. One is a useless exposure, the other is pristine with far less noise due to ETTR.Well..."far less" might be a stretch...but anyways, I'm not sure why we're concerned about the in-camera JPEG to begin with.
And what did the camera meter suggested?
Well..."far less" might be a stretch...but anyways, I'm not sure why we're concerned about the in-camera JPEG to begin with.
When metering under tungsten light, if I increase exposure by 1/3rd stop over standard exposure, I will clip the green in bright white objects. So my Nikon, at least, doesn't underexpose in tungsten light.no Graystar - I was referring to my question/note to Guillermo (see what I was answering above), who was essentially saying that spot metering is not necessary... he saying basically that matrix metering or centerweight metering will be good and I was saying that such metering will underexpose under the tungsten light... he knows that for sure, but his position is (I guess) that w/ modern sensors and raw converters that is OK and we do not need to put any expo correction for such metering in camera under such light...
Yes, 1.5 stops OVER what the incident meter suggested (ETTR) with raw normalization that shows zero blown highlights.
The JPEG I’m referring to is the camera generated JPEG that would be totally blown out using that meter suggestion. You can’t expose for raw and JPEG the same way. One major reason is the processing which I continue to ask: how can we separate that from the discussion and process. The JPEG processing is in-camera, the raw under my control. One is a useless exposure, the other is pristine with far less noise due to ETTR.
If the saturation standard is used (as with DXO measurements), exposure of an 18% reflectance card will give 12.7% sensor saturation. This allows 0.5 EV highlight headroom.
Actually, exposure of any value of reflectance will give 12.7% saturation. It's the fact that you used an 18% card that will give you 0.5 EV "headroom", which is the same as being 0.5 EV underexposed. If you used a 25% card you'd have 1.0 EV of headroom. But this idea of headroom is questionable, as the most common 18% card, the Kodak, comes with instructions to increase exposure by 0.5 EV. So anyone using an 18% card as instructed has no headroom.
If the D800's metering point provides 15% saturation then I hope there's an instruction somewhere to disregard Kodak's instructions...otherwise anyone who doesn't will be clipping his whites and not understanding why (other than the obvious "this is not a film camera.")
that is why particular camera model has to be tested and not assumed that its spot meter will be giving 12.x%That would only be for ETTR practitioners. For everyone else, all that matters is the unprocessed sRGB values.
That would only be for ETTR practitioners. For everyone else, all that matters is the unprocessed sRGB values.
When a D800 spot meters and photographs an evenly lit neutral surface...if the resulting image, demosaiced and white balanced only, reads 100, 100, 100 in the metered area, then it meters just like every other Nikon. I would be surprised if it didn't. If it does, then the Kodak gray card and instructions will work just fine.
Again, you are confusing the meter reading, which would give an f/stop and shutter speed for a given ISO. How the sensor would respond to this exposure is a different matter. According to the REI (recommended exposure index) that Japanese manufactures use, they may assign whatever sensor ISO they want, and the resulting sensor saturation may or may not be 12.7%. In your earlier posts you made some good points, but more recently you have gotten off track. Please read up on these topics before posting again. Doug Kerr has a good paper (http://dougkerr.net/pumpkin/articles/Exposure_Calibration.pdf) on the subject.
To determine the sensor saturation, one has to look directly at the raw file with a program such as Rawdigger. The first step is to determine the saturation value, which may not be 2^14 -1 (16383) for a 14 bit ADC. One then determines the raw value for the exposure (the ISO mentions an 18% card, but any uniform target (gray or white) will do as you mentioned. One may have to subtract an offset (Canon cameras). The quotient of these values is the percent saturation. The rendered sRGB value can not be used since a tone curve may have been applied.
DXO (http://www.dxomark.com/index.php/About/In-depth-measurements/Measurements/ISO-sensitivity) describes how they determine sensor sensitivity and how the manufacturer may deviate from the saturation standard. Look at the ISO measurement for the Phase One IQ180.
Metering works exactly the same way in a digital camera as it does in a film camera, as it does across digital cameras. My Canon compact, spot-metering my gray card, gives me the same exact exposure settings as my Nikon D90. This is all defined by ISO 2721. Metering has nothing to do with REI.
And yes, I've read DxO and many of Doug Kerr's papers...including the one you reference, as well as the one where his exposure test produced RGB 100, 100, 100.
http://dougkerr.net/pumpkin/articles/Scene_Reflectance.pdf
Again, you don't seem to understand that REI or any other ISO value for the sensor is needed to determine the EV of the exposure as determined by the light meter. This is why the Canon 5D Mark II yields only 7.5% sensor saturation as shown in the Libraw post (http://www.libraw.org/articles/Canon-5Dmk2-headroom.html) which you have disregarded. Have you looked at your raw files, or are you just spouting theory that you do not understand?
You have read these articles, but your comprehension appears to be deficient. Further discussion with you is pointless, and I am out of here. Where is the kill button?
I'll take that to mean that you photographed a gray card (or white wall) with your D800, saw that the metered areas of the images turned out to be 100, 100, 100 when using a neutral profile, and have no way to reconcile that result with your supposed 15% saturation...and are now looking for a way out.
BJanes - may be you can do a favor and post a raw file from D800 please...
BJanes - may be you can do a favor and post a raw file from D800 please...
BJanes - may be you can do a favor and post a raw file from D800 please...If it's just any ol' RAW file from a D800 that you want, Imaging Resource has lots of them...
If it's just any ol' RAW file from a D800 that you want, Imaging Resource has lots of them...
http://www.imaging-resource.com/PRODS/nikon-d800/nikon-d800THMB.HTM (http://www.imaging-resource.com/PRODS/nikon-d800/nikon-d800THMB.HTM)
I was talking about raw files to use for meter measurement as described in http://www.rawdigger.com/houtouse/lightmeter-calibration , so that you can see if D800 really meters like bjanes says or not...
Sorry, but I did not know what type of raw files you were seeking. I have uploaded files for calibration procedure to Usendit.
Thank you for those images. They confirm what I've been saying.
Matrix metering tends to set exposure around 1/3 stop higher than Spot/CW metering (as an aside...I think the actual value is smaller, and depending on the camera settings Matrix is usually 1/3rd stop more exposure and occasionally it's the same as Spot/CW. At least that's what I've found on my D90, and it would appear that the D800 is the same.)
As stated, the metered value is 15% of the max value. But that's with Matrix metering. If Spot metering had been used then the metered area would likely end up around 12%. It's pretty obviously to me that the D800's metering and RAW levels are much like my D90 (don't see why it wouldn't be) and spot metering a 12.7% gray card will likely get any bright whites under the same lighting near saturation without clipping.
According to my own tests with the 800e, if one is using a uniform target which occupies the whole field of view, the reading are the same regardless of the metering mode. This is also what is stated on the Rawdigger site:
"the grey target (a grey patch can also be used, as well as plain white paper, without optical whiteners). Turn off all exposure adjustments and shoot. If the target takes up the entire field of view of the shot, then the setting of the exposure meter does not matter; all settings (spot, matrix, center-weighted) will result in the same exposure. "
Not for nothing but you all realize this thread is in the Beginners' Questions forum? Ya all might want to keep that in mind...perhaps a new thread not here would be in order. Pretty sure many/most readers of the thread have run away and hidden already.You're right, but I think it's practically over now.
Technically, that's correct. However, Matrix meters the entire frame, which includes falloff, and that can cause it to increase exposure over a spot meter reading. And while the final image may not have much falloff due to the aperture use, metering always occurs wide open...so falloff is always an influence. Obviously, the lens used is also a factor. In any case, I've seen spot metering variations on my gray card from RGB 90 to RGB 110 in my D90, depending on the lens and conditions. So the RGB 108 of your Exp1.NEF file is within the ball park.
You raise a good point that I had not considered. In my previous test, I minimized falloff by using f/8 with a 105 mm lens. I repeated the tests with the Zeiss Distagon 21 mm f/2.8, which is an excellent lens but is known to have considerable falloff. Rather than upload the large raw files, I chose to merely show the results. However, the matrix metering results in lower saturation in my tests.
That's odd.
I have a different method of demonstrating the falloff effect. I set the camera on a tripod to photograph a gray card, and then I take 4 images. The first two, Spot and Matrix, are with a light source very close to the card so that the card exhibits falloff. The second two, also Spot then Matrix, are with the light source moved further away to minimize falloff. The result is that my first two images have different exposures, but the next two images don't...even though the only change to the camera, between the images of each pair, was the metering mode. So that demonstrates that the falloff is affecting Matrix metering.
However, in my shots with the light source close to the gray card, the Matrix image is clearly brighter...which, to me, would be the expected result of a metering mode that is averaging dark areas with bright areas. So I don't know why you got the opposite result. I've put the NEF files on my FTP...
ftp://graystar.tftpd.net (http://ftp://graystar.tftpd.net) (this address has to be copied and pasted into the address bar, as the forum software is adding an "http://" in front of it.)
DSC_9310 - light source close, Spot (f/8, 1/640s)
DSC_9311 - light source close, Matrix (f/8, 1/400s)
DSC_9312 - light source far, Spot (f/8, 1/20s)
DSC_9313 - light source far, matrix (f/8, 1/20s)
EDIT: The only time you get standard exposure from Matrix is when Matrix fails to match a scene in its database. Just wanted to note that with the D800's expanded scene database, it's possible that Matrix found a match between your image and one of its scenes, and applied the exposure compensation related to that scene (this is what makes Matrix so unpredictable.)