How far to the right are you?

[george013]

Urban,
Probably we are talking about different things. Via the mentioned link in the first post you come to the Michael Reichmann article. It's said that he is the grounder of that theory. Below a part of his article http://www.luminous-landscape.com/tu...se-right.shtml Bold is from me.

The simple lesson to be learned from this is to bias your exposures so that the histogram is snugged up to the right, but not to the point that the highlights are blown. This can usually be seen by the flashing alert on most camera review screens. Just back off so that the flashing stops.

Now of course when you look at the RAW file in your favourite RAW processing software, like Camera RAW, the image will likely appear to be too light. That's OK. Just use the available sliders to change the brightness level and contrast so that the data is spread out appropriately and the image looks "right". This will accomplish a number of things. The first is that it will maximize the signal to noise ratio. The second is that it will minimize the posterization and noise that potentially occurs in the darker regions of the image.

Overexposing and correcting in PP are basic elements in ETTR.
George

[Inkanyezi]

I'm sorry but I don't understand how nailing my white balance is going to give me an ideal exposure (ETTR not required).

I suppose I didn't say it would. It's entirely up to you how you expose your images.

I spot the brightest part of the scene (often the brightest patch in a cloudy sky) and overexpose the spot reading by two or two and a bit stops.

If you take a spot reading of the brightest highlight, and adjust your exposure to four times that value, you do not over-expose, but you compensate for the brightness of the image element you chose to measure. The image straight out of the camera may seem under-exposed, as most cameras blow out bright highlights when no compensation is done.

Often when taking a reading of the brightest highlight with +2 compensation, the camera will choose up to two stops less exposure than when using matrix metering without compensation. I cannot fathom why you would call that over-expose.

This takes care of the highlights, I'm not sure whether I'd call it ETTR or not.

It is indeed exposing to the right, regarding the SOOC jpeg. Whether that is the correct exposure or not depends on other factors, as the dynamic range of the scene and which ISO value you selected, the dynamic range your sensor can register, as well as your workflow. If you use the RAW file, the exposure most likely is not to the right.

I've got one or two poppy shots where the general exposure was spot on but the red channel was clipped. Exposing properly for the reds meant underexposing the rest of the scene a little.

You use the term "underexposing" a bit sloppily there. If you expose properly for the reds, you might adjust the contrast curve to lift the darker parts of the image, but it does not indicate under-exposure.

... the correct exposure depends on your interpretation of the scene, no one rule suits all occasions.

Are we discussing out of camera jpegs or RAW files with a workflow suitable for ETTR?
When you use the ETTR workflow, you can adjust the image to any possible brightness and contrast, as long as the dynamic range of the scene did not exceed the range of the sensor. Hence ETTR might be insufficient for a scene with higher dynamic range than the actual range of the sensor, but then you would have to sacrifice either shadows, highlights or both.

If you have to blow out the highlights to get the image you want, ETTR is not the way to go. If for example you choose a lower dynamic range of the sensor, by increasing ISO, you might be forced to accept both clipped highlights and noisy shadows with little detail. ETTR is not the strategy chosen by sports photographers, but many landscape photographers use it. Horses for courses.

ETTR is useful under some circumstances and not in others and personally I'm really (too) fussy about getting it spot on in camera. I get really (too) disappointed when I have to use Photoshop to correct my mistakes before processing the image.

The purpose of ETTR is to get it spot on in the camera, but it is a different idea of what "spot on" means. The workflow must coincide with the idea. With a different idea and workflow, ETTR exposure is bound to fail. With conversion in the camera, most images will be over-exposed, as highlights are almost certain to clip. With standard curves in any RAW converter you'll get similar results as in the camera.

However, if your workflow is such, that you adjust the image to your desired tonality in your RAW conversion, ETTR is spot on correct exposure. Nothing is clipped, and you get as much shadow detail as possible with the least possible noise at actual ISO setting.

And I know that most people do not subscribe to that idea, and it's fine with me. You do as you think is best, and others may have other ideas. It's a wonderful world, where we can all have our ideas. There is no conflict, only different ideas and different ways of achieving similar goals. Most people are not so picky that they bother about noise in the shadows, and many people even accept blown highlights. Lots of people are happy with the jpeg files that come out of the camera.

And some people use the ETTR strategy to get optimal image quality out of their digital exposures.

Whether you subscribe to the idea or not is entirely your own business.

[AB26]

I just read a very interesting Luminous Landscape article yesterday

Janis,

Well, Bob is not saying what camera he was using. Must be a D200 like mine because it is the only camera giving you such crappy noise at ISO800.

Take what he has to say with a little salt and try it for yourself. Always experiment to see what works for you.

In my own experience: exposing to the right is what I do when shooting with flash. Keep away from the right for the rest of my shots. If the scene is very evenly lit, in other words, a flat scene, exposing to the right is easier than shooting a scene with a high DR.

No, for most of my work I stay away from the right. I don’t like blown highlights unless it is a direct reflection from a light source.

[DanK]

I think this is simpler than the article cited by the OP suggests. It boils down to this:

1. The more light you let in, the more signal, and the higher the signal/noise ratio. So yes, you get a better signal/noise ratio if you ETTR.
2. Boosting the ISO is simply amplifying the signal from the sensor--that is, amplifying the signal AND the noise. Hence, the noisier images. This is particularly apparent in dark areas, where there is less signal but the same amount of noise as in brighter areas.
3. Does ETTR actually matter much in practice? Colin always says not. I think the answer is: sometimes, depending on the image and the camera. E.g., it matters more on older cameras and for images with a concentration of pixels at the dark end.
4. If you are going to ETTR, does it matter how? No. The sensor has no information about your metering mode. All that matters is the distribution of luminance values--the histogram. Whatever gets you the histogram you want is a good way to meter.

I personally rely a lot on spot metering because I like the control, although to be honest it may also be in part a vestige of my years with a Canon FTb, which had only a spot meter. So I sometimes ETTR by metering off the brightest area I don't want clipped and then adjusting exposure from that reading. This is not the "right" way to do it. It is a right way if it gets you the exposure you want.

[xpatUSA]

Janis,

One of my more respected authors, Guillermo Luijck, has addressed both ETTR and correct white-balancing in several articles:

http://www.guillermoluijk.com/index2_en.htm

The original site is Spanish but much of it has English options. He goes a lot deeper than most of those surface-skimming, written-for-morons tutorials that we read (not here, of course!).

UniWB was mentioned earlier; Guillermo has written the most definitive article that I know of here:

http://translate.google.com/translat...hl=EN&ie=UTF-8

Hope the above is of interest.

ETTR is important for me in many shots because my early Sigma raw-only DSLRs can produce horrible green/magenta blotching in moderate shadows in the blink of a shutter.

[Polar01]

John your post #2 is more of a finished image, mine on the other hand is only a working image in that it is still a Raw file I could care less about the people as they are unimportant. The image that I posted was only one of three I took, my posted one at +2 stops, and the others at +1.3 stops and +.7 stops. What I was wanted to know was their much difference between the blown highlight on the +.7 shot compared to the +2 stop shot, a little but not enough to worry about in this case. What was interesting to me was the increase tones and detail in the darker parts of the images, say the women's dark outfits. The tones and detail in the darks outfit was better in the +2 top exposure than in the +.7 exposure. If I brought the +.7 image up to the same brighten as the original post the quality was less, now I can darken dark areas down in raw and not lose quality but I can not brighten up dark areas as the quality was not there as much in the first place. I print a number of images on a 17" Epson 4900 printer and a lot of them are B&W so detail and tones in the blacks are very important to me so if I find can use something to increase those, I want to know how to use it, as for highlights, with the stocks I use, the printer can not tell the difference between an output of 247 and 255, to the printer it is all pure white no ink. On you monitor you probably cannot tell the difference between and output of 251 to 255 as they both appear pure white.
It is easy to find out where the blinkies come on, take a large white object (my case foam board), put into sun, set camera up so that the camera wants to shoot at 1/250s, (use those setting and set on manual), now reduce the speed by 1/3rd of a stop until your final shot is 1/8th of a second. Now set camera to show blinkies, go through images until you find the first blinkie, (mine was 1/25s). Now put images into a raw converter and set to show warning blinkies, now go through images until you get your first blinkies (mine was 1/13 s). The difference was 1 full stop from what the camera said, I now know that when the warning light comes on in the camera I still overexpose 1 more stop before I have to worry about it in post.

Cheers: Alllan

[george013]

I don't think you see a histogram based on the raw sensordata in your rawconverter. What you see is a histogram based on the rasterimage created by your rawconverter.
Also keep in mind an eventual threshold for clipping.

George

[NorthernFocus]

I think this is simpler than the article cited by the OP suggests. It boils down to this:

1. The more light you let in, the more signal, and the higher the signal/noise ratio. So yes, you get a better signal/noise ratio if you ETTR.
2. Boosting the ISO is simply amplifying the signal from the sensor--that is, amplifying the signal AND the noise. Hence, the noisier images. This is particularly apparent in dark areas, where there is less signal but the same amount of noise as in brighter areas.
3. Does ETTR actually matter much in practice? ...sometimes...
4. If you are going to ETTR, does it matter how? No...All that matters is the distribution of luminance values--the histogram. Whatever gets you the histogram you want is a good way to meter.
....

Best post yet on the topic, Dan. Nicely done.

It really is conceptually simple. Shoot so that the resultant image pixels fill the right side of the histogram as much as possible without clipping. The rest of the discussion is noise(no pun intended). We each have to figure out how to get there for our equipment, subject matter, PP software/techniques, etc. Assuming of course that one chooses to apply ETTR at all.

[ajohnw]

ETTR has a number of holes in practice. For one dslr metering systems do not meter the entire scene so highlight clipping can still occur. How much this matters depends on the shot. Delve deeply and there will be comments about altering jpg setting to try and get the histogram to show something more like a raw histogram.

A lot of the talk about ETTR has Canon cameras in mind as historically they have tended to leave what some feel is way too much raw headroom at the highlight end. Time change however and the jpg tone curves cameras use now a days are rather different. These tone curves are in real terms the same as some one might use from raw during PP. From another thread these seem to be luminosity type curves. I don't know of any other useful type.

The converse is true of Nikon. Due to the mid grey point they set the they have had a tendency to over expose. There are various arguments as to why this happened one of which is that they chose to use a different standard. DPreview on the other hand mentioned that it has shifted over time. Pass, academic really as with any camera it's a case of learning what it tends to do / checking blinkies / bracketing according to need. KR passed a comment not all that long ago on Nikon - at last a Nikon that doesn't over expose.

Dslr metering has another problem as well as not metering all of the shot. Many use the evaluation style setting. This will vary exposure based on objects in the shot and may even take account of colour. An extreme example is if an object takes up a lot of the viewfinder the camera might just more or less ignore the back ground. A histogram might be useful in this case as the user could choose to pull more of the background in, Similarly they might not like the look of it's position so can move it around or decide to clip etc. Metering of this sort is pretty complicated these days and could make all sorts of decisions for the user.

To be honest I don't see the point about many arguments about metering this one in particular. A general rule is never to clip highlights. Often when shots are taken somewhere the sky is the ruling factor. The same exposure is likely to be needed on every shot so the 1st shot can be used to sort it out via the blinkies etc. If conditions change same again but I find that often they don't so this doesn't have to be done often. It's a bit like how decent wedding photographers work. They will use the preview and if happy with that rattle of a number of shots. They might then move people around, different location and do the same again. And yet again at the reception. If some one looks at the preview and allows some clipping on the histogram it's a case of knowing what the camera leaves extra in raw. That's likely to vary according to make, model and what jpg mode the camera is in. Then there are the oddities that modern metering might introduce. Only the preview will show any instances of clipping. The need for accuracy in rough terms more relates to the dynamic range needed to take the shot. If what is there is captured then exposure doesn't matter within limits only that all of the tone range is there. I like to use jpg's when I can so things get a bit more critical then but I might set under exposure on some shots just to make sure I do capture all of the highlights along with reasonable contrast there. Part of that is down to the camera I am using but if I need very dark end detail as well I'd best bracket or guess. i can only guess from previous experiance that includes PP to see just how much can be recovered and for instance how much I need to under expose bright white to get what I want.

In short there is no magical cure for ensuring accurate metering or that ensures the shot will finish up as wanted following PP. Bracketing is one answer and that's about it really but even that might be best set say +/- 1 with 1 stop compensation. Personally when skies are involved I usually meter with it taking up tp 2/3 of the view and then recompose. Of late though m 4/3 tells me where there is clipping at either end. There often is at one or the other when I do take a shot as I seldom shoot under ideal conditions.

John
-

[arith]

Most of the time I go 2 stops overexposed, trouble is metering and it is a bit of a guess with spot metering. But also skin isn't flattered by overexposure, so here I try 1 stop overexposure and rather easier to get an accurate exposure.

The thing is, dslr's are very noisy and most information is captured in middle and highlights, if you underexpose you just don't capture the information but have to raise the image with all it's inadequate information and noise to a visible thing, that nobody wants to see.

[ajohnw]

Most of the time I go 2 stops overexposed, trouble is metering and it is a bit of a guess with spot metering. But also skin isn't flattered by overexposure, so here I try 1 stop overexposure and rather easier to get an accurate exposure.

The thing is, dslr's are very noisy and most information is captured in middle and highlights, if you underexpose you just don't capture the information but have to raise the image with all it's inadequate information and noise to a visible thing, that nobody wants to see.

If people want to use spot metering it's probably best to follow Bill's advice that he gives to students I believe. Go out and take a lot of shots to get some idea what is mid grey and what it means. From my experiences it's a method that I feel can just get too complicated. I think that thoughts in this direction stem from times when metering systems were pretty simple - just averaging the whole scene for instance. They have done more than that for a long time now. Actually apart from centre weighed average I don't think any of the evaluation styles of metering systems have ever been that simple. Accounting for colour seems to have been around for a very long time. The main changes have been ever increasing sensing points and more complications. There is a very old article on spot and others here. The comments I read on film in this sort of thing makes me smile. Nikon actually test their metering against a huge data base that probably wouldn't fit into a camera and would take far too long to do in it. It does show where a spot needs to be placed in several photo's though. No mater how a shot is metered some thought is needed. I just feel modern evaluation style needs less of it.

http://www.photozone.de/Technique

John
-

[Polar01]

I think that people need to realize that shooting one or more stop pass the blinking warning light on their digital cameras will not work when and if they only shoot jpegs. The image on the review screen, the histogram and the warning lights are all based on as if the captured image's final output format was to be a jpeg.
Now I shoot raw files and I believe this approach is all about capturing the most information you can and then using it. I do a lot of B&W now I only have 256 tones to work with and anything that will better the quality of those tones I want to use.

Cheers: Allan

[xpatUSA]

I would like to illustrate some aspects of the discussion:

It's always nice to see a RAW histogram as opposed to what the camera manufacturer thinks you should see

Here's one -



Both axes are logarithmic but it's plainly obvious that the image is blown - all three (Foveon) channels are clipped at around a raw value of ~9000. However, notice that the counts (see Y-axis) do not exceed 100 for any level over ~3850 meaning that number of "blown" pixels is actually quite small. Astute readers will suspect that the scene has some specular reflections or something like a back-lit fence with holes in it.

So, was it "over-exposed" or not? Here's the raw image with no conversion - what you are seeing is linear camera space, i.e. what the sensor captured -



Not exactly over-exposed, eh?

To make it more like what we're used to, here it is with 2.2 gamma applied -



Now we can see what the subject is: kitchen work in progress :-) Even though this is still an un-converted composite image from the raw data, I would say at this point that a little more exposure could have been used for the shot - bearing in mind that all the highlights are specular in nature.

And so we come to the converted and processed result, not a bad picture in the end and one for which the so-called ETTR (even with some blown raw data) worked quite well -



On the other hand, if the exposure had been exactly to the right but not over, it would have have been a dark and gloomy kitchen shot indeed

[george013]

I think that people need to realize that shooting one or more stop pass the blinking warning light on their digital cameras will not work when and if they only shoot jpegs. The image on the review screen, the histogram and the warning lights are all based on as if the captured image's final output format was to be a jpeg.
Now I shoot raw files and I believe this approach is all about capturing the most information you can and then using it. I do a lot of B&W now I only have 256 tones to work with and anything that will better the quality of those tones I want to use.

Cheers: Allan

If you shoot 12 bits RAW, then you have a tonal value of 4096. That's where you work with in the converter. On screen it's transferred to 8 bits.

I'm wondering if there is a difference in clipping between RAW and JPG and if so, why. The sensor is the same. Imagepixels are calculated out of the sensorpixels with the correction of a whitebalance. If the correction value for a color is more than 1, then you could get clipping if the values are near max. But that also counts for the RAW I see in the converter.
Compressing the image, I always thought that the highlights got less.
So why, or is it a story?

George

[jcuknz]

Now I shoot raw files and I believe this approach is all about capturing the most information you can and then using it.

I thought our activity was all about taking photos, not capturing information.
I also think that ETTR is a misnomer and it should be EFTR. [ '4' not '2' ]

[ajohnw]

I would like to illustrate some aspects of the discussion:

It's always nice to see a RAW histogram as opposed to what the camera manufacturer thinks you should see

Here's one -



Both axes are logarithmic but it's plainly obvious that the image is blown - all three (Foveon) channels are clipped at around a raw value of ~9000. However, notice that the counts (see Y-axis) do not exceed 100 for any level over ~3850 meaning that number of "blown" pixels is actually quite small. Astute readers will suspect that the scene has some specular reflections or something like a back-lit fence with holes in it.

So, was it "over-exposed" or not? Here's the raw image with no conversion - what you are seeing is linear camera space, i.e. what the sensor captured -



Not exactly over-exposed, eh?

To make it more like what we're used to, here it is with 2.2 gamma applied -



Now we can see what the subject is: kitchen work in progress :-) Even though this is still an un-converted composite image from the raw data, I would say at this point that a little more exposure could have been used for the shot - bearing in mind that all the highlights are specular in nature.

And so we come to the converted and processed result, not a bad picture in the end and one for which ETTR (even with some blown raw data) would have worked quite well -

Nice post Ted. It illustrates that the height of the histogram where it's clipped counts alongside the subject matter. Same applies at the dark end but many cameras offer rather a lot of scope for recovery at that end - even in jpg's actually on some. Clipped highlights are gone for ever full stop,

If some one can actually see a histogram in the viewfinder the clipping can be measured via exposure compensation. This might show a higher number of pixels are clipped at other tone levels that were previously not in the histogram. This can be useful for raw if some one knows how much spare is there. If some one can't see the histogram in the viewfinder the blinkies are more use in preview than the histograms it shows as Ted for instance could say so what my metal stuff is slightly clipped but that's jpg and there is likely enough left in raw however if they are too big that's unlikely. Little blinkies are very likely to be ok providing the camera isn't set to get more highlight range into it's jpg's.

Out of interest it seems E-M1's hardly leave any highlight head room in raw. Justified i suppose as clipping can be seen directly before the shot is actually taken but even then with iSO variations, noise and tolerance variations on the parts used I don't think there is much scope for a true raw reading even in this case.

John
-

[george013]

One remark. Photo 1 is already converted. One can't see a RAW as an image. If you do, a conversion has taken place.

The issue in this thread is the comparison betrween RAW and JPG. I just toke a picture and wrote it to card as NEF and JPG. The picture has clipping pixels. Loaded them Capture and compared the clipping places. Beside that the JPG had some very little more black clipping, they where exactly the same.

I looked for a photo with much clipping, I have lots of them. Braught it back to the original status, saved it as a JPG and opened it. Compared to tyhe NEF a lot less clipping in the highlites. So I don't know anymore.

George

[dje]

I looked for a photo with much clipping, I have lots of them. Braught it back to the original status, saved it as a JPG and opened it. Compared to tyhe NEF a lot less clipping in the highlites. So I don't know anymore.

George

George if you do a simple test such as the one Allan mentioned in post 26, you will get a good idea of the margin between white clipping point in the camera jpeg and the raw file.

I found a similar result to Allan with my D610 - about 1 stop. I find this a bit surprising and am not sure why the difference would be so great.

Incidentally, there is one raw converter that will give a raw histogram - RawTherapee.

Dave

[NorthernFocus]

I thought our activity was all about taking photos, not capturing information...

Perhaps it's just semantics but I consider the process otherwise. In the field my intent is to capture digital data that can later be processed into the desired image(aka photograph). For most subject matter, I shoot RAW with picture controls (Nikon's term for in camera processing) set to zero. One of my reasons for doing so is so that the in camera histogram represents the RAW data as closely as possible. In the field the intent is to capture as much dynamic range as possible and as far to the right as possible on the histogram. If that is accomplished maximum data is available for post with minimum noise.

In the classic argument as to whether photography is art, I'd describe myself as a technician in the field and, assuming there is any artistic quality to my work, an artist at the computer. This is a far cry, by the way, from how I viewed photography in the past.

[NorthernFocus]

...I found a similar result to Allan with my D610 - about 1 stop. I find this a bit surprising and am not sure why the difference would be so great.

If you shoot 14 bit, uncompressed RAW, that is one reason. The RAW file contains a lot more information than the 8bit jpeg. By definition the jpeg conversion process has to truncate information to accomplish the format conversion.

But I'm not and expert on the topic and have no desire to be one. As the technician/artist it suffices to recognize the phenomenon and account for it when operating the equipment.