Exposure when shooting RAW

[Scottes]

I shoot in RAW, always check my histogram, and "expose to the right" to get the most detail. I then adjust the exposure when developing so that it looks right.

So I'm not sure if I know how to expose correctly. Sure I've read the books and the web pages. I think that I even understand the Zone System. I can look at a scene and realize if the dark/bright limits are too wide, and that something will be blown out or underexposed. I shoot anyway, and fix it in developing.

Is this a bad thing? :? Am I the only one who does this?

Any suggestions for exercises to fix this? I will always shoot "for real" in RAW at least until I reach a confidence level with exposing correctly, but exercises have a purpose.

[CyberDyneSystems]

Is this a bad thing? :? Am I the only one who does this?



Nope :)

Sounds like what I do too,. I set the compensation 2/3rds or up to 2 stops HAMSTRR in certain lighting conditions and shoot raw... and I do what I can with metering, histogram, and compensation... but I will certainly always take the shot regardless of the CMOS dynamic range... :)

[Cordell]

This is pretty much what I do too. However, I think this is also what seasoned pros complain about when talking about newbies of today's technology. Sure we use these (automation and software) as tools but do we know what to expect and visualize while capturing the shot instead of after the fact. That is where I hope to be some day sooner than later. While I know I can fix some of my bad shots in C1 and/or PS, I am training myself to NOT rely on these tools to capture a great shot. My plan is to use those as small enhancements to captures.

[PacAce]

I wouldn't necessary say that I "expose for the right" (I assume you mean the right side of the histogram). AAMOF, if anything, I would tend to sigh away from the right because the closer to the right you go, the more the likelyhood of overexposing the highlight and thus losing details there. With photoshop, I can coax details out of an underexposed shadow area (even when it's so dark that it may look like there's nothing there) but nothing will bring back details in a blown highlight area.

But, as always, it also depends on the lighting situation and the desired results. If you're not interested in the highlights and care more about details in the shadows, then, yes, expose for the shadows. The opposite logic would apply for important details in the highlighted areas. And, on a mid-toned scene, you want that hump right smack in the middle.

BTW, the histogram is not just for shooting in RAW. It works very well for shooting JPEG, too. :)

[scottbergerphoto]

I wouldn't necessary say that I "expose for the right" (I assume you mean the right side of the histogram). AAMOF, if anything, I would tend to sigh away from the right because the closer to the right you go, the more the likelyhood of overexposing the highlight and thus losing details there. With photoshop, I can coax details out of an underexposed shadow area (even when it's so dark that it may look like there's nothing there) but nothing will bring back details in a blown highlight area.

But, as always, it also depends on the lighting situation and the desired results. If you're not interested in the highlights and care more about details in the shadows, then, yes, expose for the shadows. The opposite logic would apply for important details in the highlighted areas. And, on a mid-toned scene, you want that hump right smack in the middle.

BTW, the histogram is not just for shooting in RAW. It works very well for shooting JPEG, too. :)

PacAce,
You might want to read an interesting artice in Luminous-Landscape addressing this issue. "Expose (to the) Right". The CMOS sensor doesn't respond to light as film does, so that 1/2 (2048) of all tonal values are contained in the brightest f stop worth of data. That leaves only 1/2 of all the tonal values (another 2048) for all the others. Here is the link: http://luminous-landscape.com/tutorials/expose-right.shtml
This method only works if you: 1.)shoot Raw and adjust your exposure in the raw converter. 2.)Don't clip the highs by going to far to the right.
Scott

[defordphoto]

Yeah. This is where the techie/newbies kinda lose out on really learning to play with light. That's all we are doing BTW: is playing with light. I elarned the old way on a cheapo 35mm SLR, black and white film and a darkroom filled with chemicals.

But you can still learn the old way. Just without all the nasty chemicals. Set your cameras on TV/AV/M and experiment with different settings. learning to manipulate the light to your advantage helps you spend less time in the darkroom.

I grew up shooting what I wanted in the end product and that has stuck with me. Rarely do I make any major adjustments or crops on my photos. I try to make all that happen out in the field.

But, these tools are awesome and very valuable too. I rely on them, use them daily and will probably never step into another darkroom in my life, Except maybe at the Smithsonian or something. ;)

My point is to get out there and experiment. Learn your equipment, its advantages and weaknesses and use them to your advantage. Half the battle of photography these days is learning what your equipment will do.

Ignore the rules or what the camera says is the right exposure and do something different. That's how you learn. Stick that camera to a tripod and shoot the same scene 30 different ways.

[PacAce]

I wouldn't necessary say that I "expose for the right" (I assume you mean the right side of the histogram). AAMOF, if anything, I would tend to sigh away from the right because the closer to the right you go, the more the likelyhood of overexposing the highlight and thus losing details there. With photoshop, I can coax details out of an underexposed shadow area (even when it's so dark that it may look like there's nothing there) but nothing will bring back details in a blown highlight area.

But, as always, it also depends on the lighting situation and the desired results. If you're not interested in the highlights and care more about details in the shadows, then, yes, expose for the shadows. The opposite logic would apply for important details in the highlighted areas. And, on a mid-toned scene, you want that hump right smack in the middle.

BTW, the histogram is not just for shooting in RAW. It works very well for shooting JPEG, too. :)

PacAce,
You might want to read an interesting artice in Luminous-Landscape addressing this issue. "Expose (to the) Right". The CMOS sensor doesn't respond to light as film does, so that 1/2 (2048) of all tonal values are contained in the brightest f stop worth of data. That leaves only 1/2 of all the tonal values (another 2048) for all the others. Here is the link: http://luminous-landscape.com/tutorials/expose-right.shtml
This method only works if you: 1.)shoot Raw and adjust your exposure in the raw converter. 2.)Don't clip the highs by going to far to the right.
Scott

Hi, Scott, thanks for the link. It was very informative as I had not heard of the "expose to the right" technique before. You learn something new everyday! :)

But I will still stand by what I posted earlier. But that's just me based on my past experience of having blown too many highlights in what could have otherwise been good shots. However, if S/N ratio should ever become a concern for me, I will definitely give the "expose for the right" technique due consideration.

Cheers.

[maderito]

Hi, Scott, thanks for the link. It was very informative as I had not heard of the "expose to the right" technique before. You learn something new everyday! :)

After reading the "expose to the right" to optimize S/N several months ago, it definitely altered my approach to shooting.

First and foremost, as PacAce emphasizes, blown highlights is the first curse to avoid. And many average shooting scenarios really don't give you any room on your histogram after you have the highlights properly exposed.

However, there are many cases where you have room to adjust exposure. The typical scenario is a low contrast and low lit scene that does not fill all 256 luminosity levels. Normally you would try to center the histogram. After adjusting the contrast (curves/levels in PS) and after some additional processing, you may find that the shadows now have significant noise and possibly posterization artifacts. If the original shot had been exposed to the right, you'd have many more levels of luminance to work with, less posterization, and less noise.

Scott points out that this works for a RAW processing workflow in 16 bit image mode. True. But, if the theory is sound - http://luminous-landscape.com/tutorials/expose-right.shtml - it should hold up for 8 bit images also. That is, assuming the CMOS has a dynamic range of five stops to cover 256 levels, it would contain 192 discrete luminance values in the brightest two f stops and only 24 in the two darker f stop ranges. Thus you should shoot to the right to capture the additional levels (=image data) and then center and expand the dynamic range during processing.

Our common experience is that noise and posterization are most dramatic in the shadows. For some shooting conditions, you can limit these problems by adopting the "shoot to the right" approach. I'm on the bandwagon :D

[Scottes]

One thing to worry about when Exposing Right is that the histogram shows only White - it's possible to blow out a single color or two and the histogram won't show this.

Same article, under "Further Thoughts"

[dtrayers]

This may go slightly off topic, but I shoot almost exclusively in RAW format and convert with C1Rebel or Photoshop CS.

I usually don't do any exposure compensation unless in extreme conditions, like snow. But I do pay attention to the scene and try to meter off a face or mid-tone. Many times the camera histogram is to the left side.

I have noticed that when I convert the file, I have a lot of latitude to increase exposure (sometimes as much as 2 stops) without clipping any color, but very little latitude in the shadow control. It's like the camera automatically exposed "left". Am I making any sense?

In any event, I rarely get blown highlights, and with the Shadow/Highlight adjustment in CS I can easily coax detail from the shadows after conversion.

[chris.bailey]

This may go slightly off topic, but I shoot almost exclusively in RAW format and convert with C1Rebel or Photoshop CS.

I usually don't do any exposure compensation unless in extreme conditions, like snow. But I do pay attention to the scene and try to meter off a face or mid-tone. Many times the camera histogram is to the left side.

I have noticed that when I convert the file, I have a lot of latitude to increase exposure (sometimes as much as 2 stops) without clipping any color, but very little latitude in the shadow control. It's like the camera automatically exposed "left". Am I making any sense?

In any event, I rarely get blown highlights, and with the Shadow/Highlight adjustment in CS I can easily coax detail from the shadows after conversion.

I would agree with that, I would also suggest that the 10D histogram shows something a little different to that in C1 or CS Raw converter and then again the histo in CS after conversion. I have always tended to try and avoid blowing highlights by "dressing to the left", particularly with portraits where it is very very easy to blow out the skin tones. Will have to try and experiment a bit.

One final thing, if shooting 8 bit, if the DSLR has six stops of range, and half of that (128 bit) is in the upper sixth that means only four tonal ranges in the bottom sixth. That does not seem to ring true to me.

[chris.bailey]

This may go slightly off topic, but I shoot almost exclusively in RAW format and convert with C1Rebel or Photoshop CS.

I usually don't do any exposure compensation unless in extreme conditions, like snow. But I do pay attention to the scene and try to meter off a face or mid-tone. Many times the camera histogram is to the left side.

I have noticed that when I convert the file, I have a lot of latitude to increase exposure (sometimes as much as 2 stops) without clipping any color, but very little latitude in the shadow control. It's like the camera automatically exposed "left". Am I making any sense?

In any event, I rarely get blown highlights, and with the Shadow/Highlight adjustment in CS I can easily coax detail from the shadows after conversion.

I would agree with that, I would also suggest that the 10D histogram shows something a little different to that in C1 or CS Raw converter and then again the histo in CS after conversion. I have always tended to try and avoid blowing highlights by "dressing to the left", particularly with portraits where it is very very easy to blow out the skin tones. Will have to try and experiment a bit.

One final thing, if shooting 8 bit, if the DSLR has six stops of range, and half of that (128 bit) is in the upper sixth that means only four tonal ranges in the bottom sixth. That does not seem to ring true to me.

[maderito]

One final thing, if shooting 8 bit, if the DSLR has six stops of range, and half of that (128 bit) is in the upper sixth that means only four tonal ranges in the bottom sixth. That does not seem to ring true to me.

See http://luminous-landscape.com/tutorials/expose-right.shtml . The first table gives the explanation based on 5 stops of dynamic range in a 12 bit (4096 levels) DSLR image. The table is preceded by the statement:

"CCD and CMOS chips are linear devices. And, of course, each F/Stop records half of the light of the previous one, and therefore half the remaining data space available. This little table tells the tale." (table follows)

I agree - it doesn't "ring true" -- but upon reflection, it certainly seems to be true.

[chris.bailey]

One final thing, if shooting 8 bit, if the DSLR has six stops of range, and half of that (128 bit) is in the upper sixth that means only four tonal ranges in the bottom sixth. That does not seem to ring true to me.

See http://luminous-landscape.com/tutorials/expose-right.shtml . The first table gives the explanation based on 5 stops of dynamic range in a 12 bit (4096 levels) DSLR image. The table is preceded by the statement:

"CCD and CMOS chips are linear devices. And, of course, each F/Stop records half of the light of the previous one, and therefore half the remaining data space available. This little table tells the tale." (table follows)

I agree - it doesn't "ring true" -- but upon reflection, it certainly seems to be true.

But he also says most DSLR's have an effective range of closer to 6 stops. Now my understanding is that the 10D CMOS is a 12 bit device with only 8 bits "captured" unless RAW processed as a 16 bit image, in which case you get 12 bits of data in a 16 bit data "packet". If thats the case and the extra 4 bits are chucked away evenly (and why not) that would leave only 4 true tones in the darkest 6th of an image and only 8 in the next sixth ie only 12 in the darkest third. That cant be true.

Also if you process a RAW image linearly it all comes out very dark and not very light i.e. all the image data is compressed at the dark end, where, LL says none of the tonal range is. I dont get it!

[maderito]

I dont get it!

I think of it this way: Take an image at say f2. Then stop down by 5 - say f11. Compare the images. How many discrete levels of data are left in the image in the f11 shot? You've reduced your light by 2^5 = 1/32. But you're giving that light to 1/6th of your CMOS light capture range.

[chris.bailey]

I dont get it!

I think of it this way: Take an image at say f2. Then stop down by 5 - say f11. Compare the images. How many discrete levels of data are left in the image in the f11 shot? You've reduced your light by 2^5 = 1/32. But you're giving that light to 1/6th of your CMOS light capture range.

Still had to think about it but would agree, its the difference between a log scales (F stops) and a linear capture device. It does though mean that there would only be 12 shades of grey possible in the darkest third of a black and white image which covers a full range which I still find hard to believe.

[Scottes]

Anyone have a MacBeth color chart they can shoot in 12- & 8-bit modes?

[maderito]

Still had to think about it but would agree, its the difference between a log scales (F stops) and a linear capture device. It does though mean that there would only be 12 shades of grey possible in the darkest third of a black and white image which covers a full range which I still find hard to believe.

Consider a low contrast scene - say 3-4 f stops of dynamic range. Expose the scene to the left (just short of underexposure) and then to the right (just short of overexposure). Load both images into PS and adjust contrast and tonality. Which one looks better. I've done this quasi-experiment. Seeing is believing!

More generally, working with underexposed or overexposed images presents major challenges and compromises - including loss of image data. The expose to the right idea can help avoid problems in the low lit portions of an image with less than optimal dynamic range. It isn't applicable to every, or even most shooting scenarios. But it does come up from time to time. And I agree that one has to be wary of the on-camera histogram which shows a luminosity histogram and not the individual RGB histograms.

[chris.bailey]

Still had to think about it but would agree, its the difference between a log scales (F stops) and a linear capture device. It does though mean that there would only be 12 shades of grey possible in the darkest third of a black and white image which covers a full range which I still find hard to believe.

Consider a low contrast scene - say 3-4 f stops of dynamic range. Expose the scene to the left (just short of underexposure) and then to the right (just short of overexposure). Load both images into PS and adjust contrast and tonality. Which one looks better. I've done this quasi-experiment. Seeing is believing!

More generally, working with underexposed or overexposed images presents major challenges and compromises - including loss of image data. The expose to the right idea can help avoid problems in the low lit portions of an image with less than optimal dynamic range. It isn't applicable to every, or even most shooting scenarios. But it does come up from time to time. And I agree that one has to be wary of the on-camera histogram which shows a luminosity histogram and not the individual RGB histograms.

I have certainly put it on my to do list. As I do quite a lot of portrait work I tend to err towards the left (underexposure) as a blown highlight on a cheek or forehead is very easy to do and hard to hide. Interesting thread though.