Exposure Compensation

[DrWho]

What exactly happens in your camera when you adjust the Exposure Compensation feature? If you are in Aperture Priority, when you change the exposure compensation, what does the camera change? Is it the ISO, the Shutter speed, or something else. I would not think it would change the shutter speed. What if you were shooting to capture motion? Same idea with the ISO, the 'grain' would change. This would also be the same for Shutter Priority. What about Manual Mode, does it change the ISO? What happens when you change the "Exposure" in a digital editing program? What happens in the computer?

Thanks.

[Graystar]

I can only speak of Nikon DSLRs, but all that happens is that the meter shifts its reading.  The rest of the camera behaves exactly as it would normally do, had that new reading come directly from measuring light.

With Nikons this is most easily seen in Manual mode.  As you apply positive EC, which would normally lead an auto mode to increase exposure, you see the meter move to the negative side.  Being in Manual mode, this would prompt the photographer to open the aperture or slow the shutter, thereby moving the meter back to center and increasing exposure.

[DrWho]

That doesn't seem right to me. If I set the Exposure Compensation down one stop in a daylight scene, something must change in the camera.

[Graystar]

The meter changed.  When you set EC to -1 you are telling the camera to underexpose the scene.  The meter will then meter the scene and produce an Exposure Value.  To that EV, the meter will subtract -1.  This new EV is what's reported to the rest of the camera as the EV for the scene.  The rest of the camera will simply do what it would normally do with such an EV.

For example, lets take P mode on my Nikon D90 with an 85mm lens.  According to the Exposure Program outlined on page 263 of the D90 manual, an EV of 14 will give me f/5.6 at 1/500s, and an EV of 16 will give me f/8 at 1/1000s.  So...I meter my scene, and P mode is giving me f/8 at 1/1000s (EV of 16.)  However, this is a snow scene so I know I need to increase exposure by two stops.  So I set EC to +2.  The meter will subtract +2 from 16, and will report an EV of 14 as the meter reading for the scene.  P mode will now simply consult its exposure program and select the EV 14 settings, which are f/5.6 at 1/500s.  The scene will now be overexposed by two stops.

That's all EC does...it simply changes the Exposure Value that the meter will return for the scene.  It must work this way, otherwise the operation of the camera will not be predictable when using EC.

[DrWho]

Then it is changing the F Stop and the shutter speed which will completely change the depth of field and the way it captures motion. I don't really think that is acceptable. Maybe we should just wait to do exposure comp in the editing software. Then we won't mess with motion or depth.

[mahleu]

Then it is changing the F Stop and the shutter speed which will completely change the depth of field and the way it captures motion. I don't really think that is acceptable. Maybe we should just wait to do exposure comp in the editing software. Then we won't mess with motion or depth.

Photography is full of compromises, if you want to retain a certain F stop and shutter speed then you either need to change ISO, add (or block) some lights.

[Ben Rubinstein]

Built in ND filters would be nice...

[stamper]

Then it is changing the F Stop and the shutter speed which will completely change the depth of field and the way it captures motion. I don't really think that is acceptable. Maybe we should just wait to do exposure comp in the editing software. Then we won't mess with motion or depth.

If you choose aperture priority and change the EV then the aperture isn't changed. Likewise in shutter priority the shutter speed isn't changed. The poster should have explained it better by not using P mode as an example.

[RFPhotography]

In Av, the shutter speed changes.  In Tv, the aperture changes.  The fixed component remains fixed, the variable component changes.  In P, either or both may change.  In M, you change one or the other or both manually (my Nikon will do EC in M but it's a silly process and doing it manually is faster).

[Ken Bennett]

Yeah, I don't understand how EC is supposed to work in Manual exposure mode. It doesn't on my Canon bodies.

If I am in Aperture Priority, with my aperture at f/8, and the camera meter chooses 1/125 sec for the shutter speed, then I dial in 1 stop of + EC, the camera will choose 1/60 sec shutter speed (one full stop more exposure.) If I dial in one stop of - EC, it will choose 1/250 sec.

Exposure compensation is a quick and useful way to improve exposures. On the Canon, the thumb wheel on the back of the camera can directly control EC, which makes it simple to dial in a precise amount in 1/3-stop increments while looking through the viewfinder. If I feel that the camera meter is being fooled by the lighting conditions, it is trivial to dial in the appropriate compensation.

[Canon Bob]

........ Maybe we should just wait to do exposure comp in the editing software. Then we won't mess with motion or depth.

Motion and DoF may well survive but at the expense of shadow or highlight detail dependant on whether the metering was "tricked" high or low.

The metering system assumes that the portion of the image being assessed has a luminance of 18%.  This will be incorrect for scenes that are uniformly very bright or dark so the compensation is used to correct the exposure based on the shooters superior intelligence...ie, the user knows for a fact what the scene looks like as opposed to the metering system's assumption.

Maybe have a read through this  http://cpn.canon-europe.com/content/education/infobank/exposure_settings/exposure_compensation.do

Bob

[Ken Bennett]

This will be incorrect for scenes that are uniformly very bright or dark

Yes, and for scenes with a small but very bright element that may fool the meter (a bright window behind the subject, or a streetlight, or the sun in the frame, that sort of thing.)

[Canon Bob]

Yes, and for scenes with a small but very bright element that may fool the meter (a bright window behind the subject, or a streetlight, or the sun in the frame, that sort of thing.)

Correct but only if the bright element is in the area chosen for the metering. The camera assumes that the metered area has 18% luminosity and the sun or other bright light source that's not in the chosen metering region should not throw the metering off.

Bob

[Lightsmith]

Easy experiment that is very informative is to take you camera and set it for matrix metering and with aperture priority mode and take a shot at the EV the camera calculates as correct. Take another shot at -1.0 EV and a third shot at +1.0 EV, all in RAW. Open the images one by one in ACR and adjust the EV in Photoshop to bring the histogram all the way to the right without clipping.

In most situations there will be a serious loss of color fidelity with the -1.0 EV pictures and often the best picture will be the one at +1.0 EV.

As mentioned the camera wants to create a 18% gray tone and if the subject is white it will underexpose to create a picture of a gray subject in the image, and if the subject is black it will overexpose to again create a gray subject in the image. The color metering of many newer cameras is supposed to compensate better but it is not able to correct a 2 stop error as with a white subject (bride's dress or egret) or a very dark subject.

After you examine your "overexposed" RAW files you will get a better sense of why it has been long recommended to "expose to the right", and why in general this is good advice.

[Graystar]

Easy experiment that is very informative is to take you camera and set it for matrix metering and with aperture priority mode and take a shot at the EV the camera calculates as correct. Take another shot at -1.0 EV and a third shot at +1.0 EV, all in RAW. Open the images one by one in ACR and adjust the EV in Photoshop to bring the histogram all the way to the right without clipping.

In most situations there will be a serious loss of color fidelity with the -1.0 EV pictures and often the best picture will be the one at +1.0 EV. After you examine your "overexposed" RAW files you will get a better sense of why it has been long recommended to "expose to the right", and why in general this is good advice.

I performed your experiment with my Nikon D90.  Comparisons of RGB values indicate that there was no loss of color fidelity.  That is what I was expecting, as modern DSLRs can capture images at ISO 800 (-3 EV) that are clean.  In fact, I tried it with a difference of 3 EV and the results were the same.

Exposing to the right seems to embody two concepts, depending on the type of photography one pursues.  Photographers of scenes with wide dynamic ranges will underexpose images to protect highlights, and photographers of scenes with narrow dynamic ranges will overexpose to enhance signal.  But the former risks introducing noise into the subject, whereas the latter risks blowing a color channel and shifting colors.  Personally, I don’t want to risk increased noise or a blown color channel, so I always expose my subject properly.

Finally, I’m not quite sure how any of this relates to the original question.

[stamper]

Quote

In most situations there will be a serious loss of color fidelity with the -1.0 EV pictures and often the best picture will be the one at +1.0 EV. After you examine your "overexposed" RAW files you will get a better sense of why it has been long recommended to "expose to the right", and why in general this is good advice.

Unquote

At -1EV there shouldn't be any loss - imo - of fidelity, indeed there will be some increase in saturation. Some recommend underexposing to get more fidelity. If you underexpose too much - perhaps 2 stops or more - your shadow areas become blocked up and there would be some loss. As to overexposing by 1 EV your mid tones will look as if saturation has been lost but it is easily adjusted in post processing. What I - and possibly others - would like to know is how do you expose the subject properly? There is more subjectivity to exposure than there is accuracy imo. There isn't to my knowledge a proper exposure only one that is pleasing to the photographer who took the image provided that it isn't grossly underexposed or overexposed?

[hjulenissen]

There is more subjectivity to exposure than there is accuracy imo. There isn't to my knowledge a proper exposure only one that is pleasing to the photographer who took the image provided that it isn't grossly underexposed or overexposed?

Perhaps it is convenient to talk about "capture exposure" and "presentation exposure" separately?

There is no mathematical way to derive how an image should be best exposed for presentation. Some images may look best as low-key, high-key, depending on what parts of the scene intensity the photographer wants to present on a limited paper/display, and what mood she wants.

For capture exposure, it may be easier. If we can assume that there is enough light that the photographer can choose exposure time and aperture with some freedom, we would probably want the kind of exposure that maximize the precision with which the original scene is captured (minimal noise and clipping if possible). Coding in any headroom above the highlights does not seem to make any sense signal-wise for regular sensors (special Fuji "high dynamic range" sensors excluded), as one can accomplish the same in photoshop with less noise. If the scene has high dynamic range we might want to allow some clipping (and then we are back to not being able to define the ideal value).


It seems to me that it is possible to state that we ideally never want any headroom between scene highlights and sensor saturation point as long as that does not mock up motion-blur/DOF. But we cannot state that we never want sensor clipping, so it is a one-sided limit?

I think there are many similarities between analog/digital camera exposure and analog/digital audio recording. In both cases, the abilities and limitations of the two technologies cause some confusion, and the manufacturers are not always forthcoming in explaining stuff.

-h