Exposure compensation/ISO/noise and ....

[Mark L]

Generally (to start the thread), back in my film days, I used to underexpose a little (to the left of the cameras meter).
Have come across writings that suggest, as digital is a different format, exposing to the right (overexpose) gives less noise as the ISO goes up. Also offers greater dynamic range and more.
This is a short read that demonstrates what I'm starting to think ......http://www.amatteroflight.com/wordpress/?p=417

[Lance B]

There is a school of thought that suggests that you should expose to the right as much as possible without blowing the highlights. However, it still depends on the sensor and what best suits it and there are no real braod brush statements that cover all situations, you need to assess each as it arises and what you are trying to achieve My D700 was best shot slightly underexposed by a 1/3rd much of the time and my D800E is also best shot slightly underexposed as it has better shadow recovery than highlight recovery as it's high ISO ability (read good shadow noise) is superb. You really do need to assess the situation and know your camera.

[mpb]

I went for the depends option.

I am mainly using my sony A77 which has an electronic view finder, so what I actually do is expose to what I see in the view finder.
I also check the meter more as a check. If I have pushedup the iso I will try to over expose to control noise in the darker areas.

[arthurking83]

LOL!

I had to use the cheats option.

Where's the 'All of the above' option? :D

[ricktas]

I generally underexpose a tad, usually about 1/3rd of a stop. I find the metering in my Nikons is very good, but a slight underexposure tends to be more accurate to the scene at the time of shooting.

[mudman]

i tend to under expose, being a modest person:D, to avoid blown highlights. then adjust in RAW or photoshop shadows and highlights.

[Rattus79]

I move between spot metering and Center weighted.

How you run your exposure compensation will depend upon weather you're metering something bright or something dark.
If you're metering a black dog, Exposure compensation will likely be negative 1-1.3 ev
If you're metering Clouds or sand, Exposure compensation will likely be + 1-1.3 ev

[Steve Axford]

It depends. I have read the thing about there being more information at the right end of the histogram, but I've never found that to be very useful in practice. If the subject is bright, I like to expose so that no part of the subject can be overexposed (ie the whole photo may be a bit underexposed). If the subject is darker, then maybe I'll even overexpose some non-critical areas. As far as the metering goes, it will depend on the metering mode and what you really want to achieve. Even the histogram isn't always right. I find that with a pure red bright area (eg a red fungi) the camera will report a correct exposure (on the meter and the colour histogram), yet the photo can be overexposed in the red channel when processed in a RAW processor or photoshop.

- - - Updated - - -

As an update to my previous comment. The overexposure in red seems to occur when an image is converted from AdobeRGB to sRGB. This would imply that the original AdobeRGB image was not overexposed, but the conversion tries to include too much and therefore pushes the red channel too far. I'm not sure of the reason, I just know it can happen.

[Rattus79]

Adobe RGB has a larger colour space then sRGB.
It has more colour's to work with, so when you convert back, you will get some colour clipping, it should only be minimal though.

[Steve Axford]

It is minimal, except with pure red. I would have expected it to be worse with green since AdobeRGB tends to favour green, but there isn't a major problem with green. I wonder if the camera metering recognises what default colour space you have? That may fix the problem when converting to sRGB.

[arthurking83]

..... Even the histogram isn't always right. I find that with a pure red bright area (eg a red fungi) the camera will report a correct exposure (on the meter and the colour histogram), yet the photo can be overexposed in the red channel when processed in a RAW processor or photoshop.

- - - Updated - - -

As an update to my previous comment. The overexposure in red seems to occur when an image is converted from AdobeRGB to sRGB. This would imply that the original AdobeRGB image was not overexposed, but the conversion tries to include too much and therefore pushes the red channel too far. I'm not sure of the reason, I just know it can happen.


One major reason for histograms being out of sync, is that the histogram on the camera is a predetermined contrast curve that Canon have created.
If your raw converter is not from the manufacturer, and is from a thirdparty vendor, then there is a greater likelyhood that they will diverge. Thirdpary software coders don't always have full access to the manufacturers knowledge. This is how 99% of software works in Nikonland.
Your RAW converter will render the raw image as it pleases. I'd reckon(a guess, not from experience) that if you open that raw file in Canon's software, it's produce the same histogram as you saw in the camera.

My thoughts on this is that the camera is right as it's the camera historgram that I look to when shooting.

This is one of my biggest issues with thirdparty raw software, even when I use their supposedly manufacturer like contrast/colour/picture profile .. they look similar but always different.
If I open my raw files with Nikon's software they are always displayed(histogram) the same on PC as per the camera. The images themselves always have some differences in actual appearance, but the reasons for this are pretty obvious.

I reckon, if you extracted the embedded jpg from the raw file(not converted, extracted!) and open it with your raw converter software, the histogram will look identical to how it does on the camera .. not the raw converter's interpretation of the raw file.

The trick to get the histograms in sync, is to create a colour/tone curve or camera profile that prodces the same histogram, and save that as a default.

[Rattus79]

I think it may be the fact that there are more Green and blue pixels on a sensor then red ones. (at least yesterdays sensors were setup that way, I'm not sure about todays)

[Steve Axford]

One major reason for histograms being out of sync, is that the histogram on the camera is a predetermined contrast curve that Canon have created.
If your raw converter is not from the manufacturer, and is from a thirdparty vendor, then there is a greater likelyhood that they will diverge. Thirdpary software coders don't always have full access to the manufacturers knowledge. This is how 99% of software works in Nikonland.
Your RAW converter will render the raw image as it pleases. I'd reckon(a guess, not from experience) that if you open that raw file in Canon's software, it's produce the same histogram as you saw in the camera.

My thoughts on this is that the camera is right as it's the camera historgram that I look to when shooting.

This is one of my biggest issues with thirdparty raw software, even when I use their supposedly manufacturer like contrast/colour/picture profile .. they look similar but always different.
If I open my raw files with Nikon's software they are always displayed(histogram) the same on PC as per the camera. The images themselves always have some differences in actual appearance, but the reasons for this are pretty obvious.

I reckon, if you extracted the embedded jpg from the raw file(not converted, extracted!) and open it with your raw converter software, the histogram will look identical to how it does on the camera .. not the raw converter's interpretation of the raw file.

The trick to get the histograms in sync, is to create a colour/tone curve or camera profile that prodces the same histogram, and save that as a default.

Your guess is wrong, Arthur, but you did prompt me to check. Perhaps it is the same with green and to some extent with blue as well, but I've never had a pure green or blue subject, so I haven't noticed. I'll check them next. Anyway - if I get a pure red subject and take a photo that is perfectly exposed on the camera, it will be overexposed when converted to sRGB. The camera is right and so is the conversion, but the camera is taking in AdobeRGB (1998) and this is a wider gamut. There could be other ways of converting, but I doubt that this is Canon hiding something as they don't have a name for doing this and it is pointless anyway as reverse engineering is a very well tested trade. It could be that CaptureOne could have a better conversion so that they detected an overexposure in a channel and dropped back the exposure, but they do display it so you can do it yourself. It has started me thinking about what a this implies when using sRGB. Do you get the same result by exposing correctly in the camera and dropping back in post, or would it be better to underexpose in the camera and leaving it asis in post??? My feeling is the the second gives better colour Note - this only applies to pure, single colour subjects.

- - - Updated - - -


I think it may be the fact that there are more Green and blue pixels on a sensor then red ones. (at least yesterdays sensors were setup that way, I'm not sure about todays)
I don't think that is an issue. The human eye has about 10 times the number of green compared to blue rods and about 3 times when compared to red. From memory, cameras just use twice as many green pixels as red and blue, but I haven't checked.

[Rattus79]

I don't think that is an issue. The human eye has about 10 times the number of green compared to blue rods and about 3 times when compared to red. From memory, cameras just use twice as many green pixels as red and blue, but I haven't checked.

I didn't know that about the human eye!

I thought it was both green and blue that had twice as many as red - Green and blue being the most popular colours in nature: Grass & Sky

[Steve Axford]

The most useful colour for humans (and primates) is that which allows them to identify food. Most mammals have only two colour sensors, blue and a yellow/green. This is because they lost the other sensors because they were essentially nocturnal (the dinosaurs ruled the day). When primates evolved they regained a red because it helped to identify ripe fruit, etc. Our eyes are very insensitive to blue,

[Steve Axford]

Update to previous comment on overexposed reds. It seems that the histogram in camera is based on a jpeg which is converted from the RAW file using the defaults set in the camera. If I use AdobeRGB as the default, then the histogram will be wrong for any imags converted to sRGB. So, it can be important to set the colour space to that which you commonly use. This is only rarely important as it will not effect photos with a mix of colours.

[Jad]

I have always been a Black and White photographer. With film I used a densitometer and created a density step chart for use with the zone system. My film exposure and development times were calibrated and I was able to maintain of nice consistency with my negatives. The switch to digital has me relying more on the camera than I had ever done in the past. I check the histogram and set my LCD screen to Black and White to act like a Polaroid back for a preview. I know that with camera RAW and PS I can do a lot with the image , but do I fee l comfortable with my exposures in the field as I did with film? No. With film it was the concern with grain and with digital it is a concern with noise.

[outstar79]

I bracket shots, being predominately Landscape photographer so most of the above! :D

[swifty]

When you have a controlled environment and you can repeat the shots over again with little time contraint, then pushing the exposure to the right without blowing highlights can yield better results sometimes.
But the key there is 'without blowing the highlights'.
Since with many subject matters, you only get one chance to capture a particular moment, I tend to play it slightly conservatively and underexpose a tad (if I'm working at lower ISO's).
At higher ISO's its a different story and you really have to be on the ball not to over nor underexpose - it becomes a balancing game and sometime you need to sacrifice some highlights or shadows in the unimportant areas.

Furthermore, certain subjects such as skin tones, even in a controlled environment such as a studio I won't expose to the right (ETTR). For some reason that I'm yet to understand well, I can't get as pleasing processed results by ETTR (without blowing highlights) as I do by exposing the skin tone how I want the final image to appear.