Upload & Sell: Off
Great lenses cannot overcome a crappy sensor and at several stops below mid-tone the 5D3 has a crappy sensor. Look for comparisons with the D800 to see where I'm coming from with that statement.
The 5D mark III is not particularly good at letting you retrieve clean image data out of deep shadows / dark areas. In fact it is pretty poor. Mind you, even the D800 starts looking poor at high ISO. Because the camera is so bad in that regard you need to consider increasing exposures and give the NR software a fighting chance. It will make a huge improvement if you can do it without losing important highlight data.
So if we can agree that with whichever camera we have it might help to increase the exposure, then we just need to consider how to achieve it:
A large aperture (smaller f/number) would be great if you can do without as much DOF. It might also allow a lower/cleaner ISO and more dynamic range. You might be able to overcome the DOF limitation with a TS-E lens but it could be too hard to get the stars sharp as well as the ground in a wide angle image. Easier with a tele lens such as the TS-E 90 or even the 45 than with a 17 or 24.
A longer shutter exposure would cause more obvious star trails. On a windy night the clouds and trees would show signs of motion blur too. Therefore probably not much good.
With a few cameras I have found that simply increasing the ISO helps while retaining the same aperture and shutter speed. Even after reducing the image exposure to equivalent levels in post processing the shadows are far cleaner than capturing less exposure at lower ISO. In some ways this is counter-intuitive but it works if the scene does not have too much dynamic range that all needs to be captured.
The really tricky thing is knowing how far you can increase the exposure without doing damage to important highlight details. The in-camera histogram is borderline useless at default camera settings because it is so greatly affected by jpg camera settings even when taking raw files. You have to know how to work around the limitations but doing so will make the jpg previews look rather poor while improving the underlying raw data:
1. Use neutral profile; No hidden contrast or saturation enhancements are wanted.
2. Use minimum contrast to get data away from the edges of the histogram.
3. Use low saturation, for the same reason.
4. Don't use too much sharpness. Some helps you better assess focus but too much brightens small bright highlights enough to give a false impression of overexposure.
5. Use an appropriate WB. This can have a big impact on the histogram especially at the edges (bright and dark limits).
6. Use the R, G and B histograms to show individual channels. Forget about overall luminance and concentrate on the individual channels. It's the channels that burn out.
7. Apply the over-exposure blinkies. Preferably to a single colour channel that is most likely to blow out. In this case I'd expect the blue channel is at most risk. In another image it might be the red channel. Green is usually safer than red or blue.
The result so far will look rather bland but only in the LCD preview. Don't panic; the raw data is still there. If any of the preview image still looks overexposed then the capture probably really is at or near true overexposure - as in the bright details are about to be lost even in the raw data. If they are not showing signs of overexposure with this setup then it is safe to increase the exposure so go ahead and do it and shoot again.
8. If you are sufficiently familiar with this approach then it may suffice. Otherwise it is time to start throwing in some exposure bracketing.
9. If you are really keen then you can also look up uni-WB and find or create a settings file that works best for your camera. Image previews will look terribly green but the histogram will be as close as you'll ever get to representing what the sensor captured in raw data. To be honest, I tend to stop at step 8 rather than go all the way to using uni-WB. Nobody likes the look of uni-WB previews.
10. Know that most of these steps need to be undone in post processing.
Get hold of a utility called rawdigger to see what the raw data histograms really look like. Unfortunately it needs to run on a computer - great if you can do a tethered capture or upload in the field. Unfortunately it does not know exactly what the head room is for each camera but if you see peaks at the bright end of the raw histogram and there are no specular highlights in the scene then you know that something is being lost due to overexposure. It also helps by allowing you to develop a better feel for what the steps above do to your raw data.
Edited on Feb 01, 2013 at 08:22 PM · View previous versions