A question of compensation - August
View Single Post
07-08-10, 10:50 AM
Join Date: Jul 2009
Originally Posted by
Since this runs contrary to all the advice I have previously received and also since it doesn't seem to make sense, perhaps you could help me understand.
It is contrary to previous advice, in that data lost in highlights is so terminal that the recommendation has been to shoot "Left of centre" on the histogram, since there is more chance of recovering data in shadows. (My experience reinforces that -a picture with a bright window and harsh shadows can only be recovered in the shadows).
It's not advocating blowing out your highlights. It's advocating shooting to the right whilst not blowing them out. So you still retain data in all regions of the image.
The bit that doesn't somehow seem to make sense in the essay is the assumption that the 4096 discrete tonal values are shared between the F-stops. That suggests that if I shoot at f2.8 I will get 2048 levels, whilst at F32 I will have 128levels. Or have I misread it? If I haven't misread it then frankly I find it hard to believe. The F stop is a mechanical thing, changing the amount of light that reaches the sensor. So the amount of light hitting it will increase / decrease, but why would that change the dynamic range (The frequency / colour) of the received light?
It has nothing to do with the aperture you set your lens to. It relates to the difference in light intensity in darkest and brightest regions of an image (the shadows and highlights). And their estimate is that there is a maximum of 5 - 6 stops of difference between these regions that can be recorded by a 12 bit sensor. Irrespective of what aperture, iso or shutter speed you use to expose the image.
The 4096 tonal values relate to the bit depth of the image - 12 bits is 2 to the power of 12, which equals 4096. What they say is that 50% of those tonal values lie in the brightest (un-blown) stop available from that 5 - 6 stop exposure range that it was possible to record. Whilst the darkest (un-clipped) stop available in the range will hold in the region of only 128 tonal values.
You can put this to the test by shooting a shot where the histogram is bunched up to the left, try without clipping it, and then increasing the exposure in Lightroom, ACR or DPP etc. You'll very quickly see a lot of noise and banding appear as a result of lack of data. Whereas if you shoot to the right, without clipping, and then decrease the exposure you will find that you have far more tonal data.
View Forum Profile
Send a private message to ap4a
Find all posts by ap4a