What your talking about is commonly refered to as "pushing" the exposure in post. I can't speak for the 50D but I can say that I've "pushed" quite a few exposures on images recorded with my 5D.


I would say it is safe to say to use higher ISO and slightly over expose rather than under expose and push in post.


I find my 40D does extremely well at 1600 ISO if I overexpose .3-.5 ev and pull down the overexposure when developing the raw file. By comping up +.3ev I am shooting at a 1250 ISO equivalent. I am loosing 1/3 stop shutter for a given aperture. Or I could just shoot 1250 ISO at 0ev comp. Is there a difference? Not really. I'll splain..........


The 30D & 40D (I have a feeling this goes for the 50D as well) emulate interum ISO values in 1/3 stop increments. Not sure about 1/2, never used it. ( maybe someone could add to this if experienced using 1/2 stop ISO values) When shooting at 160, 320, 640 & 1250 ISO the camera shoots at 200, 400, 800, &1600 but mathematically offsetsthe raw data before writting to the CF card. This is the same as the 1/3 stop of overexposure and results in a much cleaner image than the normal ISO settings. For 125, 250, 500, & 1000 ISO the inverse is true. 100 is pushed mathematically up to 125, 200 to 250 and so forth which introduces more noise into the image. When shooting with the 40D, I use 160, 320, 640, & 1250. If I need to go higher I go to 3200 and over expose by .3-.7 which will give me an eqivalent ISO of 2400 at +.3ev or 2000 at +.7ev (this 2000 is cleaner than 1600 ISO underexposed and pushed in post) Oh, I forgot to mention that there is usually +1ev worth of headroom in a Canon Raw file at 3200 ISO. That is information un reported by the histogram. Simply compensate down during post processing and viola!! normal exposure, withless noise and optimized shutter speed.


The 5D is straight up! I hope the 5D MII is as perfect. 100, 200, 400, 800, 1600, & 3200 are as clean as it gets. Emulated interum values don't offer any benefit from a noise cancelling perspective.


I discovered all this from a post on dpreview a few years ago. I have experimented and shown this to be very true. Here is an excellent study.


http://www.pages.drexel.edu/~par24/rawhistogram/5DTest/5DTest.html


http://www.pages.drexel.edu/~par24/rawhistogram/30DTest.htmlThis is applicable to the 40D. I'm wondering if anyone can confirm this for the 50D?


Hope this helps,


Chuck