Welcome to round two of the Canon 50D High ISO pixel-peep show...
In round two - we'll be looking at what Canon's Digital Photo Professional can do with the files from the 50D when compared to files from the 40D... the previous post looked at "standard" picture style JPGs w/ no noise reduction done in camera. For this test, I used the RAW files (of the RAW+JPEG) that were shot for the test. These were also set w/ "standard" picture style, so they have sharpening set to 3. I'll get into this later, but for those of you looking for the setup, read my previous post Canon 50D - High ISO - JPGs.
With the increase to 15.1 MP, (a jump of 50% in sensor resolution) there enters a larger degree of difficulty. Some sites have up-converted 40D images to the 50D size then cropped... This doesn't make sense to me as you will now distort the 40D image. Others have down-converted the 50D image down to the 40D resolution and then they crop, but again - you may lose detail and hide (or magnify) issues w/ the 50D issue...
I'm going to leave the images posted at their orig resolution, so the 50D images (consistently on the bottom of the paired images) will appear slightly magnified. Just seemed this was the best route. This way you get 100% crops w/ zero scaling applied to either camera's image.
For this test, the RAW files were opened in DPP, exported as 16 bit TIFFs, imported into PhotoShop CS3, pixel aligned manually (to avoid any "stacking" distortion from CS3), 100% cropped to match the JPG images, and then exported as JPGs as quality 100 files for posting to the site. Some might argue that I should drop the NR to zero on the RAW files, but I disagree - I want to show a canon baseline for NR. I know that lots of folks have their favorite methods of reducing noise - a combo of camera and software processing; a favorite plugin for your post processing software; whatever. While that may work for you - it may not be accessible to all. Since DPP comes with all Canon cameras, I figured this would be a good method.
Here is a baseline at 100 ISO:
With that as a benchmark, here are the comparison images between matching ISOs. So 1600 and 3200 are no big deal - each is a valid setting on both cameras (tho you are really pushing the ISO on the 40D to get 3200 - thus the (H) in the LCD when you use it), but with 6400 and 12,800 (H1 and H2 on the 50D) there is no setting on the 40D to capture images at these ISOs. Since those ISOs don't exist on the 40D, I've left the 3200 ISO (H) image in for comparison - I think this will give you a great way to look for the advertised improvements in the 50D images:
Okay - while looking at the higher ISO images, I noticed distinct over sharpening. While I've not had many problems with how the .jpg sharpening works on Canon images, their routines in RAW have never looked good to me. There seems to be a herringbone like pattern to the images. I usually use a sharpening of zero, so I went back into DPP and set sharpening to 0 and then re-exported (you've got to love RAW for that, no?). Here are those images - starting with a baseline with zero sharpening, then the higher ISOs:
As with the previous post - here are the images above, but sorted by camera. I've even included the low to mid ISOs as well...
From the 40D: RAW ISOs of 100-3200(h) with sharpness of 3; 1600 and 3200 (H) with sharpness of 0:
From the 50D: RAW ISOs 100 - 12800 (H2) sharpness of 3; and RAW ISOs 1600 - 12800 (H2) sharpness of 0:
So there you have it. I think perhaps I've pummeled the pixel peeping to the proverbial pulp.
Thanks for reading!


















Thanks for the great post. I'm curious if you could also post the non-sharpened ISO 100 "baseline" for the 50D. The non-sharpened ISO 1600 50D image appears soft in comparison but I think it would be helpful to see both non-sharpened versions.
I'm debating between this and the 5D Mark II (currently still using a Rebel XT) and while I'm sure the quality in the Mark II will be better the price difference is significant (the cost of a couple lenses I want). This post goes a long way towards making me think I should go for the 50D and a couple lenses...
Thanks again for the great post.
Brian
Posted by: Brian | 13 October 2008 at 01:05 PM
Brian,
I've added the 100 ISO baseline w/ zero sharpening into the post above. Glad the post has been of assistance.
Before jumping into the 5D Mk II - be sure to see if you really have a justification for it. Meaning will you really need the better ISO performance, or need the FF sensor. If so, why, and will you need that more than better glass for your XT or 50D?
As much as the 5D Mk II has me drooling over the move to FF (for better bokeh for any given focal length), the better (hopefully) ISO given the samples posted out on the net, and the even better resolution. I recognize that it would mean I'd need to replace some of my current lenses to take advantage of it. There are also the costs of another battery grip, extra batteries, and so on. In my case, as I was moving from the 40D (well - keeping it to be my second camera, selling my 30D) the grip and batteries were the same, so no additional cost for me for those items to go to the 50D...
As an example, I use my 10-22 a bunch for my interior shots; that would need to be replaced by the 16-35 f/2.8 L II ($1,500). Plus battery grip, a spare battery (costs tbd at this point), so the cost over the 50D would be at least $4,000 (tho less if you don't want the battery grip, etc.)
Just make sure you know the total real cost of "upgrading" to FF.
- Will
Posted by: WillShootPhotos | 13 October 2008 at 03:05 PM
Thanks Will - the ISO 100 unsharpened photos are definitely helpful.
I actually have no real "need" for the 5DII but I wouldn't mind more DOF control and very likely better night ability...not that I wouldn't get the latter with the 50D. The total cost, as you suggest, is why I started looking at the 50D again. I only have the XT body, the 10-22 lens, and a bunch of batteries so I'd be starting from scratch - and with a big tab to prove it.
Brian
Posted by: Brian | 13 October 2008 at 08:34 PM
Will,
How do you align pixels manually (to avoid any "stacking" distortion from CS3)?
Posted by: Ralph | 14 October 2008 at 11:14 AM
Ralph,
What I was doing was trying to ensure that the stack of images that I was comparing (different ISOs of the same shot) were lined up properly with eachother.
I was manually adjusting the ISO and shutter speed between photos, and between switching cameras the tripod kept the battery grip mounted and I just unscrewed one camera, lifted, then set the other back down on the BG and screwed it back in.
The resulting photos were shifted slightly from image to image... to keep a crop of all of them the same, I copied and pasted them all into one PS file, I then hid and un-hid each subsequent layer - nudging images as needed - to ensure they were all lined up so when I picked my 500px wide sample area, all ISOs would show the same detail.
There is a way to import a group of images into CS3 and have the software stack them for you (for HDR work), but I believe it may do subtle rotation and distortion to ensure that the entire image stays aligned. I didn't want to use this as I didn't want CS3 to modify the image at all.
Sorry for the confusion.
- Will
Posted by: WillShootPhotos | 14 October 2008 at 02:08 PM
thank you very much; i've been trying to find a basic comparison like this everywhere. i'm sticking with the 40d and will wait to see what comes after the 50d. if they made the 50d a 12mp camera, assuming that would have created less noise and sharper images than the 40d, i would have gone for it.
Posted by: kyle | 23 October 2008 at 08:22 PM