I can't see any appreciable differences at higher ISO's between the two bodies. But, two caveats - I only shoot RAW, and I don't still have the 40D to compare side-by-side. Just looking at older 40D files and current 50D, the noise looks about the same under similar shooting conditions. With any of the Canon bodies, thoughtful exposure makes all the difference, even going back to my ancient 10D.
I only shoot RAW and only in full stops with ISO... therefore I cannot tell you much about in camera JPEG processing or shooting with ISO 1000.
What I can tell you though, is that I tested these cameras very well in particular on the S/N front, as this was a very hot topic at some point in time.
Here is my take on things then:
1. Up to ISO 800 and below (800, 400, 200, and 100) the sensors behave equal on all accounts concerning noise and signal dispersion. I.e. blacks and highlights are the same in both chroma and luminance noise. The out-of-focus areas look the same on pixel level.
2. When underexposing the 50D as apposed to the 40D, low light recovery of the 50D is better as far as detail is concerned, but adds more noise than the 40D. The reverse is true for the highlights though.
3. Color rendition on both cameras are very good and very similar, but I do reserve some speculation about the 50D's red channel, which I find a bit over the top towards magenta.
DXO Mark comparison of S/N between the 40D, 50D, and D300. As you can see, they are extremely close, if not equal in most respects. (Whole link not accepted. Access their database.)
However take into account, and as you mentioned, the higher resolution of the 50D, then I can only applaud the performance of the 50D sensor and Canon's implementation thereof.
Just remember that higher resolution does come at some price. You will have to tweak your sharpening technique somewhat. Your exposure should be to the right of the histogram in order to take full advantage of the highlight tone and higher mid-value. Also your focus will be more critical, and you will probably be better off stopping down your lenses somewhat to deal with diffraction.
Other than this, these cams are very equal, and where they are not, the better for the 50D. It is a brilliant piece of hardware, and will perform exactly like you want, given that you treat her right that is.
In real world shooting (and printing) the difference is somewhere between bugger all and nothing. Differences at 100% on screen just don't translate through to prints.
Aragosh wrote:
you will probably be better off stopping down your lenses somewhat to deal with diffraction.
Actually, you will be better off not stopping down your lenses as much, to avoid seeing the effects of diffraction. The diffraction begins to be visible in a 40D from f8.5 while with the 50D it is already at f7.0!
droopy1592 wrote:
Meh, the diffraction myth again!
The 50D still gets more overall image detail @ f/13 than the 40D does at it's optimum aperture.
No, it's not a myth. It's simple physics. Diffraction means that even perfect optics cannot produce smaller detail than Airy Disk (see http://en.wikipedia.org/wiki/Airy_disc ).
Let us then examine the case of the 40D at f8.5 (diffraction-limited aperture) against the 50D at f13.
The combined system resolution of a camera + lens can be calculated as follows: 1/Sr = 1/Cr + 1/Lr, where Sr = system resolution, Cr = camera resolution and Lr = lens resolution.
50D resolution equals 106.3 lp/mm (=4752pixels/22.3mm/2). Let us assume a perfect lens that is diffraction limited. This basically means that at f7 you can get at most 106.3 lp/mm as well (remember, f7 was calculated to be the aperture where a perfect lens matches the 50D pixel pitch - that is why the number is exactly the same).
The system resolution for the 50D case (at f7.0) is then:
Sr = 1/(1/106.3+1/106.3)=1/(2/106.3)=53 lp/mm.
40D resolution equals 88 lp/mm (=3888/22.2/2). Taking the same lens as above but stopping it to f8.5 (diffraction limit at 40D pixel size), we'll get 88 lp/mm (=106.3/(8.5/7)) for the lens - not surprisingly, as we are using the diffraction limit as the criteria and we assumed the lens to be diffraction-limited. Now the system resolution is 44 lp/mm.
When we stop the same lens to f13, the resolution drops to 58 lp/mm (=106.3/(13/7)). With the 50D the system resolution is now 1/(1/106.3+1/58)=38 lp/mm.
Thus, 40D wins in this case: 44 lp/mm vs 38 lp/mm.
All this assuming a perfect lens that is only limited by the diffraction of light (a physical fact). Any issues in the real world only make the results worse. Consider e.g. lens aberrations (especially outside the center), camera antialias filter and resolution reduction by Bayer matrix.
At the 40D diffraction-limited aperture (f8.5) the system resolutions would be as follows:
40D: 44 lp/mm (as calculated above)
50D: 1(1/106.3+1/88)=48 lp/mm. Slightly more than 40D but not hugely (9%).
Let us still use the same diffraction-limited lens but stop it to f4 only (and still assume it to be diffraction-limited - a very high-quality lens, indeed). The lens resolution would then be 106.3/(4/7)=186 lp/mm.
Thus, the theoretical resolution improvement of 50D over 40D would be 21% (=5.7um/4.7um). In practice we were able to get only 13% (=68/60). Even at f2.8 (show me a lens that performs at diffraction-limited resolution there), it would be 15% (=76lp/mm / 66lp/mm).
Diffraction limitations are here to stay - they have already long been a reality with the P&S cameras and now DSLRs are also starting to notice the effect.
I hope Canon will concentrate on improving the noise, dynamic ratio and shooting features of the digital cameras, instead of continuing on the megapixel bandwagon.
astrolucida wrote:
I hope Canon will concentrate on improving the noise, dynamic ratio and shooting features of the digital cameras, instead of continuing on the megapixel bandwagon.
