Ever since the Canon EOS-10D was released, I’ve been hearing about the wonderful high ISO performance. People have been raving on the forums that 10D images are clearer than many digital SLR’s on the market today, including the Canon EOS-1D. I never had the opportunity to test this theory until last night when Mike came over to return my 28-70 lens.
I’m most interested in evaluating a camera’s low-light, high ISO performance, since these are situations I face on a daily basis. By low-light, high ISO situations, I mean lighting conditions that necessitate setting a high ISO speed in order to maintain a reasonable hand-holdable shutter speed. Most high ISO tests on the Internet don’t replicate these types of real-world conditions; they usually involve photographing the same well-lit scene, varying only the ISO speed. For instance, they test using ISO100, f/8.0, 1/50s all the way to ISO3200, f/8.0, 1/3200s. Unless you’re shooting sports, you’re rarely finding yourself in a situation where you’re thinking, “Hmm… should I shoot ISO100 or ISO3200 if I can handhold the shot?” More often than not, you’ll find yourself thinking, “I must use ISO1600 in order to handhold this shot.”
Methodology (or lack thereof)
We used a tripod and hand-holding to test the performance of the 1D and 10D. We used the 24/1.4 lens at f/2.8 and the 17-40/4.0 lens at 4.0 for our testing. The tripod was not moved during testing, which means the 1.6 crop factor and sensor pixel density (6MP vs. 4MP) of the 10D makes its images look larger than the 1D’s. When hand-holding, I tried to keep the relative size of the image the same. The 1D’s noise reduction was set to NR1, meaning all exposures longer than 1/15 second would use a dark-noise subtraction algorithm to clean up the image (for more information on the 1D’s noise reduction settings, read this article).
Here are full-size crops of the images that we had taken. The crops were saved in Photoshop at JPEG 11. You judge for yourself which camera produces the least noise.
The 1D may have more noise, but the noise appears smaller and overall cleaner than the 10D’s. The 10D’s noise looks to be larger and more clumped together. In my opinion, this is not as appealing as the 1D’s noise. My conclusion is that for these lighting conditions, the 1D has much better noise characteristics at high ISO than the 10D. I have no idea what the results would be in a hockey arena where you have to use ISO1600 or ISO3200 in order to freeze the action at 1/250 second.
As to people’s claims that the 10D’s high ISO performance is amazing, Mike hypothesizes that these people are probably former D30 or D60 users. Compared to the D30 and D60, the 10D has very good high ISO performance. When I had the D30, I rarely went above ISO400 because of what I perceived to be high noise at 800 or 1600. In retrospect, ISO 800 wasn’t that bad on the D30. On the 10D, I wouldn’t mind going to ISO1600, but 3200 is unacceptable for anything but web images (but remember, everything looks good at web resolution!).
Here are web-sized images of the samples that we took. You can see even in these images that the 1D’s images look better than the 10D’s. They might not be noiser or cleaner, but they look nicer.