cdifoto wrote in post #8097391
That's why looking at 100% crops is the only way to compare cameras and their sensors. It gives you a worst-case scenario.
Sorry, I'm having a really hard time following this logic. Here's a 100% crop:
It's from a 14-year-old Powershot 600 at base ISO. (And wasn't cropped at all, since that was the entire image, but since it fits on the screen, it's still "100%".)
When you compare that with the ISO 3200 shots on the 40D, it looks like the powershot is far superior. But in fact, if you resample both images to the same spatial frequency (e.g. same print size, or same crop size), it becomes clear that the 40D has has less noise, even at ISO 3200, than the 14-year-old digicam.
I hope that helps illuminate the flaw with 100% crop methodology. (Some people actually believe the Powershot 600 was the height of technology and things have gone *downhill* since then, so sometimes posting this image doesn't help.)
cdifoto wrote in post #8097448
To my mind there's no point in comparing "normalized" images. If you're gonna do that, there's no sense in upgrading equipment in the first place. Might as well just use what you have. If you're comparing because you're upgrading, you probably want to know how much MORE you can get out of the new kit. Thus, 100%.
In other words, if my gear is already great for the sizes I print and I know I won't be printing any bigger than I already do, I'm not going to buy more pixels.
First of all: that's fine. If you wont upgrade unless you can get the same noise power at *higher* spatial frequencies, that's fine. You are expecting more performance from the new camera than the old camera, and that's OK. But it's not correct to say that the 50D is noisier than the 40D. It can do everything the 40D can do (same noise at the same spatial frequency), but it has the option of doing more (higher spatial frequencies), if desired by the photographer.
Second, there will always be higher noise power at higher spatial frequencies because of photon shot noise caused by the quantum nature of light. Given that quantum efficiency is already near the maximum theoretical limit, and Canon has only improved it 30% in the last three years (5% in the last 12 months), new cameras will always have higher noise power at higher spatial frequencies. (But, again, they will still retain the same low noise power at lower spatial frequencies as the old cameras.)