Originally posted by builttospill The K-1 has a much larger sensor so this should provide better noise control. But ISO 100 images from the larger sensor aren't as clean as the CCD sensor. CMOS is a newer technology so a 10 MP CMOS sensor wouldn't make anything fair.
We're not discussing a fair comparison, we're talking about the differences and if those are substantial enough to take notice. My eyes notice the difference whether it's a fair comparison or not. Strengths and weaknesses are found in each tool. Use the best tool for the job.
Maybe I can make this a little easier to understand...
The issue here is the size of the pixel site. The larger the pixel site, the more likely noise is to be drowned out by the larger number of photons that react with that site. One noisey photon out of two or three has more effect than one noisey photon out of 25. It's pixel size that affects noise, not sensor size.
Quote: The K-1 has a much larger sensor so this should provide better noise control.
So, not necessarily.
Quote: But ISO 100 images from the larger sensor aren't as clean as the CCD sensor.
Because the pixel sites are smaller on the larger sensor.
Quote: My eyes notice the difference whether it's a fair comparison or not.
I suspect only at a resolution that can't be achieved with the K10D.
The K10D pixels pitch is 6.1 microns producing 3872 x 2592 images.
The K-1 pixel pitch is 4.88 microns, producing 7360x 4912
What I'm suggesting here, is that the K-1 natively should have more noise.
But, if you treat the K-1 image as a 4x oversample and reduce the image to 3872x2592, you will probably have the same or better noise, at the same or probably better lw/ph resolution.
I'd be the last one to suggest it actually is that way, as opposed to just being a theoretical construct, but to compare systems. reducing the K-1 to K10D res, might provide a better image, even if the noise is worse at full size, which is what i would expect.
It's unreasonable, given the pretty constant and predictable nature of noise to expect the K-1 to do anything but capture it accurately. But if you use it as a 4x oversample, you should get better images, more dynamic range etc. than a K10D image. And if you use pixel shift, also a 4x oversample, and then add a size reduction 4x oversample, you should be able to exceed your K10D output by a considerable margin.
After all, if you can live with the 3872x2592 of a K10D and you are interested in better noise reduction and IQ, why wouldn't you go that route with your K-1, at least see if it's comparable. I'd do it, but I don't have a K10D.
I'm guessing, using pixel shift and reducing the size to say even 4500 x3000, you can get both clean images and more resolution than your K10D.