We test the K10D at night and for dark field image noise.
To look at longer exposure image noise we setup te Pentax K10D on a solid tripod and sot our normal city view with 8 second exposures and aperture varying from f5.6 to f22 as we varied the ISO.
PEF raw files for all these images are available from the camera test page.
Below are minimal compression JPEGs of the central section of the RAW image.
100ISO
200ISO
400ISO
800ISO
1600ISO
As you can see, image noise is minimal at 100ISO, rising noise quickly from 200 to 1600ISO, which is quite heavy and objectionable by 1600ISO.
But there is more to it than that. Below are reduced resolution, full images:
100ISO
400ISO
1600ISO
Now it may not be so visible on the web, but as the ISO is increased, apart from the rise in noise, is a brightening on the right of the image compared to the left.
To see what is going on, I took 30 second exposures with the lens cap on for a dark field image. These are reproduced below, along with a processed version using AutoLevels in Photoshop, to make the noise more visible. These were shot at home in low light with the lens cap well in place.
100ISO
processed
400ISO
processed
1600ISO
processed
You can see that at 100 and 400ISO there really is just the odd hot pixel. But at 1600ISO there is a strong rise in noise, especially on the right of the image, which accounts for the lightening we see in the images. This is thermal noise likely generated by circuitry located on or close to the right of the sensor. Now, of course, this is stronger in the 30 second exposures than the actual 8 second ones of the skyline, but it does seem to explain what is visible in the image. Also of interest is how the noise preferentially affects different color pixels as we go from left to right. My guess is from the quite linear nature of the noise that it is more strongly impacting on certain columns, and thus certain filtered pixels in the Bayer array.
So the K10D produces a pretty low noise image at 100ISO, but is subject to substantial noise as the ISO rises, especially at 1600ISO.
wow.. good indepth analysis.. really helpful