I want to know how my K-3 compares at high-ISO with my K-5IIs. Does it have more, less, or about the same noise?
I had previously downloaded DNGs from the Imaging Resource web site for both of these cameras. I had used the "NRAD" DNGs, (I think this means "auto high ISO noise reduction"), showing a studio with clothes and bottles, etc:
My A3+ prints of those showed much worse noise performance from their K-3 at high ISO. This manifested itself in the need to use more extreme noise reduction values in Lightroom to make the K-3 prints match the K-5IIs prints. For example, at ISO 51200 I needed Luminance 75 and Colour 50 for their K-3 to match the print for their K-5IIs using Luminance 50 and Colour 25.
Method and discussion
I put my DA* 60-250mm f/4 lens on a tripod, pointing a simple target. I used 60mm and f/5.6 throughout. So (as far as possible) this lens stayed still during the test. I shot with the curtains closed and a single tungsten light illuminating the scene.
I set the same operating values in my K-3 and my K-5IIs, then attached them in turn to the lens. Values included: white balance Tungsten; 12 second delay; matrix metering; Av mode with aperture f/5.6. I shot DNG throughout. I then imported all of these into Lightroom 3.6 at the same time using a "Neutral" preset that ensured the use of embedded profiles, and by default no sharpening or noise reduction. (And no use of a lens profile). I also took a virtual copy of the ISO 51200 DNGs and set noise reduction parameters for those copies.
I opened each DNG in Photoshop, added a Levels adjustment layer with input values "1 1.0 140" to lighten up the images, then printed them. I printed at A3+ on an Epson R3000 using Permajet Gloss 271 paper and FotoSpeed inks. I then resized them and saved them as JPEGs, width 1400 pixels because SXGA+ is commonly used for club photography. There were 8 such prints in all, corresponding exactly to the camera and processing setting identified below for the 8 JPEGs.
In spite of using the same camera settings, and importing them into Lightroom at the same time to ensure common parameters, the colour balance of the images is different between cameras. Perhaps the embedded profiles are not designed to a common colour balance, or the meaning of "Tungsten" has changed. Or perhaps there are camera features, such as the new scene processor in the K-3, that has caused this. Or perhaps I made a mistake somewhere. This makes comparison harder.
I think the prints show slightly more noise from the K-3 than from the K-5IIs, but not a lot more. I also think the K-5IIs noise is slightly cleaner (which perhaps means less colour noise), but not a lot of difference. There may be a trace of banding from the K-5IIs at the highest ISO. I'll get a second opinion.
Why doesn't this agree with my prints from the Imaging Resource web site?
I don't know. But in that case I had got a second opinion, and agreement that my prints showed what I stated above. The prints from the DNGs from the Imaging Resource web site simply don't tell the same story as the prints from own tests. My tests were carried out at nearly the same time for both cameras. The Imaging Resource tests were done at different times for the two cameras, and some key factors may have changed.
I've leave the readers to judge the JPEGs:
|Camera||ISO||Luminance NR||Colour NR||Link|