This is one in a series of posts about the Nikon D5. The series starts here.
Now we’re going to look at the histograms of a 200×200 pixel central patch in a series of dark field images made with the D5 at various ISO settings. Because the noise behavior in the preceding post was so interesting and curious, we’re going to look at more of these than I usually show.
I am used to seeing gaps in the red and green raw histograms with Nikons, becasue of the digital white balance prescaling, but I’m not used to seeing gaps in green channels. Even with the eye, you can see that the green channels are tighter than the red and blue.
ISO 125 has the same input referred read noise as ISO 100. It is apparent that there is some digital processing at this ISO setting. Is it possible that all the gain from 100 to 125 is digitally created?
ISO 160 also has the same input referred read noise as ISO 100 and 125.
At ISO 200, the read noise actually drops compared to ISO 160. The histograms also look less processed.
These histograms look cleaner than the ISO 125 ones. The input-referred read noise at ISO 200 and ISO 250 is identical.
Looks a lot like ISO 160, doesn’t it?
Again, the read noise drops with increasing ISO.
The input-referred read noise at ISO 400 and ISO 500 is the same.
The read noise drops as you go from ISO 500 to ISO 640, breaking the one-stop periodic pattern that we’ve seen up to now.
Same input-referred read noise as ISO 640.
Also same input referred read noise.
At ISO 1250, the read noise drops again.
ISO 1600 has the same input referred read noise as ISO 1250/
ISO 2000, 1600, and 1250 all have the same input-referred read noise.
It’s hard for me to tell what kind of processing Nikon is doing. If you look at the progression of the bottoms of the sawteeth, the noise doesn’t rise with ISO setting anywhere near as fast as you’d expect it to in a modern Nikon camera.