See also Jan 22 entry. Tonight I made some longer exposures on the Nikon D3x, D3 and the Canon 5D Mark II (mostly in the 30 second range). My findings will be written up in DAP soon, covering a range of ISOs with and without long exposure noise reduction (LENR). My interest is not just long exposures per se, but whether night-time images can stand up to scrutiny: streaking and hot pixels and other nasties are real issues to deal with.
Not so expected was the “christmas tree lights” performance of the Canon 5D Mark II, no matter what the ISO. This is one “dirty” sensor! (firmware version 1.0.7, I observed no “black dots”). You get what you pay for evidently; the 5DM2 is 1/3 the price of the D3x. Exposure times are effectively double with the 5D Mark II because without long exposure noise reduction (dark frame subtraction), you get stuff like this, all over the frame.
Canon 5D Mark II, ISO 200, 30 seconds, actual pixels
Hundreds of such such sprinkles over a 5D Mark II frame are not manageable. They do go away by turning on long exposure noise reduction, but this doubles the time it takes (eg 30 second exposure followed by 30 seconds of dark noise subtraction for a total time of 60 seconds). Not so fun for 15 minute exposures which are rather amazing on the D3x even without LENR.
The D3x can only be described as ridiculously “clean”, as might be expected based on the findings in Pushing the Blacks. The D3x doesn’t even activate long exposure noise reduction for less than 8 second or so exposures (at least not in tonight’s testing). I didn’t find even one such hot pixel in the same image from the D3x—not one, and that’s at the same 30 seconds with all noise reduction disabled.