I have the images that you sent to me privately.
I would say you are accidentally comparing the D800 with the D800E(DX) rather than with the D800E
I have not written my results up yet but I studied this recently and the PDR results are very repeatable between multiple cameras and with the same camera.
Oops! I see that now. How stupid of me!
Cool! Thanks for including that Bill.
More likely this comes down to imprecise gain values. I suspect the read noise in electrons ought to match. (But not those in ADUs!)
That brings up an interesting notion. To get better specs on read noise, would we need a Peltier plate behind the sensor, to chill it? That might require a heat sink, too. Or maybe a liquid nitrogen reservoir, hmm?
If you haven't tried the D800 at ISO6400 and up, you might want to give it a look. The thermal noise component is very strong, even at 1/80th sec, and much worse if you're using live view. Though you can mitigate this with a dark frame subtraction, it otherwise impacts the overall performance in a way that is not represented in the DR measurements.
Good illustration can be found here http://mansurovs.com/nikon-d800-vs-d800e
A better illustration is here:
I rarely use high ISO, so don't have much experience in this area. How did you identify the effect as thermal noise? Because it worsens when Live View is left on?
Because PDR is based on a fixed output size so the DX Crop Mode image must be enlarged more to fill the same area.
You are welcome !
I took black frames at 1/80th and ISO6400, one after one minute of live view. I've been using these frames to do black frame subtraction on images taken at those settings. This added step yields images of surprisingly good quality even after 1-3 stops of additional digital gain. But without it, your images will be awash in a blue-magenta cast. Even at 1/80th. The dark frame taken after 1 min of live view is quite a bit worse.
Interesting. I guess this may be more of an issue now that the other noise sources are so well controlled.
Right, and this illustration also shows the usual set of 2 AA filters.
This speculation doesn't make sense. They did not remove the AA filter. In the D800 there's 2 birefringent filters, one that acts horizontally, and one that acts vertically. In the D800E, they rotate one of those filters so that the second reverses the effect of the first. Same amount of glass, same amount of attenuation. I think the difference shown might just be sample variation. Were multiple D800's sampled so that a standard deviation could determined? The increase is measurable but is it significant? It is possible that it's just one camera's a little more sensitive than the other, it is possible there could have been slight fluctuation in the lighting for the test, it is possible that the D800E is more sensitive (maybe they use a slightly different micro-lens array because of the AA difference).
What is output size or resolution of the output area? I believe DxO is 8MP. What is yours?
