Noise... camera ISO sweet spots?

The D800(E) sensor is not ISOless. If it was, then we would see the DXOmark dynamic range plot drop by 1 Ev for each stop increase in ISO. In fact it drops only about 1/2 Ev from 100 to 200 ISO and drops 2.7 Ev instead of 3 Ev from 200 to 1600 ISO. Above 1600 ISO the D800(E) is ISOless. Another way to check for an ISOless sensor is to check the read noise in e- derived from the full SNR data. For the D800E that is the noise n in the following table (m is full well count and g% is PRNU):

NikonD800E

ISO: 100 200 400 800 1600 3200 6400 12800 25600

m = 54924 26293 13414 6722 3483 1776 833 438 237

n = 5.16 3.25 3.27 2.95 2.72 2.78 2.41 2.81 2.83

g% = 0.30 0.37 0.42 0.50 0.51 0.74 0.00 0.49 0.00

Note: Data at 6400 and 25600 ISO is unreliable because of implausible g% value.

We see the read noise starts at 5.16 e- then drops by about 1.9 e- at 200 ISO. Then it drops about another 0.5 e- to about 2.75 e- at 1600 ISO and above.

Compare this to the Canon 1DX table below:

Canon1DX

ISO: 100 200 400 800 1600 3200 6400 12800 25600 51200 102400 204800

m = 90101 46728 23720 11969 6296 3072 1524 738 332 163 104 52

n = 38.51 19.23 9.98 5.28 3.58 2.19 1.89 1.62 1.27 1.70 1.65 1.17

g% = 0.44 0.43 0.52 0.59 0.71 0.83 0.84 0.76 -0.00 -0.00 5.92 8.38

Note: Data above 12800 ISO is unreliable because of implausible g% value.

Here the read noise starts much higher at 38.5 e- at base ISO, then steadily drops to about 1.6 e- at 12800 ISO. So ISO gain does have a much greater effect on read noise for this Canon camera than for the Nikon D800(E).