Up to Nikon D3 index

Nikon D3 summary of findings

UnitsRedGreenBlue
Read noise at ISO 20014-bit DN*4.3*
Read noise at ISO 160014-bit DN*9.84*
Full well capacity at ISO 200electronsnot measurable65,600not measurable
Maximum mean DN14-bit DNs16,38315,907.616,383
Firmware scale factor of DNs at ISO 200->1.11~1.1
Gain at ISO 200electrons/14-bit DN*4.1*
Gain at ISO 1600electrons/14-bit DN*0.498*
Max signal to noise ratio at ISO 200-231242199
Max signal to noise ratio at ISO 1600-79.485.379.7
Midgrey signal to noise ratio at ISO 200-537252
Midgrey signal to noise ratio at ISO 6400-7.711.27.5
Unity gain ISO (12-bit DNs)--3,280-
Max dynamic range at ISO 200 (Smax/read noise)stops-11.9-
Practical dynamic range at ISO 200 (to S/N=3dB)stops10.811.010.7
Practical dynamic range at ISO 1600 (to S/N=3dB)stops8.99.38.9

Unlike Christian Buil here, my measurements suggest that the higher ISO speeds on the Nikon D3 are correctly rated and closely comparable with my Canon EOS 5D (which of course does not go so high).

Unlike Canon cameras, the Nikon D3 has no offset to the voltage applied to its analogue-to-digital converter, so zero exposure of the sensor should produce a zero DN in the raw file (apart from noise).

In a dark frame taken at ISO 200, 74% of the green pixels had value zero. It was thus not possible to measure the read noise by looking at the standard deviation of subtracted dark frames.

Saturation of the green wells of the sensor is measurable, because the values never reach the 16,383 level at which they would be clipped either by the analogue-to-digital converter or by the 14-bit raw file. Saturation of the red and blue wells is not, because the values of these pixels are clipped at 16,383 after stretching in firmware.

Peter Facey, Winchester, England
20080217 corrected units of scale factor in table 20080131 updated
20080126 originated