Dark signal comparison of the Nikon D70 and the Canon 10D

COMPARISON
NIKON D70 AND CANON 10D
Christian Buil - Marc-André Besel
Jean-Marc Mari - Christian Servière

Histogram analysis of a nearly uniform scene
The histogram is constructed only with the green pixels population of the raw image
(the Iris software is used for decode the raw files)
To be confirmed ! To be confirmed ! To be confirmed !

NIKON D70
ISO 400

CANON 10D
ISO 400

Dots representation

Numerous digital code seem to be loss. The number of Analog Digital Unit between two valid counts is near 10, but this value is not very stable and it is not a power of two (not an analog digital converter error!).
But, a problem concerning the raw decoding algorithm of Iris software is not excluded (Iris use the dcraw library)...

Dots representation

All digital counts are valid.

Impulse representation

Dark signal (1)

RAW format - ISO 400 - 900 sec. exposure at 20°C
The user interface noise removal procedure is off
Part of the dark frame scale by a factor 7

Display threshold: HI= 600, LOW= 0

Numerous double "hot-pixels" are visible (aligned along the horizontal axis). The two components are always equal.
(possible result of an internal masked processing for remove dark signal ?)

Dark signal (2)

This Nikon images are from Marc-André Besel
See also the Marc-André analysis (D70, D100, D2H and 10D comparison)
The analyse is based on Raw format images. The software Iris is used for load and process these data.

NIKON D70 ISO400 - EXPOSURE TIME: 180 seconds - TEMPERATURE: 19.7°C	CANON 10D ISO 400 - EXPOSURE TIME: 180 seconds - TEMPERATURE: 20.5°C
Full frame Size reduced by a factor 0.15 - Display threshold: HI= 80, LOW= -70	Full frame Size reduced by a factor 0.15 - Display threshold: HI= 210, LOW= 60
Full frame Size reduced by a factor 0.15 - Display threshold: HI= 20, LOW= -20	Full frame Size reduced by a factor 0.15 - Display threshold: HI= 150, LOW= -110
Detail Original scale (crop of the RAW CFA image) Display threshold: HI= 40, LOW= -30	Detail Original scale (crop of the RAW CFA image) Display threshold: HI= 170, LOW= 100
Histogram Subframe of 900x900 pixels - Amplifier glow is excluded of this frame (note the partial loss of some counts)	Histogram Subframe of 900x900 pixels - Amplifier glow is excluded of this frame
Full frame - Two successive 180 seconds dark frame subtracted Display threshold: HI= 30, LOW= -30	Full frame - Two successive 180 seconds dark frame subtracted Display threshold: HI= 30, LOW= -30

Comment

The observed D70 thermal signal level is incredibly low. I don't know CCD technologie equivalent if no internal processing is done by the internal firmware. For this test the internal noise reduction function was turned off, but the result is strange... A real time correction of the hot pixels is very probable! For the mass consumer, internal processing or not, the result is impressive, but the consequences for astronomers is to evaluate.

The thermal dark histogram of the D70 is not excellent because the signal is clearly clipped for the zero level (possible loss of low light level information). The 10D histogram (and 300D histogram) is much more gaussian and coherent with a normal electronic and signal noise distribution (but a problem concerning the raw conversion algorithm of Iris software is also a possibility for the D70).

The glow of the amplifier and the output amplifier stage is evident for the D70 and very weak for the 10D. For the D70, the peak intensity of the glow is given by the formula (in ADU):

S = 0.0019 . ISO . EXPO

ISO is the ISO sensibility and EXPO is the exposure time in seconds. For example: is ISO=800 and exposure=300 seconds, the parasitic light level produced by the amplifier is equal to 450 ADU (remember the dynamic range of 4096 ADU).

Some very faint horizontal patterns are visible for the D70 and can not be entirely removed after difference of two successive frame.

The electronic gain of the D70 is of 2.9 e-/ADU at ISO400. Also for the D70, the RMS noise is of 3.6 ADU for the 180 seconds exposure (or 11 e-). The measured gain of the 10D is 2.6 e-/ADU at ISO400.

More test are necessary for evaluate the global performance of the D70 : quantum efficiency, Ha sensibility, response linearity, spatial resolution (more punchy and sharp images for the D70 compared to the 10D because the antialiasing filter seem less severe - but photometric error is also to evaluate for this reason). Some points are irritating with D70 (no wire remote control, amplifier glow, zero clipped offset) but this first view shows a camera valuable for the astronomical imagery.

Back