This was a bare sensor test as described by Emil Martinec here.
The two cameras were placed side by side with no lens fitted. An 8 second exposure was used in near total darkness at ISO 200. During the exposure a Canon 580EX flash was fired directly at the cameras at 1/128 of its full power from a distance of 6.3 metres.
I found this a difficult test to conduct without an assistant. The 8 second exposure was necessary to give me time to press the shutter releases and then run from one end of my sitting room to the other and fire the flash. At 6.3m distance, anything brighter than 1/128 of the flash's full power caused the sensors to saturate. I found to my surprise that my Sekonic L-758D could not measure this flash when placed in the position of the cameras, although it was quite bright enough to make the cameras clearly visible to the naked eye.
I did not, as Emil Martinec did, take a second measurement with the cameras interchanged, in order to verify that the illumination did not vary between camera positions. I ought to have done so, but I had enough trouble taking one measurement.
The cameras were set to write raw files (14-bit for the D3, 12-bit for the 5D) and all in-camera noise reduction options were turned off. The raw files were processed in such a way as to extract the numbers (referred to as DN for data number) without any alteration. The raw files were split into their red, green and blue components, but no Bayer interpolation (or 'demosaicing') was done.
The table below shows figures for a 200 x 200 pixel area near the centre of each image for the three different colours: For each camera, the mean value of each colour component is first calculated as a percentage of its saturation or maximum possible value. Then these percentages are compared between the two cameras (bottom line of the table). This comparison of course encompasses both the sensitivity of the sensor and the analogue and/or digital scaling performed by the cameras before writing the raw files.
I have some doubt about the reliability of this test, due to my failure to verify that the two cameras received essentially the same illumination from the flash.
Peter Facey, Winchester, England
20080217 typos corrected
20080124 originated