Last Modified:
 
email dave@allmon.com

There are some images made with the KAF which seem to have a large number of hot pixels which did not appear in the dark frame. It does not happen often, at least not with either of my devices, but when it happens, it is fairly annoying. These hot pixels produce trails in an averaged image.

The cause of these pixels is unknown. It would seem to make sense that, because the detector should produce a constant output for a constant light input, that the problem must not be associated with light, but something else which affects only certain pixels. Yet these pixels are hot when light strikes them, but not otherwise.

Since it can be generally said that CCDs are linear, using the average value of each frame as a reference point, any measured difference between the slope of the frame average and the individual pixel can be considered a non-linearity.

Test program

The test program looks at a frame to determine the average pixel value. It then looks at the difference between this average, and the average for the next frame. This becomes the reference slope, or gamma. It then looks at each pixel in the array to find the value, and the slope between this value and the value of the same pixel in the next frame. This is the slope of the individual pixel. The two slopes are compared. The difference is the non-linearity of the individual pixel.

Test Data

To get a curve for the device, flat frames from several points within the range of the detector are needed. The resolution of the resulting curve is dependent on the number of points and the consistency of the spacing between them.

The test program is unable to discriminate between signal and noise. For this reason, the flat frames needed to be of short, constant duration. The amount of illumination was varied, by pulsing a light source for varying durations during the constant exposure time. A special fixture was made to facilitate the test.

A number of frames were made at each illumination level, and averaged together to reduce noise. These frames were passed through the routine, and the results printed to a file.

An area of the detector was selected which did not include the two edge rows and columns. The software I used to read the camera seems to read the array incorrectly, and returns two bad rows, and one bad column. Since the hot pixels I have had problems with fairly evenly distributed across the detector, the loss of these areas should not skew the results.