COLOUR IMAGES WITH THE KAF-0400C CCD
The KAF-0400C is a colour version of the classic KAF-0400. The two are identical electrically, but the Audine camera design requires slight modification to the heat sink to be used with the KAF-0400C. This is due to the fact that the colour camera is housed in a smaller box. The modification consists of adding a small mechanical part at the top of the sink to allow the sink to reach the back of the KAF-0400C. It is also necessary to solder the colour chip to a printed circuit board for it to fit the support housing of the KAF-0400.
An Audine camera fitted with the KAF-0400C.
 
The mechanism by which the KAF-0400C generates the colours consists of tiny red, green and blue colour filters placed just infront of the CCD array pixels. The organisation of these colour filters adopted by Kodak is so-called CFA-form, as illustrated below :
 
B G B G B
G R G R G
B G B G B
G R G R G
 
To produce a colour image, it is necessary to process the image generated by the KAF-0400C using suitable interpolation algorithms, to estimate, for example, the intensity of blue and green at the point of a red-filtered pixel using the signal values for the blue and green pixels surrounding the red pixel under analysis. The result is a set of three images each of the size of the original (760x480 pixels) and each corresponding to one of the primary colours of red, green and blue. These can then be recombined (e.g. using the command TRICHRO in QMiPS32, or with a graphics package such as Paint Shop Pro once converted from PIC to BMP format) to produce a 24-bit true colour image.

The following image shows the relative sensitivity of the three CFA-matrix colours of the KAF-0400C. The curves represent the spectral response of a pixel resulting from the spectral transmissivity of the its associated filter.

Sensitivity with respect to the three filter colours of the KAF-0400C.

It is clear from this figure that the the blue filter is very poor at filtering the red and infrared parts of the spectrum; in other words, the blue filter passes as much infrared light as blue, which is clearly less than ideal for reconstituting a well colour-balanced image. It is therefore recommended to use a filter at some point in the optical path to suppress the infrared part of the spectrum. The following curves show the spectral sensitivity of the KAF-0400C when used with a BG18 filter or a KG3 filter respectively.

Spectral sensitivity of the KAF-0400C combined with a BG18 filter.
Spectral sensitivity of the KAF-0400C combined with a KG3 filter.

Transmission characteristics of a 1mm-thick BG18 filter.

  Lambda(µm) Transmission
    .30         .000
    .32         .005
    .34         .174
    .36         .402
    .38         .521
    .40         .603
    .42         .658
    .44         .713
    .46         .759
    .48         .786
    .50         .813
    .52         .813
    .54         .795
    .56         .722
    .58         .594
    .60         .411
    .62         .274
    .64         .146
    .66         .064
    .68         .027
    .70         .009
    .72         .003
    .74         .001
    .80         .000
    .85         .000
    .90         .000
    .95         .000
 
Transmission characteristics of a 2mm-thick KG3 filter

  Lambda(µm) Transmission
    .30         .018
    .32         .284
    .34         .654
    .36         .823
    .38         .859
    .40         .836
    .42         .815
    .44         .819
    .46         .819
    .48         .835
    .50         .846
    .52         .840
    .54         .842
    .56         .858
    .58         .849
    .60         .830
    .62         .793
    .64         .740
    .66         .674
    .68         .601
    .70         .513
    .72         .423
    .74         .333
    .80         .124
    .85         .042
    .90         .013
    .95         .004
 
The BG18 suppresses the infrared successfully, but to the point where a large part of the red spectrum has been suppressed too, including the H-alpha ray region, of interest in astronomy. The KG3 filter is more satisfactory in this regard, but has a poorer spectral purity for everyday image capture:

The STAR WARS film poster shown to the left was taken with an Audine camera fitted with the KAF-0400C and a BG18 filter. The image to the right is of the same poster taken with a KG3 filter. The colour fidelity is better with the BG18 for this kind of image (approximately uniform spectrum). For observing nebulae emitting radiation however, it is far better to use a KAF-0400C with a KG3 filter.
 
Having taken an image with a KAF-0400C, the colours do not appear immedately on the screen: it is first necessary to produce the three interpolated red, green and blue composite images, and then to combine them, to produce the true colour image.The illustrations below were generated using QMiPS32 for the primary colour images and Paint Shop Pro for the combined true-colour image.
A section of a typical raw image of M57 taken with a KAF-0400C with a KG3 filter, from a 190mm telescope, with a 2 minute image capture cycle.
Enlargement of the previous image. The difference in photometry for pixels with different associated filters is clearly discernible (the darkest pixels correspond to the blue filter). As a consequence of the spatial separation of pixels of the same colour, the spatial resolution degradation of the KAF-0400C relative to the KAF-0400 can be noted.
Enlarged segment of a flat-field image produced with the KAF-0400C. The CFA matrix structure is clearly discernible. This image was generated by pointing the camera at a matt white piece of paper in daylight, allowing the spectral response of each pixel to be determined in advance. Using an image taken in sunlight was not by hazard: this results in a colour balance that means that stars of the same spectral characteristics as the sun have a white spectral response at the end of the day.
Division of the raw image of M57 by the flat-field image. As expected, the resulting pixel intensities are more uniform, but a small residue remains which in effect allows the colour image of M57 to be produced.
At this point, the interpolation for each of the three primary colours is performed. The image shown is the resultant of interpolating the green pixels. To perform the interpolation, the QMiPS32 or Iris CFA command can be used.
Resulting combination of the 12 original raw images after splitting into individual colour images and interpolation, and combined using Paint Shop Pro.
Enlargement of the previous image.

A summary of the procedures used to produce the image above is now given:

Step 1: take a large number of images of the object of interest, ensuring that the object is displaced a few pixels in X and Y for each image taken.
Step 2: remove the offset and dark photometry values from each image.
Step 3: divide each image by the flat-field.
Step 4: apply the CFA command to each image.
Step 5: Recentre each of the images using one colour image only as reference.
Step 5: produce three BMP files containing the combined individual colour images. If a command such as SAVEBMP is used, make sure not to "erode" the signal in order to ensure that the final quality of each image is maintained and to keep as much of the relative spectra in the final colour image.
Step 7: generate a 24-bit image, e.g. using software such as Paint Shop Pro to combine the three BMP images.

The nebula M27 taken with an Audine camera fitted with a KAF-0400C and a KG3 filter. The image is a composite of 14 images each of 2 minute collection cycles, using a 190mm diameter telescope. Image taken on the 25/06/1999.

The basic advantage of the KAF-0400C is that it can produce colour images without requiring three separate image capture sequences and colour filter changes, i.e. more simply. For precise photometry analyses, however, the individual filter approach is still preferable (better spatial resolution and knowledge of the individual filter characteristics).

It is important to restress that the KAF-0400C is significantly less sensitive than the KAF-0400, typically of a factor of 3 to 4 over the average of the RGB filter range. It is therefore not ideal for routine deep-sky observations.