The Leitz table described above is one key component in the design of a measuring machine with the required accuracy. The second key component is the system which images the emulsion onto the detector. This has to be of high image quality (as measured by the MTF and by the amount of scattered light) and free from any type of distortion which can vary, either randomly due to mechanical defects, or systematically with table position.
The design chosen is based around a commercially-available telecentric lens. A telecentric lens has a stop placed so that only rays parallel or nearly parallel to the optical axis are allowed to form an image. This means that the magnification of the image formed is insensitive to the distance between the object (i.e the emulsion) and the lens. Since we require positional stability of m in the image plane, the system is required to have a magnification which is fixed to about . Without a telecentric lens, for an object-lens distance of 20cm, this would place an upper limit on the amount of variation which could be allowed in the object-lens distance of m! With the current telecentric lens, the amount of allowed variation is m, comparable to the depth-of-field required to maintain an acceptable MTF.
The telecentric lens system also has the advantage that it rejects off-axis scattered light very efficiently, to the extent that scans taken with the room lights left on show no discernible degradation in image quality from scans taken with the room darkened.
One disadvantage, however, is that the effective range in optical density of the system is reduced. This arises because much of the light which is transmitted through a dark photographic emulsion is scattered off-axis. A diffuse densitometer can collect that light. But SuperCOSMOS is a highly specular system, and the measured transmission of dark emulsions is much lower than would be measured by a diffuse densitometer. The ratio of specular to diffuse densities is called the Callier Q-factor, and has a value of about 1.6 for SuperCOSMOS. Hence an emulsion with a diffuse optical density of 2 has a SuperCOSMOS density of 3.2, and the amount of light measured is thus a factor 16 times less than would be expected from the diffuse density value.