In practice, there are gaps and possibly steps between the CCDs in a mosaic and each CCD maybe tilted differently. To obtain a good flatness, the steps and tilts of the surfaces must beminimized. The tolerance for this deviation of each detector surface from the ideal image planedepends on the size of the pixels of the detector and the focal ratio of the instrument. A usualrequirement is for the image blur to be less than the pixelsize. For a f/2 instrument with15 mm pixels the CCDs must be within ± 30 microns of aflat surface. Some spectrographs have a much stricter requirement, such as half a pixel blurlimit (all CCD surfaces ± 15 mm from a flat plane).

To achieve the alignment within these limits it is necessary to measure the deviation of thedetector surface with higher accuracy; good practice is to have the accuracy of the measuring device5 to 10 times finer than the tolerance of the surface. This means a determination of the surfacewithin 3 to 6 mm is necessary. A significant constraint to thetechnique used for the planarity measurement is the gaps between the single CCDs; a narrow bandpassinterferometer cannot be used. Absolute distance measurement with a range of the largeststep between the single devices in a mosaic is necessary.

In order to quickly iterate measuring, aligning and re-measuring, it must be possible to measurewarm CCDs with the CCDs not connected electrically. Also, measurements of the cold CCD surfaceshould be possible, which automatically implies the measurement through a dewar window in frontof the CCD. The measurement itself and the analysis should run automatically and be fast inorder to avoid deviations due to temperature changes.

This long list of requirements reduces possible designs to few non-contact technologies such asbroad bandpass interferometers, focus sensors and triangulation techniques. For interferometry,only broad band pass devices or white light interferometers are usable to measure discrete stepsbetween CCDs. There are only a few broad bandpass interferometers on the market, and these areexpensive (approx. 20.000 to 50.000 Euro).

Focus sensors usually focus a light beam onto a target and senses whether the surface isexactly at this focus point or how far the surface is from focus. The most common applicationfor focus sensing is in the reading heads of CD players. The accuracy or measurable range ofcommercial devices does not fit to the requirements derived above.

There are several possibilities for measuring the distance to a surface by triangulation techniques.One very common device is the triangulation sensor. Since there are a large variety of such deviceswith reasonable prices available on the market, we decided to do some tests with such devices.