next up previous
Next: Sensitivity limits Up: Technical considerations Previous: Technical considerations

Filled arrays vs. fully efficient pixels

The discussion of the relative merits of filled arrays, with instantanous Nyquist sampling, and sparsely sampled arrays, with ``fully efficient'' pixels, lingers on in the bolometer community. A fully efficient pixel is one that has optimum coupling to a point source, representing the well known compromise between aperture and beam efficiency for coherent receivers. The bolometer group at MPIfR kept favouring the latter approach. Fully efficient pixels have a spacing that is about four times that required for Nyquist sampling. They therefore require scanning in order to achieve Nyquist-sampled maps. However, scanning is necessary anyway if the area to be mapped is larger than that of the array. For a large hexagonal array, each subscan will provide Nyquist sampling, exept for a very narrow angular range around the lattice directions. The situation is even more favorable for a Cassegrain telescope, where the array can be oriented for optimum sampling in a single subscan. Several authors derived a factor of two speed (i.e. time) advantage of the filled array. In a much simpler argument, it can be shown that the aperture efficiency of a closely packed array of fully efficient pixels is about 70%, leading to a 50% speed disadvantage compared with filled arrays. However, such arguments do not address the problem rigorously. S. Withington of Cambridge University, UK, convincingly argues that the coherence properties of the fields must also be taken into account, but he has not yet completed this task. Even if one accepts the value for the speed advantage, one must realize that a filled array has 16 times the number of pixels on the same area of the focal plane, a non-trivial increase in complexity. Furthermore, the thermal background per pixel is only 1/16 of that of the fully efficient pixel, requiring an noise-equivalent power (NEP) per pixel that is four times lower to reach equivalent sensitivity. This is also not a trivial task to achieve, and might require lower operating temperatures, which again adds cryogenic complexity. The filled array has no intrinsic selectivity with regard to the acceptance angle. The beam definition has to be provided externally by a cold Lyot stop. This means that after the first large (warm) re-imaging mirror, a cold stop (aperture) will be placed at the image of the pupil. The stop and all following optical elements must be at low temperature inside the cryostat. The cryostat window is near the stop and must be necessarily larger than if it were at the final focus, compounding the IR filter problem. In summary, we do not see any realistic advantages of filled arrays over fully efficient arrays for mapping observations.
next up previous
Next: Sensitivity limits Up: Technical considerations Previous: Technical considerations
Frank Bertoldi 2002-08-21