Atmosphere and number of detectors

From CMB-S4 wiki
Jump to navigationJump to search

August 31, 2020 - Colin Bischoff and Reijo Keskitalo


The way atmospheric noise will average down in CMB-S4 will have an immediate impact on our ability to meet our science goals. Atmospheric fluctuations have been shown to produce extreme noise correlations in SPT. In the same work we demonstrate that our atmospheric simulation module is able to reproduce those correlation properties.

In this post we use the reference baseline simulations to forecast the atmospheric averaging for a SAT observing at the South Pole. The study relies on the approximation that the atmospheric model calibrated against SPT data is applicable to a SAT. There will be a follow-up study to check the atmospheric simulation module against BICEP/Keck data.


We simulate 10 independent passes of the Pole Deep field using one MFL optics tube. A pass is a full observation of the field, comprising 11 hours and 21 separate observing elevations 0.25 degrees apart. Each elevation was scanned for 30 minutes. The deck angle was rotated by 45 degrees after each pass. Two binned maps of the atmospheric simulation are made: one with a full complement of all MFLS1 bolometers (3556 in total) and one using 1/4 of the available detectors (890) evenly distributed across the 28.6-degree field-of-view (FOV).


The binned atmospheric noise maps (without filtering) are shown as the top panels here:

Atmosphere vs ndet.png

The parallels in the grid are 5 degrees apart. The meridians are 15 degrees apart. There is an obvious difference in the two top plots. The right side plot with the sparsely sampled focalplane has more scan direction striping. This is evident from the difference plot shown in bottom left. The striping is likely caused by undersampling of the patch. Absent in the difference plot are the coherent degree-scale noise modes that are identical between the two top panels. At least at these scales, the atmospheric modes do not seem to benefit at all from a fully-populated focal plane. Due to undersampling, the simulation results are inconclusive at sub-degree scales.

Angular power spectra of all the three maps are shown on the lower right. They also demonstrate that the super-degree scales () are dominated by noise modes that do not average down at all by the introduction of 4X more detectors across the same FOV.


These results challenge the model where the atmospheric noise scales like regular noise by the square root of the number of detectors. In our model, adding detectors that observe the same correlated structures of the atmosphere does not reduce the level of atmospheric noise in the final map. If the additional detectors enlarge the FOV, it will reduce the effective correlation factor across the focalplane and lead to some gains in the noise level. These projections are *not* definitive, as the correlation length across the SAT FOV may turn out entirely different than for the LAT, owing to the differences in the geometry of the near and far fields and the distribution of the atmospheric water vapor in that geometry.

The striped structures in the difference map also suggest that our current South Pole scan strategy (based on stepping the observing elevation in 0.25 degree increments and rotating the deck angle in 45 degree increments) complements poorly the gaps between detector rows on the focal plane. The South Pole SAT scanning strategy should be optimized to avoid aliasing large scale atmospheric noise into smaller scales as seen in the study.