Simulation And Analysis Of The Planck Full Focal Plane

Julian Borrill, Christopher Cantalupo and Theodore Kisner
in collaboration with the Planck Sky Model and US Planck teams,
and using the Level-S, Springtide, MADmap, M3, GCP and HEALPix software packages.



Using the NERSC High Performance Computing systems we have
  1. simulated one year of time-ordered data for all 74 detectors in the Planck focal plane,
  2. produced intensity and, where appropriate, polarization maps at each of the 9 Planck frequencies,
  3. produced intensity and polarization maps of the data at all frequencies combined.
This exercise is intended both to test our computational capability and to provide a complete simulated data set for a range of post-processing analyses. It follows in the tradition of our previous single frequency analysis.

The input data at each frequency include
  1. the polarized CMB sky smoothed with a symmetric beam common to all the detectors at that frequency (but varying between frequencies),
  2. the polarized diffuse foreground sky smoothed with the same symmetric beam,
  3. a catalogue of point sources,
  4. correlated detector noise with the current best-estimates of the amplitude and knee at that frequency,
with the diffuse and point-source foreground components coming from the Planck Sky Model.

The time-ordered data were generated at each detector's full sampling rate by the Level-S multimod tool in 1-day files, consistent with the expected data downloads from the satellite. For each detector the CMB+noise and foreground timestreams were simulated seperately to support analyses both with and without the foregrounds. The resulting data set comprised 54,168 files containing 750 billion samples and occupying 3TB of disk space. Note that we did not generate individual detector pointing, which would have required an additional 9TB of disk space.

Individual frequency maps - both with and without the foregrounds - were made using the Springtide destriping and MADmap maximum likelihood codes. Maps combining the CMB+noise and the CMB+noise+foregrounds data at all frequencies were made using MADmap. All analyses read their data through the M3 data abstraction layer, which in turn used the GCP library to generate on the fly dense detector-specific from sparse boresight-generic pointing.

Map images were generated using the HEALPix map2gif routine.







Animated sequence of single-frequency CMB+noise+foreground temperature maps.




Intensity maps for the CMB+noise and CMB+noise+foregrounds combining the data at all frequencies.