The mantra for CMB-S4 is to keep the assumptions consistent with "scaling from achieved performance". There has been a long line of simulated "Data Challenge" (DT) SAT sims made by the \(r\) forecast group which scale from BICEP/Keck published numbers. The latest example scaling from BK15 in a maximally simple manner is 20191016_dc06_dsr.
Recently Reijo and co are making "Design Tool" (DT) ab initio sims starting with calculated per detector NET's. Many times in the past ab initio have lead to significant over estimates of experimental performance versus what has subsequently been achieved. We would thus like to check the \(N_\ell\) spectra of these new sims versus what one expects by scaling from achieved performance.
To provide such a cross check we will simply make a DC07 set using the latest calculated NET ratios and and detector-year numbers for the preliminary Baseline Design (pBD).
The first thing we need is NET's for S4 and for BK configurations calculated with the same code. John Ruhl has provided these in this google spreadsheet - see tabs "Pole SATs" and "BICEPKeckComparisons". We have:
Band: 30 40 85 95 145 155 220 270 BK: 265.6 289.4 723.8 S4: 169.2 204.3 290.0 248.2 280.4 297.4 609.1 1434.9 Ratio: 0.637 0.769 1.092 0.935 0.969 1.028 0.842 1.983
where the ratio is versus the closest BK band. The ratio factors are the numbers by which the map noise should scale - the factors to scale the BK15 \(\mu\)K-arcmin numbers by. Ben has provided fits to the BK15 published \(N\ell\)'s in 20190220_noiseparams_bk15_forS4 - at the bottom he provides the params.dat the DC sim generation code eats. Applying the above ratios we get the following params.dat for DC07 which can be compared with params.dat for DC06.
The second thing we need is the ratio of detector-years S4 verus BK15. The S4 numbers below come from this spreadsheet under tab "SATs", line 46. The BK15 numbers are the same as in 20191016_dc06_dsr.
The ratios are the numbers of times more detector-years S4 will have than the BK15 closest band. These are thus the numbers to scale up the total of the hit map by versus the BK15 hit map. This is provided in DC sim generation code in the reldetyrs.dat file.
The third thing we need to scale from BK15 is the hit map for the pBD and we take this from the DT sims provided by Reijo.
Below are the \(N_\ell\) spectra for the above ("traditional") re-scaling recipe from BK15 to pBD configuration. The \(N_\ell\) spectra which Ben fitted were post correction for tod filtering. The spectra below thus need no further correction.
Fig 1
Below are the \(N_\ell\) spectra for the Design Tool (DT) sims described at this link. These sims not corrected for filtering towards low \(\ell\).
Fig 2
Below is the ratio of the high ell "white noise" map noise levels (Procedure: take the mean of \(N_\ell\) for \(700<\ell<950\)), take the square-root, and take the ratio.) We see close to agreement in polarization for 30/40GHz but lower noise in DT sims at higher frequencies. The \(TT\) noise is pretty crazy high for 85GHz and above. Colin already pointed this out in the preliminary DT run last summer - see fig 4 in 20200817_DM_AWG_sim.
Fig 3
The plot below shows the absolute levels and adds the DSR/PBD measurement requirement values.
Fig 4
