This post follows from 20210323_dt1_vs_bk15_2. In that posting we saw BB noise spectra which matched scaling from BK15 quite well. However, as discussed in that post it was not clear why the agreement would not be even better since both calcs are using the same NETs (from this google spreadsheet tabs "Pole SATs" and "BICEPKeckComparisons"), and the DT sims are applying efficiency factors which are supposed to be the ones which are implicit in the BK15 \(N_\ell\)'s. See this google spreadsheet tab "Observation_Efficiencies", and this google doc for some reasoning behind these numbers and links. The BK15 scaling is described in 20210303_dt1_vs_bk15.
Since the 20210323_dt1_vs_bk15_2 post Reijo and Andrea have produced an update to their noise maps with files dated Apr 22. The changes from last time include a small additional penalty factor intended to account for the non-uniformity of detector NET's (see 20210318_bk15_obseff). In addition focal plane thinning was being applied in the DT sims for higher frequencies which affected the residual \(1/f\) in polarization - this thinning has now been effectively turned off.
Reproducing Fig. 2 from the previous post we get the below. With only one realization of the DT sim the sample variance is large at low \(\ell\). However, it looks like the \(1/f\) is quite well reproduced in the lower bands, but is under produced in the higher ones. Note that agreement is expected between the blue and the red below. The Green curves which are taken from the PBDR table 2-1 (same as the program level requirements doc) also come from BK15 scaling but use (for example) older NETs.
Fig 2:
Measuring the level of the spectra in the range \(800<\ell<950\) and converting from power to temperature units we get the below. The \(T\) panel should not be taken seriously since the BK15 scaling with \(N_\mathrm{det}\) simultaneous is clearly over optimistic in this case as shown in 20210330_noise_int_down_t and 20210406_noise_int_down_B32017jack02.
Fig 3:
It is a bit hard to see the relative levels above so the ratios are plotted below. It is not clear to me why the 220GHz point is more discrepant? Remember that the same NETs are being assumed in both calcs so it must be a difference between the efficiency factor which is implicit in the B15 \(N_\ell\)'s and the efficiency factor used in Reijo's sims.
Fig 4:
As already mentioned the BK15 scaling is expected to under predict low \(\ell\) \(T\) noise. The measurement requirements no longer include SAT \(T\) noise numbers. Below is the plot showing \(TT\) spectra. It is not clear to me if the DT sims are consistent with BK performance for \(T\) - we should run the machinery for BK15 config - mainly \(N_\mathrm{det}\) simultaneous - to find out.
Fig 5:
