DSR Sim Map Set 06 - some spectra and map check plots
2019 Nov 1 - Clem Pryke
Summary
The first 10 realizations of set 06 (DSR style) have been generated
and are transfering to NERSC.
The next step is for someone to run foreground cleaning on the
high res bands and produce cleaned maps for input to lensing reconstruction.
Delensers are then invited to run their best algorithms on the result
and provide lensing templates similar to those which were provided
here.
These can be taken through to power specra and ML r values similar to what was
done here and
here.
Spectra plots for "r bands"
In 20191016_dc06_dsr the setup for
sim set 06 was described.
Here are some check plots of the first realization.
06b is "Pole deep",
06c is Pole wide",
06d is "Chile deep".
We are using two foreground models: 07 amplitude modulated Gaussian dust+sync, and
09 Vansyngel model.
Below are component spectra of the "r bands". A few notes:
- Color order is red-yellow-green-blue lowest to highest frequency.
- In all the column we see that \(\ell<30\) has been cut out of
all components to simulate loss of these modes to timestream filtering.
- Noise is plotted as \(C_\ell\) so we can see the white + \(1/f\) form
which is being assumed.
Higher ell knee for 20GHz since this is assumed to be on LAT.
- In lLCDM column we can see beam size getting smaller from 30 to 270GHz
with 20 GHz (LAT) beam size similar to 270 GHz (SAT).
Simce these are simple anafast spectra we see E to B mixing causing
an aparent excess over the lensing B at low \(\ell\).
- In the 07 amplitude modulated Gaussian dust+sync column we can see
foreground getting lower with increasing frequency and then higher again.
- In the 09 Vansyngel model column we see a big bump at low \(\ell\).
This is an artifact caused by applying the \(\ell<30\) cut to a
galactic map which has a very bright galactic plane.
Even cutting out the plane with an \(l<10\) deg cut before going to
harmonic space there is still aliased power at the cutoff which spreads
over the whole sky.
At high latitude it dominates over the actual high latitude structure
which is in the model.
We can see this in the maps below.
Suggestions for how to reduce or elliminate this problem are welcome.
- In all cases each hit pattern has spectra taken using its "natural"
mask - just the hit pattern itself.
For "Chile deep" one would throw out the high foreground parts in
reality.
Fig 1: "Pole deep"
Fig 2: "Pole wide"
Fig 3: "Chile deep"
Spectra Plots for Delensing Bands
There are six high res delensing bands:
Fig 1: "Pole deep"
Fig 2: "Pole wide"
Fig 3: "Chile deep"
Map Plots
For what they are worth here are some map plots - plotting such small maps leads to
heavy pixel aliasing.
All have \(\pm\)100\(\mu\)K for T and \(\pm\)5\(\mu\)K for Q/U.
Fig 4: Common full sky components: LCDM and the two foreground models
Fig 5: Noise for the three hit patterns
Fig 6: Combined for the three hit patterns