Winter 2021 design tool run

[zonca]

Progress to release:

• finalize instrument config
• fix observing efficiencies
• fix atmosphere scalings
• verify telescope years in 02_atmosphere
• verify tube years in 03_noise
• verify hitmaps and wcov in 04_hitmap_wcov
• run SAT
• run LAT

Verification:

• verify SAT foregrounds
• verify SAT CMB
• verify SAT atmo noise
• verify SAT atmo noise splits
• verify LAT foregrounds
• verify LAT CMB
• verify LAT atmo noise
• verify LAT atmo noise splits

For later:

• run beyond vanilla cosmology
• run Chile SAT

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[zonca]

@keskitalo I am verifying the TOAST hitmaps, if I load:

'/global/cscratch1/sd/keskital/s4sim/reference_tool_round_2/out/00000000/pole_noise_SAT_HFS1_telescope_all_time_all_hits.fits.gz'

and I check sampling frequency with:

sampling_rate = input_map.sum() 
sampling_rate /= 10 * 24 * 3600 * 0.642
sampling_rate /= 8592 # HF detectors per tube
sampling_rate *= 8 # focal plane thinning factor

I get 16.8 Hz instead of 20 Hz. Is this expected?

[keskitalo]

@zonca That is roughly consistent with the turnaround time. A sweep takes less around 30 seconds at 1.5deg/s and the turnaround is 3.1 seconds. However, that particular loss of observing efficiency is already included in Sara's overall observing efficiency factors. If you can scale it out from the noise map, that would simplify the interpretation.

[zonca]

in the previous round of simulation this effect was not present, something changed in flagging?

[keskitalo]

In the previous round we scanned slower and used a higher acceleration in turnarounds. The new numbers match BICEP/Keck parameters.

[zonca]

@keskitalo ok, can you write down exactly how do I compute the scaling parameter and how do I scale it out from the noise maps? also what about atmosphere?

[keskitalo]

You can replace the scheduling efficiencies I gave you by:
eff = nhit / (10 * 86400 * fsample). You can do this for all cases. This of course means that you cannot apply the sampling rate check anymore, except to check for rough agreement. I would apply the same efficiency to both atmosphere and noise. The one place where the efficiencies could diverge is applying the yield factor but that is well within our atmospheric calibration uncertainty.

[zonca]

@keskitalo I also needed to consider detectors per tube, the thinning factor and the simulated tubes.
I get:

Pole ULFPL1 Efficiency from hitmap 0.71, scheduling efficiency 0.75, ratio 94.1%
Pole LFPL1 Efficiency from hitmap 0.71, scheduling efficiency 0.75, ratio 94.1%
Pole LFPL2 Efficiency from hitmap 0.71, scheduling efficiency 0.75, ratio 94.1%
Pole MFPL1 Efficiency from hitmap 0.71, scheduling efficiency 0.75, ratio 94.1%
Pole MFPL2 Efficiency from hitmap 0.71, scheduling efficiency 0.75, ratio 94.1%
Pole HFPL1 Efficiency from hitmap 0.71, scheduling efficiency 0.75, ratio 94.1%
Pole HFPL2 Efficiency from hitmap 0.71, scheduling efficiency 0.75, ratio 94.1%
Pole LFS1 Efficiency from hitmap 0.54, scheduling efficiency 0.64, ratio 84.3%
Pole LFS2 Efficiency from hitmap 0.54, scheduling efficiency 0.64, ratio 84.3%
Pole MFLS1 Efficiency from hitmap 0.54, scheduling efficiency 0.64, ratio 84.3%
Pole MFHS1 Efficiency from hitmap 0.54, scheduling efficiency 0.64, ratio 84.3%
Pole MFLS2 Efficiency from hitmap 0.54, scheduling efficiency 0.64, ratio 84.3%
Pole MFHS2 Efficiency from hitmap 0.54, scheduling efficiency 0.64, ratio 84.3%
Pole HFS1 Efficiency from hitmap 0.54, scheduling efficiency 0.64, ratio 84.3%
Pole HFS2 Efficiency from hitmap 0.54, scheduling efficiency 0.64, ratio 84.3%
Chile LFL1 Efficiency from hitmap 1.04, scheduling efficiency 1.04, ratio 99.5%
Chile LFL2 Efficiency from hitmap 1.04, scheduling efficiency 1.04, ratio 99.5%
Chile MFL1 Efficiency from hitmap 1.04, scheduling efficiency 1.04, ratio 99.5%
Chile MFL2 Efficiency from hitmap 1.04, scheduling efficiency 1.04, ratio 99.5%
Chile HFL1 Efficiency from hitmap 1.04, scheduling efficiency 1.04, ratio 99.3%
Chile HFL2 Efficiency from hitmap 1.04, scheduling efficiency 1.04, ratio 99.3%

Do you confirm it is reasonable?

[keskitalo]

Yes. Looks reasonable. The CHLAT scan is very wide, making the turnaround loss less significant.

[zonca]

ok, thanks @keskitalo , last thing I am checking the white noise covariance matrices,
these tests were passing fine in the last release, but are off by a factor of 10 now,

https://nbviewer.jupyter.org/gist/zonca/613386a7a90330e27afedd3a82cd5cbe

starting at cell 56. can you please take a look? I try to estimate the channel NET from the white noise covariance matrix
and I get 7093 instead of 610 uK sqrt(s)

[keskitalo]

You seem load hits in cells 17 and 57 but use the hits from cell 17 in 63. Could that be the issue?

[zonca]

no, 57 is the scaled hitmap, I am doing the check just on the simulation so it is right I use the raw hits from 17

[keskitalo]

Thanks. I found the issue: https://github.com/hpc4cmb/toast/blob/master/src/toast/todmap/filterbin.py#L562-L571

I'll scale the SAT noise matrices. The rest are fine. Will be done in 30min.

[keskitalo]

The SAT wcov files are now fixed.

-R

[zonca]

can you please fix permissions?

[keskitalo]

Done!

[zonca]

thanks @keskitalo all tests pass,
I have run the SAT and notified the simulations mailing list, now started verification,
I have the gnomviews of the foregrounds, https://github.com/CMB-S4/s4mapbasedsims/blob/master/202102_design_tool_run/plot_gnomview/README.md, the only thing that stands out to me is some boundary effects on LFS2 in polarization, also in HFS1 and HFS2 in the bottom there is a small gap in coverage?

[keskitalo]

Those seem benign and must follow from gaps in the pixel distribution and the fact that we observe at fixed elevations and boresight rotations.

[zonca]

@keskitalo @cbischoff I've completed verification of SAT atmo+noise,
there is some variability compared to DSR, can you check it is all expected?

https://github.com/CMB-S4/s4mapbasedsims/blob/master/202102_design_tool_run/plots/SAT.md

[zonca]

run completed, will add Chile SAT and beyond-vanilla cosmology later on.

see https://github.com/CMB-S4/s4mapbasedsims/tree/master/202102_design_tool_run

updated the docs of the design tool itself https://cmb-s4.github.io/s4_design_sim_tool/

releasing 1.1.0