Discrepancies between input and output maps in noiseless simulations (scanmap and conviqt)

[matteobilli]

Comparison with scanmap
I noticed some differences between the input CMB maps and the output maps generated with scanmap. Considering the CMB intensity map, the absolute difference is at the level of 1e-5 uK.

Setup and Configuration
As input map I used a CMB map smoothed with a Gaussian beam with FWHM=1deg, nside=256, no noise, TQU maps with Q and U set to 0.
I have attached the following files:

• The scanning strategy schedule. schedule_2days_deep56.txt
• The script used to run the simulation. so_sim_scanmap.sh

Simulation Parameters:

• nside = 256
• lmax = 512
• wafer = 'w16' ## LAT wafer
• thinfp = '25' (Thinning of the focal plane -- 36 pairs)
• fsample = '100 Hz'

I was wondering if a difference of 1e-5 uK is expected behavior. Since I am running the simulation without noise, I expected the residuals to be much smaller (closer to numerical precision).

Comparison with conviqt
I also performed simulations using conviqt instead of scanmap. Same schedule and input alm extracted from the map used to make the comparison. Here the script used to run the simulation. salloc_so_sim_conviqt.sh

In this scenario, I am observing a difference between output and input on the order of ~1-2 uK, assuming the same Gaussian beam for every detector as input for conviqt (ideal case without systematics).

Are the magnitudes of these residuals (output vs input) considered normal for each tool respectively, given the noiseless setup?