Problem and solution to deal with high biased sources

[fjaviersanchez]

Hi @damonge,

When I try to generate sources with bias ~2 or higher with a relatively high density (~30 mill sources total in full-sky) the sources tend to accumulate in certain positions so the resulting map is not correct (here I am showing a masked version of the map):

This is solved when I use some gaussian smoothing following the suggestion by @slosar. In this case the smoothing sigma is 5 Mpc/h which is smaller than the resolution of the simulation, 6.2 Mpc/h.

I just wanted to leave this solution written for other users and we can try to solve this issue (in the case it can be totally solved) in the future.

[damonge]

Thanks a lot! This is the main motivation for the r_smooth parameter. I'll leave this open since other users may find it useful

[slosar]

I think the map above is not strictly wrong in the sense that the power spectrum averaged over many realisations will likely be correct.
What is funny is that you can get power spectra that don't look too bad, but the correlation function explodes at small r -- I think this is what we are seeing here.