Very simple code to compute the pesky normalization factor in super-sample covariance w/ trapz rule for a simulation box
where
which is appropriately normalized to
Run this code with julia sigmab.jl
The integration range and number of points appear to be decent at a
Default ouput on my machine is:
Is the window close to 1? true
Power spectrum variance in a cubic box of side length 625.0 Mpc/h is 6.756502209257932e-5.
This is in close agreement with directly computing the variance of linear field on a grid but has the benefit of being faster and not requiring seed averaging.