There are two bugs in the CAR implementation that cause the effective sky fraction for CAR to be different from that in the healpix implementation, thus causing them to pull different noise curves from the SO noise model. The bugs are:
The default sky fraction calculation assumes equal area pixels. For CAR, one can switch this to a calculation that sums over a map of the pixel areas in the unmasked region.
When converting an existing CAR hitmap to a different geometry, it uses enmap.project which by default uses order=3 interpolation. This should be switched to order=0 (which is what the original healpix -> CAR and a healpy.ud_grade operation does).
There are two bugs in the CAR implementation that cause the effective sky fraction for CAR to be different from that in the healpix implementation, thus causing them to pull different noise curves from the SO noise model. The bugs are: