order of C_ell different from HEALPIX!

[zonca]

Order of spectra in HEALPIX is .... TE, TB, EB:
http://healpix.jpl.nasa.gov/html/facilitiesnode7.htm

while the order in healpy is .... TE, EB, TB:
https://github.com/healpy/healpy/blob/master/healpy/sphtfunc.py#L79

I fixed the test case on anafast, see related commit.
What is strange is that healpy uses HEALPIX C++ which uses the same convention as the fortran code.

Cyrille, do you have an idea of what might causing this?

[crosset]

Hi,

The order of the spectra is not fixed by the healpix library, but is internal to healpy.

The cross power spectra are put into a symmetric correlation matrix and decomposed into Cholesky form (one for each ell). The matrix can be represented either in diagonal order (first diagonal, then second diagonal, etc.) or in row (=column) order. I chosed the diagonal order (TT, EE, BB, TE, EB, TB), which is the closest to the healpix order.

Note that few months ago, I prepared to standardize to this diagonal order all the related functions (map2alm, alm2map, anafast, synfast, ...), see dicumention of synalm https://github.com/healpy/healpy/blob/master/healpy/sphtfunc.py#L249 (before, the different functions had different conventions....). Actually, internally, the cholesky function still use the row order (hence the need for reordering function : https://github.com/healpy/healpy/blob/master/healpy/sphtfunc.py#L823)

I think we should keep this order, as it allows to compute the cross-spectra of any number of maps (and not necessarily polarization maps, see the 'pol' keyword in anafast : https://github.com/healpy/healpy/blob/master/healpy/sphtfunc.py#L40), or make simulation of correlated maps.

We can add a warning message to make it clear to the user.

What do you think ?

Cyrille.

Le 6 nov. 2012 ? 20:35, Andrea Zonca a ?crit :

Order of spectra in HEALPIX is ....TE, TB, EB:
http://healpix.jpl.nasa.gov/html/facilitiesnode7.htm

while the order in healpy is .... TE, EB, TB:
https://github.com/healpy/healpy/blob/master/healpy/sphtfunc.py#L79

I fixed the test case on anafast, see related commit.
What is strange is that healpy uses HEALPIX C++ which uses the same convention as the fortran code.

Cyrille, do you have an idea of what might causing this?

?
Reply to this email directly or view it on GitHub.

[zonca]

internally we use row order anyway, so I think it would be better to use HealPix order for input/output, as it makes more sense to list them in order of amplitude or importance, and better be compatible with HealPix.

[zonca]

see #687