Comparison of CAMB with CMB fast

[Daniel Grin]

I am modifying cosmological recombination histories to include new atomic physics affects, and as a test case decided to compare CMBfast with CAMB with a standard recombination history.

I made sure that I am using the same values of all densities, the Hubble parameter, neutrino parameters, lmax, and (k\eta)_{max}. I used the more up-to-date version of RECFAST in the most recent CMBFAST version and force-fed it (through a file output and read) to CAMB, to do a 'fair comparison', with both codes using the same recombination code.

I get a 1% discrepancy between my TT and EE power spectrum (scalars only), unlike the quoted discrepancy on the CAMB web site of 0.1%-0.3%, when CAMB runs with the highest possible precision settings.

This discrepancy seems to hold for the COBE-unnormalized LOS integrals.

Has anyone run into a similar problem, and if so, has anyone identified the origin of the discrepancy?

[Antony Lewis]

Maybe difference in optical depth -> z_re mapping?