camb vs cosmomc matter power spectrum

[Paolo Serra]

Hello,
I need to calculate the linear matter power spectrum for very small scales;
the output from camb (in output_matterpower.dat) is the same as the one in cosmomc
(given by the variable mpk_lin(i)=MatterPowerAt(Theory,k_scaled(i))/a_scl**3 in mpk.f90),
but they start being different on very small scales, starting from k~3, even though the cosmology used is the same. Any ideas? Maybe the variable "mpk_lin" is not the correct one to look at?
Thanks
p

[Antony Lewis]

Possibly a k range issue - see the settings at the top of cmbtypes.f90 in cosmomc/source. cosmomc will do log-linear extrapolation if out of the computed range, which is obviously only OK if in small tails of window functions.