[CAMB] Tuning timesteps

Use of Healpix, camb, CLASS, cosmomc, compilers, etc.
Post Reply
Luke Hart
Posts: 42
Joined: July 13 2015
Affiliation: University of Manchester

[CAMB] Tuning timesteps

Post by Luke Hart » June 25 2018

So I am currently looking at the modules.f90 file and noticing that the timesteps during recombination are set by appropriate limits of the visibility function.

Adding small perturbations during recombination at different redshifts has meant I have had to change the size of the timestep in CAMB. I did an analysis on this, for a given width of these perturbations and found that you start to see the behavior of the CMB power spectra becoming less erratic the more finely tuned your steps are (which seems obvious right?)

The problem is that I have now, for efficiency reasons, tried to localize these changes. (i.e. if the perturbation is a smooth, small change near given redshift, then the finer settings are set for a small number of redshifts either side of the perturbation.) This has still led to erratic behavior in the power spectra at low l's.

So my question is: why is a coarser sampling of the time steps outside of my perturbation affecting the quality of my CMB output power spectra, even when the recombination history is the fiducial background model outside of this range? I've checked the visibility, the differential opacity and the opacity itself, [tex]\tau[/tex] and still, there are no noticeable deviations in these whether I use the finer precision in the whole recombination regime or just the precise epoch.

Please let me know when you can :)


Antony Lewis
Posts: 1394
Joined: September 23 2004
Affiliation: University of Sussex

Re: [CAMB] Tuning timesteps

Post by Antony Lewis » June 26 2018

The time steps affect the sampling for the line-of-sight integration, and possibly also indirectly on other accuracy-related parameters related to the step size. Why you are seeing effects mainly at low L I'm not sure off hand.

Post Reply