GammaLib

Dear Jurgen,

we have tried to do some basic analysis with recently released 1dc GPS data, but encountered certain issues during the fitting step.

Namely, executing

ctlike inobs=p.evt inmodel=psrb1.xml outmodel=tst_1dc.xml chatter=4 debug=yes | tee -a log_min.log

in the attached Example folder (close-to-minimal working example) we get a lot of errors, similar to one below

*** ERROR: cta_nroi_diffuse_kern_phi::eval(phi=4.4915): NaN/Inf encountered (nroi=-nan, intensity=3.4392e-14, cos_phi=-0.219101, sin_phi=-0.975702)
*** ERROR: GCTAResponseIrf::npsf: NaN/Inf encountered (value=-nan, roi_radius=0.0872665, roi_psf_distance=0.0859906, r=[0,inf])
*** ERROR: GCTAResponseIrf::nroi: NaN/Inf encountered (nroi=-nan, theta=0.0859906, phi=0)

We’ve noticed, that changing the IRFs from 1dc to prod3b in p.evt file (=p_good.evt) makes ctlike to work fine.

Exclusion of the IEM background from psrb1.xml region model also makes ctlike to work even with 1dc IRFs.

We’ve tried also replace cube_iem.fits.gz file with our own “Map Cube” file based on galprop code, but this did not work and ctlike still failed with 1dc IRFs.

With the other diffuse/extended sources, e.g. Fermi Bublbes (=SpatialMap source type), or HESSJ1303 (=EllipticalGauss source type) ctlike works fine with 1dc IRFs (if no IEM is in the model)

Probably ctlike fails to work with new 1dc IRFs if a source with MapCubeFunction spatial model is present in ROI model.

May you take a look into this issue and let us know?

Cheers,
Denis and Masha