In L815 in grizli.multifit.GroupFLT, flt.compute_model_orders effectively passes x=None, y=None when extracting by ID. This means the centre of the object is recalculated for each FLT - is this actually the desired behaviour? This already causes issues if a number of pixels contain negative flux (gbrammer/grizli#299), but may lead to small shifts in the wavelength solution for each blocking filter.
Should test with a handful of bright objects (in e.g. Par028) to see whether fixing the object centre in all filters changes the extracted 1D spectrum.
In L815 in
grizli.multifit.GroupFLT,flt.compute_model_orderseffectively passesx=None,y=Nonewhen extracting by ID. This means the centre of the object is recalculated for each FLT - is this actually the desired behaviour? This already causes issues if a number of pixels contain negative flux (gbrammer/grizli#299), but may lead to small shifts in the wavelength solution for each blocking filter.Should test with a handful of bright objects (in e.g. Par028) to see whether fixing the object centre in all filters changes the extracted 1D spectrum.