Equidistant backward interpolation from the comoving real grid onto the inertial real grid using the two fine grids. Real knots (•) are marked on the real and the fine grids; virtual knots (°) are marked on the fine grid. The inertial real knot i is projected onto the inertial fine grid using the map array (m_C), and corrected for the shift s to obtain the related nearest-right neighbor n_C in the comoving fine grid. If n_C< 1 or n_max ≤ n_C, then constant extrapolation is done. If n_C is an internal knot, then we set n_L = n_C−1 and n_R = n_C and check whether they both are real. If not, we find the nearest real knots using prior_real (indicated by the solid orange curve) or next_real (indicated by the orange dotted curve). Then n_L and n_R are projected onto the comoving real grid to get the corresponding knots j and j + 1. The desired l.h.s. ξ (sloped solid black line) and r.h.s. 1−ξ (sloped dotted black line) weights are computed using n_L, n_C, n_R, and 4.