In Pedersen et al. [1] we proposed a method to utilize a temporally dependent superlattice potential to mediate spin-selective transport, and thereby probe long and short range magnetic correlations in optical lattices. Specifically this can be used for detecting antiferromagnetic ordering in repulsive fermionic optical lattice systems, but more generally it can serve as a means of directly probing correlations among the atoms by measuring the mean value of an observable, the number of double occupied sites. Here, we provide a detailed investigation of the physical processes which limit the effectiveness of this "conveyer belt method". Furthermore we propose a simple ways to improve the procedure, resulting in an essentially perfect (error-free) probing of the magnetic correlations. These results shows that suitably constructed superlattices constitute a promising way of manipulating atoms of different spin species as well as probing their interactions. PACS numbers: 37.10.Jk, 73.21.Cd, 74.25.-q, 75.50.Ee arXiv:1509.00644v2 [cond-mat.quant-gas] 4 Sep 2015