In various fields of neutron scattering there is a tendency to use smaller and smaller samples. There are various reasons for this, e.g. the limited size in high pressure cells, the restrictions given by growth methods of thin films, or the impossibility to grow larger single crystals. With conventional guides this leads to the situation that a white beam with some 50cm 2 cross-section and a broad divergence is to illuminate a sample of some mm 2 area. Thus more than 99% of the neutrons leaving the guide are not needed and cause background and radiation problems.It is suggested to change the order of the optical elements and the design of the guide section to filter neutrons not intended to hit the sample as early as possible. As an example a set-up for specular reflectivity on small samples is presented. A double monochromator some meters behind the source cuts away all neutrons of the wrong wavelength even before they enter the guide. The guide itself is one branch of an ellipse. It maps the divergent beam from the monochromator to a convergent beam at the sample position. An entry aperture at the first focal point, a bit larger than the sample, guarantees that just enough neutrons enter the guide to bath the sample. There is no direct line of sight to the source and the guide ends far away from the sample position, so that there are only few spacial restrictions.Detailed McStas calculations and a design study for a down-scaled test device, both for reflectometry and diffraction, are presented.