Recently, a novel scanning principle (ESAD - Extended Shear Angle Difference) has been presented for the ultra-precise determination of slope and topography of near-plane surfaces which is based on the measurement of differences of reflection angles with an autocollimator. It uses large lateral displacements (shears) of the order of mm to cm, and a newly developed algorithm allows the exact reconstruction of the slope from the angle differences with two different shears. The topography is then derived from the slope by an accurate integration procedure. Reproducibility of the reconstructed topography of the order of below 1 nm for scans over a near-flat surface has already been achieved. In contrast to interferometric techniques, this scanning method does not rely on an external reference surface of matched topography. The measurands are directly traced back to the base units of angle and length. Therefore, this technique is well suited for creating a very precise standard for straightness and flatness. A central element considerably influencing the overall uncertainty is the autocollimator itself, making a high-accuracy traceable calibration of the device necessary. Autocollimators are calibrated at the PTB by the use of a primary standard angle comparator which is traced back to the natural standard of the full angle of 2Pi rad. The traceable calibration is realized by an incremental circular scale, incorporated into a highly precise rotary table as a measuring system, consisting of a radial optical phase grating and several photoelectric reading heads, thus reaching an angular resolution of approximately 0.001 arcsec. With this comparator, calibration of an electronic autocollimator with 0.001 arcsec resolution and a measurement range of (150 arcsec with an uncertainty of 0.007 arcsec has currently been achieved. The autocollimator type to be used in the ESAD scanning facility was calibrated and investigated with respect to its resolution, reproducibility and accuracy in the range of 0.01 arcsec over a variety of measurement ranges. These experiences will be described. Methods to avoid or control error influences are discussed.