The precise determination of the SAR (synthetic aperture radar) antenna pointing is an essential task initially performed during the commissioning phase of a spaceborne SAR system and is permanently monitored during the whole mission life-time. Besides a correct illumination of the scene during data acquisition, antenna pointing is required for proper compensation of the radiation pattern for radiometric correction during SAR data processing. The Amazon rainforest is a well-established target area for antenna pointing estimation in elevation as proven by many past and current SAR missions. Several new SAR systems are now proposed which are using long wavelengths, i.e., L- and P-bands, which will be implemented using reflector-based antenna systems. These reflectors have, in contrast to planar phased array antennas, no completely rigid connection to the satellite body and, hence, a more volatile antenna pointing. Due to the huge dimensions of such reflector antennas required for the envisaged long wavelengths and the finite stiffness of the boom, the antenna pointing may change significantly along the orbit. Such variation cannot be tracked using the common Amazon rainforest approach only, as this measurement opportunity exists only at two positions along the orbit (ascending and descending). Here, the performance of an alternative technique is presented which mitigates the influence of the underlying SAR scene by employing two coherent SAR datasets acquired simultaneously with different antenna patterns. This allows the use of amplitude and phase information for pointing estimation. No assumption upon the homogeneity of the underlying scene is required and, hence, pointing estimation becomes feasible at nearly any point along the orbit. This paper outlines the technique, describes simulation results and presents outcomes from first experimental acquisitions performed with the TerraSAR-X satellite.