Optical diffraction tomography (ODT) is a non-invasive method for quantitative measurement of micrometre-sized samples. In ODT a series of multiple holograms captured for various illumination directions with respect to a sample is processed using a tomographic reconstruction algorithm. The result of tomographic evaluation is 3D distribution of refractive index. Data acquisition in ODT is commonly realized in two ways, either by rotating a sample under fixed illumination and observation directions (object rotation configuration - ORC), or by scanning the illumination direction of a fixed sample (illumination scanning configuration - ISC). From the purely theoretical standpoint, the ORC configuration is superior to ISC due to larger (in terms of volume) and more isotropic optical transfer function. However, the theoretical maximal resolution achievable with ORC is lower than that provided with ISC. Moreover, the quality of tomographic reconstructions in ORC is significantly degraded due to experimental difficulties, including problematic determination of location of the rotation axis. This applies particularly to displacement of the rotation axis from the infocus plane that is either disregarded or detected with object-dependent autofocusing algorithms, which do not provide sufficient accuracy. In this paper we propose a new ODT approach, which provides solution to the both mentioned problems of ORC – the resolution limit and the rotation axis misalignment problem. The proposed tomographic method, besides rotating a sample in a full angle of 360°, uses simultaneous illumination from two fixed, highly off-axis directions. This modification enables enlarging the ORC optical transfer function up to the ISC limit. Moreover, the system enables implementation of an accurate, efficient and object-independent autofocusing method, which takes advantage of the off-axis illumination. The autofocusing method provides accurate and reliable detection of axial location of the rotation axis, enabling precise alignment of the tomographic data.