Nowadays, CubeSat missions grow more and more complicated. Such tasks as telecommunication, Earth observation and astronomy attract attention of nanosatellite developers. One of the main requirements for the success of a complex mission is the precision and reliability of satellite's attitude determination and control system. Better pointing accuracy and better stabilization may be achieved by using a star tracker (ST) as a main attitude sensor. Since it's method of operation is based on capturing images of stars, star tracker can provide a pointing accuracy better than 1 angular minute. In the last couple of years several laboratories and companies performed huge work on star tracker miniaturization, designing and delivering first prototypes that comply with size, mass and power restrictions of 3U CubeSats. Newly developed miniature star trackers while preserving core functionality are noticeably different compared to existing large-sized star trackers. The differences might be found in their optics, image sensors, algorithms and processing hardware. Newly developed miniature star trackers have a set of hardware similar to a modern smartphone. At the same time fast improving application program interfaces (API) of smartphone operating systems give developers today a better control over the smartphone internals. It becomes possible to implement a complete star tracker algorithm in a form of a smartphone application. During the tests, even with an expected overhead of OS stack, lost-in-space task was solved in less than one second. That defined the choice of a smartphone as a hardware platform for star tracker performance study. In this work we analyze the performance of a polestar algorithm for autonomous attitude determination. By changing capture parameters, such as sensitivity, resolution etc. the flaws and bottlenecks of the algorithm are exposed. Subsequently, algorithmic and hardware solutions are proposed to mitigate discovered performance losses.