The optimal power ow problem (OPF) aims at constructing optimal op- erating regime of a power system that minimizes generation cost or network losses subject to engineering and physical network constraints. The OPF prob- lem is a part of System Operator's daily routine, however most of the industrial solvers use the linearized (so-called DC) formulation. Original quadratic formu- lation (AC OPF) is more precise, but harder to solve. Among dierent optimiza- tion approaches we investigate semidenite relaxations [1], [2] that demonstrated their eciency for the number of industrial benchmarks. However, for transmis- sion networks there is no theoretical validation of the conditions that guarantee the exactness of relaxation. Motivated by the simple illustrative example [3], we investigate the situations when the relaxed formulation is unable to retrieve feasible solution.