This paper analyzes the errors entailed by the signal amplitude inequality when measuring the phase shift among low frequency harmonic signals for algorithmic sum-difference methods. It is shown that additional phase shifts of summary and difference signals caused by the inequality of compared signals in some methods affect the errors, whereas in others they do not. A method in which there is no additive component of the error has been found. This method has been admitted to be the best to measure small phase shifts. The method for measuring the phase shift, which enables to avoid errors due to amplitude inequality of compared signals has been proposed.