Abstract: The exploitation of GPS differential technology for engineering surveys differs in anumber of respects from more conveational geodatic and cadastral applications. Among theprincipal considerations are the greater accuracy and reliability requirements. To satisfythese demands further improvement of existing data processing methods becomes necesary,which leads to this investigation. In this study different models are utilized to processGPS carrier phase observations, which were acquired from a test network with interstationdistances ranging from 0.6km to 30km using GPS receiver TI——4100. The experimentsshow:——the standard deviation of double differnce of phase observations is dependent ofthe baseline length;——introduction of the mathematical correlations among the double differences maynot improve the results;——the dual frequency correction for the ionospheric effect gives worse results thannot applying the correction at all;——careful modelling of tropospheric delay is of much importance for accurate base-line determination, especially for height component;——the solution will be greatly strenghened if the estimated ambiguities are properlyforced to iategers.