GPS is often used in local tie observations. Precision of GPS measurement is usually evaluated by looking at repeatability from time series of baseline components. It is, however, not enough for local tie observations which require high accuracy in absolute sense. In this study, two kind of error sources are examined for absolute length measurements by GPS.

The error by difference of antenna phase characteristics are evaluated by using several Choke ring antennas of the same type. The observations were carried out by putting the antennas on the tripods which are set within a distance of about 10m, and by rotating and swapping the antennas. The phase center variation (PCV) of antennas relative to a reference antenna is estimated by BERNESE software. The results shows that baseline solution changed by the rotation and swapping of the antennas by 1 mm level even if using the antennas of the same type. The situation can be improvement to 0.3 mm level by applying the PCV correction.

The scale error caused by ionosperic delay is usually neglected for short distance up to 1 km. Approaching to next period solar maximum, the TEC value have been reached to be a few TECU which should cause a few mm scale error for a 1 km baseline. The use of Ionosphere free linear combination has disadvantage to amplify measurement noises. Use of L1 solution with ionosphere model correction is realistic solution. The results of a test analysis by applying global ionosphere map produced by IGS analysis centers and that by applying regional ionosphere map derived by surrounding GPS station shows clear improvement of the bias.