AbstractThe DORIS (Doppler Orbitography and Radiopositioning Integrated by Satellite) data are processed in the framework of the International DORIS Service (IDS) at several analysis centers. Each analysis center follows its own processing strategy and models. This manuscript presents the results of the processing based on a development version 5.0 of the Bernese GPS Software at the Geodetic Observatory Pecný analysis center (GOP). The complete period 1993.0–2009.0 was processed applying the free-network approach in order to estimate main parameters namely station and polar motion coordinates. A significant improvement has been achieved in the estimated station coordinates and polar parameters by processing the data from the satellites equipped with the second generation of the DORIS receiver, SPOT-5 and Envisat, launched in 2002. The RMS of polar coordinates in 2003.0–2008.0 shows a decreasing trend over the entire analyzed time period. The transformation parameters between the DORIS solution and ITRF2005 were subjected to a spectral analysis, confirming the domination of the annual and semiannual periodicity. The behavior of the terrestrial reference frame scale is quite stable with a few exceptions. The analysis of a major scale shift at the end of 2004 revealed SPOT-5 and Envisat satellites as the source of the problem. However, the termination of the TOPEX/Poseidon DORIS data processing at the end of 2004 did not influence significantly the overall scale level.Another objective of the paper is a detailed analysis of relations between the value of the observation residuals and the length of the observation Doppler count interval. A simple empirical model considering the observation noise as a sum of the constant and time-dependent terms is propounded and discussed. The estimated troposphere total zenith delays are compared to the corresponding values derived from GNSS (IGS PPP products) and the source of the differences as well as their systematic behavior and dependence on several factors are discussed. A significant DORIS–GNSS bias was detected when processing SPOT-5 data for the South American stations Santiago, Cachoeira Paulista and Arequipa. The location of these stations is close to the South Atlantic Anomaly (SAA). An important effect of SAA on the DORIS observation noise is well known from Jason-1 data, but it has not been detected for any other satellite yet. Additional tests of the estimated station coordinates and the beacon frequency offsets, based on single-satellite solutions, confirm the possible relation between the SPOT-5 bias and the SAA.