For the calibration of vacuum gauges in the range of medium and high vacuum, the method of static gas expansion is widely adopted as high-accuracy primary method. In this method, gas is expanded from a small vessel into a large one, whereby the pressure is reduced according to the volume increase. Uncertainties of the generated pressure arise from uncertainties of initial pressure, volume ratio of both vessels, real gas behavior, and thermal effects. As comes out from the uncertainty budget of a conventional expansion apparatus, the dominant contribution stems from temperature effects. For example, a typical temperature uncertainty of 0.3K gives a relative pressure uncertainty of 0.1% . In order to obtain a substantially improved accuracy, the whole apparatus was immersed in a circulated, temperature controlled water bath. This measure resulted in a tenfold improved temperature stability of the vessel walls, i.e. better than 0.03K. Furthermore, thermal effects caused by the thermodynamic process of gas expansion were investigated. In fact, the gas in the small vessel cools by as much as -200K. As our experimental and theoretical studies reveal, the thermal equilibrium between the gas and the vessel walls is reached with time constants of the order of seconds, depending on gas species and pressure. Accordingly, after a typical waiting time of about 1 minute after filling the small vessel and after expansion into the large vessel, the equilibrium is almost fully achieved even in worst case. Other disturbing effects were also investigated in detail. The uncertainty budget of the generated calibration pressures (2@sigma@-level) gives a total uncertainty below 0.1% in the pressure range 0.1-to- 10 mbar (single step expansion) and below 0.15% in the pressure range 0.001 to 0.1 mbar (two step expansion).