II. CONSIDERACIONES PRELIMINARES
2.1. Sobre la importancia de la Inspección General del Trabajo
The station clocks (frequency standards) are used to generate the UTC time- tags for the observations used in the correlation. These time-tags need to be extremely accurate in order to be able to correlate the observations and model the delays accurately. Although the frequency standards at the stations are nor- mally highly accurate hydrogen masers, these clocks are usually independent of one another and can exhibit relative drifts and jumps. The delay model needs to account for these variations in the station clocks. Synchronization of the station clocks can be monitored to some extent by utilizing the frequency information from GPS satellites. However, the accuracy of GPS clocks is not sufficient to realize the frequency standard for the VLBI observations. Typically the station clocks are modelled as quadratic functions (sometimes appended with piece- wise linear modelling) during the observation interval. The clock delay on the
3.5 VLBI delay model 55
baseline between telescopes 1 and 2 is then given by
τc= a0+ a1(t − t0) + a2(t − t0)2, (3.32)
where the coefficients a0, a1, and a2are the constant, linear, and quadratic terms
of the clock polynomial, respectively. The clock performance can be divided into two parts: stability in time and stability in frequency. A highly stable clock can exhibit large drifts over time while still having small error in frequency. This behaviour must be accounted for in different ways depending on the length of the observing session, by e.g. estimating the clock polynomial for smaller intervals. The clock delay constitutes one of the largest contributions to the total observed delay, thus it is important to handle it appropriately as unmodeled clock behaviour will propagate into the estimated geodetic parameters (Sovers et al., 1998).
Chapter
4
VLBI data processing
VLBI data processing is a procedure that involves multiple steps in which geode- tic parameters and physical phenomena are estimated and investigated using the VLBI observables as input. The correlation and the related post-processing also involve various steps of processing and analysing the data. In the context of this chapter VLBI analysis refers to the analysis of post-processed VLBI data which is performed with a VLBI data analysis software.
VLBI data analysis is an intricate subject which involves modelling a com- plex set of phenomena related to signal-propagation, geophysical processed on Earth, and the evolution of geodetic parameters in time. It is usually in the VLBI analysis software where models are implemented to test and validate different hypotheses and experimental configurations. Furthermore, VLBI analysis soft- ware are also used by the correlator to compute the theoretical delays, which are needed as a priori values in the correlation process to align observations from the telescopes in the network (McCarthy and Petit, 2004).
Historically the available VLBI software have always had components that require human interaction in order to reach the highest accuracy in the geodetic products. Automation and simplification of the VLBI analysis procedure can be said to always have been a subject of interest within the software develop- ment. However, with the upcoming VGOS observations with greatly increased data quantities and continuous operations, the automation of the analysis pro- cess has moved from being an object of technical interest into a necessity. While many of the VLBI software packages offer good possibilities in processing the sessions in a semi-automatic fashion, they still require human input or valida- tion in some of the key steps in the analysis chain. These include e.g. resolving ambiguities, detecting anomalous clock behaviour (clock breaks), or establish- ing whether the station had problems during observations, which are stated in the correlation report. Real-time analysis also introduces the aspect of avail- ability of external data. VLBI analysis typically requires some a priori informa-
58 VLBI data processing
tion about station positions, EOP, instrumental delays, and weather conditions. Real-time processing means either these values have to be supplied in real-time (could be feasible for instrumental delays) or estimated from prediction mod- els. In some cases even if the data, such as meteorological data, are available, they need to be amended by parameters that are generated e.g. via Numerical Weather Models (NWM). Since VLBI is the primary method used to determine the EOP the a priori information needs to be predicted or modelled to some degree. Thus it is necessary to investigate which factors are crucial for the so- lution quality when the time resolution and accuracy of the available external information is limited.
In a multi-layered system changes in the observation equipment may cause hard to model, unexpected biases or changes in instrumental precision. These changes can be investigated on an instrumental or system-wide levels. Even though the technical specifications of the components are normally known with exceptional detail it is important to also investigate whether there are any un- expected effects on the estimated parameters. In 2011, as a part of the prepara- tion in the transition to VGOS, the Onsala Space Observatory installed a Digital Base-band Converter (DBBC). It was run in parallel with the then operational system consisting of the analogue Mark 4 rack. During a three-year period these systems were run in parallel to collect data with DBBC while the old and proven analogue system remained the operational system. During this period the per- formance of the DBBC could be assessed by running local zero-baseline tests. Additionally, the sessions recorded in parallel could be compared in order to investigate possible discrepancies in the observed delays and geodetic parame- ters.
These issues connected to the general process of transitioning into VGOS compliant operations are studied in Papers I and II appended to this thesis. The transition into operational use of DBBC at OSO is investigated in Paper I, where the results from the DBBC-analogue comparison are presented. The near-real time automated observation and analysis aspects are investigated in Paper II, presenting the results from investigation of the requirements for a priori infor- mation for automated near-real time analysis of IVS-INT1 sessions.
This chapter will give a general introduction to the various aspects of VLBI data processing such as the analysis software packages, VLBI data formats, aux- iliary data, steps in the analysis process, differences in analysis approach when analysing different session types (e.g. Rapid-turnaround and INT-sessions), and automation. Moreover, the specific issues related to the VLBI analysis proce- dures, associated software, and analysis approaches used for the results of these two papers are discussed in more detail.