The output from the correlator was converted into the Flexible Image Transport System (FITS) format, using the B2FITS task of the Caltech VLBI package [Pearson 1991] and then taken into the National Radio Astronomy Observatory (NRAO) Astronomical Image Processing Software (AIPS) reduction package for further processing. The first processing task was to accurately determine the amplitude and phase of the fringes at each of the 2 second correlator periods. This process is known as fringe-fitting. Fringe-fitting algorithms search for and remove any residual fringe rate from the data so that the fringes can be integrated in time, allowing an accurate amplitude and delay to be obtained. The delay on the i - j baseline, 
, is directly related to the baseline phase, 
,
The amplitudes and phases determined by fringe-fitting are the (somewhat corrupted) real and complex components of the visibility function, V, in 
2.1.
The fringe-fitting task FRING in the AIPS package has been used to fringe-fit all of the SHEVE observations correlated at the Block II processor. During fringe-fitting, any information on the absolute phases of the visibility function was sacrificed in favour of increasing the sensitivity to weak fringes on some baselines. The powerful technique of global fringe-fitting was implemented to achieve this [Schwab & Cotton 1983]. In global fringe-fitting the residual phase rates are separated into antenna based components which satisfy closure relations around sets of three or more antennae. A rudimentary model for the source structure is therefore required for this method to proceed, a single point source being sufficiently accurate for most applications. Thus, from the detection of fringes on sensitive baselines the fringe rate can be predicted for the less sensitive baselines, giving a better chance for the detection of weak fringes on those baselines.
Any information on the absolute phase of the visibility function is lost since, with this method, the phase on a designated reference antenna is set to zero during each period that the fringe rate is evaluated. All baseline phases are determined relative to the reference antenna.
After fringe-fitting, the fringe amplitude and fringe phase relative to the reference antenna exist at each 2 second correlator output interval. For a 12 hour observation with 5 or 6 antennae this usually produced a file of the order of several megabytes in size. This FITS format file was then converted into the MERGE format, standard for the Caltech VLBI package tasks. To reduce the size of the dataset and to improve signal to noise, the data were integrated coherently over a 30 second period, well within the coherence time typical for SHEVE observations [King 1994]. This was achieved with the Caltech package task AVERAGE, using a coherent average and with errors derived from the scatter in the amplitudes and phases.
At this point the data were edited of bad visibilities. Editing was achieved within DIFMAP [Shepherd, Pearson & Taylor 1994] or with the Caltech task IED. Occasionally the task of editing on some baselines was a substantial one since the Rubidium frequency standards are not as robust at the higher frequency of 8.4 GHz as at 2.3 or 4.8 GHz.