The correlators at JIVE

Summary form only given. For many years, the heart of operations at JIVE has been the MkIV hardware correlator, a custom-built high-performance data processor. Initially tape reels were used, followed by hard disks; later the correlator was retrofitted to handle e-VLBI data streams in real time. At this time the Mk4 has been replaced by the EVN software correlator (SFXC), a locally developed software correlator that runs on a moderately-sized Intel-based computer cluster. This development has vastly improved the science capacity of the EVN, by providing far higher spectral resolution and polarization accuracy, but most notably, by enabling completely new observing modes. Observing multiple simultaneous field centers allows for wide-field imaging, while a phased-array mode has made it possible to do pulsar time series with the EVN. For this technique calibration data have to be fed back into the correlation process. A latest highlight is the combination of these two methods, allowing simultaneous time series for multiple objects. New algorithms have been developed for near-field VLBI, making it possible to focus on objects within our solar system. This has been used to track the RadioAstron satellite, and applying the derived orbital parameters to improve subsequent space VLBI observations. For the time being, the capacity of the software correlator is sufficient for most VLBI processing done at JIVE. However, new digital baseband convertors are being rolled out across the EVN, allowing far higher data rates; already 4Gbs real-time correlation with five EVN stations has been demonstrated. In view of this development, future mm-wavelength VLBI observations, and larger global arrays (at some point including the AVN, the SKA precursors and the SKA itself), we are investigating more powerful and economical solutions. The EC-sponsored UniBoard project deliverd an FPGA-based generic computing platform. The VLBI correlator personality engineered at JIVE is now reaching mat- rity, allowing us to assess the scalability and flexibility of this platform. The new UniBoard² project, also sponsored by the EC, will skip two generations of FPGA technology and deliver enormous processing power at a fraction of the power consumption of previous FPGA devices. We will discuss how the development of these different technologies is transforming the way VLBI is done. Maybe just as importantly, we will consider how we can make sure that future users will have the software tools to take full advantage of the greater capacity that VLBI observations will offer.