One of the great strengths of the LOFAR system is its capacity for enhancement. It is of course common for astronomical facilities to increase their capabilities through continued software development. For LOFAR however, the system design is suffciently flexible that scientific capacity can be added relatively straightforwardly at both the software and hardware levels. In the simplest case, this capacity increase can be achieved through the addition of more stations to the array resulting in improved uv coverage, longer baselines, and increased sensitivity. Obviously, such extensions do also require the addition of significant additional compute capacity.
Similarly, the capabilities of individual LOFAR stations can also be expanded. With minimal modifications, the data-stream from a given station can be replicated and processed independently of the standard LOFAR processing. A number of such “stand-alone” or single station enhancements are already available and in further development (users proposing to use these observing modes should adhere to the regulations given here). The first of these, called ARTEMIS, implements a real-time dedispersion search engine to detect pulsars using the data-streams from one or more LOFAR stations (Serylak et al. 2012). The hardware consists of 4 12-core servers hosting high-end NVIDIA GPU cards. These servers are all fed data through a broadband (10 Gigabit Ethernet) switch, which is also responsible for sending the data back to the Netherlands during normal international LOFAR telescope (ILT) operations. The data being processed adds up to a stream of ~3.2 Gbits/s, which consists of a sky bandwidth of approximately 48 MHz, sampled at 5.12 ns intervals, in two polarisations. Real-time processing generally ensures that the ~ 400 MB of data per second are reduced to manageable rates both for storage and further processing.
The ARTEMIS servers can perform in real-time all the operations necessary to discover short duration radio pulses from pulsars and fast transients, thanks to a modular software structure operating in a C++ scalable framework developed at the University of Oxford. ARTEMIS includes processing modules for receiving the data, for further channellisation in finer frequency channels using a polyphase filter, generation of Stokes parameters, excision of terrestrial radio frequency interference, temporal integration, real-time brute force dedispersion using at least 3000 trial DM values and detection of interesting signals, in high-throughput CPU and GPU code
Data from all German LOFAR stations can be recorded in GLOW-mode whenever the stations are not used for observations with the full array. Current capabilities include recording of the raw, beam-formed data for 24h from four stations at full bandwidth or all stations at reduced bandwidth. It is also possible to do pulsar folding or generation of dynamic spectra on-the-fly, which allows continuous, extended observations. Upgrade plans include recording of the full bandwidth from all stations and an offline correlator.
Another, EU-funded project named AARTFAAC expands upon LOFAR’s ability to monitor radio transients by correlating the signals from all dipoles on the Superterp in real-time (Prasad & Wijnholds 2012).
Finally, a design for an expanded station concept is being developed by the French LOFAR consortium. This design (NenuFAR) adds significant numbers of additional dipoles as well as computing capability to the current French station at Nancay (FR606) resulting in a “SuperStation” optimized for beam-formed observations with high instantaneous sensitivity in the 10-80 MHz range (Zarka et al. 2012).