Description
The Expanded GMRT (eGMRT) is a proposal to look at three expansions to the
GMRT - increasing the field of view, increasing angular resolution and
improving the sensitivity to the extended radio emission. In this talk, we
would focus on increasing the field-of-view using Focal Plane Array (FPA)
and the ongoing development of the FPGA-based prototype FPA beamformer. A
prototype is being developed to understand the challenges in building
real-time beamformer with multiple inputs and multiple beams. As part of
this project, we have calculated the computational complexity of
implementing FPA beamformer in real-time and the various tradeoffs at the
architectural level. This tradeoff analysis helped in arriving at an
optimal number of inputs, spectral channels, and beams for Virtex-5 FPGA.
In case of an FPA system, the computation of beamformer weights through the
correlation operation is only needed at distinct time intervals. This led
to a further increase in the number of inputs for the same hardware setup,
by separating the correlation and beamforming designs, and exploring FPGA
reconfigurability. An alternate approach through a fully offline
correlation design using recorded raw voltages would also be described. The
current version of the multi-element, multi-beam design of the narrowband
(32 MHz bandwidth, 16-input, 4-beam) and wideband (300 MHz bandwidth)
beamformers have been developed using the CASPER (Collaboration for
Astronomy Signal Processing and Electronics Research) methodology. The
recent test results from the above- mentioned beamformer designs conducted
in a free-space test range (at the GMRT site) would be described. As a step
towards migration to contemporary signal processing platforms and the final
prototype beamformer, we have started beamformer development using Xilinx
RFSoC (RF System-on-Chip).
