In the fall of 2006, the National Radio Astronomy Observatory (NRAO) hosted a Green Bank Telescope (GBT) Future Instrumentation workshop and invited staff, friends, and collaborators to discuss the direction of instrumentation development for the Robert C. Byrd GBT. For three days, an auditorium filled with radio astronomy experts (and at least one student) discussed science drivers and design goals to optimally use the GBT's unique features and capabilities.1
As its principal investigator, Dan Werthimer presented the work of the Center for Astronomy Signal Processing and Electronics Research (CASPER) and outlined how they were able to build eight instruments in roughly two years.2 Scott Ransom (NRAO) presented a case for a new GBT pulsar instrument. At the conclusion of the workshop, NRAO's senior management recognized the potential of the many great ideas presented at the meeting. This is where the Green Bank Ultimate Pulsar Processing Instrument (GUPPI),3 GBT's latest instrument for pulsar studies, has its roots.4
Rick Fisher (NRAO) captured a strong sentiment of the workshop's participants when he said, “We have a tendency to take everyone's wish lists and put them in our pockets before starting the design of a new instrument…” Instead, he suggested, we should focus on one or two features at a time, which would incrementally improve their overall functionality and capability.
For GUPPI, Scott Ransom's wish list for the ‘next-generation dream pulsar machine’ outlined a digital-signal-processing instrument to digitize and process the GBT's analog signal, with specifications to make optimal use of the telescope's performance (and operational structure) while satisfying science drivers for pulsar studies.5 At the time, we asked ourselves a big question: could we turn this dream machine into a fully operational instrument for the GBT (within the near future)? Today, the answer is ‘yes,’ the success of which I attribute to the CASPER community and platform, and our collective ability to partition ‘the dream-machine problem’ into steps we could take in reality, one at a time.
Excerpt from a design using MSSGE.6
Developers can use high-level block diagrams to implement powerful signal-processing functionality.
CASPER simplifies the design flow of radio-astronomy instrumentation through the development of platform-independent, open-source software and hardware.7 Although it does not deliver turnkey instrumentation, the center provides an actively developed suite of hardware and software designs, libraries, tools, and documentation to enable institutions to adopt this design approach. The community8 meets regularly at workshops hosted by the University of California at Berkeley, discusses designs and challenges on wikis and mailing lists, and also meets at various institutions to focus on a particular design challenge. This approach is more than just portability of design: it is portability of expertise.
At NRAO, we procured a set of hardware to begin development using the CASPER design approach.9 GUPPI consists of an iADC (an analog-to-digital converter that captures input), an IBOB (an internet break-out board that ‘packetizes’ input), and BEE2 (the second Berkeley emulation engine, for processing). CASPER is developing a new generation of hardware, Reconfigurable Open Architecture Computing Hardware (ROACH).10
The CASPER toolflow allows high-level design, simulation, testing, and iteration using Matlab/Simulink integrated with Xilinx tools, referred to as MSSGE6 (see Figure 1): Matlab/Simulink/System Generator/EDK (based on the Xilinx system generator and embedded development kit). The hardware uses Xilinx field-programmable gate arrays, providing for flexible implementations (i.e., no design is permanent) and designs that can be ported to other instruments or upgraded to new hardware. The platform provides a wealth of useful examples in a variety of formats, e.g., ready-to-run design files, screencasts, and memos. With MSSGE, we are building GUPPI to specification and continually improve its feature set. Although it has a nontrivial learning curve, our engineers find the toolflow to have notable advantages over traditional digital-system-development tools.
Figure 2. A 3.5min Green Bank Telescope observation in the L band using GUPPI, looking at the original millisecond pulsar (PSR) B1937+21. Here, we use a full-Stokes, 2048-channel, 40.96 microsecond sampling mode. DM: Dispersion measure.
The CASPER platform and community have provided NRAO with a modular, portable, open-source design process that has delivered results. GUPPI recorded its ‘first light’ in April 2008, and it has since been used for various GBT observations (under a shared-risk development mode).11 In Figure 2 white horizontal lines show where radio-frequency interference has been removed, while the dark patches are ‘scintles,’ caused by diffractive scintillation in the interstellar medium. Figure 2 also includes a pulse profile, showing the total intensity (black) as well as linear (red) and circularly (blue) polarized components.12
The GUPPI team is planning a full-scale instrument release to GBT users in June 2009, and NRAO is applying this instrumentation approach to other projects.
GUPPI would not have been possible without a broad collaboration. We thank the CASPER group and its community. The GUPPI team includes Patrick Brandt, Paul Demorest, Ron DuPlain, John Ford, Glen Langston, Randy McCullough, Scott Ransom, and Jason Ray at NRAO and Duncan Lorimer, Maura McLaughlin, and Mitch Mickaliger at West Virginia University (WVU). GUPPI receives further support from many of the staff at NRAO, including Walter Brisken, Rick Fisher, Rich Lacasse, and Amy Shelton, and has been supported in the past by Mike Borowczak, Andy Severyn, Michael Teuschler, and Philip A. Wilsey at the University of Cincinnati, and Brandon Rumberg at WVU. Special thanks are due to Xilinx Inc. for its support and generous donations to the NRAO. The NRAO is a facility of the National Science Foundation, operated under cooperative agreement by Associated Universities Inc.
Office of End-to-End Operations
National Radio Astronomy Observatory
Ron DuPlain is a software and systems engineer. He has a BS in computer engineering from the University of Cincinnati, where he completed a minor in very large scale integration. At Cincinnati, he and three classmates helped start the GUPPI family of instruments.