Green Bank Ultimate Pulsar Processing Instrument (GUPPI)

Introduction

The GBT Spectrometer, which took approximately 10 years to design, build, and deploy, is now more than 15 years old. Owing to the complex nature of its design, any significant enhancements to its pulsar search and precision timing capabilities (i.e., increased dynamic range) would prove to be prohibitive both in terms of hard costs and manpower requirements.

When Scott Ransom first introduced his concept of the ultimate “Dream Pulsar Machine” during the GBT Future Instrumentation Workshop held at Green Bank in the fall of 2006, it was clear that we needed a new approach. As luck would have it, at that same workshop, Dan Werthimer of U.C. Berkeley’s CASPER Group (Center for Astronomy Signal Processing and Electronics Research) introduced their open-source, platform independent approach to FPGA-based instrument design for astronomy applications.

Subsequent NRAO efforts aimed at evaluating the suitability of CASPER’s approach to our applications all yielded positive results and the decision was made to launch a comprehensive program for the design and development of advanced digital backends based upon CASPER’s hardware and software platforms. The GUPPI is actually the second such backend to be developed under this program (the first being the “Transient Event Capture device” which was designed for capturing transient events with the 43m Telescope).

Project Goals

The goal of this project is to provide a publicly available, state-of-the-art pulsar backend for the GBT based on FPGA technology. The idea is that that backend (name to be determined) will provide the following two capabilities:
  1. "Super-SPIGOT": Provide a much improved "basic" pulsar capability for "standard" (i.e. using an incoherent filterbank) pulsar observations, including Full-Stokes.
  2. "Super-GASP": Provide wide-bandwidth (200-800 MHz) coherent-dedispersion observation modes primarily for high-precision timing applications.
The specifications for this machine are found here. The project will be considered a success if both of the above listed capabilities are implemented for expert-user-level use on the GBT by end of summer 2008. However, even if Super-GASP capabilities turn out to be too difficult to implement on the current FPGA hardware, the Super-SPIGOT capabilities alone will result in dramatically improved observational capabilities over the GBT's current instrumentation.

Project Charter

Project Justification

The GBT Spigot and the Spectral Processor are currently the only observatory-supported and publicly available pulsar backends for the GBT. Both of these machines have serious limitations. The Spectral Processor is very old and fragile and can only be operated in narrow bandwidth modes to produce folded profiles of known pulsars. The Spigot has been a very successful instrument for the GBT due to its wide bandwidth modes (800MHz). However, it has relatively coarse spectral resolution (1024 channels over 800 MHz for 16-bit modes and 2048 channels for new 8-bit modes) and large pulse profile systematics for strong pulsars due (primarily) to the 3-level sampling in the GBT Spectrometer. In addition, while it was originally designed to be able to handle cross-correlation modes (and therefore polarization), it was realized in 2005 that these modes would never be implemented due to several technical issues with the system. Given the nature of these two systems, there are many pulsar experiments (including scintillation studies, single-pulse studies, and most importantly, high-precision timing observations) that simply cannot be done with the current instruments.

One of the highest profile experiments that can (and currently is) being undertaken with the GBT is to attempt to directly detect nano-Hz gravitational waves via the ultra-high-precision timing of an "array" of millisecond pulsars. The current efforts, which include several members of the NGPP project team, use the Gasp and CGSR machines (both of which are PI insstruments) to coherently dedisperse relatively narrow bandwidths (<100 MHz). It is highly likely that using much larger bandwidths would allow
  1. Better signal-to-noise in the measured pulse profiles and correspondingly improved arrival time precision
  2. "Instantaneous" determination of the dispersion measure towards any pulsar to high precision, allowing much better correction of the arrival times for interstellar medium effects
Since the detection of gravitational waves using current telescopes depends vitally on both of the above, building a new instrument that directly addresses these issues is a high priority. This is the primary reason why we are building a NGPP.

Primary Stakeholders(and their roles)

  • Scott Ransom NRAO, Charlottesville (Project Scientist)
  • Walter Brisken NRAO, Charlottesville(Project Team Member, scientific and development support as required)
  • Ingrid Stairs Dept. of Physics and Astronomy, University of British Columbia (User community collaborator)
  • Duncan Lorimer Dept. of Physics, West Virginia University (User community collaborator)
  • Maura McLaughlin Dept. of Physics, West Virginia University (User community collaborator)
  • Joeri van Leeuwen Dept. of Astronomy, University of California, Berkeley (User community collaborator)
  • Richard Prestage NRAO, Green Bank Site Director (sponsor. GBT program management)
  • Karen O'Neil NRAO, Green Bank GBT Program Manager (GBT program management)
  • Randy McCullough NRAO, Green Bank Electronics Division (Project Engineer, Project Team)
  • John Ford NRAO, Green Bank Electronics Division Head (Project Manager; engineering support as required)
  • Glen Langston NRAO, Green Bank Scientific Division (Project Team; scientific support as required)
  • Jason Ray NRAO, Green Bank Electronics Division (Project Team; engineering support as required)
  • Paul Demorest NRAO, Charlottesville (Project team; scientific and development support as required)
  • Electronics Division NRAO, Green Bank (Additional engineering and technical support as required)
  • Software Development Division NRAO, Green Bank (Additional software support as required)
  • Operations Division NRAO, Green Bank (Additional support as required)
  • CASPER Group University of California, Berkeley (Principal collaborator on toolflow, design blocks, etc.)

Project Personnel

  • Project Manager: JohnFord
  • Project Scientist: ScottRansom
  • Project Engineers: RandyMcCullough
  • Project Team: Patrick Brandt, Walter Brisken, Paul Demorest, Ron Duplain, Glen Langston, Randy McCullough, Jason Ray, Brandon Rumberg

Project Charter Approval

Approval of the charter means that the program is authorized to proceed, and commit funds and labor hours to the project.

  • Approved by Pulsar Project team members
    • DONE Walter Brisken
    • DONE Paul Demorest
    • DONE John Ford
    • DONE Glen Langston
    • DONE Randy McCullough
    • DONE Scott Ransom
  • Approved by GBT Program Management
    • ALERT! Karen O'Neil
    • ALERT! Richard Prestage

Symbols:
  • Use %X% not approve (will display ALERT!)
  • Use %Y% approve (will display DONE)

Project Scope

To design, implement, test, document, and deploy a new cutting-edge digital pulsar backend for use on the GBT utilizing the hardware and software platforms developed by U.C. Berkeley’s CASPER group. The resultant machine will reside in the GBT Equipment Room and will be readily re-configurable and useable (within minutes) to perform either pulsar search or pulsar precision timing missions by means of a “Lite RPC Server” with appropriately defined parameters, etc. Its capabilities are defined here, but will include: dual polarization, power, full stokes, coherent dedispersion, pulsar period folding (on multiple sources if possible), and data output streaming in standard PSRFITS format to a large capacity data collection machine (or cluster of machines).

Scope Management Plan

Changes in mission will be accepted only by the project manager in consultation with GBT Program Management (Currently K. O'Neil and R. Prestage), along with the project's team members. Increases in scope will not be accepted unless a clear source of additional resources to handle the increased scope is provided with the request.

Approach

The Project will be delivered in phases, each phase being dependent upon the previous phase. Phase 1 will include the items comprising the incoherent pulsar processor, along with software to allow expert use of the machine. Phase 2 will release the remaining capabilities defined in the specification, also for expert user mode. Finally, Phase 3 will seamlessly integrate the system into the M&C system, Provide proper end user documents, and provide appropriate data processing software to enable common user mode.

Expert use of the machine implies an in-depth knowledge of the machine, how to operate it, and how to handle the data output from the machine. It also requires that the observer control the machine, rather than leave it to an operator or staff scientist. Conversely, a common-user mode implies that the machine can be run from Astrid, and the data is in a form that is understood by standard Pulsar or GBT data analysis software. It is conceivable that some modes of operation are inherently expert user, and will never be released in a common user mode.

Key Assumptions

  • An additional development system will be procured and deployed by the CICADA group by the end of September, 2007
  • 1.5 FTE's will be provided for software development (Required for common user mode to be developed)
  • Electronics Division or Scientific staff will provide software resident in the pulsar machine for expert user control and data collection
  • Scott Ransom, Paul Demorest, and Walter Brisken will be available for questions, implementation help, and testing of the designs .
  • Site time vs GPS/UTC time measurements are accurate enough for precision pulsar timing

Out of Scope

  • Operator interface with the instrument is out of scope until Phase 3
  • Astrid integration is out of scope until Phase 3
  • Data processing/reduction is out of scope for this project, but will be developed in parallel by S Ransom and P Demorest
  • Data archiving will not be provided ever (unless change in NRAO policy)

Work Breakdown Structure (WBS)

The primary deliverable provided by the project is a common user pulsar machine capable of doing coherent pulsar timing observations as well as incoherent observations, both timing and search. This overall deliverable is provided by the following interim deliverables, delivered in three phases:
  • Phase 1
    1. Detailed specifications document(s)
    2. Detailed test plan document(s)
    3. Detailed deployment plan document(s)
    4. NRAO’s version of CASPER’s “Pocket Spectrometer” (design only)
    5. Next Generation EXPERT User NRAO Pulsar Backend including: dual polarization, detected power, “Lite RPC Server”, and test / demonstrator data collection machine (scaled down)
    6. Next Generation EXPERT User NRAO Pulsar Backend (same as Phase 1, item 5 with PSRFITS formatter, full stokes added)
  • Phase 2
    1. Next Generation EXPERT User NRAO Pulsar Backend (same as Phase 1, item 6 with coherent dedispersion added)
    2. Next Generation EXPERT User NRAO Pulsar Backend (same as Phase 2, item 1 with pulsar period folding added)
  • Phase 3
    1. Next Generation COMMON User NRAO Pulsar Backend (same as Phase 2, item 2 with full-blown data collection machine having large disk storage capacity and control software integration into the GBT M&C System) NOTE: This item represents what is intended to be the final version of the pulsar backend and will be released as a primary deliverable (see below) once the test and deployment plans are successfully completed

As a secondary deliverable item an RFI Rejection Spectrometer will be provided to the 43 Meter project as part of Phase #1, Item 5.

Project Schedule Summary

  • Phase 1
    1. Detailed Phase 1 specifications document(s)
      • 1 week
      • Scott Ransom, Randy McCullough, Paul Demorest
    2. Detailed test plan document(s)
      • 1 week, spread over phase 1 development
      • Randy McCullough, Science staff
    3. Detailed deployment plan document(s)
      • 2 weeks, spread over Phase 1 development
      • Randy McCullough
    4. NRAO’s version of CASPER’s “Pocket Spectrometer” (design only)
      • 2 weeks
      • Jason Ray, Scott Ransom, Paul Demorest, Randy McCullough, Glen Langston
    5. Next Generation EXPERT User NRAO Pulsar Backend including: dual polarization, detected power, “Lite RPC Server”, and test / demonstrator data collection machine (scaled down)
      • 6 weeks beginning end of September
      • Jason Ray, Randy McCullough, Paul Demorest, Walter Brisken, Scott Ransom
    6. Next Generation EXPERT User NRAO Pulsar Backend, including PSRFITS formatter and full stokes parameters
      • 3 weeks beginning after end of previous item
      • Jason Ray, Randy McCullough, Paul Demorest, Walter Brisken, Scott Ransom
    7. Complete specifications for Phase 2
      • Scott Ransom, Paul Demorest, Walter Brisken, Randy McCullough, Jason Ray
    8. End of Phase 1 approximately January 1, 2008
  • Phase 2
    1. Prototype and design coherent dedispersion modes
      • 14 weeks (finish April 1, 2008)
      • Scott Ransom, Paul Demorest, Walter Brisken, Randy McCullough, Jason Ray
    2. Next Generation EXPERT User NRAO Pulsar Backend (same as 1.6 with coherent dedispersion added)
      • 6 weeks
      • Jason Ray, Randy McCullough, Science effort
    3. Next Generation EXPERT User NRAO Pulsar Backend (same as 1.7 with pulsar period folding added)
      • 6 weeks
      • Jason Ray, Randy McCullough, Science effort
    4. End of Phase 2 approximately June 30, 2008

  • Phase 3 Begins as soon as software effort is available
    1. User documentation
    2. M&C Integration (cleo screen, Manager, Taskmaster, etc.)
    3. Data analysis software
    4. Completion of test plan and deployment plan
    5. Data collection machine with large disk array in place

Estimated Costs

  1. 2 iBOB $14K
  2. 2 iADC $05K
  3. 1 BEE2 $35K
  4. 1 10 Gb Ethernet switch $15K
  5. Backend data collection machine $20K
  6. miscellaneous $10K

Total cost is $99K We already have all this hardware, but we will need more to continue development of other machines if we deploy our development systems.

Project Plan Approvals

  • Approved by Pulsar Project team members
    • DONE John Ford
    • DONE Randy McCullough
    • DONE Scott Ransom
  • Approved by GBT Program Management
    • ALERT! Karen O'Neil
    • ALERT! Richard Prestage

Request for effort allocations

We are expecting Randy McCullough and Jason Ray to be committed to the project at the 50% level. In order to make progress on the common-user part of the project, software division resources on the 1.5 FTE level will be needed for the next 9 months. It also assumes that Scott Ransom, Paul Demorest, and Walter Brisken are providing support and help as needed.
  • Management
    • John Ford 0.1 FTE
  • Engineering
    • Randy McCullough 0.50 FTE
    • Jason Ray 0.50 FTE
  • Software 1.50 FTE
  • Science effort 1.5 FTE (Ransom, Demorest, Brisken, Langston)

-- JohnFord - 22 Aug 2007
Topic revision: r13 - 2010-09-19, RandyMcCullough
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding NRAO Public Wiki? Send feedback