Home » Science » Technology in the Next Decade

Technology in the Next Decade

Home » Science » Technology in the Next Decade

Science in the Next Decade

The GBT Today

The Robert C. Byrd Green Bank Telescope (GBT) is a unique resource for the US and global astronomical communities. The combination of its fully steerable 100-m unblocked aperture which provides an extremely clean point spread function allowing high dynamic range observations of diffuse emission, active surface that can maintain an RMS surface accuracy of 230 microns , 0.29–116 GHz nearly continuous frequency coverage with low-noise radio receivers, flexible instrumentation, the ability to observe declinations as low as -47°, and location in two different interference protection zones are not found in any other single telescope. This makes it one of the world’s premier telescopes for studying low-frequency gravitational waves, multi-messenger astronomy, fundamental physics, fast radio transients, cosmology, star formation, astrochemistry, gas in galaxies, and in the search for technosignatures.

The GBT was built to be flexible and continuously upgraded to meet the needs of the astronomical community. In the next decade, there are several projects planned which will expand the GBT’s performance in five key areas: survey speed, point source sensitivity, radio frequency interference (RFI) protection, accessibility, and preservation of legacy data. These projects are:

  1. Expanding the instantaneous field of view (FoV) of the GBT with advanced radio cameras.
  2. Increasing the instantaneous bandwidth of the GBT by developing ultrawideband analog and digital instruments.
  3. Preserving scientific data quality while sharing the radio spectrum with a growing number of private, commercial, and civil users through better identification and excision of RFI.
  4. Ensuring long-lasting public access to GBT data through a multi-petabyte data archive and high-performance processing tools.
  5. Providing increased funding for peer-reviewed use of the GBT for the U.S. astronomy community.

As part of the Astro2020 Decadal Survey the following white papers on activities and projects for the GBT were submitted:

Ultrawideband Systems

Recent advances in feed design and broadband low noise amplifiers have led to the deployment and scientific validation of so-called ultrawideband (UWB) receivers with bandwidth ratios approaching 6:1 or greater. Green Bank is actively developing a 0.7–4 GHz UWB receiver for the GBT with the primary science goal of improving sensitivity to low-frequency Gravitational Waves and broad-band fast transients. Secondary science goals include the study of radio recombination lines and molecular spectroscopy.

The first phase of operations with the UWB receiver will make use of existing signal transport infrastructure and back-end instruments. However, Radio Frequency Interference (RFI) is prevalent at these frequencies, and there is already evidence that some of these sources of RFI can cause non-linear behavior in the GBT’s analog signal path. We have started research on new high-speed and high-dynamic range analog-to-digital converters that can sample up to 5 GHz of bandwidth at 12-bits, along with related technologies. This will enable direct digitization of the entire bandwidth provided by the UWB receiver with a minimum of analog components, creating a much more robust and stable system in the presence of strong RFI.

The UWB receiver and accompanying digital upgrades will serve as a pilot program for future upgrades. Receivers operating above 18 GHz could provide >>8 GHz of instantaneous bandwidth but are limited by existing analog filters to 4–6 GHz in most cases. The analog signal transport system is also susceptible to environmental changes that can be a limiting factor for very deep observations. New digital technology being researched and developed at GBO has the potential to eliminate nearly all of these restrictions.

Bypassing the analog restrictions on bandwidth and digitally sampling the full available frequency range of the GBT’s receivers could lead to as much as a factor of 6.6 increase in survey speed when observing widely spaced spectral lines (see below) and improve spectral baselines. This would be a transformational modernization of the GBT, analogous to the upgrades of the Jansky Very Large Array, and would revolutionize all areas of GBT science.

Table 1: Expansion of GBT Receiver Bandwidth

Receiver Frequency Range (GHz) Current Max. BW (GHz) Potential Max. BW (GHz) Factor Increase
7-pixel KFPA 18:0–27.5 1.8 9.5 5.3
Ka-Band 26.0–31.0 4.0 13.5 3.4
Q-Band 39.2–49.8 4.0 10.6 2.6
W-Band 67.0–74.0 6.0 26.3 4.4–6.6
W-Band 73.0–92.0 4.0 26.3 4.4–6.6
Argus 74.0–116 1.5 8 5.3
Note — GBT receivers operating below 18 GHz are not bandwidth limited. A wide-band 8 GHz mode exists for KFPA but can only use one pixel. The maximum BW of Argus is determined by integrated receiver components.

Key Performance Requirements: For the UWB receiver: Measurements of pulsar dispersive delays with fractional uncertainty ≤ 10-5 and sensitivity over frequencies of interest for fast radio bursts (anywhere from 0.3–8 GHz). For UWB digital upgrades: Digital sampling of up to 26 GHz bandwidth (in multiple subbands) with high dynamic range.

Technical Requirements: For the UWB receiver: Tsys≤30K and total efficiency ≤ƞ=0.5–0.7 from 0.7–4 GHz. For the UWB digital upgrades: Sampling rates up to 10 Gsps with up to 12-bit precision, support for data rates up to 500 Gbps.

Technology Drivers: The UWB receiver is made possible by new quad-ridge feed designs with excellent performance over the frequencies of interest, as well as wide-band low noise amplifiers. The Analog Devices AD9213 ADC and Xilinx Vertex UltraScale+ FPGA VCU118 kit are representative of the next generation of technology that will enable digital sampling closer to the output of the receivers (see tables below).

Comparison of Current and Next Generation ADCs
ADC Feature/SpecEV8AQ160 (Current Generation)AD9213
Max. Sampling Rate5 Gsps10.25
Bit-depth812
Spur-free Dynamic Range56 dBc68 dBc
Power Consumption4.2 W5.1 W
Effective Number of Bits7.17.7
Comparison of Current and Next Generation FPGAs
FPGA ResourceROACH2 (Current Generation)UltraScale+ VCU188
FPGA System Logic Cells476K2586K
FPGA DSP Slice20166840
FPGA BRAM36 Mb6840 Mb
Onboard DDR42 GB8 GB
High-speed Ethernet8 x 10-GbE3 x 100-GbE
FPGA Silicon Feature Size40 nm16 nm
Expansion Bus2 x ZDOK1 x FMC, 1x FMC+
Max FPGA Tranceiver Speed6.6 Gbps32.75 Gbps

Current Status and Schedule: The UWB receiver is finishing preliminary design and a prototype will be built and tested by the end of 2019. Scientific commissioning will begin in late 2020 or early 2021. GBO has started the first phase of R&D with the VCU118 development kit. Development will begin with the AD9213 starting in Q4 of FY2019. The second and third phases would integrate a digital signal transport system with the UWB receiver, followed by additional GBT receivers.

Acknowledgments: The UWB receiver is funded in part by the Gordon and Betty Moore Foundation through Grant GBMF7576 to Associated Universities Inc. to support the work of the Green Bank Observatory and NANOGrav Physics Frontiers Center.

Research and development of new digital hardware and RFI excision techniques is supported by the National Science Foundation through Advanced Technology and Instrumentation grant number 1910302. GBO is grateful to the Breakthrough Listen project for providing hardware in support of digital research and development.

Radio Cameras

Currently the GBT has three multi-pixel receivers in use: the K-Band Focal Plane Array (KFPA), the Argus 3-mm receiver, and the MUSTANG2 90 GHz bolometer array. FLAG, the Focal L-Band Array for the GBT, is a pathfinder cryogenically cooled phased array feed receiver that currently holds the sensitivity record for a PAF, was also developed for use by the GBT but is currently not available for observing.

Looking toward the future, the GBT is an idea instrument for radio camera development. Over the next decade, that development will be broken into two areas – traditional feed horn arrays and phased array feeds.

Traditional Feed Horn Arrays

Argus+: Argus+ is a proposed 144 feed-horn camera operating within the 74–116 GHz band that would provide wide-field imaging of key molecular transitions for the study of star formation and astrochemistry. It will include a dedicated spectrometer providing a total velocity coverage of 2000 km s-1 with 0.015 km s-1 resolution at 90 GHz. Argus+ is a natural extension of the existing Argus receiver and would improve the mapping speed within this band by an order of magnitude. You can find the Argus+ Astro-2020 white paper here.

Phased Array Feeds

KPAF: The K-Band PAF receiver will be capable of forming 225 independent, Nyquist-sampled beams which will dramatically increase the mapping capability of the GBT from 18–26 GHz. This instrument will be ideally suited to the size scales found in star-forming regions and will complement continuum studies such as Herschel’s SPIRE program with kinematic and accurate temperature measurements. It will provide ≤0.1 K RMS noise in 0.1 km s-1 channels, with a system temperature ≤50 K and formed beam efficiency of 0.61. The project is in the preliminary design phase but builds off of experience with the existing FLAG instrument on the GBT. You can find the KPAF Astro-2020 white paper here.

FLAG2: FLAG is the first operational cryogenically cooled PAF receiver and the most sensitive PAF receiver in the world. It samples the focal plane of the GBT using 19 dipole elements, and its digital beamformer produces seven Nyquist-sampled beams on the sky with a bandwidth of 150 MHz. FLAG2 will have four times the survey speed of the original FLAG and more bandwidth, providing a powerful survey instrument for pulsars, FRBs, and HI.

Optimized Receivers

The GBT is already one of the world’s most sensitive radio telescopes, and straightforward upgrades of the existing GBT receivers would lead to a 30– 50% improvement in survey speed, even without adding additional pixels. This can be achieved by replacing older LNAs and other receiver components.

Sharing the Radio Spectrum

Spectrum occupancy will continue to grow for the foreseeable future, including at frequencies that were once comparatively free of RFI. Digital systems that effectively share the spectrum for scientific, civil, and commercial use are therefore critically important. Green Bank Observatory has been actively testing several techniques for automated RFI detection and excision. These include the use of median absolute deviation of complex voltage samples, spectral kurtosis, robust recursive power estimation, and a new exploration of machine learning algorithms that is in its early stages.

A demonstration of real-time RFI identification and excision using a robust recursive power estimator. Red data show an unmitigated polarization channel in which an airport radar is clearly visible every 12-seconds. Green data are from the mitigated polarization channel. Approximately 97% of data impacted by the radar are rejected and replaced with statistical noise. These data were collected with the 20-m telescope at the Green Bank Observatory. Further testing and astronomical verification using the GBT are underway. The firmware blocks will be released through the CASPER collaboration.

However, existing digital signal processing (DSP) hardware lacks the on-board memory and processing power to implement most of these techniques in real-time along side channelization, detection, and data formatting functions. The newest generation of hardware overcome these limitations. They offer significantly more on-board resources, processors specialized for machine learning, and support for high data rates. RFI excision methods can also be included in processing that occurs on GPUs. The next generation of wideband digital backends will be built using these technologies, and RFI mitigation will be included in DSP designs.

An example of RFI mitigation using machine learning. Data are passed through a classifier that assigns a confidence that each sample is RFI. Samples above some confidence level are replaced with zeros.

Key Performance Requirements: The ability to automatically identify and remove data affected by RFI in individual frequency and time samples, without any compromise of data quality.

Technology Drivers: FPGAs and GPUs with the processing power needed to include different RFI mitigation strategies in traditional DSP chains.

Current Status and Schedule: GBO has developed offline implementations of median absolute deviation and spectral kurtosis methods of identifying RFI in voltage data, and demonstrated their
efficacy using archival pulsar data. A proof-of-concept real-time robust recursive power estimation algorithm has been implemented on the 20-m telescope, and its impact on astrophysically relevant data products is being evaluated in detail. A machine learning approach is currently underway, beginning with carefully curating a training data set. The second phase (not yet funded) would focus on deploying these techniques as part of the real-time signal processing system.

Acknowledgments: Research and development of new digital hardware and RFI excision techniques is supported by the National Science Foundation through Advanced Technology and Instrumentation grant number 1910302. GBO is grateful to the Breakthrough Listen project for providing hardware in support of digital research and development.

Archiving & High-Performance Processing Tools

Large area surveys provide extremely valuable legacy data sets. The discovery of FRBs is an excellent example of unexpected phenomena that can be uncovered with new and improved analysis techniques applied to old data. This requires archiving low-level data products that can easily grow to PetaBytescales. These data must be easily accessible in a well documented and commonly used format, and come with tools that allow for easy reprocessing. High-level data products should also be available for researchers who are exploring the data in different ways. Finally, an archive must contain complete meta-data and processing pipelines to ensure reproducibility.

While Green Bank archives spectral line and low-data-rate pulsar data, they have not archived high-time-resolution pulsar data because of the associated data volumes. A multi-PetaByte data archive will ensure that all GBT data resulting from open-skies projects will be preserved.

The scientific value of the archive will be further enhanced by developing a new suite of data reduction software. GBO currently supports gbtidl for spectral line data reduction, and a Python-based pipeline for creating maps and spectral cubes. We will port gbtidl to Python and optimize the mapping pipeline for large, multi-pixel maps.

Technical Requirements: A 2-PB filestore that can be expanded as needs arise and storage costs decrease. Access to subsets of the data over internet, and computational resources sufficient to support local processing of full data sets.

Technology Drivers: Low-power, fast I/O, affordable solid state storage; open-source data processing packages upon which processing pipelines can be collaboratively built.

Current Status and Schedule: Green Bank is completing an evaluation of current and future requirements for construction of a complete data center, which could be completed in less than five years.