Beamformer and Calibration Performance for the Focal-plane L-band Array feed for the Green Bank Telescope (FLAG)

Transcription

1 Beamformer and Calibration Performance for the Focal-plane L-band Array feed for the Green Bank Telescope (FLAG) B. D. Jeffs 1, K. F. Warnick 1, R. A. Black 1, M. Ruzindanna 1, M. Burnett 1 1 Brigham Young University D. J. Pisano 2, D. R. Lorimer 2, N. Pingel 2, K. Rajwade 2 2 West Virginia University R. M. Prestage 3, S. White 3, B. Simon 3, L. Hawkins 3, W. Shillue 4, D. Anish Roshi 4 3 Green Bank Observatory, 4 NRAO Central Development Laboratory PAF Workshop 217, Sydney National Science Foundation Award

2 The FLAG BYU & WVU Graduate Student Team BYU: Richard Black Mark Ruzindana Mitchell Burnett WVU: Nicholas Pingel Kaustubh Rajwade

3 FLAG Specifications Overview 19 dual polarized elements,.7λ hex grid spacing 15 MHz instantaneous BW in the range of 1.2 to 1.5 GHz Cryogenically cooled LNAs, room temperature antennas 7 real-time beams with XX *, YY * & XY * pol. Many beams using postcorrelation beamforming (PCB) FPGA - GPU real-time digital receiver/correlator/ beamformer back end.

4 FLAG Specifications Overview 5 coarse frequency channels 33 khz BW each, 15 Mhz total for PCB and real-time beamformer Fine polyphase filter bank: 32 channels, 9.5 khz BW each, 3 MHz total for PCB Commensal transient search and HI survey modes 9 arcmin beamwidth, 3 beamwidth field of view diameter Element design optimized for best sensitivity and average active array impedance match across beams

5 FLAG Specifications Overview Digital back end: 5 ROACH II FPGAs for DDL digital fiber communication and PFB frequency channelization 5 high performance PCs 1 GPUS for fine PFB, real-time beam-former, and real-time correlator Ideal beamformer architecture for RFI mitigation

6 8 Fiber Digital Optical Rcvr Card 4 x 1 Gbe I/F Card Cryostat Ethernet Switch Melanox SX X 4 GbE ports 8 Fiber Digital Optical Rcvr Card 4 x 1 Gbe I/F Card FLAG Block Diagram Ant. LN A I-Q mix, ADC, Serialize & Optical Xmit LO Ch. 1 Ch. 8 ROACH II FPGA 4 X 1 GbE 4 Gbe 4 X 1 GbE 4 GbE CPU/GPU (Blade server + 2 nvidia GTX79) ( 5) Ant. LN A ( 4) I-Q mix, ADC, Serialize & Optical Xmit LO Array aperture, Antenna elements, LNAs, Cryo system, Down converters Ch. 33 Ch. 4 Signal Transport: GBT prime focus to Jansky Lab fiber link ( 5) ROACH II FPGA In Mezzanine I/F Slot F Engine: DDL deserialization, boundary alignment, polyphase filter bank and 1 Gbe I/O Rack Mount PC 12 TB SATA RAID Disk Array System control and data storage CPU/GPU (Blade server + 2 nvidia GTX79) XB Engine: Correlator/Beamfor mer, Spectrometer NRAO DDL System BYU Correlator Beamformer

7 Back End, Correlator Beamformer

8 Ethernet Switch Melanox SX 112 Channel Selection (k =... ) Channel Selection (k =... ) FITS Formatter Lustre Disk Array Storage First-of-its-kind Architecture Packets from 5 ROACH IIs: 25 out of 5 channels from all 4 input ports 4 X 1 GbE Hashpipe instance 2 Hashpipe instance 1 3 All 25 Coarse Channels, 7.5 MHz total BW All 25 samp/s Channels, 7.5 MHz total samp/s Fine PFB 32 point FFT, 256 tap 32 point Filter FFT, 256 tap Filter Fine PFB 5 Selected Coarse Channels 5 Selected Coarse Channels Integrator Real-time c S Beamformer k,( j, j ) = Integrator N-1 b k, Real-time j [n] = 1 * åc N S b k, j b k, j Beamformer k,( j, j ) = w H n=1 k, j x k [n] N-1 b k, j [n] = N = for.1 ms * åb k, 1 j b k, j N Coarse/ w H n=1 k, j x k [n] Fine Correlator / Integrator Coarse/ (XGPU code) Fine x k [n Correlator R ] k = 1 N-1 / Integrator (XGPU åcode) x k [n] x H k [n] N n= N-1 x k [ncoarse: RN 3 for.1 ms k = 1 åx k [n] x H k [n] Fine: N = 4,75 N for 5 ms ] n= 16 Fine Channels, 1.51 MHz total samp/s Coarse: N = 3 for.1 ms Fine: N = 4,75 for 5 ms 16 Fine Channels, 1.51 MHz total samp/s N = 3 for.1 ms R k Channel Selection (k =... ) Coarse: 5 of 5 coarse channels Fine: Channel All 16 fine Selection channels (k =... ) RPost- k Coarse: Correlation 5 to 25 coarse Beamformer channels f S Fine: k,( j, j ) = All 16 fine w H k, channels j R k w k, j 1 2 c S k,( j, j ) R k Nvidia GTX79 Ti GPU, 1 of 2 per HPC To 4 other HPCs Nvidia GTX79 Ti GPU, 2 of 2 per HPC

9 First-of-its-kind Architecture PCB is the primary mode for all observations: f S k,( j, j ) = w H k, j R k w k, j Array covariance, R k, is the output data product Correlator STI dump rates: T STI 1 ms: store R k for all channels T STI < 1 ms: store R k for fewer selected channels PCB permits: Revisiting observations with different beampatterns Recovery from bad calibration, poor beams Arbitrarily dense beam spacing for radio camera Adaptive beamforming RFI mitigation after the fact!

10 BYU s FLAG Array Element Design

11 217 Phase II Commissioning Goals On-GBT: May, July, & Aug Full system hardware, firmware, software, and control: integration and functional evaluation. Test major operational modes: calibration correlator Real-time beamformer / transient detection HI fine channelized postcorrelation beamforming First science observations

12 217 Phase II Commissioning Goals On-GBT: May, July, & Aug Full system hardware, firmware, software, and control: integration and functional evaluation. Test major operational modes: calibration correlator Real-time beamformer / transient detection HI fine channelized postcorrelation beamforming First science observations

13 A productive year for the FLAG project! PROGRESS IN COMMISSIONING AND DEVELOPMENT, 217

14 Resolving Commissioning Challenges 216 / 217 Some dead elements, primarily due to problems with the DDL digital optical downlink Overheating, nighttime operation only Limited observing time, just a few data sets Not able yet to nail down true T sys and sensitivity Verified operation for most system components and got some early data Limited bandwidth: 3 out of 15 MHz.

15 Resolving Commissioning Challenges 216 / 217 DDL links all working, no dead X-pol elements, full array beamforming possible Overheating, nighttime operation only Limited observing time, just a few data sets Not able yet to nail down true T sys and sensitivity Verified operation for most system components and got some early data Limited bandwidth: 3 out of 15 MHz.

16 Resolving Commissioning Challenges 216 / 217 DDL links all working, no dead X-pol elements, full array beamforming possible New water cooling system, No overheating stoppages Limited observing time, just a few data sets Not able yet to nail down true T sys and sensitivity Verified operation for most system components and got some early data Limited bandwidth: 3 out of 15 MHz.

17 Resolving Commissioning Challenges 216 / 217 DDL links all working, no dead X-pol elements, full array beamforming possible New water cooling system, No overheating stoppages Full schedule of observations: Calibration, Functional, HI, Pulsar Not able yet to nail down true T sys and sensitivity Verified operation for most system components and got some early data Limited bandwidth: 3 out of 15 MHz.

18 Resolving Commissioning Challenges 216 / 217 DDL links all working, no dead X-pol elements, full array beamforming possible New water cooling system, No overheating stoppages Full schedule of observations: Calibration, Functional, HI, Pulsar Best yet reported for a PAF: T sys / η ap (See Shillue talk this p.m.) Verified operation for most system components and got some early data Limited bandwidth: 3 out of 15 MHz.

19 Resolving Commissioning Challenges 216 / 217 DDL links all working, no dead X-pol elements, full array beamforming possible New water cooling system, No overheating stoppages Full schedule of observations: Calibration, Functional, HI, Pulsar Best yet reported for a PAF: T sys / η ap (See Shillue talk this p.m.) Verified operation for ALL system components and collected science performance demonstration data Limited bandwidth: 3 out of 15 MHz.

20 Resolving Commissioning Challenges 216 / 217 DDL links all working, no dead X-pol elements, full array beamforming possible New water cooling system, No overheating stoppages Full schedule of observations: Calibration, Functional, HI, Pulsar Best yet reported for a PAF: T sys / η ap (See Shillue talk this p.m.) Verified operation for ALL system components and collected science performance demonstration data Full 15 MHz beamformer and correlator demonstrated

21 216 Results: Element patterns 1391 MHz

22 217 Results: Element patterns.2 1X Element Patterns - X-Polarization, 144 MHz 2X 3X 4X X Elevation (degrees) X 7X 8X 9X 1X X 12X 13X 14X 15X X 17X 18X 19X Cross-Elevation (degrees) -3

23 216 Results: Beam Patterns Max SNR beamformer 1391 MHz w q = R -1 off a q b q (f) 2 = w q H R f w q

24 217 Results: Beam Patterns AGBT16B_4_13 - X-Polarization, 144 MHz.2 Beam 1 Beam Elevation (degrees) Beam 3 Beam 4 Beam Beam 6 Beam Cross-Elevation (degrees) -4

25 216 Results: Sensitivity over FOV S(q) = 2k b 1-26 F s SNR(q) SNR(q) = w H q R q w q - w H q R off w q w H q R off w q m 2 / K

26 217 Results: Sensitivity over FOV Formed Beam Sensitivity Map - Xpol, MHz Elevation Offset (degrees) m 2 / K Cross-Elevation Offset (degrees)

27 216: Beamformed T sys vs Freq. Tsys (K) Minimum Tsys in FoV Assumes h a =.65 May be corrupted by sources in the only off-pointing available Higher due to fixable DDL bit errors and missing elements These hardware failures cause poor Frequency (MHz) illumination pattern

28 217: Beamformed T sys / η ap vs Freq.

29 LCMV Response-Constrained Beams Max-SNR-like beams with point constraints Controls mainlobe shape to reduce coma Zero response constraints in undistorted ring of nominal first beampattern null Weight calculation:

30 LCMV Real Data Results

31 LCMV Real Data Results

32 HI Velocity Channel Map D.J. Pisano and Nicholas Pingel NGC6946 Single beam physical scan Fine PFB channel mode Grey field and red contours from FLAG. Blue contours from prior single pixel GBT scan, longer integration: Pisano, D.J. 214, Astronomical Journal, 147, 4

33 Pulsar Observation Kaustubh Rajwade and Duncan Lorimer Giant Pulse detection, single pulse FLAG Beam (boresight) real-time beamformer PSR B From IAU 337 Symposium Proceedings, Sept. 217

34 FLAG Real-time RFI Nulling Project MS student thesis project starting now Real-time subspace projection on FLAG Use the rapid dump FLAG real-time correlator Calculate weights, w, in HPC, rapid weight update in GPU beamformer Operate only on affected coarse channels Demonstrate see-through performance with moving RFI during science observations, e.g. BeiDou COMPASS navigation satellites

35 FLAG Real-time RFI Nulling Project MS student thesis project starting now Real-time subspace projection on FLAG Use the rapid dump FLAG real-time correlator Calculate weights, w, in HPC, rapid weight update in GPU beamformer Operate only on affected coarse channels Demonstrate see-through performance with moving RFI during science observations, e.g. BeiDou COMPASS navigation satellites

36 Conventional Subspace Projection (SP) Zero forcing, deeper nulls than with max SNR, LCMV, etc. Must assume interference is the dominant source. Use eigenvector decomposition to identify interference subspace. Partition eigenspace. Largest eigenvalues(s) correspond to RFI. R ˆ k [U int U sig+noise ] = [U int U sig+noise ]L Form perpendicular subspace projection matrix: H P k = I - U int U int Compute weights and beamform: w SSP,k = P k w nominal, H b(n) = w SSP,k x(n), ê k = n ë ê N ú û ú

37 Deep, Broad Nulls for Moving RFI Method developed by Richard Black. Moving RFI smears the interference subspace estimate. SP nulls go shallow. Short-term integration for freezes motion, stops smearing, but few time sample cause sample estimation error. SP nulls go shallow. Two-stage eigen-decomposition eliminates smearing and high sample estimation error. SP nulls are deeper!

38 Deep, Broad Nulls for Moving RFI

39 Deep, Broad Nulls for Moving RFI Cancelation Performance - Without Auxiliary Unmitigated Motion-Agnostic Motion-Cognizant Broad Null Post-Mitigation INR (db) Pre-Mitigation INR (db)

40 Conclusions Successful commissioning. Next steps: science observations in January, then shared risk facility instrument. PCB works! We are excited for its future promise for RFI canceling and beam control. We have a sensitive, stable instrument!