Preliminary Report on Preparation and Execution of First Tests of thePhased Array Feed on the GBT
R. Fisher, B. Simon, B. Shillue
Jan 2014
I. Modification of receiver for fiber link signal transmission
Before 2013 all tests of the L-Band phased array feed wereperformed on the 20-meter telescope where the distance between thefront-end box and the data acquisition rack (located at the base of the 20m telescope),was short enough to usecoaxial cables for signal transmission between the two.Installationof the data acquisition rack in the GBT receiver room was impractical, due to weight and available rack space.Also, it was inadvisable because of the possibility of interference to otherGBT receivers from the PAF digital electronics.This required thatthe 1.4 GHz signal from each of the 38 PAF elements be transmitted tothe Jansky Lab via optical fiber. Analog fiber transmitters were installed in the front-end box, and fiberreceivers were installed in the Jansky Lab shielded space adjacent tothe GBT control room.Fiber cable connections were established to theJansky Lab from three possible PAF locations: the GBT Prime focusarea, the outdoor test building behind the Jansky Lab, and the20-meter telescope. Analog fiber modems have a rather high noise figure, on the order of3x10^6 Kelvin, so they require a lot of RF gain in preceding amplifiersstages to overcome fiber link noise which meant operating severalcomponents in the signal path near the upper limit of their dynamicrange.This made setting the signal levels at several locations ineach of the 38 channels a fairly tedious procedure.All channels needto operate linearly when the array is looking at cold sky and at a300K absorber for calibration.
II. LNA failure investigation and oscillation problems
The cryogenic PAF was successfully tested on the 20-meter telescope,but in the process of testing it for GBT operation,LNA failures beganto occur, presumably due to cryogenic temperature cycling of theplastic chip packages.The exact failure mode has yet to beestablished, but there were enough spare chips to proceed with the GBTtests.The chips installed in the PAF are no longer available so along-term solution to cryogenic amplifiers for the next generation PAFremains to be found. In the process of diagnosing higher-than-average noise temperatures ina few receiver channels, an out-of-band high frequency oscillation wasdiscovered in the cryogenic amplifiers.This was traced to a resonancein the LNA metal enclosure.The spurious oscillation was suppressed bymodifying the metal enclosure for all LNAs.Sample LNAs wereextensively tested in both Green Bank and Charlottesville to establishan accurate noise figure for the devices in the PAF installedon the GBT.
III. Modification of GBT Monitor and Control for control of PAF andfield-of-view mapping procedure
Efficient operation of the PAF on the GBT required some additions tothe GBT Monitor and Control system to synchronize data acquisition withtelescope control.The PAF control software was rewritten to acceptEthernet socket messages from GBT M&C for setting the receiver skyfrequency and starting data scans.The new PAF software also providesa graphical user interface for controlling the PAF in stand-alone mode.Paul Marganian from the GBT M&C group in Green Bank modified andtested the GBT end of this connection so that the PAF could becontrolled through the standard astronomer interface. Mapping of the PAF field of view requires a telescope positioning gridthat offsets the telescope boresight in a rectangular grid that has oneaxis that is parallel to the horizon whose origin tracks a fixed pointin the sky.The customary secant(Z) cross-elevation offset is notaccurate enough for a large grid, particularly at high elevations andlarge grid fields so Joe Brandt of the GBT M&C group implemented andtested the algorithm for large-scale photographic plate coordinatetransformation and added this to the GBT user interface.
IV. Software for data processing and archiving
The data product from the PAF tests on the 20-meter was raw ADCsamples of the 625 kHz receiver passbands which were then directlyprocessed with Matlab software.Documentation of the data and dataprocessing was pretty sparse.For the GBT tests the data archivingsystem used by the GBT, based on the FITS standard, was adopted by thePAF project and new software was written by the PAF group to producereceiver status and cross-correlation FITS files for each GBT scan.Each scan corresponds to one point on the PAF field of view grid.These PAF FITS files are noow written to the GBT data archivedirectories to take advantage to existing data preservation protocols.
V. Receiver tests in the outdoor test building
Due to the increased fiber noise and poor dynamic range of the IF downconverters, extensive tests to measure the backend noise contributionand linearity of the system were done. Power levels and gain atdifferent stages were adjusted to get the best signal-to-noise ratio and dynamic range. A preliminary receiver temperaturemeasurement using the digital backend was made by observing a hot load and cold sky. We plan to do more such measurements this monthwhen the PAF is back in the outdoor test building.
VI. Summary of GBT measurements of Dec. 19 - 26
The PAF tests on the GBT were scheduled for five sessions ofapproximately four hours each on the following dates:
Date / Hour / Source, DescriptionDec 19 / LST 0730-1130 / Virgo A grid, 1700 MHz
Virgo A 1700 MHz GBT focus track from 500mm to 1100 mm
Dec 21 / LST 0908-1308 / Virgo A on-off 1250 to 1800 MHz
Virgo A grid, 1700 MHz
Dec 22 / LST 0742-1142 / Virgo A on-off 1250 to 1800 MHz
3C286 on-off 1250 to 1800 MHz
3C286 grid, 1700 MHz
Virgo A 1450 MHz GBT focus track from 500mm to 1100mm
Dec 23 / LST 0931-1346 / Virgo A on-off 1250 to 1800 MHz at boreside beam position and 4 offset beam positions.
Virgo A grid, 1450 MHz
Dec 26 / LST 2000-2400 / 3C353 on-off 1420.866 and 1420.639 MHz
3C348 on-off 1420.866 and 1420.639 MHz
NGC6822 1420.639 MHz
DDO221 1420.866 MHz
3C48 on-off 1420.866 and 1420.639 MHz
3C48 on-off 1250 to 1800 MHz
3C123 on-off 1250 to 1800 MHz at boreside beam position and 4 offset beam positions.
The primary objectives of the measurements were to test beam-formingalgorithms across the array's field of view while testing for G/T andbeam shape of each beam.This data processing will take some time todo carefully.Initial sampling of the data shows that the strongestresponse at the central grid location is from the central arrayelement and that the next strongest six responses are from the innerring of six elements.So, it looks like we have a useful data setwith good signal to noise ratio.The measurements of NGC6822 andDDO221 were mainly for fun to see a spectral line in the passband,and, indeed, the line profile from NGC6822 in the center elementchannel looks like the one measured with the GBT's L-band horn. Since one element of the array over-illuminates the GBT, which has anopening half-angle of about 45 degrees at prime focus, we expect theaperture efficiency for just one element to be rather low and thespillover noise to be quite high.Indeed, this is the case soquantitative evaluation of the array will need to wait for moreextensive data processing and beam forming trials in the coming days.