NATIONAL RADIO ASTRONOMY OBSERVATORY
FROM: Ronald J. Maddalena and Dana Balser
SUBJECT: Tests of the 600-900 MHz section of the GBT prime focus receiver
On November 5 through 7, 1996 we tested the 600-900 MHz part of the GBT 300-900 MHz Prime focus receiver on the 140-ft telescope. The tests we conducted measured:
The receiver performed well although we found some problems, many of which were known to exist beforehand. One of the two receiver channels was known to have high system temperatures due to a problem with its amplifier. The second channel, although it did not have world-class performance, is noticeably better than the receiver traditionally used by 140-ft observers at these frequencies.
We also saw problems resulting from the receiver's feed, but this was
expected. The feed, originally designed for the f-ratio of the GBT, had its last
few segments removed in order to work at the f-ratio of the 140-ft. The
stripped-down feed should have under illuminated the 140-ft dish and had
measurably poor properties.
2. Noise Diode Values or Aperture Efficiencies
George Behrens and Rich Hall measured system temperatures and the intensity of the receiver's noise diode using the traditional hot-cold load method. The tests, made in the receiver test building, used the sky as the cold load. The derived diode values have an uncertainty of about 10-20% due to such things as contamination from interference or not knowing precisely the sky's temperature (~5-8 K). We will use the engineer's values derived for a sky temperature of 5 K.
If one assumes an aperture efficiency, then measurements with the 140-ft of astronomical sources with known fluxes can be used to measure the intensity of the noise diode. Conversely, if one assumes noise diode values, then one can use 140-ft observations to measure the efficiency. However, one cannot measure independently the efficiency and the noise diode values. We, therefore, first assumed an aperture efficiency to calculate diode values, and then assumed diode values to calculate efficiencies.
The 140-ft measurements used the same observing and data reduction technique as described in previous memos by RJM on the testing of the GBT Cassegrain receivers. Note that the method provides relative noise diode values (or efficiencies) to an accuracy of 2-5 % and gives absolute values to about 10%, about the same as the engineer's hot-cold method. We made continuum, on-the-fly measurements with a 1.25-MHz wide bandpass toward Cygnus A, 3C123, and 3C147 . We also assumed the frequency-dependent fluxes reported in Baars et al (1987; A&A, 61, 99) for Cygnus A and Ott et al (1994; A&A, 284, 331) for the other two sources.
Table 1 and 2 give the engineer's noise diode measurements (Col. 2), for the two (X and Y) linearly-polarized receiver channels. If we assume an aperture efficiency of 50% (about what one would expect for a scalar feed that under illuminates the dish), the 140-ft measurements give the noise diode values in Columns 3-5 for the three astronomical sources. Note that the 3C123 and 3C147 values agree but that the Cygnus A values are noticeably different. (It is also important to keep in mind that interference might contaminate the results; as such, one must concentrate on the trends in the data rather than individual results.) Since Cygnus A has a much stronger flux (2500-3100 Jy) than the other sources (~75 and 32 Jy), our suspicion is that some component of the receiver system is nonlinear or is saturated by a strong source. We suggest that someone look into the possibility of receiver saturation or nonlinearities. Since we are very suspicious of the Cygnus A data, Column 6 of Tables 1 and 2 is an average of only the 3C123 and 3C147 measurements.
Note that the noise diode values as measured on the 140-ft (Col. 3-6) are
consistently higher than that measured by the engineer. The probable cause of
the difference is an efficiency that is lower than the assumed 50%. If one uses
the noise diode values provided by the engineer (i.e., Col. 2), then Column 7 is
the average of the measured efficiencies for 3C123 and 3C147. The observed low
efficiency and the significant changes of efficiency with frequency are probably
due to the feed. If future tests are made with this receiver on the 140-ft, then
we may want to characterize better the expected performance of the stripped-down
|Table 1: Noise Diode Values and Efficiencies for channel X|
|Noise Diode values (K)||Measured Aperture
|Engineer's||CygnusA||3C123||3C147||< 3C123 + 3C147 >|
|Table 2: Noise Diode Values and Efficiencies for channel Y|
|Noise Diode values (K)||Measured Aperture
|Engineer's||CygnusA||3C123||3C147||< 3C123 + 3C147 >|
3. System Temperature Measurements
We measured the system temperature by pointing the telescope at the zenith (at a time when the galactic plane was not transiting) and used the methods outlined in RJM's previous memos. The determination of system temperatures requires knowing the noise diode values. Since there is some uncertainty in noise diode values, we used both the engineer's values (Col. 2 in Tables 1 and 2) and our values (Col. 6 in Tables 1 and 2, which assumes a 50% aperture efficiency) to derive the system temperatures listed in Table 3 for the two channels. As with all of our measurements, the data at some frequencies probably are contaminated by interference.
As we knew beforehand, the system temperature was high on channel X due to a problem with the amplifier. The channel Y system temperatures are noticeable better than the current 140-ft receiver that covers the same band. However, it is our understanding that better amplifiers could be developed and used.
Table 3 also gives the engineer's measured values for the system temperature,
as derived from hot-cold tests. Note that the engineer's values should be lower
than that measured on the 140-ft since ground spillover and scattering does not
add to the system temperature with the engineer's test setup but does add to the
system temperature with the 140-ft measurements. The difference between the
engineer's system temperatures and the 140-ft measurements should be constant if
spillover and scattering on the 140-ft are independent of frequency. If we
assume spillover and scattering is constant, then the data in Table 3 imply that
the engineer's values for the noise diode are correct, (and further suggests
that the telescope's efficiency changes with frequency). Nevertheless, without
more tests of the stripped-down feed, we cannot rule out the unlikely
possibility that both the efficiency and the spillover are changing with
|Table 3: System Temperatures|
|Channel X||Channel Y|
|Measured on 140-ft||Measured by engineer||Measured on 140-ft||Measured by engineer|
|Using Table 1, Col. 2 diode values||Using Table 1, Col. 6 diode values||Using Table 2, Col. 2 diode values||Using Table 2, Col. 6 diode values|
4. Bandpass Checks
To examine the bandpass of the receiver's channel X, we divided the Mark IV Autocorrelator into four sections, each with a bandwidth of 10 MHz. The I.F. of each section was offset from its neighbor by 10 MHz, allowing us to sample simultaneously 40 MHz of the receiver's bandpass. We chose an integration time of two minutes and pointed the telescope at the zenith. The first spectral-line observation for channel X covered the sky frequencies of 695-735 MHz, the next observation covered 735-775 MHz, and so on until 900 MHz was reached. We then repeated the measurements for channel Y.
A quick examination of the spectra did not show any strange bandpass shapes,
although some spectra were severely contaminated by interference. The system
temperatures measured with the Autocorrelator agreed with that derived from the
continuum measurements. Since the number of spectra is large, we cannot
reproduce them in this report but they are available if anyone is interested.
5. Noise as a Function of Integration time
To test whether the amplitude of spectral-line noise followed the radiometer equation, we needed long integrations. The bandpass measurements suggested that 816 MHz was clear of interference. We divided the Autocorrelator into four sections of 1.25 MHz bandwidths. Channel X fed two of the sections and channel Y the other two. For the two channel X sections, the I.F. of the second section was offset 1.25 MHz from the I.F. of the first section. Thus, the two sections covered from ~815.3 to 817.8 MHz. The channel Y sections were similarly set up.
We wanted to examine the shape of the instrumental bandpass under typical observing conditions. Since strong continuum sources observed with the total power mode often produce noticeable baseline problems, we decided to observe Orion A. We integrated for 10 minutes on source and used as a reference observation a 10 min. observation at a position 10min of Right Ascension to the west of Orion A. We then repeated on and off source measurements four times and averaged the four difference spectra. The resulting spectra (Figs. 1-4) show flat, pleasing baselines and a hint of two weak, highly polarized interference lines.
We then removed a second order baseline from the data, measured the rms of
the data and calculated the theoretical rms (Figs. 5-8). For all of the
Autocorreltor sections, the measured and theoretical rms agree, implying that
the receiver's performance follows the radiometer equation up to the limits of
6. Stability and 1/F Power Spectra
The receiver engineer had no opportunity to measure in the lab the receiver's noise power spectrum. To make these measurements at the 140-ft, we pointed the telescope at the zenith at a frequency (815 MHz) near where no RFI had been detected by our previous bandpass checks.
We used four of the continuum detectors and 10 MHz bandpass filters. Two were fed by channel X and differed in I.F. by about 10 MHz. The other two detectors were similarly configured for channel Y. Using an integration time of 0.2 sec., the fastest the detectors can run, we collected data for about 27 min for a total of 8096 samples.
Figures 9-12 are the first 4096 points of the original data with a simple dc-offset removed from the data. The abrupt change in the detected signals near sample 700 with all four detectors may show a problem with the receiver or interference. In channel X (Figs. 9-10), the change lasted a few seconds and was a significant drop in power level. Although most RFI appears as a jump up in power, sometimes a drop in power can be caused by strong, out-of-band RFI. In channel Y (Figs. 11-12), the change lasted a few minutes and was an increase in power levels.
With UniPops we generated the power spectra given in Figures 13-16. The clean power spectra suggest that the receivers have very good stability. The refrigerator spike, normally seen in the power spectra from other receivers at about 1 Hz, is noticeably weak or absent.
7. Beam Shape and Polarization Properties
We spent some time measuring the beam shape and polarization properties of the receiver with its stripped-down feed. To map the beam, we used a 10 MHz bandpass on channels X and Y and the strong point-source Tau A. On-the-fly maps were made at three frequencies (706, 815, and 905 MHz) suspected to be clean of RFI at the ends and middle of the receiver's band. At each frequency, we made two maps, one with the telescope slewing in Right Ascension and the other with the telescope slewing in Declination. The data from the two maps were interweaved, producing an image with a dynamic range greater than 30 dB. Figure 17 is an average of all the beam maps, regardless of frequency or polarization.
Figures 18-20 are the beam maps at the three frequencies with the data from channel X and Y averaged together. The extremely low sidelobe levels (about -24dB) are probably due to the under illumination of the telescope's dish by the feed. Toward the northwest and southeast edges of the main beam there is an unusual boxing of the lower contour levels. This is not an artifact of the data reduction but was seen in all maps. We guess that the stripped down feed might be responsible.
The X-channel maps showed a main beam that is elliptical in the north-south direction while the Y-channel maps showed a main beam that is elliptical in the east-west direction. The ellipticity of the beam is easily seen after averaging the channel X maps (Fig. 21) and channel Y maps (Fig. 22). To investigate this ellipticity further, we examined a continuum pointing observation toward Tau A with the telescope slewing east-west (Fig. 23) and north-south (Fig. 24) through the source. The data for channel X is plotted with a solid line and the data for channel Y with a dashed line. In addition to the easily-seen ellipticity, Figures 23 and 24 show some evidence that the channel X and Y beams are not coincident by about 1'; that is, the data show some evidence of beam squint.
Thus, we see many different problems with the stripped down feed, both with
its shape and with its polarization properties.
The major conclusions we come to are that, except for the (expected) high
system temperatures for the channel X, the receiver performs well. We would like
to see some effort into reducing the system temperature for both channels to
make the receiver a world-class instrument. We suggest that someone look into
whether we are correct about the possibility of saturation or nonlinearities for
strong sources. The problems or difficulties we encountered mostly stem from the
stripped-down feed. If further tests of the receiver are to be performed on the
140-ft, we suggest that someone investigate the properties of the feed.
Figure 1: Channel X spectral-line data toward Ori A at a sky frequency of 815.9 MHz. Except for a DC-offset, no baseline has been removed. The data, in units of TA, assume the engineer's value for the noise diode.
Figure 2: Same as Fig. 1 but for sky frequency of 817.1 MHz.
Figures 3-4: Same as Figs. 1 and 2 but for channel Y.
Figures 5-8: Same as Figs. 1- 4 except a low-order polynomial baseline has been removed. The values for the measured and theoretical rms are given on the plots.
Figure 9: Channel-X continuum data toward the zenith. 10 MHz bandwidth and centered at 810 MHz. Sampling was 0.2 seconds; only the first 4096 samples out of 8192 are shown. The data, in units of TA, assume the engineer's value for the noise diode. Only a DC-offset has been removed.
Figure 10: Same as Fig. 9 but centered at 820MHz.
Figures 11-12: Same as Figs. 9 and 10 except for channel Y.
Figures 13-16: The power spectrum of the data in Figs. 9-12. Note that the Y-axis is logarithmic.
Figure 17: Average of the beam maps for channel X and Y at 706, 815, and 905 MHz. To show both the main beam and side lobes, two plots are given which differ only in their grey-scale. One plot shows the sidelobes best while the other emphasizes the structure of the main beam. The superimposed contours are separated by 3 dB. The map has had a two-dimensional background level removed and has been normalized to a value of one.
Figure 18: Average of the channel X and Y beam maps at 706 MHz. Details the same as Fig. 18.
Figure 19: Same as Fig. 19 except at 815 MHz.
Figure 20: Same as Fig 19 except at 905 MHz.
Figure 21: Average of the beam maps for channel X at 706,815, and 905 MHz. Details the same as in Fig. 18.
Figure 22. Same as Fig. 22 except for channel Y.
Figure 23. Continuum observation using a 10-MHZ bandwidth with the telescope slewing east-west through Tau A. The data have a DC-offset removed and have been normalized to a value of one. The data from channel X are represented by the solid line and the data from channel Y by the dashed line.
Figure 24: Same as Fig. 23 except with the telescope slewing north-south.