GBT S-band RFI survey, part 1: Jan.30, 2002

by F. Ghigo (Feb.6, 2002)


On January 30-31, 2002, a survey for interference was done with the GBT 2_3 GHz receiver over the frequency range 1.7 to 2.6 GHz using the Spectral Processor. Data were acquired with the GBT positioned at four different azimuths (az = 6, 96, 186, and 276 degrees), all at an elevation of 55 degrees. Linear polarization was used. The laser power and computer systems were on for these measurements. The experiment will be repeated with the laser systems off, in order to find any RFI due to the laser system.

Plot of all data from Jan 30-31 run averaged. (See Results section below for more details.)

Very strong and variable RFI from a satellite is seen between 2320 and 2346 MHz. This part of the band is essentially unusable. The 2487-2500 MHz RFI may also be due to a satellite. A summary of the strongest RFI components is given in Table 1. (click on the first column to see a plot of that feature.)

Table 1. Strong RFI, 1.7-2.6 GHz
F(MHz) Peak power (db) HP width (MHz) Comments
1814.6 0.9 0.05
1844.0 1.4 1.0
2118.0 1.8 0.8
2285-2292 1.0 5.0
2320-2345 15.3 6.0 Satellite Digital Audio Radio Services
2361.5 1.2 1.0
2487-2498 4.0 5.0 Non Geostationary Mobile Satellite Service?
2561.3 0.8 1.0
Note: Column 2 is the approxmate peak relative to the baseline.
In the plots, the X polarization is in red, the Y in green.

A table of all significant narrow band components is listed in Table 2. (see Results section for more details.) This table excludes the components from the satellite bands listed in Table 1.

We note signals that may be coming from the direction of the Jansky Lab at frequencies of about 1815, 2200, 2227, and 2244 MHz.


To cover the whole band from 1.68 to 2.65 GHz using the Spectral Processor, which has a maximum bandwidth of 40 MHz, a series of 28 scans was done spaced at 35 MHz intervals. The lowest band center was 1695 MHz, and the highest was 2640 MHz. A two-minute integration was done at each frequency. The sequence took about an hour.

We experimented with the azimuth positioning of the GBT to find the azimuth that pointed directly towards the Control Building. This was done by visual sighting using the binoculars on the Control Room Deck. The azimuth of the control room turned out to be 96 degrees (+/- 0.5 deg).

Observations were done at 4 azimuths, 96, 186, 276, and 6 degrees, with the elevation set to 55 degrees in all cases. Table 3 summarizes the observations: the hour-long sequence of 28 scans was done at the 4 azimuths, then the whole series was repeated.
Table 3: Observations 31 Jan 2002, project: rfiS_jan30
Scan #s Start Time (UT) Azimuth (deg)
7-34 01:49 96
35-62 03:17 186
63-90 04:24 276
91-118 05:35 6
119-146 06:51 96
147-174 08:02 186
175-202 09:14 276
203-230 10:22 6


The spectra were processed with the "uni2" package of aips++. It was found that the frequency information for the first and last spectra in each sequence was missing because the nominal band for the S-band feed, as supplied to the M&C software, was too narrow. Thus the first and last spectra of each sequence were not used. This means that the data presented here span from 1712 to 2623 MHz. The channel spacing was 39 kHz.

The 26 process-able spectra were pasted together to make a spectrum for the whole band. Since the spectra were spaced 35 MHz apart, the central 35 MHz of each spectrum was selected and the rest discarded. The result was a raw spectrum which showed the Spectral Processor bandpass repeated 26 times. This raw spectrum is useful for identifying the strongest RFI components.

A filtered version of the the spectrum was produced by doing a median filter of 7 channels in width on each individual spectrum, then dividing this filtered spectrum into the raw data. This produced a spectrum with a flattened baseline showing the narrow RFI features. These were pasted together as described in the last paragraph to produce a filtered spectrum for the whole band. The spectral processor has artifacts that occur in its bandpass at 0.25, 0.5, and 0.75 of the bandpass. For the 40 MHz band width this means that the artifacts occur at -10, 0, and +10 MHz with respect to the band center. These features were edited out by the filtering software.

It is useful to look at the raw spectrum as well as the filtered version, because wide band features are eliminated by the filtering process.


Raw Spectra

First, we present all the spectra averaged together, i.e., all eight of the scan sequences averaged. This means that the total integration time per spectral channel is 16 minutes, and that spectra from all four azimuth pointings were averaged together.

A plot of the averaged spectrum is shown in Figure 1.

The spectrum is dominated by the satellite RFI in the range 2320 to 2345 MHz.

To look for weaker features, the vertical scale is expanded and shown in Figure 2.

In Figure 2, the periodic appearance is due to the bandpass shape of the spectral processor repeated 26 times and exxagerated by the expanded scale. In this plot, one can pick out several strong RFI spikes that cannot be seen in Figure 1. Close-up views of these RFI spikes are displayed in plots that can be viewed by clicking on the frequency column in Table 1.

Another thing to note is that the noise level is much higher in X polarization (red plot) that in Y (green) at frequencies above 2200 MHz. Also the noise level is particularly high in X between 2250 and 2400 MHz. Perhaps there is some broadband effect of the satellite RFI, or the spectral processor is driven non-linear in the presence of these strong signals.

Filtered Spectra

The filtered, baseline-flattened spectra were produced as described above by dividing each spectrum by the median-filtered version of itself. This brings out the narrow band features.

In Figure 3 we show the filtered spectrum for the X polarization, averaged over all scans.

Figure 4 shows the same information for the Y polarization. Note that the spectrum is cleaner for frequencies less than 2200 MHz in the Y polarization.

The list of RFI peaks is given in Table 2. The spectra shown in Figures 3 and 4 are normalized to a baseline of 1.0. Any peaks exceeding 1.002 are listed in Table 2. We have removed all the peaks in the list between frequencies of 2486-2500 and 2317-2346 MHz, all presumably due to satellite RFI. The power (P) in Table 2 is the y-coordinate from the plot (Figures 3 and 4) transformed to 0.001 units above 1.0; i.e., P = 1000(Y-1). The maximum of the 2 polarizations is listed.

Directional dependences

To look for RFI that may be coming from a particular direction, we have averaged the 2 spectra that were taken at each azimuth pointing and from it subtracted the spectrum averaged over all directions. The difference spectra are given in Figures 5-8.
Figure 5: difference spectrum for az=6 (North)
Figure 6: difference spectrum for az=96 (East: toward the Jansky Lab)
Figure 7: difference spectrum for az=186 (South)
Figure 8: difference spectrum for az=276 (West)

In these spectra it is probably best to discount any of the peaks between about 2300 and 2500 MHz as having something to do with one of the satellite RFI signals. Aside from these, we can suspect signals that are large and positive as likely coming from a particular direction.

Looking North (az=6 degrees), there are apparently no significant signals.

Looking East (az=96 degrees), towards the Jansky lab, there are four significant signals, at frequencies of about 1815, 2200, 2227, and 2244 MHz. The signal at 2200 seems quite significant.

Looking South (az=186 degrees), there are no obvious signals.

Looking West (az=276 degrees), which might include RFI from Snowshow or Cass, there is perhaps one significant signal at 2226 MHz.