15mm RFI Survey 2016-AUG-01

Recent observations have suggested the possibility that RFI is becoming a problem in the ATCA 15mm band. This survey was conducted to determine what parts of the 15mm band might be affected by RFI, and where the RFI is coming from. The hypothesis is that the high-frequency RFI is coming from satellites, and particularly the recently-launched Sky Muster satellites that provide wireless internet for the Australian NBN.

The survey was conducted in three parts. For each part, the entire 15mm band, from 16 GHz to 26 GHz was surveyed.

The band was covered with ten frequency tunings, with centre frequency IF1/IF2 combinations as shown in the following table.

Tuning	Centre Frequency [MHz]
	IF 1	IF 2
1	17000	17000
2	16700	17300
3	19000	19000
4	18700	19300
5	21000	21000
6	20700	21300
7	23000	23000
8	22700	23300
9	25000	25000
10	24700	25300

The tunings are made this way so that we get multiple values per 1 MHz frequency bin. By taking the median of these values, we remove any channel-stationary signals arising due to self-generated RFI. Each tuning was observed for 5 minutes. Three antennas were used: CA01, CA03 and CA04. The array was in its H75 configuration, with CA01 on W104, CA03 on N5 and CA04 on W106.

During the integrations, the correlator phase centre was set to be the south celestial pole, with coordinates RA = 0°, Dec = -90°. This is done to prevent fringe rotation, which has the effect of making the effect of the RFI as bad as possible.

The antennas themselves were pointed in three different directions (hence the three parts of the survey).

• For the first part, the antennas were pointed directly at the south celestial pole, where Az = 180°, El ≅ 30°.
• For the second part, the antennas were pointed at the celestial equator at transit, where Az = 0°, El ≅ 60°.
• For the third part, the antennas were stowed, where Az ≅ 90°, El ≅ 85°.

During data reduction, the XX and YY cross-correlations on the three baselines were averaged together to give total intensity. As mentioned before, for each 1 MHz frequency channel, the median value of all the observations of that channel was calculated, and this value is plotted below for each of the three different directions (all plots have the same axes, and the y-axis is shown as a log₁₀ scale.

It would seem that our hypothesis is correct, and that the RFI seen in the 15mm band does come from satellites in the equatorial satellite band. While pointing at the south celestial pole, the total intensity received does not vary much with frequency. In contrast, there is very obvious emission seen while the antennas are pointing towards the equator. The same emission is seen while the antennas are stowed, but at a much lower level.

Where do these emissions come from? One thought is the Sky Muster satellite used by the NBN. This satellite outputs numerous beams to cover the country to provide broadband internet. The beam location pattern can be seen here. From this beam pattern we can see that ATCA would be covered by beam 23, and further details about that beam can be found here. This document (page 48) shows that the Sky Muster to VSAT (the receivers at each user's premises) downlink frequency ranges are 17.7 - 18.2 GHz, 18.8 - 19.3 GHz and 19.7 - 20.2 GHz.

If the emissions are from the Sky Muster satellite, the emission should be strongest from the direction Az = 341.5°, El = 53.1°, as seen from the ATCA. This is approximately 18.9° away from the equatorial pointing location we used for this experiment. This location would be at a declination of +6° 35', at an hour angle of +1h 14m (i.e. west of transit).

The following three figures show the emission seen while pointing at the equator, but with narrower frequency ranges that bracket the three expected Sky Muster bands.

These plots suggest that we are seeing emissions from the Sky Muster satellite, as the emission falls nicely into the expected bands. The emission also seems to be channelised, which is as one might expect given the nature of the service. Not all the channels are being received with the same power, suggesting that some of the emission may be coming from other beams that are not pointing straight at us, like beams 27 and 28.

There is one obvious exception here, with the emission seen slightly above 19.3 GHz; this is unlikely to be from the Sky Muster satellite, and it doesn't look like the other channels.

The recommended continuum observing frequencies in the 15mm band covers the frequency range 15.976 to 20.024 GHz, in two chunks 15.976 - 18.024 GHz and 17.976 - 20.024 GHz. If we moved the lower band down by 300 MHz to cover 15.676 - 17.724 GHz (16.7 GHz centre frequency) we could avoid the lowest downlink frequency band. According to the sensitivity calculator, this move would actually improve the band sensitivity by about 3%. To avoid the highest downlink frequencies, we would have to move the upper band up by 2200 MHz to cover 20.176 to 22.224 GHz (21.2 GHz centre frequency). This, strangely, also improves the sensitivity by about 10%!

A downside to moving the recommended continuum frequencies from 17 & 19 GHz to 16.7 & 21.2 GHz would thus be the resultant non-contiguous coverage, although that is not a huge problem, as can be seen from 4cm observations (5.5 & 9 GHz). Perhaps more importantly though, it would make comparisons between data taken at 17 & 19 GHz and data taken at 16.7 & 21.2 GHz a little more troublesome.