Radio-Frequency Interference at the McGill Arctic Research Station
T. Dyson, H. C. Chiang, E. Egan, N. Ghazi, T. Menard, R. A. Monsalve, T. Moso, J. Peterson, J. L. Sievers, S. Tartakovsky
JJournal of Astronomical Instrumentation © World Scientific Publishing Company
Radio-Frequency Interference at the McGill Arctic Research Station
T. Dyson , † , H. C. Chiang , , E. Egan , N. Ghazi , T. M´enard , R. A. Monsalve , , , T. Moso , J. Peterson ,J. L. Sievers , , S. Tartakovsky Department of Physics, McGill University, Montr´eal, Qu´ebec H3A 2T8, Canada School of Mathematics, Statistics, and Computer Science, University of KwaZulu-Natal, Durban 4000, South Africa School of Earth and Space Exploration, Arizona State University, Arizona 85287, USA Facultad de Ingenier´ıa, Universidad Cat´olica de la Sant´ısima Concepci´on, Alonso de Ribera 2850, Concepci´on, Chile Department of Physics, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, USA School of Chemistry and Physics, University of KwaZulu-Natal, Durban 4000, South Africa
Received (to be inserted by publisher); Revised (to be inserted by publisher); Accepted (to be inserted by publisher);The frequencies of interest for redshifted 21 cm observations are heavily affected by terrestrial radio-frequencyinterference (RFI). We identify the McGill Arctic Research Station (MARS) as a new RFI-quiet site and reportits RFI occupancy using 122 hours of data taken with a prototype antenna station developed for the Array ofLong-Baseline Antennas for Taking Radio Observations from the Sub-Antarctic. Using an RFI flagging processtailored to the MARS data, we find an overall RFI occupancy of 1.8% averaged over 20–125 MHz. In particular,the FM broadcast band (88–108 MHz) is found to have an RFI occupancy of at most 1.6%. The data were takenduring the Arctic summer, when degraded ionospheric conditions and an active research base contributed toincreased RFI. The results quoted here therefore represent the maximum-level RFI environment at MARS.
Keywords : radio astronomy; site testing
1. Introduction
One of the greatest challenges facing contemporary radio astronomy experiments is terrestrial radio-frequency interference (RFI), which has steadily worsened over time as the globe has been populatedwith an increasing number of transmitters and other radiating sources. Radio astronomy experiments areoften forced to operate from remote locations, where the RFI background is minimized, but the remotenessrequirement is directly at odds with the simultaneous need for accessibility and logistical infrastructure.The need for RFI-quiet locations is especially important for cosmic dawn experiments measuring globallyaveraged 21-cm emission of neutral hydrogen (DiLullo et al., 2020; Nhan et al., 2019; Philip et al., 2019;Bowman et al., 2018; Singh et al., 2018). Because these experiments measure total power, the require-ments on background RFI levels are far more stringent than the typical requirements for interferometricexperiments, which benefit from cross-correlation. The problem is further compounded by the observingfrequency range for cosmic dawn ( ∼ a as a new location with an exception-ally quiet RFI environment that can serve as an observing site for future low-frequency radio astronomyexperiments. This paper presents spectral measurements below 125 MHz that were taken from MARS dur-ing July 2019, the methodology for identifying RFI in the data, and an assessment of the RFI occupancy.
2. Instrument
The primary data presented in this paper were recorded by a single pathfinder antenna station thatwas developed for the Array of Long Baseline Antennas for Taking Radio Observations from the Sub- † Corresponding author: [email protected] a a r X i v : . [ a s t r o - ph . I M ] D ec T. Dyson et al.
LWA antenna Activebalun35 dB 20-m coaxRG-58 Bias teeZFBT-4R2GW-FT+ ⨉ Fig. 1: Block diagram of the ALBATROS pathfinder antenna installed at MARS. A dual-polarization LWAantenna, equipped with a front-end active balun, connects via 20-m coaxial cables to the back-end readoutelectronics, housed in a Faraday cage denoted by the dashed lines. Each of the two antenna signals is passedto a second-stage electronics chain consisting of filters and further amplification. The signals are digitizedat 250 Msamp/s by a SNAP board, which computes channelized baseband data and spectra. The systemis powered by 12 V batteries that are manually recharged.Antarctic (ALBATROS; Chiang et al., 2020). Figure 1 shows a schematic diagram of the ALBATROSstation. Incoming signals are received by a dual-polarization Long Wavelength Array (LWA) dipole an-tenna (Ellingson & Kramer, 2005), outfitted with an active-balun front-end electronics (FEE) module thatprovides ∼
37 dB gain (Hicks et al., 2012). The FEE is powered with 16 V, which is passed through thecoaxial cable via bias tees. The back-end RF electronics consist of high- and low-pass filters from Minicir-cuits that band-limit the signal to 1.2–155 MHz, and amplifiers and attenuators that together provide anadditional ∼
44 dB gain. A Smart Network ADC Processor (SNAP; Hickish et al., 2016) board digitizesthe RF signals at 250 Msamp/s, and the ADCs are locked to a 10-MHz reference produced by a TrimbleThunderbolt E GPS-disciplined clock module. The SNAP FPGA computes auto- and cross-spectra of thetwo inputs over the full 0–125 MHz frequency range, with 2048 frequency channels and accumulation overfew-second intervals. (The FPGA also computes channelized baseband data for each polarization over tun-able frequency windows within the 0–125 MHz operating range, but these data are not used in the analysispresented here.) The low-pass filter cutoff in the back-end is intentionally set higher than the Nyquistfrequency to alias in the 137–138 MHz downlink signal from the ORBCOMM satellite constellation. Thus,any RFI observed above 95 MHz may have been aliased from 125–155 MHz. A Raspberry Pi 3B+ singleboard computer controls the SNAP board and receives the auto- and cross-spectra via GPIO connections,and the spectra are saved to an on-board SD card. The back-end electronics (within the dashed box inFigure 1) are portable, but the LWA antenna and front-end are not. At the sites surrounding MARS (Fig-ure 2), RFI measurements were taken with a LoWavz LW-10K60M antenna outfitted with its impedencematcher, a Mini-Circuits ZFL-500LN+ amplifier providing ∼
24 dB gain, and a bias tee to block the DC adio-Frequency Interference at the McGill Arctic Research Station Site Label Latitude Longitude( ◦ N) ( ◦ W)MARS 79.415 90.7491 79.402 91.2092 79.370 90.9533 79.343 90.6014 79.425 90.6545 79.355 90.7266 79.457 90.801Table 1: Locations of RFI survey sites Fig. 2: Google Earth satellite map of Expedition Fjord with themain MARS base and survey sites marked. Survey sites are num-bered in the order visited.voltage supplied by the back-end electronics. While the LoWavz antenna is not sufficiently sensitive forcosmological observations, especially above 60 MHz, it suffices for an RFI survey.
3. Observations
MARS is a small research base located at 79 ◦ (cid:48) N, 90 ◦ (cid:48) W, approximately 40 km inland at the headof Expedition Fjord on Axel Heiberg Island, Nunavut. MARS is accessible during April–August via smallchartered aircraft from Resolute Bay (540 km south). The closest radio transmitters are Eureka (120 kmnortheast), Grise Fjord (360 km south), Resolute Bay, and Alert (590 km northeast), suggesting that MARSand its surroundings should have low levels of RFI. Furthermore, MARS is attractive for low-frequency radioastronomy because favourable ionospheric conditions are expected at high latitudes, especially during theArctic winter when the polar ionosphere is only weakly ionized in the prolonged absence of solar radiation.Simulations using the International Reference Ionosphere model (Bilitza, 2018) suggest that the summerand winter plasma cutoff frequencies differ by about a factor of two at MARS. Additionally, measurementstaken during the Arctic winter are free of solar RFI during the continuous night. Thus, due to its isolationfrom radio transmitters, its polar latitude, and its reasonable accessibility, MARS is a promising candidatelocation for ground-based low-frequency radio astronomy and is being investigated as a potential site forthe ALBATROS and MIST b experiments. To assess the RFI environment at MARS, observations weretaken using the instrumentation described in § MARS base
The RFI measurements discussed in this paper are derived from data taken at the MARS base duringJuly 9–21, 2019. A total of 122 and 106 hours of data were recorded from the east-west and north-southantenna polarizations, respectively. During the summer period, the sun remains above the horizon continu-ally, raising the ionospheric plasma cutoff frequency in comparison to winter conditions. Furthermore, when b T. Dyson et al. the base is actively used by researchers, power is supplied primarily by solar panels and a power inverter,which generate a large amount of RFI. Data were therefore recorded during overnight periods while theMARS power system was turned off to eliminate the largest source of local RFI. Some weaker sources oflocal RFI (e.g., personal electronics) remained present, particularly as the base population increased duringthe summer period. Because of slightly elevated local RFI and degraded ionospheric conditions, the datapresented here likely represents the worst-case RFI environment at MARS.
Survey sites
To investigate potential local variation in the RFI environment, data were also recorded at several siteswithin 10 km of MARS, as shown in Figure 2 and listed in Table 1. These sites are candidate locations forfuture ALBATROS antenna installations because they have suitable terrain and are accessible by foot. Alldata at these sites were taken while the MARS power inverter and electronics were on; however, the RFIemitted at the base is not visible at few-km separation distances. The observing period at sites 1–5 was ∼
20 minutes each, and data were recorded at site 6 for ∼
4. RFI identification methodology
The MARS RFI environment was characterized by analyzing autospectra with the processing steps de-scribed below, which focus on identifying narrow-band features. Each of the two antenna polarizationsfor each night of data, comprising ∼ . σ , and the flagging threshold of 5 MADs therefore corresponds to 3.35 σ .The expected false flagging rate, corresponding to the percentage of noise > . σ , is thus 0.04% at aminimum (variations in noise level during an observation due to changing Galactic brightness could causehigher false flagging rates).The above flagging procedure has two main limitations. First, the flagging breaks down for RFI-saturated frequency channels, where the true RFI occupancy is above 50%. During the Arctic summer,shortwave and HF radio reflected from the ionosphere cause this saturation at frequencies below roughly20 MHz. Although RFI occupancy is calculated over the full frequency range, the values below 20 MHz arelikely underestimated and are omitted when computing the average levels. Second, the flagging procedureis insensitive to low-intensity broadband RFI features, which cannot be distinguished from broadbandnon-RFI signal variation (such as gain fluctuations). The removal of these types of signals can be seenin Figure 3, where faint horizontal striping features are subtracted between the second and third panels.Because the flagging procedure can potentially remove bright broadband RFI as well, we searched for thesehigh-amplitude events separately by examining the distribution of root mean square (RMS) values of each adio-Frequency Interference at the McGill Arctic Research Station Fig. 3: This figure shows a segment of east-west polarized autospectra recorded at MARS on the night ofJuly 11 th that is representative of the entire dataset. The top panel shows the raw power spectra, while theother panels show the data after each step of the flagging process (see § §
5. Results and conclusion
Figure 4 shows the RFI occupancy as a function of frequency for all data recorded at the MARS base.Excluding frequencies below 20 MHz, the overall RFI occupancy is 1.6% in east-west polarized data and2.0% in north-south polarized data.At frequencies above the 20 MHz ionospheric cutoff, the RFI environment of MARS is exceptionallyclean. The RFI occupancy over the FM band (88–108 MHz), averaged over each polarization, is 1.6%.Because the low-pass filter in the readout electronics has a 155 MHz cutoff that lies above the Nyquist
T. Dyson et al.
Fig. 4: Overall RFI occupancy at each frequency channel from data taken at the MARS. The green andblue lines correspond to the RFI occupancy in east-west and north-south polarized data, respectively. Thehorizontal blue and green dashed lines show the mean occupancy across all channels above 20 MHz foreach polarization. The red line indicates the 0 .
04% minimum false flag rate expected from Gaussian noise.The dashed line at 95 MHz indicates that any signal at higher frequency could have been aliased from125–155 MHz (see § (cid:38)
100 MHz arelikely air-to-ground sources such as aeronautical radionavigation or weather and mobile satellites c . The fullMARS data set exhibits a single meteor event, when distant radio transmitting sources reflect off of themeteor’s ionization trails and appear briefly in the observed spectra (Mallama & Espenak, 1999). During c adio-Frequency Interference at the McGill Arctic Research Station the observed meteor event, which encompassed only one time sample, a dense cluster of RFI lines appearedin the range of 70–100 MHz, corresponding to frequencies allocated to ground-based broadcasts.The RFI occupancy at the remote sites surrounding MARS (Figure 2) is qualitatively consistent withthe measurements recorded at the base, taking into account greater uncertainty from the significantlysmaller data volumes at the remote sites and the different response of the antenna used during surveying.The results suggest that locally-generated RFI at MARS is not visible at few-km distances (measurementswere recorded while the MARS power system was turned on) and that there are no additional major localsources of RFI.The results presented here indicate that MARS is exceptionally clear of terrestrial RFI, even duringsummer conditions when the RFI levels are expected to be at their maximum. Observations that takeplace throughout the Arctic winter will further benefit from improved ionospheric conditions and reducedlevels of local RFI from the absence of human activity at the research base. The quiet RFI conditions,in combination with well developed infrastructure and regular summer access, make MARS an excellentcandidate site for the development of radio astronomy experiments. MARS is well suited for the deploymentof antennas that occupy relatively small footprints, and the location is therefore particularly promisingfor future observations of cosmic dawn, exploratory measurements for the cosmic dark ages, and radiotransients from ultra-high energy neutrinos. Acknowledgments