1 Frequency occupation
2 RFI statistics for MeerKAT
- 2.1 UHF Statistics
- 2.2 L-band statistics
- 2.3 Time dependence
- 2.4 Baseline dependence
  - 2.4.1 UHF and L band time dependent summary
3 S-band statistics

SARAO conducts regular Radio Frequency Interference (RFI) monitoring, both by analysing telescope data as well as independent on-site monitoring. However, some RFI may only emerge after deep integrations and may not be detected by standard RFI flaggers. Users who find any anomalies not noted below are encouraged to inform the observatory by raising a MeerKAT Service Desk ticket under ‘RFI’. With the MeerKAT extension work that commenced in 2023, RFI has continued to pose a challenge for MeerKAT telescope operations. To describe the changing RFI environment the RFI statistics reported are generally for the period August 2024 to April 2025. The RFI statistics below are a representation of both environmental factors and instrumental.

Frequency occupation

Frequency ranges of the most persistent sources are described in Table 1 and subsequent figures.

Table 1: Summary of major RFI contaminated regions in the frequency range covered by the MeerKAT receivers.

RFI source	Frequency range (MHz)
Digital TV (UHF)	8 MHz “rectangular” broadband: (306 +Digital TV Channel# x 8) +/- 4 MHz 768 - 778 Vodacom downlink 801 - 811 MTN downlink 811 - 821 Telkom downlink
GSM (Mobile phones) (UHF + L-band)	880 - 915 uplink^* 925 - 960 downlink
Aircraft transponders	Multiple <1 MHz bandwidth intermittent signals between 962 and 1213 MHz
GPS	L1: 1565 - 1585 L2: 1217 - 1237 L3: 1375 - 1387 L5: 1166 - 1186
GLONASS	L1: 1592 - 1610 L2: 1242 - 1249 L3: 1202.025
Galileo	E1: 1575.42 E5a: 1176.45 E5b: 1207.14 E5 AltBOC: 1191.795 E6: 1278.75
Iridium	1616 - 1626
Inmarsat	1526 - 1554
Globalstar	2483.5 - 2495.0
Wi-Fi^*	2400 - 2495
Bluetooth^*	2400.0 - 2483.5

The GSM uplink, Wi-Fi and Bluetooth bands should be clear of RFI on the MeerKAT site and are rarely seen. We request that users who do note RFI in these bands report the frequency range and date and time by raising a MeerKAT Service Desk under ‘RFI’.

RFI statistics for MeerKAT

UHF Statistics

There are no satellite-based transmitters in the UHF band. The primary sources of RFI are GSM, and Digital TV transmissions (DTV). All DTV transmissions are beamed away from site, with the exception of distant transmitters which are occasionally detected - they are generally transmitted over the horizon by anomalous tropospheric ducting. It is expected that the situation will improve in future as further sites discontinue terrestrial transmission.

Figures 1–3 below show the aggregated results of UHF (4K mode) observations carried out between 2024-12-24 10:01:39.381 to 2025-04-12 12:26:17.633, totaling about 51 hours of integration, in random pointing directions. Flags generated by the SDP ingest flagger (prior to calibration) are used for this analysis.

The very faint RFI detections are under investigation. The following frequency ranges are affected by faint but persistent RFI: 647.1, 653.1, 875 MHz.

These often result in positive / negative localized spikes in the target source spectra.

Figure 1: Fraction of the time flagged for all baselines. The grey shaded areas indicate expected RFI bands. Red solid line is the weighted median, the blue solid line is the weighted mean and the blue shaded area is the weighted standard deviation.

Figure 2: Fraction of the time flagged for long baselines. The grey shaded areas indicate expected RFI bands. Red solid line is the weighted median, and the red shaded area is the weighted standard deviation.

Figure 3: Fraction of the time flagged for short baselines. The grey shaded areas indicate expected RFI bands. Blue solid line is the weighted median, blue shaded area is the weighted standard deviation.

SARAO has a couple of in-house tools to monitor and analyse RFI, one particular tool, which will be used here for environmental RFI patterns is KATHPRFI framework. This is a framework that uses offline Data Science Processer (DSP) flagger on MeerKAT Visibility Format (MVF) datasets. In general, the framework works by constructing MASTER and COUNTER arrays, these are multi-dimensional arrays of time, frequency, baseline, elevation and Azimuth for a chosen MeerKAT science observation. For a given MeerKAT science observation, a descriptive statistics about the RFI is contained in the mentioned arrays. The MASTER array for a given dimension counts the number of RFI samples per voxel and COUNTER counts the total number of observations per voxel. For compactness and relevance, in our case we focus only on time and frequency dimensions.

We start by consolidating the RFI statistics for the period of August 2024 to December 2025 (this amounts to approximately 760 total observation hours) for time of the day dependence and frequency band dependence.

First, we consider time-binned fractional RFI flagged for multiple science observations over a given month: Let us provide a precise mathematical description of the computation: Let

be a matrix representation of hourly fractional RFI flagged values for a given observation:

where

is fractional RFI flagged value of an observation x at hour h for month m.

Now for each monthly

, an hourly median is computed to get the robust RFI representation for a given month m. This can be formulated as follow:

Now, we then determine the final profile by smoothing variability across the month by averaging the medians across all the months M as follow:

Figure 4: The above plot presents the fractional RFI flagging (using cal_rfi flags) for the period August 2024 to April 2025, focusing on UHF horizontal cross-pol observations. The flagged RFI IS shown as a function of time of the day (UTC). We observe that RFI levels are relatively high in the early hours then remain relatively low in the late mornings to early afternoon, then rise again irregularly in the afternoon towards evening across all the months.

Figure 5: The above plot presents the fractional RFI flagging for the period August 2024 to April 2025 UHF horizontal cross pol observations, shown as a function of Frequency in MHz. The frequency spectra above appears consistent with the MeerKAT frequency occupation ranges table.

Figure 6: RFI occupancy as a function of time and frequency. Bright regions indicate persistent RFI contamination, showing distinct irregular patterns with elevated interference during the morning and afternoon to evening period. Horizontal strips correspond to known frequency transmitters (see Table 1), while vertical features reflect time-dependent environmental sources.

L-band statistics

There are three major contributors to RFI in the MeerKAT L-band frequency range: Global System for Mobile Communication (GSM), Distance Measurement Equipment (DME) on aircraft and Global Navigation Satellite System (GNSS) satellites.

We show an example of the typical RFI environment, as observed during a 30 hour run on primary calibrators over Christmas Day 2019, in Figures 7 through 9. The science data processor (SDP) pipeline flagger output is also shown.

Note that, in the interests of speed, the online flagger uses a very aggressive static mask (based on long term observation statistics), which masks all known transmission bands on baselines out to 1 km. This enables reasonable calibration solutions to be produced. Users may wish to disable these flags on their target fields and adopt their own flagging strategies. A second pass at flagging after applying calibration solutions also improves the detection of fainter RFI signals.

Figure 7: RFI frequency occupancy for a 30 hour continuous observation. The blue line shows flagging on core baselines (<1 km), which includes a static mask as well as outlier detection, while the orange line shows flagging of the longer baselines. Only a single polarisation (XX) is shown. The YY polarisation shows similar behaviour.

Figure 8: As for Figure 1, but zoomed in on the quietest portion of L-band. Galactic HI can be seen (and is often flagged by the calibration pipeline) at 1420 MHz. The GPS L3 signal is sporadic but seen fairly often.

Time dependence

Activity from aircraft transponders is at a minimum from ~23:00 to 06:00 SAST under typical operating conditions. Figure 9 is of an observation that includes this window.

Figure 9: Fraction of baselines flagged per scan. The time range shown here runs from 2019-12-24 22:59 UTC (bottom) to 2019-12-26 04:29 UTC (top).

Baseline dependence

The frequency ranges in Table 1 are included in the SDP pipeline static mask but are only applied on baselines shorter than 1 km (see Figure 10) since most of the RFI decorrelates on longer baselines. Note that this is an aggregate mask built up over time, and it is possible to recover more channels on an individual observation if manual or outlier-based flagging is used instead.

Figure 10: Fraction of time that baselines of a particular length are flagged. Note that the antenna separation axis is not linear. The time range of the observation is 2019-12-24 22:59 to 2019-12-26 04:29 UTC. The SDP calibration flag report also shows flags from calibration solutions. Horizontal lines of flags would correspond to baselines to an antenna that has been flagged either for a fraction of time or for the entire observation.

The below Figures (11 -13), depicts time and frequency dependence statistics for the MeerKAT L-band receiver, our datasets represents ~ 1610 hours of observation starting from August 2024 to April 2025.

Figure 11: The above plot presents the fractional RFI flagging (using cal_rfi flags) for the period August 2024 to April 2025, focusing on L-band horizontal cross-pol observations. The flagged RFI is shown as a function of time of the day (UTC). We observe that RFI levels are relatively high in the early morning hours, decrease to lower levels from the late morning to early afternoon, rise again in the evening, and drop after 17h00 UTC - a pattern consistent across all the months.

Figure 12: The above plot presents the fractional RFI flagging (using cal_rfi flags) for the period August 2024 to April 2025, focusing on L band horizontal crossipol observations. The flagged RFI shown as a function of Frequency in MHz. The frequency spectra above, for the most parts appears consistent with the MeerKAT frequency occupation ranges table, except for frequencies between 1240 MHz to 1320 MHz, 1440 MHz and 1680 MHz.

Figure 13: RFI occupancy as a function of time and frequency. Bright regions indicate persistent RFI contamination, showing distinct irregular patterns with elevated interference during the morning and evening period. Horizontal strips correspond to known frequency transmitters (see Table 1), while vertical features reflect time-dependent environmental sources.

UHF and L band time dependent summary

The RFI levels for MeerKAT science observation remain relatively stable during the construction phase, as evidenced by data from our in-house SDP flagger using cal_rfi flags. This analysis presents a high level overview of RFI impacts on MeerKAT science based observations; note that the intrinsic short-term, observation-specific RFI variations may not be apparent as the figures show central tendencies aggregated over nine months (August 2025 - April 2025). The presented statistics reveals distinct patterns in RFI occurrence across UHF and L-band and times of the day. In the L-band, we observe consistently higher RFI probabilities ranging between 29-35%, while the UHF band shows lower interference levels of 16-24%. Both bands exhibit similar diurnal patterns, with RFI peaks occurring during early morning and evening hours, likely corresponding to periods of increased human activity.

The frequency-dependent characteristics of the RFI align with known MeerKAT interference patterns across the radio spectrum. The spectral distribution shows concentrations in expected regions, including satellite communication bands and terrestrial transmitter frequencies. These results confirm that MeerKAT's data quality remains well-protected against RFI under current construction conditions.

S-band statistics

We have started to build statistics for S-band (4K mode). The results in Figure 8 are obtained by aggregating several datasets collected between 2023-02-22 to 2023-08-25. Some of the transmitters are known, however there are those that are still under investigation, e.g. detections around 2200 MHz, and those above 2600 MHz.

Figure 14: Fraction of the time flagged for all baselines. The grey shaded areas indicate expected RFI bands. The blue, red, black, cyan and green solid lines are the means of S0, S1, S2, S3 and S4 respectively. The colour coding of the standard deviation of the sub-bands is similar to that of the mean.