Satellites are radiating into protected bands, and the Radio Regulations do not cover the mechanism

The measurement, not the model

Radio telescopes in Western Australia have recorded emissions from Starlink satellites in bands where those satellites are licensed to transmit nothing at all. Over roughly 29 days with an SKA-Low station prototype, Grigg, Tingay and Sokolowski logged 112,534 detections of 1,806 unique Starlink satellites between 73 MHz and 235 MHz ^[1]. 703 were detected inside 150.05–153.00 MHz, protected for radio astronomy; thirteen inside 73.00–74.60 MHz, also protected ^[1].

The levels are not marginal. Mean flux density was 93 Jy/beam, peaking at 312 Jy/beam at 170.3 MHz, and in the worst datasets a Starlink satellite appears in about 30 per cent of all images ^[1]. Detecting the epoch of reionisation requires sensitivity at roughly the millijansky level, which the authors state places such science beyond reach without mitigation ^[1].

These emissions are not downlinks. They are unintended electromagnetic radiation: leakage from power supplies, clocks and digital subsystems, plus passive intermodulation in spacecraft structure ^[2]. An earlier all-sky survey at the same site — 13 bands of roughly 1 MHz each, about 20 days, 1.6 million images — identified 152 satellites and found several decommissioned spacecraft apparently radiating when their solar panels were illuminated ^[3]. A satellite that no longer has an operator can still be an emitter.

The rules cover two of three signal classes

Balthasar Indermuehle of CSIRO, who chairs ITU-R Working Party 7D, and Federico Di Vruno of the SKA Observatory, its vice-chair, divide satellite emissions into three categories ^[4]. Intended emissions are deliberate transmissions, powerful by design and closely regulated. Unwanted emissions are out-of-band and spurious residue; Appendix 3 of the Radio Regulations already limits these, and the authors argue the need there is enforcement, not new rules. The third category is unintended radiation, and on this they are explicit: these signals are not explicitly addressed in most spectrum management frameworks ^[4]. Spacecraft electromagnetic compatibility standards exist to stop a satellite interfering with itself, not with a ground receiver ^[4].

The criteria are old. Recommendation ITU-R RA.769-2, which sets interference thresholds for radio astronomical measurements and is incorporated by reference in the Radio Regulations, dates from May 2003 ^[5]. It permits no more than a 10 per cent loss of sensitivity ^[2]. Resolution 739 sets per-system equivalent power flux-density limits, but only up to 22 GHz ^[6].

How far past the threshold

Two analyses published in the last two months quantify the exceedance, both from measurements rather than filings.

The first reinterprets all-sky interferometric imaging as a set of simultaneous beamformed measurements, which allows the ITU-R equivalent power flux-density method to be applied to observed data instead of modelled satellite contributions ^[7]. Using detected unintended radiation from Starlink satellites at 150 MHz, it finds 50–70 per cent of samples exceed radio astronomy protection thresholds, and calculates that 16.3–18.6 dB of uniform attenuation would be needed to meet the 2 per cent single-system criterion in Recommendation ITU-R RA.1513-2 ^[7]. Exceedance differs by polarisation: 63–72 per cent in YY against 48–63 per cent in XX ^[7]. The stated reason for the approach is that unintended radiation cannot be reliably modelled, because the emitting subsystem's characteristics are unknown ^[7].

The second measures per-satellite emission levels directly, then scales them. Harmonic emission from the Starlink direct-to-device downlink at 5240 MHz already causes 59 per cent data loss in geodetic very long baseline interferometry, and the 2620 MHz downlink would need only about nine satellites to breach the VLBI threshold ^[8]. At 300,000 satellites, a figure the authors treat as plausible mid-term, second-harmonic data loss reaches 100 per cent. Compliance would demand per-satellite limits of 65–70 dB(µV/m) at 10 m; present spacecraft exceed that by roughly 20 dB ^[8]. The proposed remedy is an inverted criterion, under which permitted per-satellite radiation tightens as the constellation grows, aimed at WRC-31 ^[8].

The population changed faster than the framework

Indermuehle records that active satellites in low Earth orbit have risen roughly tenfold since WRC-19, from about 1,500 to over 15,000, and that credible ITU filings point above 100,000 spacecraft within the operational lifetime of systems now being commissioned ^[6]. The consequence he identifies is a change in the nature of the problem: interference has moved from sparse, identifiable transmitters whose position can be predicted and avoided, to dense, time-varying ensembles of emitters across the visible sky ^[6]. Aggregate equivalent power flux-density has displaced single-entry interference as the figure of merit ^[6].

The sensitivity gap explains why leakage matters at all. Radio telescopes work at roughly 10⁻²⁹ W m⁻² Hz⁻¹; satellite downlinks operate at 10⁻¹³ to 10⁻¹⁶ W m⁻² Hz⁻¹ ^[2]. Below 30 GHz, less than 2.5 per cent of spectrum is allocated to the radio astronomy service and only 1.3 per cent is exclusive to passive use ^[2]. Indermuehle states that multi-band campaigns at Australian facilities expose receivers to levels exceeding RA.769 thresholds by many orders of magnitude even where the transmitters operate in nominal compliance with their assigned masks ^[6].

The regulatory clock, and the regional track

Five WRC-27 agenda items bear on this. Item 1.16, held by Working Party 7D, covers provisions to protect radio astronomy from aggregate interference caused by non-geostationary systems. Item 1.18 covers protection above 76 GHz and is shared between WP 7C and WP 7D. Items 1.12, 1.13 and 1.14, all held by WP 4C, concern new mobile-satellite allocations, including direct connectivity between space stations and IMT user equipment ^[9]. Di Vruno and Piero Benvenuti note that 1.16 and 1.18 are new items for this cycle ^[10].

Asia-Pacific positions are formed before Geneva, not at it. The third meeting of the APT Conference Preparatory Group for WRC-27 met in Sapporo from 27 to 31 July 2026, handling more than 142 contributions ^[11]. Regional preliminary views developed there carry into the conference. Malaysian policy on non-geostationary systems and direct-to-cell allocations is shaped in that room, not after the fact.

What this means for practitioners outside astronomy

Radio astronomy hosts the most sensitive receivers, so it sees the problem first. The mechanism is general. Any organisation doing RF survey work, emitter geolocation or interference attribution now works beneath a noise floor containing thousands of incidental emitters that appear in no filing, change as spacecraft are launched and retired, and sometimes belong to objects nobody operates ^[3]. Three consequences follow. Unintended radiation cannot be characterised from documentation, because the emitting subsystem is not what was filed ^[7]. Measured baselines have a shelf life, because the population grew tenfold in one conference cycle ^[6]. And attribution requires the three signal classes to be separated before cause is assigned, since an emission inside a protected band may come from a transmitter in full compliance with its mask ^{[4] [6]}.

Actions

• Treat wide-field or all-sky imaging as a spectrum monitoring instrument. The measurement-based EPFD method exists because modelled contributions understate the observed case, and measured baselines should be recorded per band with timestamps and correlated against satellite ephemerides, so a detection can be tied to an object rather than filed as unexplained interference ^{[3] [7]}.
• Extend electromagnetic compatibility test plans to passive science bands, covering filter roll-off, amplifier non-linearity, harmonics and passive intermodulation, not only in-band performance ^[2].
• Pursue boresight avoidance and ephemeris sharing with operators where coordination is possible. Steering spot beams away from sensitive sites is in use already and needs no new regulation ^{[2] [10]}.
• For administrations: positions on items 1.13 and 1.16 form through the APT preparatory process, and the per-satellite limit proposal is aimed at WRC-31, so this cycle determines the next decade ^{[8] [9] [11]}.

Domain note

Electronic intelligence rests on knowing which emitters exist, where they are, and what they are supposed to be doing. Measurements from one site in Western Australia now show hundreds of spacecraft radiating in bands they are not licensed for, at levels orders of magnitude above the protection criteria, through a mechanism no spectrum management framework regulates. These objects are catalogued as orbital hardware and uncatalogued as signals. That is a gap in the electromagnetic picture, and the instrument that found it was a radio telescope rather than a monitoring receiver.