1
Signal Mode Transition Detection in Starlink LEO
Satellite Downlink Signals
Mohammad Neinavaie and Zaher M. Kassas
Department of Electrical and Computer Engineering
The Ohio State University, Columbus, OH, USA
Abstract—A receiver architecture for detection and tracking
of Starlink orthogonal frequency division multiplexing (OFDM)-
based signals is proposed. The proposed receiver enables ex-
ploiting all the transmitted periodic beacons of Starlink low
Earth orbit (LEO) signals to draw carrier phase, code phase,
and Doppler observables. The reference signals (RSs) of modern
OFDM-based systems contain both always-on and on-demand
components. These components can be unknown and subject
to dynamic transmission modes. Thanks to a matched subspace-
based detection algorithm, the proposed receiver is shown to
be capable of cognitive detection of both always-on and on-
demand components in the Starlink OFDM-based RSs. It is
shown that despite the dynamic nature of Starlink RSs, the
proposed matched subspace detector senses the transition between
the transmission modes of Starlink RSs, and detects all the
accessible RSs with a predetermined probability of false alarm.
Experimental results are provided to validate the performance of
the proposed receiver in transmission mode detection in Starlink
downlink signals.
Index Terms—Positioning, navigation, signals of opportunity,
low Earth orbit satellite, Starlink, OFDM, 5G, on-demand.
I. INTRODUCTION
Abundant man-made terrestrial and extraterrestrial signals
of opportunity (SOPs) have been shown to possess promising
features for positioning, navigation, and timing [1]–[4]. High
bandwidth and diverse synchronization signals of orthogonal
frequency division multiplexing (OFDM) signals in cellular
fourth-generation (4G) long-term evolution (LTE) and 5G new
radio (NR) systems enabled meter-level and decimeter-level
navigation on ground vehicles [5], [6] and unmanned aerial
vehicles (UAVs) [7], [8], respectively. Similarly to 4G LTE
and 5G NR, Starlink low Earth orbit (LEO) space vehicles
(SVs) also adopt OFDM [9] signals with considerably high
bandwidth [10]. While a single LTE channel has a bandwidth
of up to 20 MHz, the bandwidth of a single 5G NR channel
goes up to 100 MHz and 400 MHz for FR1 and FR2,
respectively [11]. On the other hand, Starlink downlink signals
occupy 250 MHz bandwidth of the Ku band to provide
high rate broadband connectivity [12]. The OFDM reference
signals (RSs) are spread across the whole bandwidth, which
promises good correlation properties, leading to high ranging
and localization accuracy.
This work was supported in part by the Office of Naval Research (ONR)
under Grant N00014-22-1-2242, in part by the Air Force Office of Scientific
Research (AFOSR) under Grant FA9550-22-1-0476, and in part by the U.S.
Department of Transportation (USDOT) under Grant 69A3552047138 for the
CARMEN University Transportation Center (UTC).
SOP-based navigation receivers typically rely on known
synchronization sequences or beacons transmitted by SOP
sources to draw time-of-arrival (TOA), direction-of-arrival
(DOA), and frequency-of-arrival (FOA) measurements [13].
Due to the unknown and dynamic nature of modern commu-
nication signals in private networks, such as Starlink, a naviga-
tion receiver that is based on reverse engineering the downlink
signals either (i) fails to exploit the whole available bandwidth
unless all RSs get determined or (ii) fails to operate if the
operator changes their signal. As such, designing receivers that
can cognitively acquire partially known, unknown, or dynamic
beacon signals is an emerging need for the future of cognitive
opportunistic navigation [14]–[17].
Cognitive opportunistic navigation [17] has recently been
introduced to address the following challenges of navigation
with SOPs in modern and private networks. First, opportunistic
navigation frameworks usually exploit the broadcast RSs for
navigation [13]. In public networks, these signals are known
by the user equipment (UE) and are universal across net-
work operators. Hence, they can be exploited for positioning
without the need for the UE to be a network subscriber.
However, in private networks, the signal specifications may
not be available to the public or are subject to change, which
makes acquiring and tracking these signals impossible for
conventional opportunistic navigation receivers [17]. Second,
conventional cellular networks broadcast RSs at regular and
known time intervals, regardless of the number of UEs in the
environments (e.g., the cell-specific reference signal (CRS)
in LTE). Modern communication systems, such as 5G NR,
minimize the transmission of always-on signals, by adopting
an ultra-lean design which entails transmitting some of the
RSs only when necessary or on-demand [18].
Matched subspace detectors have been widely adopted to
solve the detection problem of sources with unknown param-
eters in the presence of other interfering sources [19], [20]. In
the signal processing literature, matched subspace detectors
were used to detect the unknown signal activities in multiple-
input multiple-output (MIMO) radars, passive bistatic radars,
and blind array signal processing [21]–[23]. Recently, machine
learning approaches have also been proposed for unknown
transmitter detection, identification, and classification [24],
[25]. In the navigation literature, the detection of unknown
signals has been studied to design frameworks that are capable
of navigating with unknown or partially known signals. The
problem of detecting Galileo and Compass satellites signals
