Remotesensing 14 00250

remote sensing
Article
An eLoran Signal Cycle Identification Method Based on Joint
Time–Frequency Domain
Wenhe Yan 1,2,3, * , Ming Dong 4 , Shifeng Li 1,2,3 , Chaozhong Yang 1,3 , Jiangbin Yuan 1,3 , Zhaopeng Hu 1,3
and Yu Hua 1,2,3
1 National Time Service Center, Chinese Academy of Sciences, Xi’an 710600, China; lishifeng@ntsc.ac.cn (S.L.);
ycz@ntsc.ac.cn (C.Y.); yuanjiangbin@ntsc.ac.cn (J.Y.); huzhaopeng@ntsc.ac.cn (Z.H.); hy@ntsc.ac.cn (Y.H.)
2 University of Chinese Academy of Sciences, Beijing 100049, China
3 Key Laboratory of Precise Positioning and Timing Technology, Chinese Academy of Sciences,
Xi’an 710600, China
4 Beijing Institute of Tracking and Telecommunications Technology, Beijing 100094, China;
dongmingdmdm1985@126.com
* Correspondence: ywh@ntsc.ac.cn
Abstract: The eLoran system is an international standardized positioning, navigation, and timing
service system, which can complement global navigation satellite systems to cope with navigation
and timing warfare. The eLoran receiver measures time-of-arrival (TOA) through cycle identifica-
tion, which is key in determining timing and positioning accuracy. However, noise and skywave
interference can cause cycle identification errors, resulting in TOA-measurement errors that are
integral multiples of 10 µs. Therefore, this article proposes a cycle identification method in the joint
time–frequency domain. Based on the spectrum-division method to determine the cycle identification
range, the time–domain peak-to-peak ratio and waveform matching are used for accurate cycle
identification. The performance of the method is analyzed via simulation. When the signal-to-noise

ratio (SNR) ≥ 0 dB and skywave-to-groundwave ratio (SGR) ≤ 23 dB, the success rate of cycle
Citation: Yan, W.; Dong, M.; Li, S.; identification is 100%; when SNR ≥ −13 dB and SGR ≤ 23 dB, the success rate exceeds 75%. To
Yang, C.; Yuan, J.; Hu, Z.; Hua, Y. An verify its practicability, the method was implemented in the eLoran receiver and tested at three test
eLoran Signal Cycle Identification sites within 1000 km using actual signals emitted by an eLoran system. The results show that the
Method Based on Joint method has a high identification probability and can be used in modern eLoran receivers to improve
Time–Frequency Domain. Remote TOA-measurement accuracy.
Sens. 2022, 14, 250. https://
doi.org/10.3390/rs14020250 Keywords: eLoran; cycle identification; skywave interference; time of arrival; spectrum division
Academic Editor: José Darrozes
Received: 27 November 2021

Accepted: 4 January 2022
1. Introduction
Published: 6 January 2022
With continuous improvements in informatization, positioning, navigation, and tim-
Publisher’s Note: MDPI stays neutral ing (PNT) systems have become indispensable for social development and modern na-
with regard to jurisdictional claims in tional defense infrastructure, reflecting a country’s international status and comprehensive
published maps and institutional affil-
strength [1–3]. Although global navigation satellite systems (GNSS) are currently the most
iations.
widely used type of PNT system, GNSS signals are weak, ineffective at penetration, and
susceptible to interference. GNSS systems are thus not suitable for applications in special
scenarios, including underground, underwater, and indoors [4–6]. Therefore, relying solely
Copyright: © 2022 by the authors.
on GNSS involves serious risks.
Licensee MDPI, Basel, Switzerland.
The enhanced long-range navigation (eLoran) system evolved from the Loran-C
This article is an open access article
system. It is an internationally standardized medium and a long-range land-based radio
distributed under the terms and system [7,8]. It can satisfy the requirements of PNT applications in most fields in terms of
conditions of the Creative Commons accuracy, reliability, and continuity. It can also provide a time service better than 100 ns
Attribution (CC BY) license (https:// and a position service of 20 m after differential correction [7,9]. When a GNSS signal is
creativecommons.org/licenses/by/ unavailable, the eLoran system can be used as a backup, which reduces the risk caused
4.0/). by relying on GNSS; this also allows for coping with timing warfare and navigation
Remote Sens. 2022, 14, 250. https://doi.org/10.3390/rs14020250 https://www.mdpi.com/journal/remotesensing

Remote Sens. 2022, 14, 250 2 of 20
warfare in modern war [10–12]. Therefore, several countries are developing and upgrading
their eLoran systems. In August 2019, the US Department of Defense (DoD) released a
report titled “Strategy for the DoD Positioning, Navigation, and Timing (PNT) Enterprise—
ensuring a US military PNT advantage”, which stated that Loran-C should be listed as the
PNT source, second only to GNSS [13]. China also planned to build three eLoran stations
in its western region in 2018 to achieve national coverage of eLoran signals together with
the existing eLoran system [14]. The construction of the three stations officially started in
2020. Russia has developed a new land-based radio navigation system called “Scorpion,”
with the aim of achieving balanced development of satellite-based and land-based PNT
systems [15]. South Korea began to develop and deploy its own eLoran system in 2013
due to the deliberate interference of North Korea on its GPS; in 2021, it was announced
that the developed eLoran system had commenced operation [16,17]. Although eLoran can
be used as to supplement GNSS, the current global eLoran application technology is not
sufficiently advanced. Compared with those of GNSS, the accuracy and reliability of the
application terminals of eLoran remain inadequate. Therefore, eLoran cannot fully satisfy
the requirements of users of PNT services.
The eLoran signal is a long-wave (LW) signal with a frequency of 100 kHz. According
to the propagation theory of this frequency band, the eLoran signal is a composite wave
consisting of a groundwave signal and a skywave signal [18,19]. The eLoran receiver
provides timing and positioning services by measuring the time-of-arrival (TOA) of the
groundwave signal in the composite wave. Cycle identification is an important task for
the eLoran receiver. It involves correctly identifying the standard zero-crossing (SZC)
of the eLoran ground wave signal to measure the TOA [20,21]. Because the SZC point
of the eLoran signal has no prominent characteristics, the skywave interference (SWI)
signal generated by ionospheric reflection arrives at the receiver a certain cycle after the
groundwave signal does, which affects the cycle identification result. The carrier frequency
of the eLoran signal is 100 kHz. If the cycle identification is incorrect, it produces errors
in the TOA measurement that are integral multiples of 10 µs; this severely affects the
positioning and timing accuracy. Moreover, the cycle identification is affected by not
only SWI but also noise [19,21]. As the receiving distance increases, the signal amplitude
decreases, and the signal-to-noise ratio (SNR) deteriorates, which considerably affects the
accuracy of cycle identification.
To solve the problems with eLoran signal cycle identification, in [22], a delayed-
addition method was proposed, which uses an eLoran pulse signal to subtract the signal
after a certain time delay to achieve phase inversion of the signal at 30 µs. This is achieved
by detecting the position of the phase inversion point to realize SZC detection. However,
this method exhibits poor anti-noise performance, and SWI causes the result to have
multiple phase-reversal points. In [23], the polarity-judgment method was proposed, in
which a positive zero crossing is selected, and the signals are accumulated at t − 32.5 µs,
t − 22.5 µs, and t − 12.5 µs for a long duration. If the accumulated result exceeds the
threshold, then t is the SZC; otherwise, t is moved forward or backward by 10 µs, and
the process is repeated. This method is time-consuming, and it is challenging to find a
suitable threshold. In [24], a cross-correlation method was proposed, which uses the local
standard eLoran signal to correlate with the received signal and calculates the SZC by
detecting the maximum value of the correlation result. This method is only applicable to
situations where the amplitude of the groundwave is greater than that of the skywave.
In [25], the peak-to-peak ratio detection method was proposed, in which the ratio of the
signal amplitude at t + 2.5 µs to that at t − 7.5 µs is calculated. If the ratio is close to a
standard value, then it is considered as the standard zero-crossing. However, this method
is affected by the skywave and noise, which can produce multiple values, resulting in
identification errors. In [26], a delay-locked loop (DLL) method was proposed, wherein the
lag envelope is subtracted from the advance envelope to obtain the envelope with a peak
point of 0, and the SZC is obtained by detecting the zero point of the envelope. This method
uses the peak value of the envelope of the ground wave signal; thus, the premise of the
Remote Sens. 2022, 14, 250 3 of 20
method is to eliminate the skywave. Bian and Last [27] proposed the method of spectrum
division to estimate the starting positions of the skywave and groundwave to realize cycle
identification. However, this method has a poor estimation resolution under the condition
of low SNR. Other researchers have proposed the autoregressive moving average model
(ARMA) and multiple signal classification algorithm (MUSIC) methods based on spectrum
division to improve the resolution of estimation; however, these methods are not only
difficult to implement but are suffer from inaccurate parameter estimation [28–30].
Based on the foregoing account of previous research, cycle identification methods can
be divided into two types: signal-waveform reconstruction and parameter estimation. In
signal-waveform reconstruction methods, cycle identification is realized by constructing
the characteristics of SZC. This method can fail under low SNR, and multiple identification
results under SWI can be produced; hence, relying solely on this method is hazardous.
The parameter-estimation method is mainly based on spectrum-division theory, and the
estimation error is high at under SNR. Therefore, a single-cycle identification method
is unreliable and has poor anti-noise ability. Hence, this paper proposes a joint time–
frequency domain method to realize cycle identification of an eLoran signal. On the basis
of improving SNR and repairing the envelope using a digital bandpass filter (BPF) and
linear digital averaging (LDA), the range of SZCs is determined via spectral division in
the frequency domain, and accurate identification of the SZC is realized through peak-to-
peak ratio detection and waveform matching. In this study, the performance of the cycle
identification method was simulated, and the method was verified via field testing. The
methods, findings, and conclusions of this study are presented in the following sections.
The rest of the paper is organized as follows. Section 2 presents the format of the eLoran
signal and skywave interference and describes the impact of the skywave interference on
cycle identification. The principle and model of the joint cycle identification method in
the time–frequency domain are also described. Section 3 describes the verification of the
effectiveness of the cycle identification method and the simulation of the performance of
the period identification method under different signal-to-noise ratios (SNRs) and skywave
interference. The eLoran signals actually received in three places were used to verify the
cycle identification method, and the test results were verified. Section 4 summarizes the
main conclusions of this study.
2. Materials and Methods

In this section, we first describe the single-pulse signal format and pulse group signal
characteristics broadcast by the eLoran system. Second, along with the propagation channel
of the eLoran signal, the influence of SWI on cycle identification is analyzed. Finally, the
principle of the cycle identification method based on the joint time–frequency domain is
presented, and the key parameters of the algorithm are discussed.
2.1. eLoran Signal Format

The working frequency allocated to the eLoran system by the International Telecom-
munication Union (ITU) is 100 kHz, and the bandwidth is 90–110 kHz. It is required that
99% of the transmitted signal energy be concentrated within the specified bandwidth. At
the beginning of the system design, the center frequency, signal-waveform, transmission pe-
riod, and data-modulation mode of the transmission signal are considered comprehensively,
and the existing pulse-phase signal system is formed. The single-pulse signal waveform of
the eLoran system is defined by the United States Coast Guard (USCG) as [31].


 0h i f or t < τ
2 −2( t − τ )
s(t) = A(t − τ ) exp 65 sin(2π f t + Pc ) f or τ ≤ t ≤ 65 + τ , (1)

unde f ined f or t > 65 + τ

where t is time (µs), A is a normalization constant related to the peak amplitude, f is

the carrier frequency of 100 kHz, τ is the envelope-to-cycle difference (ECD) (µs), and Pc
0 <
0 ( ) <
( )= ( − ) exp[ ( ) ]sin(2 + )
≤ ≤ 65 + , (1)
( )= ( − ) exp[ ]sin(2 + )
≤ ≤ 65 + , (1)
Remote Sens. 2022, 14, 250 > 65 + 4 of 20
> 65 +
where is time (μs), A is a normalization constant related to the peak amplitude, is the
where is time (μs), A is a normalization constant related to the peak amplitude, is the
carrier frequency of 100 kHz, is the envelope-to-cycle difference (ECD) (μs), and is
carrier
is the frequency of
phase-code 100 kHz, which
parameter, is the
is envelope-to-cycle
0 for positive difference(PC)
phase-coded (ECD) (μs), and
pulses and π for is
the phase-code parameter, which is 0 for positive phase-coded (PC) pulses and for neg-
the phase-code
negative parameter,
PC pulses. whichpulse
is 0 for positive phase-coded (PC) pulses and (1). for neg-
ative PC pulses. The The eLoran
eLoran pulse signal signal is generated
is generated according
according to Equation
to Equation The
(1). The time–
ative PC pulses.
time–domain The
and eLoran pulse signal
frequency–domain is generated according
characteristics of a single to Equation
eLoran (1).are
signal Theshown
time–
domain and frequency–domain characteristics of a single eLoran signal are shown in Fig-
domain
in and frequency–domain characteristics of athe
single eLoran signalrapidly,
are shown in Fig-
ureFigure 1. Itbe
1. It can can be that
seen seenthe
that the
leadingleading
edge of edge
theofsignalsignal first rapidly,
first rises rises reaching
reaching a peak a
ure 1.atIt65
peak can
µs,beand
seen thatfalls
then therapidly.
leading edge of the signal first rises rapidly, reaching a peak
at 65 μs, and then falls rapidly.
at 65 μs, and then falls rapidly.
Normalized
Normalized Amplitude
Amplitude
(a) (b)
(a) (b)
Figure 1. Characteristics of a single eLoran signal: (a) eLoran pulse signal; (b) eLoran signal ampli-
Figure
Figure 1. Characteristics
1. Characteristicsofofaasingle
tude spectrum. singleeLoran
eLoransignal: (a)(a)
signal: eLoran pulse
eLoran signal;
pulse (b) (b)
signal; eLoran signal
eLoran amplitude
signal ampli-
spectrum.
tude spectrum.
The eLoran system
The system can be be a timing station
station or aa navigation
navigation chain. TheThe timing station
station
The eLoran
eLoran system can can be aa timing
timing station or or a navigation chain.
chain. The timing
timing station
generally
generally has only
has only one
onlyone master
onemaster station,
masterstation, and
station,andandthe navigation chain generally consists of one
generally has thethe navigation
navigation chain
chain generally
generally consists
consists of oneof
master
one station
master and
station two
and to
twofive
to secondary
five stations.
secondary The
stations.master
The station
master is generally
station is repre-
generally
master station and two to five secondary stations. The master station is generally repre-
sented by M, and
represented the secondary station is represented by letters such as W, X, Y, and Z.
sented by M,byand M, the
andsecondary
the secondary station
station is represented
is represented by letters
by letters suchsuch as W,
as W, X, Y, X, and
Y, andZ.
To distinguish
Z. between the signals of thethemaster and secondary stations, each group of
ToTo distinguish
distinguish between
between thethe signals
signals of of master
the master andand secondary
secondary stations,
stations, each
each group
group of
the
of master station transmits nine eLoran pulse signals. The interval between the first eight
thethe master
master station
station transmits
transmits ninenine eLoran
eLoran pulsepulse signals.
signals. The interval
The interval between
between theeight
the first first
pulsepulse
eight signals is 1 ms,
signals is 1 andand
ms, thatthat
between
between thethe
eighth
eighthand
andninth
ninthpulse
pulse signals
signals is
is 22 ms.
ms. The
The
pulse signals is 1 ms, and that between the eighth and ninth pulse signals is 2 ms. The
secondary
secondary station transmits only eight pulses. The pulse-group signals of the master and
secondary station
station transmits
transmits only
only eight
eight pulses.
pulses. TheThe pulse-group
pulse-group signals
signals ofof the
the master
master and and
secondary stations
secondary stations are
are shown
shown in in Figure
Figure2.2.
secondary stations are shown in Figure 2.
1 1
1 1
Amplitude
Amplitude
0.5 0.5
Amplitude
Amplitude
0.5 0.5
0
Normalized
0
Normalized
0
Normalized
0
Normalized
-0.5
-0.5 -0.5
-0.5
-1
-1 0 1 2 3 4 5 6 7 8 9 10 -1
0 1 2 3 4 5 6 7 8 9 10 -1 0 1 2 3 4 5 6 7 8 9 10
Time (ms) 0 1 2 3 4 Time5 (ms) 6 7 8 9 10
Time (ms) Time (ms)
(a) (b)
(a) (b)
Figure 2. Characteristics of an eLoran pulse group signal: (a) pulse-group signal of master station;
(b) pulse-group signal of secondary station.
The interval between the eLoran pulse groups is called the group-repetition interval
(GRI). The GRI cycle range is defined as 40,000–99,990 µs, with a resolution of 10 µs. The
Figure 2. Characteristics of an eLoran pulse group signal: (a) pulse-group signal of master station;
(b) pulse-group signal of secondary station.
Remote Sens. 2022, 14, 250 5 of 20
The interval between the eLoran pulse groups is called the group-repetition interval
(GRI). The GRI cycle range is defined as 40,000–99,990 μs, with a resolution of 10 μs. The
eLoran system continuously emits pulse-group signals according to its own GRI. The
eLoran system continuously emits pulse-group signals according to its own GRI. The
phase-code cycles are listed in Table 1. The two code cycles are alternately modulated with
phase-code cycles are listed in Table 1. The two code cycles are alternately modulated with
the signal; “+” indicates a phase code of 0 radians, and “−” indicates a phase code of π
the signal; “+” indicates a phase code of 0 radians, and “−” indicates a phase code of π
radians. Each pulse signal in the eLoran pulse group of the master and secondary stations
radians. Each pulse signal in the eLoran pulse group of the master and secondary stations
is phase-coded. The third to eighth pulse signals of each pulse group are used for the
is phase-coded. The third to eighth pulse signals of each pulse group are used for the Loran
Loran data channel (LDC), which is used for transmitting system status, time information,
data channel (LDC), which is used for transmitting system status, time information, etc.
etc. The modulation mode adopts Eurofix data-modulation technology [32,33].
The modulation mode adopts Eurofix data-modulation technology [32,33].
Table 1. Phase codes of eLoran signals.
Table 1. Phase codes of eLoran signals.
Ground Master Secondary
Ground Master Secondary
A ++−−+−+−+ +++++−−+
AB + ++ −
− −−+++−+++−+ +
− + −+ ++ −+ +−+−−+−
++
B +−−+++++− +−+−++−−
In the eLoran system, the sixth zero crossing (30 μs) of the first pulse of the pulse
group the
In eLoran system,
is designated the
as the sixthThe
SZC. zero crossing
SZC (30 key
has two µs) of the first pulse
functions: of the pulse
(i) it serves as thegroup
pulse-
istime
designated as the SZC. The SZC has two key functions: (i) it serves as the
reference (PTR) for measuring the signal specification of the eLoran broadcasting pulse-time
reference (PTR)
station; (ii) for measuring
it allows the eLoranthe signal specification
receiver to measure the of TOA
the eLoran
of the broadcasting
signal to realizestation;
high-
(ii) it allowspositioning
precision the eLoranandreceiver to measure the TOA of the signal to realize high-precision
timing.
positioning and timing.
2.2. Skywave Interference and Cycle Identification
2.2. Skywave Interference and Cycle Identification
The eLoran signal belongs to an LW frequency band with a carrier frequency of 100
The eLoran signal belongs to an LW frequency band with a carrier frequency of 100
kHz. According to the propagation theory of this frequency band, the eLoran signal
kHz. According to the propagation theory of this frequency band, the eLoran signal
mainly reaches the receiving point through two paths: (i) as a groundwave signal (surface
mainly reaches the receiving point through two paths: (i) as a groundwave signal (surface
wave) transmitted along the surface of the Earth, and (ii) as a skywave signal (space wave)
wave) transmitted along the surface of the Earth, and (ii) as a skywave signal (space wave)
produced by one or more reflections between the ionosphere and the ground when the
produced by one or more reflections between the ionosphere and the ground when the
signal is projected to the ionosphere at a certain elevation angle. Generally, only one-hop
signal is projected to the ionosphere at a certain elevation angle. Generally, only one-
skywave is considered for receiving points with a receiving distance of less than 2000 km
hop skywave is considered for receiving points with a receiving distance of less than
[34]. The propagation path of the eLoran signal is shown in Figure 3.
2000 km [34]. The propagation path of the eLoran signal is shown in Figure 3.
Groundwave
Earth Surface
Transmitter Receiver
Figure3.3.Schematic
Figure Schematicdiagram
diagramofofeLoran
eLoransignal
signalpropagation
propagationchannel.
channel.
Thecomposite
The compositewave
wavecomprising
comprisingthe
theeLoran
eLorangroundwave
groundwaveand
andskywave
skywavesignals
signalsre-
re-
ceivedininspace
ceived spacecan
canbebeexpressed
expressedas:
as:
( )= ( )+ ( − ∆ ) + ( ), (2)
x (t) = As(t) + As s(t − ∆T ) + n(t), (2)
where ( ) is the standard eLoran signal with a normalized amplitude, is the ampli-
where
tude ofs(tthe
) is groundwave
the standard signal,
eLoran signal with
is the a normalized
amplitude of theamplitude, A is the∆ amplitude
skywave signal, is the de-
oflay
the groundwave
difference of thesignal,
skywaveAs relative
is the amplitude of the skywave
to the groundwave, and (signal, ∆T is the delay
) is noise.
difference of the skywave relative to the groundwave, and n(t) is noise.
The skywave-to-groundwave ratio (SGR) is the amplitude ratio of the skywave relative
to the groundwave, expressed as:

As
RSGR = 20 log(rSGR ) = 20 log , (3)
A
Remote Sens. 2022, 14, x FOR PEER REVIEW 6 of 20
The skywave-to-groundwave ratio (SGR) is the amplitude ratio of the skywave rela-
Remote Sens. 2022, 14, 250
tive to the groundwave, expressed as: 6 of 20
= 20 log( ) = 20log( ), (3)
The time-delaydifference
The time-delay difference∆T∆of the
of the skywave
skywave relative
relative to the groundwave
to the groundwave is ex-
is expressed as:
pressed as:
∆T∆ ==Ts −−TTOA ,, (4)
(4)
where TTOA and
where and Ts are the propagation delays of the the groundwave
groundwave and andskywave,
skywave,respec-
respec-
tively.
tively.
Figure 44shows
Figure showstwo twotypical
typicaleLoran
eLoranreceived-signal
received-signal waveforms.
waveforms. TheThe first
first case
caseisiswhere
where
thephase
phasedifference
difference is ◦
is 0°
0 due
duetotothe
thetime
timedelay
delayofofthe
theskywave
skywaverelative
relativetotothe
thegroundwave,
groundwave,
the
whichincreases
which increasesthetheamplitude
amplitude of of
thethe composite
composite signal.
signal. The second
The second case iscase
where is where the
the phase
phase difference
difference is 180° is
due180to◦ the
duetime
to the timeofdelay
delay of the skywave
the skywave relativerelative to the groundwave,
to the groundwave, which
which reduces
reduces the amplitude
the amplitude of the composite
of the composite signal. Insignal. In both
both cases, the cases, the groundwave
groundwave is distorted,is
distorted, and the signal amplitude exhibits two extremes. This significantly
and the signal amplitude exhibits two extremes. This significantly affects the cycle identi- affects the
cycle identification
fication result of theresult
eLoran of groundwave.
the eLoran groundwave.
(a) (b)
Figure 4. Two typical eLoran received-signal waveforms: (a) composite wave with a phase differ-
Figure 4. Two typical eLoran received-signal waveforms: (a) composite wave with a phase difference
ence◦of 0°; (b) composite wave with a phase difference of◦ 180°.
of 0 ; (b) composite wave with a phase difference of 180 .
Cycle
Cycleidentification
identificationisisthe theprocess
processof offinding
findingthe theSZC
SZCof ofthe
thefirst
firstpulse
pulseof ofthe
theground-
ground-
wave
wave after
afterthetheeLoran
eLoranpulsepulseacquisition.
acquisition.Ideally,
Ideally,only thethe
only pure groundwave
pure groundwave signal is re-
signal is
ceived, and the SZC can be detected by measuring the peak value
received, and the SZC can be detected by measuring the peak value (65 µs) of the eLoran (65 μs) of the eLoran
groundwave
groundwaveenvelope.
envelope.However,
However, thethe
skywave
skywave is generated
is generatedduedue to reflection fromfrom
to reflection the ion-
the
osphere,
ionosphere, and it always arrives at the receiver a certain cycle after the groundwavedoes.
and it always arrives at the receiver a certain cycle after the groundwave does.
Therefore,
Therefore, thethe eLoran
eLoran groundwave
groundwave signalsignal isis susceptible
susceptible to to interference
interference by by the
the skywave
skywave
delay
delay [21,35]. According totothe
[21,35]. According theminimum
minimum receiver
receiver performance
performance standards
standards for marine
for marine eLo-
eLoran receiving
ran receiving equipment
equipment issued
issued by the
by the Radio
Radio Technical
Technical Commission
Commission for Maritime
for Maritime Ser-
Services
vices (RTCM),
(RTCM), thedelay
the time time delay
rangerange ∆t skywave
∆t of the of the skywave
relativerelative to the groundwave
to the groundwave is 37.5–
is 37.5–1500 µs,
1500 μs, and the amplitude ratio SGR of the skywave relative to the
and the amplitude ratio SGR of the skywave relative to the groundwave is 12–26 dB [36]. groundwave is 12–26
dB [36]. Therefore,
Therefore, to consider to consider
both the both the stability
stability and the SNR, and thetheSNR,
positivethe zero
positive zero (30
crossing crossing
µs) in
(30
theμs) in week
third the third week isselected
is usually usuallyasselected
the SZCas[37].
the SZC [37].
2.3. Time–Frequency
2.3. Time–Frequency Domain
Domain Cycle
Cycle Identification
Identification Method
Method
Because the
Because the phase
phase of
of the
the first
first pulse
pulse signal
signal in
in the
the eLoran
eLoran pulse
pulse group
group is
is “+”
“+”and
andthere
there
is no data modulation, cycle identification is performed using the first pulse
is no data modulation, cycle identification is performed using the first pulse in the eLoranin the eLoran
pulse group.
pulse group. ToTo overcome
overcome the the low
low identification
identification probability
probability and
and poor
poor anti-interference
anti-interference
ability of
ability of existing
existing cycle
identification methods,
methods, we we propose
propose aa time–frequency
time–frequency domaindomain
method to identify the SZC of the eLoran signal. The frequency–domain
method to identify the SZC of the eLoran signal. The frequency–domain method uses method uses
spectrum division to estimate the approximate starting positions of the skywave and
groundwave, which is also the cycle identification range; the time–domain method uses the
peak-to-peak ratio and waveform matching to determine the SZC accurately. The specific
spectrum division to estimate the approximate starting positions of the skywave and
Remote Sens. 2022, 14, 250 7 of 20
groundwave, which is also the cycle identification range; the time–domain method uses
the peak-to-peak ratio and waveform matching to determine the SZC accurately. The spe-
cific implementation process of the method is shown in Figure 5. We assume that the re-
implementation
ceived process
signal is as of the method is shown in Figure 5. We assume that the received
follows:
signal is as follows:
x (t() =
) =s(t (− − ) +rSGR s(t( −−T2 ))++n((t),),
T1 ) + (5)
(5)
where
where rSGR isisthe
theamplitude
amplituderatio ofof
ratio thethe
skywave
skywave to the groundwave,
to the groundwave, isT1the is delay dif-
the delay
ference of the groundwave relative to the local standard eLoran signal, is
difference of the groundwave relative to the local standard eLoran signal, T2 is the delay the delay
difference
difference of
of the
the skywave
skywave relative
relative to
to the
the local
local standard
standard eLoran
eLoran signal, ∆ =
signal, ∆T = T2 − − T1 is
is the
the
delay
delay difference
difference of
of the
the eLoran
eLoran skywave
skywaverelative
relativeto
togroundwave,
groundwave,and and n((t)) is noise in the
signal.
RF Signal SZC
Analog-to-digital Standard Waveform Matching
Conversion eLoran Signal
Ditital Bandpass Filter Fast Fourier Window Peak-to-peak Ratio

（B=30kHz） Transform Function and Error Calculation
Estimation of the
Linear Digital Average Fast Fourier Spectrum Windowed Positive Direction Zero
Approximate Location
(M=64) Transform Division IFFT Crossing Detection
of Sky-Ground Waves
interference suppression Frequency Domain Time Domain
Figure 5. Principle of time–frequency domain joint cycle identification method.

Figure 5. Principle of time–frequency domain joint cycle identification method.
2.3.1. Digital
Digital Bandpass
Bandpass Filtering
Filtering and Linear Digital Average
Before the cycle identification,
identification, the
the received
received signal
signal must
must be
be filtered to suppress noise
and interference and improve the SNR, to obtain relatively clean composite signals of the
groundwave and skywave.
From the eLoran frequency band, the cycle identification results are mainly affected
and continuous
by noise and continuous wave
wave interference
interference (CWI).
(CWI). We
We used BPFs to suppress
suppress out-of-band
noise and
and interference.
interference.The
TheBPF
BPFincludes
includesa Hamming
a Hamming window
windowwith a bandwidth
with of 30ofkHz
a bandwidth 30
and and
kHz an order of 128,
an order andand
of 128, its its
out-of-band
out-of-bandsuppression
suppressioncapability
capabilitycan
canexceed
exceed30
30 dB. The
bandpass-filtered signal is:
x BPF (t() =
) =s(t(−− ) +rSGR s((t −−T2 ))++n1 ((t)),
T1 ) + , (6)
(6)
where ( ) is the residual noise in the bandpass-filtered signal.
where n1 (t) is the residual noise in the bandpass-filtered signal.
In the eLoran frequency band, the cycle identification results are mainly affected by
In the eLoran frequency band, the cycle identification results are mainly affected
noise. After bandpass-filtering, we used LDA to suppress in-band noise, which not only
by noise. After bandpass-filtering, we used LDA to suppress in-band noise, which not
suppresses the noise and improves the SNR of the received signal, but also repairs the
only suppresses the noise and improves the SNR of the received signal, but also repairs
eLoran signal and reduces the signal distortion. The LDA is the cumulative average pro-
the eLoran signal and reduces the signal distortion. The LDA is the cumulative average
cessing of the
processing of received signal
the received for M
signal times
for according
M times to thetoGRI
according thecycle. Assuming
GRI cycle. that there
Assuming that
are M signals with a length equal to the GRI, the LDA result can be expressed as:
there are M signals with a length equal to the GRI, the LDA result can be expressed as:
(t) = ∑ (t) = ∑ [ ( − ) + ( − )] + ( ), (7)
1 M M
x LDA (t) = ∑
M i=noise 1
x BPF ( t ) = ∑ i =1
[ s ( t − T 1 ) + r SGR s ( t − T2 )] + n 2 ( t ) , (7)
where ( ) is the residual in the signal after LDA.
whereWhen
n2 (t)the number
is the of LDA
residual noiseisinM, thethe gainafter
signal of the SNR can be increased by 10 log
LDA. .
We chose the linear digital average of M = 64 to ensure that the
When the number of LDA is M, the gain of the SNR can be increased by 10 log10 M. SNR of the signal can be
improved by 18
We chose the dB. A
linear relatively
digital clean
average ofcomposite
M = 64 to signalensurewiththat athe
length
SNRequal
of thetosignal
the GRI cancan
be
be obtained using the linear digital average. The cycle identification
improved by 18 dB. A relatively clean composite signal with a length equal to the GRI method in the time–
frequency domain
can be obtained was the
using subsequently
linear digital used to identify
average. Thethe SZC.
cycle identification method in the
time–frequency domain was subsequently used to identify the SZC.
Remote Sens. 2022, 14, 250 8 of 20
2.3.2. Spectrum Division in Frequency Domain

The
Remote Sens. 2022, 14, x FOR PEER REVIEW fast Fourier transform (FFT) was used to obtain the signal spectrum 8after of 20 LDA,
and the approximate starting positions of the skywave and groundwave were determined
via spectrum division. The FFT of x LDA (t) is:
2.3.2. Spectrum Division in Frequency h Domain i
The fast Fourier X
transform f ) e j2π
( f ) = S((FFT) f T1
was used+ rSGR e j2π f Tthe
to obtain
2 + N ( f ),
signal spectrum after LDA, (8)
and the approximate starting positions of the skywave and groundwave were determined
where S( f ) is the
via spectrum FFT result
division. of the
The FFT (t) is:
of eLoran signal and N ( f ) is the FFT result of the noise.
The spectrum-division method, proposed by Bian and Last, is based on high-resolution
( ) = ( )[ +r ] + ( ), (8)
time delay estimation under multipath conditions [38]. When Equation (8) is divided by
Fourier(transform
the where ) is the FFT
S0result
( f ) ofofthe
thestandard
eLoran signal ( ) isthe
andsignal,
eLoran thespectrum-division
FFT result of the noise.
result is:
The spectrum-division method, proposed by Bian and Last, is based on high-resolu-
tion time delay estimation under
X ( f ) multipath
h conditions [38]. When i Equation (8) is divided
Y ( f ) = ( ) of=the
by the Fourier transform e j2π f T1 +
k gstandard eLoran
j2π f T2
rSGR esignal, +W
the ( f ),
spectrum-division re- (9)
S0 ( f )
sult is:
where k g is the amplitude ratio of ( )the received groundwave signal to the standard signal,
( ) =
and W ( f ) is the result of dividing = [ + ] + ( ), (9)
( )the noise by the standard eLoran signal. According to
Euler’s formula:
where is the amplitude ratio of the received groundwave signal to the standard signal,
e j2πT f = cos(2πT f ) + jsin(2πT f ), (10)
and ( ) is the result of dividing the noise by the standard eLoran signal. According to
Euler’s formula: (10) is a sine wave with variable f , and the frequency of sine wave is T.
The Equation
Therefore, the result of the spectrum = division
cos(2 of)the + received
(2 ),signal in Equation (9)(10) becomes
an estimate of two continuous wave frequencies T1 and T2 . The amplitude ratio rSGR is the
ratio of The Equation (10)ofisthe
the amplitudes a sine
twowave with variable
continuous waves., and the frequency of sine wave is
. Therefore, the result of the spectrum division of the received signal in Equation (9) be-
Figure 6 shows the results of spectrum division without noise and with SNR = 5 dB.
comes an estimate of two continuous wave frequencies and . The amplitude ratio
The waveform of the spectrum division result in the absence of noise was composed of
is the ratio of the amplitudes of the two continuous waves.
continuous waves
Figure 6 showsformed by different
the results time division
of spectrum delays, which
withoutisnoise
consistent with
and with the= theoretical
SNR 5 dB.
analysis.
The waveform of the spectrum division result in the absence of noise was composed ofby the
However, after noise was added, the out-of-band noise was amplified
spectrum division,
continuous waveswhich
formed increased thetime
by different out-of-band noise
delays, which is of the eLoran
consistent withsignal and reduced
the theoretical
the analysis.
SNR sharply.
However, The after
waveform of the
noise was two the
added, continuous
out-of-bandwaves
noisewas
wasdifficult to distinguish.
amplified by the
Thespectrum
Hamming window
division, offered
which the advantages
increased the out-of-bandof low
noiseside-lobe attenuation
of the eLoran andre-strong
signal and
duced thesuppression.
out-of-band SNR sharply. Therefore,
The waveform of the twowindow
a Hamming continuouswithwaves was difficult
a bandwidth to dis-
of 50 kHz was
tinguish.
selected The Hamming
to suppress windownoise
out-of-band offeredinthe
theadvantages of low side-lobe
spectrum-division attenuation
results. and
After windowing,
strong (9)
Equation out-of-band
became: suppression. Therefore, a Hamming window with a bandwidth of 50
kHz was selected to suppress out-of-band noise in the spectrum-division results. After
windowing, Equation (9) became:
n h i o
Z ( f ) = H ( f )Y ( f ) = H ( f ) k g e j2π f T1 + rSGR e j2π f T2 + W ( f ) , (11)
( )= ( ) ( )= ( ) [ + ]+ ( ), (11)
where H ( f )( is) the
where window
is the windowfunction.
function.
10 5
4
1 Spectrum division results (without noise) Spectrum division results (SNR = 5 dB)
3
2
0.5
Amplitude
Amplitude
0
0
-1
-0.5
-2
-1 -3
0 200 400 600 800 1000 0 200 400 600 800 1000
Frequency (kHz) Frequency (kHz)
(a) (b)
Figure 6. Results of spectrum division under two conditions: (a) spectrum-division results without
Figure Results
6. (b)
noise; of spectrum results
spectrum-division division under
with SNR two conditions: (a) spectrum-division results without
= 5 dB.
noise; (b) spectrum-division results with SNR = 5 dB.
Remote Sens. 2022, 14, 250 9 of 20
Finally, the inverse fast Fourier transform (IFFT) was used to estimate the parameters
of the skywave and groundwave [39,40]. The IFFT of Equation (11) is:
F −1 ( Z ( f )) = δ(t − T1 ) + rSGR δ(t − T2 ) + F −1 [ H ( f )W ( f )], (12)
where F −1 represents the IFFT.

It can be seen from Equation (12) that δ(t) is the impulse response function resulting
from the IFFT of the synthetic signals of the groundwave and skywave. The time delays
T1 and T2 and amplitude ratio rSGR can be estimated by detecting the time–domain pulse
peak value after IFFT. However, the time delay only represents the approximate starting
positions of the skywave and groundwave, and the estimated position has a bias; thus, the
SZC had to be further determined in the time domain.
2.3.3. Peak-to-Peak Ratio and Waveform Matching in Time Domain

The result of spectrum division determined the starting position range T1 –T2 of the
groundwave and the skywave in the composite wave. Therefore, we determined the
position of the SZC using the peak-to-peak ratio and waveform methods in the time
domain.
The peak-to-peak ratio method first determined all positive zero crossing points (t)
in the range of T1 –T2 and then calculated the ratio of the positive peak value in the next
week (t + 2.5 µs) and the previous week (t − 7.5 µs) of each positive zero crossing. If the
calculated peak-to-peak ratio was close to the peak-to-peak ratio of the standard eLoran
signal a certain range, then it was considered as a possible SZC [23]. The theoretical formula
for calculating the peak-to-peak ratio is:
s(t + 2.5) 2 −10/65

s(t + 2.5)
h(t) = = e , (13)
s(t − 7.5) s(t − 7.5)
The peak-to-peak ratios of positive zero crossings in the theoretical calculation are
displayed in Table 2. It can be seen that the peak-to-peak ratio at 30 µs was 1.5338.
Table 2. Theoretical calculation results of peak-to-peak ratio at different zero-crossing points.
Time (µs) 10 20 30 40 50 60 70
h(t) 18.3790 2.3819 1.5338 1.2571 1.1218 1.0419 0.9892
The peak-to-peak ratios of all positive zero crossings in the range T1 –T2 after LDA are
calculated first; then, the errors of the peak-to-peak ratios are calculated:
∆h(n) = |hn (t) − 1.5338|, (14)
where hn (t) is the peak-to-peak ratio of zero crossings, and n is the number of zero crossings.
Because the eLoran signal was distorted during propagation and signal processing,
which can lead to deviations in the peak-to-peak ratios, when ∆h(n) < 0.3, we determined
that the zero crossing n was the closest to the SZC. Thus, the peak-to-peak ratio results
contained multiple possible SZCs. Therefore, we used waveform matching to calculate the
root mean square (RMS) error between the received signal and the standard waveform over
a cycle of time. The minimum value of the RMS error corresponded to the zero crossing to
determine the SZC. The waveform-matching calculation formula was as follows:
v
uR t
u 2 [s(t) − sn (t)]2
t t = t1
σ(n) = , (15)
t2 − t1
where s(t) is the standard eLoran signal, sn (t) is the received eLoran signal, and t1 –t2 is
the waveform-matching range. To prevent the influence of SWI, we adopted a waveform
Remote Sens. 2022, 14, 250 10 of 20

where ( ) is the standard eLoran signal, ( ) is the received eLoran signal, and –
is the waveform-matching range. To prevent the influence of SWI, we adopted a wave-
form matching
matching range range of µs.
of 10–50 10–50 μs. Therefore,
Therefore, when
when σ(n ( )smallest,
) was the was thethe
smallest, the corre-
corresponding n
sponding n
was the SZC.was the SZC.
3. Results
In this section,
section, we
weanalyze
analyzethe
theeffectiveness
effectivenessand
and performance
performance of the
of the joint
joint time–
time–fre-
frequency
quency cycle identification method through simulation. Then, we validate the cycle cycle
cycle identification method through simulation. Then, we validate the iden-
identification method
tification method by receiving
by receiving actual
actual eLoran
eLoran signals
signals at different
at different locations.
locations.
3.1.
3.1. Validation
Validation of
of Cycle
Cycle Identification
Method
To
To verify the effectiveness of the proposed
verify the effectiveness of the proposed method,
method, wewe performed
performed aa simulation,
simulation, the
the
parameters
parameters of which were set as follows. The GRI of the eLoran signal generated by
of which were set as follows. The GRI of the eLoran signal generated by the
the
simulation was 60 ms. After adding noise to the signal, the SNR of groundwave was −5 dB;
simulation was 60 ms. After adding noise to the signal, the SNR of groundwave was −5
the amplitude ratio of the skywave to the groundwave was SGR = 10 dB; the time-delay
dB; the amplitude ratio of the skywave to the groundwave was SGR = 10 dB; the time-
difference between the skywave and the groundwave was ∆T = 62.5 µs; and the sampling
delay difference between the skywave and the groundwave was ∆ = 62.5 μs; and the
rate of the simulation process signal was f s = 2 MHz. The simulation results are shown in
sampling rate of the simulation process signal was = 2 MHz. The simulation results are
Figure 7.
shown in Figure 7.
1
Band-pass filtering results
Normalized amplitude
0.5
-0.5
-1
0 500 1000 1500 2000
Sampling time (1/fs )
(a) (b)
1
LDA result
Normalized amplitude
0.5
Amplitude (dB)
-0.5
-1
0 500 1000 1500 2000
(c) (d)
Figure 7. Cont.
Remote Sens.
Remote 2022,
Sens. 14, 14,
2022, x FOR
250 PEER REVIEW 11 of1120
of 20
1.5
Peak-to-peak ratio error Received eLoran signal (third zero crossing)
Standard eLoran signal
3
Peak-to-peak ratio error
2.4
Amplitude
0.5
1.8
0
1.2
-0.5
0.6
0 -1
1 2 3 4 5 6 10 15 20 25 30 35 40 45 50
Zero crossing Sampling time (1/fs )
(e) (f)
2
Received eLoran signal (fourth zero crossing)
Standard eLoran signal
1.5
1
Amplitude
0.5
-0.5
-1
-1.5
0 10 20 30 40 50 60 70 80
(g)
Figure 7. Simulation
Figure results
7. Simulation of cycle
results identification:
of cycle (a) the(a)signal
identification: generated
the signal by theby
generated simulation; (b)
the simulation;
bandpass-filtering result; (c) linear digital averaging result; (d) spectrum-division result; (e) peak-
(b) bandpass-filtering result; (c) linear digital averaging result; (d) spectrum-division result; (e) peak-
to-peak ratio error; (f) third zero-crossing waveform match; (g) fourth zero-crossing waveform
to-peak ratio error; (f) third zero-crossing waveform match; (g) fourth zero-crossing waveform
match.
match.
As can be seen from Figure 7a, the composite wave of the skywave and groundwave
As can be seen from Figure 7a, the composite wave of the skywave and groundwave
generated in the simulation is submerged in noise, and the groundwave signal is severely
generated in the simulation is submerged in noise, and the groundwave signal is severely
distorted. Figure 7b presents the bandpass-filtered signal. It can be seen that although the
distorted. Figure 7b presents the bandpass-filtered signal. It can be seen that although the
noise in the
noise signal
in the is suppressed,
signal is suppressed, thethe
signal distortion
signal is relatively
distortion large.
is relatively Figure
large. Figure 7c 7c
shows
shows
thethe
signal after LDA. It can be seen that the signal not only became cleaner,
signal after LDA. It can be seen that the signal not only became cleaner, but the signal but the signal
waveform
waveform also
alsobecame
became complete
complete through
through repair. The
repair. spectrum-division
The spectrum-division method
method waswasthen
then
used to obtain the initial positions of the groundwave and skywave on the
used to obtain the initial positions of the groundwave and skywave on the linearly averaged linearly aver-
aged signal,
signal, as as shown
shown ininFigure
Figure7d.7d.The
The range
range ofof the
the SZC
SZCwas wasobserved
observedtotobebe320–445.
320–445. TheThe
results of calculating the peak-to-peak ratio errors of all zero crossings in
results of calculating the peak-to-peak ratio errors of all zero crossings in the determinedthe determined
range
rangeareare
shown
shown in Figure
in Figure 7e.7e.
AsAsobserved,
observed, thethe
peak-to-peak
peak-to-peak ratio errors
ratio of zero
errors of zerocrossings
crossings
3 and
3 and4 were
4 wereless than
less than0.3.0.3.
Waveform
Waveform matching
matching was performed
was performed ononthethe
twotwozero-crossing
zero-crossing
points, and
points, andthetheresults areare
results shown
shown in in
Figure
Figure7f,g. The
7f,g. waveform-matching
The waveform-matching coefficients
coefficientsof of
thethe
two zero-crossings were calculated to be 0.0456621 and 0.1117095, respectively.
two zero-crossings were calculated to be 0.0456621 and 0.1117095, respectively. Thus, Thus,
zero crossing
zero crossing 3 was thethe
3 was SZC.SZC.TheTheSZCSZC determined
determined using
usingthethe
cycle identification method
method
was consistent with that set based on the simulation, which proves the
was consistent with that set based on the simulation, which proves the effectiveness effectiveness of the
of the
method.
method.
3.2.3.2. Performance
Performance Analysis
Analysis of Cycle
of Cycle Identification
Method
Noise
Noise andand
SWISWI
are are
the the
mostmost important
important interference
interference sources
sources that affect
that affect cycle cycle iden-
identifi-
tification,
cation, and are
and they theyinevitable.
are inevitable. According
According to the to the minimum
minimum performance
performance standard
standard for
Remote Sens. 2022, 14, 250 12 of 20
eLoran receivers, the requirements for an eLoran receiver to function normally were SNR
≥ −10
for dB, SGR
eLoran < 12 dB,
receivers, and
the ∆ > 37.5 μs
requirements for[36]. Therefore,
an eLoran to verify
receiver the performance
to function normallyof the
were
SNR ≥ −10
proposed cycle
dB,identification and ∆T >
SGR < 12 dB,method, we analyzed
37.5 µs [36].the skywaveto
Therefore, and groundwave
verify param-
the performance
eter-estimation
of the proposedperformance of the spectrum-division
cycle identification method, we analyzed method
the and the probability
skywave of cycle
and groundwave
identification based on
parameter-estimation the joint time–frequency
performance domain under
of the spectrum-division different
method SNRs.
and the probability of
cycleWe
identification
simulatedbased on the joint time–frequency
the parameter-estimation domain under
performance of thedifferent SNRs.
spectrum-division
method under SNR = 5, 0, and –10 dB. For SGR = 5–10 dB and ∆ = 37–150 μs, themethod
We simulated the parameter-estimation performance of the spectrum-division ampli-
under SNRerror
tude-ratio = 5, 0,
andand –10 dB. For SGRerror
delay-difference = 5–10
after and ∆T =estimation
dBspectrum 37–150 µs,are
theshown
amplitude-ratio
in Figure
error
8. and delay-difference error after spectrum estimation are shown in Figure 8.
2 1
1.5 0.8
0.6
1
0.4
0.5
0.2
0
0
-5
50 -5 50
0
0
5 100 100
SGR (dB) SGR (dB) 5
10 150 ∆T (μs) △T (μs)
10 150
(a) (b)
3 1
0.8
2
0.6
0.4
1
0.2
0
-10 0
50 -10 50
-5 -5
0 100 0 100
SGR (dB) 5 5
∆T (μs) SGR (dB) 10 150 ∆T (μs)
10 150
(c) (d)
Amplitude ratio estimation error (dB)
30
2
25
1.5
20
15 1
10 0.5
5
0
120 -5
25 50
100 20 0
80 15 100
10 5
∆T (μs) 60 5 SGR (dB)
0 SGR (dB) 10 150 ∆T (μs)
40 -5
(e) (f)
Figure 8. Performance simulation of spectrum-division method: (a) delay-difference estimation er-
Figure 8. Performance simulation of spectrum-division method: (a) delay-difference estimation error
ror at SNR = 5 dB; (b) amplitude-ratio estimation error at SNR = 5 dB; (c) delay-difference estimation
at SNR = 5 dB; (b) amplitude-ratio estimation error at SNR = 5 dB; (c) delay-difference estimation error
error at SNR = 0 dB; (d) amplitude-ratio estimation error at SNR = 0 dB; (e) delay-difference estima-
at SNR
tion = 0atdB;
error SNR(d)=amplitude-ratio estimation estimation
−10 dB; (f) amplitude-ratio error at SNR = 0at
error dB; (e) =delay-difference
SNR −10 dB. estimation
error at SNR = −10 dB; (f) amplitude-ratio estimation error at SNR = −10 dB.
Remote Sens. 2022, 14, 250 13 of 20
Figure 88shows
Figure showsthatthatwith
witha adecrease
decrease inin SNR,
SNR, thethe estimation
estimation error
error of the
of the time-delay
time-delay dif-
difference and the amplitude ratio of the skywave to the groundwave
ference and the amplitude ratio of the skywave to the groundwave based on the spectrum- based on the spec-
trum-division
division method method increased.
increased. When SNR When = 5SNR = 5 dB,
dB, the the amplitude-ratio
amplitude-ratio estimationestimation
error waserror
less
was less
than 1 dB,than
and 1the
dB, and the delay-difference
delay-difference estimation estimation
error was less error was
than lessWhen
2 µs. than SNR
2 μs.=When
0 dB,
SNR
the = 0 dB, the amplitude-ratio
amplitude-ratio estimation error estimation
was lesserror
than 1wasdB,lessandthan 1 dB, and the delay-differ-
the delay-difference estimation
ence estimation
error was less thanerror10wasµs. less
When thanSNR10 μs.=− When
10 dB, SNRthe=amplitude-ratio
−10 dB, the amplitude-ratio
estimation error esti-
mation error was less than 2 dB, and the delay-difference estimation
was less than 2 dB, and the delay-difference estimation error was less than 10 µs in most error was less than
10 μs however,
cases; in most cases; ∆T > 50 µs
whenhowever, when
and SGR∆ ><50−3μsdB, and theSGR < −3 dB, the delay-difference
delay-difference estimation error
estimation
was greatererror
thanwas greater
10 µs. In thethan 10 received
actual μs. In thesignal,
actual thereceived
SNR maysignal, the SNR may
deteriorate, dete-
resulting
riorate,
in resulting
a greater in a greater
estimation estimation
error in the delayerror in the delay
difference. difference.
Therefore, basedTherefore, based on
on the simulation
the simulation
results, results, the spectrum-division
the spectrum-division method cannotmethod be usedcannot be usedcycle
for complete for complete cycle
identification;
identification;
more accurate more
cycle accurate cycle based
identification identification
on spectrumbaseddivision
on spectrumusingdivision using the
the time–domain
time–domain
method methodToisaddress
is necessary. necessary. To address
the large the large
estimation errorestimation error of the spectrum-
of the spectrum-division method
under
division a low
methodSNR, we proposed
under a low SNR, a cycle identification
we proposed a cyclemethod based on
identification the combined
method based on
time–frequency domain. Through
the combined time–frequency spectral
domain. division,
Through we further
spectral improved
division, the accuracy
we further improved of
the accuracy of cycleinidentification
the time domain. in the time domain.
We
Wesimulated
simulatedthetheparameter
parameter estimation
estimationperformance
performance of the of
cycle
theidentification method
under
method SNRunder −5, −
= 0,SNR = 10, and
0, −5, −13
−10, dB.−13
and Figure 9 shows
dB. Figure the error
9 shows therate of rate
error cycleofidentification
cycle identi-
when
ficationSGR
when= 5–10
SGRdB and ∆T
= 5–10 dB =and ∆ =µs.
37–150 37–150 μs.
Cycle identification error rate (%)
0.6 12
10
0.4 8
0.2 4
0 0
120 120
100 25 100 25
20 20
80 15 80 15
10 10
∆T(μs) 60 5 60 5
0 SGR(dB) ∆T(μs) 0 SGR(dB)
40 -5 40 -5
(a) (b)
30
30
25
20
20
15
10
10
5 0
120
140
25 120 25
100 20
15 100 20
80 15
10 80 10
60 5 ∆T(μs) 5
∆T(μs) 0 60
40 -5
SGR(dB) 0 SGR(dB)
40 -5
(c) (d)
Figure 9. Performance simulation results of cycle identification method: (a) SNR = 0 dB; (b) SNR =
Figure 9. Performance simulation results of cycle identification method: (a) SNR = 0 dB; (b) SNR = −5 dB;
−5 dB; (c) SNR = −10 dB; (d) SNR = −13 dB.
(c) SNR = −10 dB; (d) SNR = −13 dB.
The simulation
The simulation results
resultsin
inFigure
Figure99show
showthat
thatwhen
whenSGR
SGR≤≤ 23
23 dB
dBand
andSNR
SNR≥≥ 00 dB,
dB, the
the
success rate of cycle identification was 100%. When SGR ≤ 23 dB and SNR ≥ −10
success rate of cycle identification was 100%. When SGR ≤ 23 dB and SNR ≥ −10 dB, thedB, the
success rate was better than 75%. When SGR ≤ 23 dB and SNR ≥ −13 dB, the success
success rate was better than 75%. When SGR ≤ 23 dB and SNR ≥ −13 dB, the success rate rate
Remote Sens.
Remote Sens. 2022, 14, x250
2022, 14, FOR PEER REVIEW 1414of
of 20
20
was
was better
better than
than 55%. The
55%. Thecycle
cycleidentification
identificationwas
was mainly
mainly affected
affected by by
thethe relative
relative ampli-
amplitudes
was
tudes better
of the than 55%.and
skywave Thegroundwave,
cycle identification was
especially mainly
under a affected
low SNR. by the relative
Although the ampli-
success
of the skywave and groundwave, especially under a low SNR. Although the success rate
tudes
rate of
of cyclethe skywave and
identification groundwave,
decreased especially
withananincreaseunder
increasein a
inSGR, low SNR.
SGR, accurate Although
accurate cycle the success
of cycle identification decreased with cycle identification
identification
rate ofbecycle
could identification decreased with an increase in SGR, accurate cycle identification
could be achieved via multiple cycle identification calculations. Therefore,the-simulation
achieved via multiple cycle identification calculations. Therefore, the-simulation
could be
results achieved via multiple cycle identification calculations. Therefore, the-simulation
resultsshowed
showedthat thatthe
theproposed
proposedcycle
cycleidentification
identificationmethod
methodisiseffective
effectiveand
andoffers
offersaahigh
high
results showed
identification that the proposed cycle identification method is effective and offers a high
identificationprobability.
probability.
identification probability.
3.3.Experimental
3.3. ExperimentalVerification
VerificationofofCycle
CycleIdentification
IdentificationMethod
Method
3.3. Experimental Verification of Cycle Identification Method
Wevalidated
We validatedthethecycle
method withwith actual
actual signals
signals transmitted
transmitted by a
by a real
eLoran Wesystem.
real eLoranvalidatedThethe
system. cycle
The
test testidentification
scheme scheme
is shown method
is shown in with
in Figure 10.actual
Figure signals
10. eLoran
The The eLorantransmitted
receiver
receiver by a real
obtains
obtains the
eLoran
the signal
signal system.
transmitted Thebytest
transmitted byscheme
the isstation
the eLoran
eLoran shown in Figure
station the10.
through
through The
the eLoran
antenna,
antenna, receiver
filtersfilters andobtains the
amplifies
and amplifies the
signal
signal transmitted
the signal
received by by
received by the
the the eLoran station
radiofrequency
radiofrequency through
(RF)(RF) theand
unit,
unit, and antenna,
converts
converts filters
it itinto and
into amplifies
a adigital
digitalsignalthe
signal
signal
through received
the A/D by the
chip. radiofrequency
Next, the (RF) unit, and
digital-signal-processing converts
unit of it
the
through the A/D chip. Next, the digital-signal-processing unit of the receiver captures the into a
receiver digital
capturessignal
the
through the A/D
eLoranpulse-group
eLoran chip.
pulse-groupsignal Next, the
signaltotoobtain digital-signal-processing
obtainthe theapproximate
approximateposition unit of
positionofofthethe receiver
thepulse captures
pulsesignal. the
signal.Finally,
Finally,
eLoran
the pulse-group
theproposed
proposed cycle signal to obtain
cycleidentification
methodthe approximate
wasused
was usedto toposition
identifythe
identify of
thethe SZC
SZC pulse signal.
ofthe
of the Finally,
firsteLoran
first eLoran
the
signalproposed
signal ofthe
of cyclegroup,
thepulse
pulse identification
group, and the
and the method was used
identification
identification resulttowas
result identify
was saved
saved the
ininSZC of the first
aacomputer.
computer. The
TheeLoran
digital
dig-
signal
signal of the pulse
processing group,
in Figure and10 the
was identification
completed onresult
an was
FPGA saved
chip.
ital signal processing in Figure 10 was completed on an FPGA chip. The test platform builtin
Thea computer.
test The
platform dig-
built
ital signal
according processing
to the test in Figure
scheme is 10 was
shown
according to the test scheme is shown in Figure 11. completed
in Figure on
11. an FPGA chip. The test platform built
according to the test scheme is shown in Figure 11.
Pulse group signal

Acquisition
Pulse group signal
Acquisition
eLoran station
eLoran station
signal amplification Noise and interference Cycle identificationin Data collection

AD
andamplification
signal filtering suppression
Noise and interference joint
Cycletime-frequency
identificationin and storage
Data collection
AD
and filtering suppression joint time-frequency and storage
Digital signal processing （FPGA）

Digital signal processing （FPGA）
Figure 10. Test scheme of cycle identification method.
Figure 10.
Figure Test scheme
10. Test scheme of
of cycle
identification method.
method.
Data collection Computer

Data collection Computer
eLoran antenna
eLoran antenna
eLoran receiver
eLoran receiver
(a) (b)
(a) (b)
Figure 11. Test platform for cycle identification method: (a) outside of the test platform; (b) inside
Figure
of 11.platform.
the test Test platform for cycle identification method: (a) outside of the test platform; (b) inside
Figure 11. Test platform for cycle identification method: (a) outside of the test platform; (b) inside of
of the test platform.
the test platform.
The actual received signal is the signal emitted by the BPL timing system in China.
The actual received
The actual
The coordinates received
of the BPL signal
signal isisthe
timing the signal
signal
system emitted
(034
emitted
are by
by∘ 57 the
theN, BPL
109
BPL ∘ timing
Itssystem
33 E).system
timing GRIin andin China.
China. The
trans-
(034 ∘ ∘
The coordinates
coordinates of theof the
BPL BPL
timingtiming
system system
are ( are
034 57
◦ 570 N, 109 N, 109
◦ 33 0 E). 33
Its
mission power are 60 ms and 2 MW, respectively. In July 2021, we selected three test sites E).
GRI Its
andGRI and trans-
transmission
mission
power1000
within power
are 60
kmmsare
ofand60 ms
the andrespectively.
2system
MW, 2toMW, respectively.
receive In July
actual In July
2021,
signals. we 2021,
The we selected
selected
coordinates, threegreatthree
test sitestest sites
within
circle dis-
within
1000 km
tances, 1000
and km of the
of electric-field system
the system tostrengths to receive
receive actual actual
of thesignals.
test sites signals.
The The
arecoordinates, coordinates,
listed in Table great3. circle distances, dis-
great circle and
tances, and electric-field
electric-field strengths ofstrengths
the test sitesof the
aretest sites
listed in are
Table listed
3. in Table 3.
Remote Sens.
Remote Sens. 2022,
2022, 14,
14, 250
x FOR PEER REVIEW 15
15 of 20
of 20
Table3.3. Information
Table Information of
ofthe
thethree
threetest
testsites.
sites.
Table 3. Information of the three test sites.
Great
GreatCircle
CircleDistance Electric-Field
Electric-Field Strength
Strength
Test
Test City
City Coordinates
Coordinates Great Circle Distance Electric-Field StrengthSNR
SNR(dB)
(dB)
Test City Coordinates Distance(km)
(km) (dBμV/m)
(dBµV/m) SNR (dB)
(km) (dBμV/m)
Lintong City,City,
Lintong Shaanxi
11 Lintong (34◦∘22
City, Shaanxi (34 220 06.84
06.8400 N,N, 109
109 ∘0
◦ 1313 018.32
018.32 00 E) E) 71.271.2 60.460.4 9.8
9.8
1 Shaanxi Province
Province (34∘ 22 06.84 N, 109∘ 13 018.32 E) 71.2 60.4 9.8
Province
Fuyang City,City,
Fuyang Anhui
22 Fuyang City, Anhui (32 (32◦∘49
490 37.00
37.0000 N,N, 115
115 ∘ 0 118.0500 E)
◦ 5555 118.05 E) 635.1
635.1 65.565.5 12.1
12.1
2 Anhui Province
Province (32∘ 49 37.00 N, 115∘ 55 118.05 E) 635.1 65.5 12.1
Province
Nanjing City,City,
Nanjing Jiangsu
33 Nanjing City, (32◦∘45
Jiangsu (32 450 54.54
54.54 N, 118∘ 29
00 N, 118◦ 29 0 030.4900 E)
030.49 E) 862.5
862.5 45.145.1 −−4.9
4.9
3 Jiangsu Province
Province (32∘ 45 54.54 N, 118∘ 29 030.49 E) 862.5 45.1 −4.9
Province
The environments
The environments ofthe the three selected
testtest points
werewere different.
Test 1Test
was 11inwas in a
The environmentsofof three
the selected
three selected points
test points different.
were different. Test a closed
was in a
closed
indoor indoor environment, and Figure 12 shows the collected eLoran signal. It can be
closed environment, and Figure
indoor environment, and12 shows12the
Figure collected
shows eLoran signal.
the collected eLoran It can be seen
signal. It canthat
be
seen that the
although although the distance
receiving receiving distance was there
short and onlythere was only a groundwave
seen that although the receivingwas short and
distance was shortwas and therea was
groundwave signal, the
only a groundwave
signal,was
signal the signal was full of pulse interference.
Test 2 was inTest
the 22surrounding
was in the surrounding open envi-
signal, the full of pulse
signal interference.
was full of pulse interference. Test open environment,
was in the surrounding and
open envi-
ronment,
Figure and
13 shows Figure 13 shows
the collected the
eLoran collected eLoran
signal. It eLoran signal.
can be seen It can
thatItthere be
wasseen that
SWIthat there
in the was
signal,
ronment, and Figure 13 shows the collected signal. can be seen there was
SWI
which in the signal, which made the second half of the groundwave distorted. Test 3 was in
SWI inmade the second
the signal, whichhalf
made of the
thesecond
groundwave
half of distorted.
the groundwaveTest 3 was in a mountainous
distorted. Test 3 was in
a mountainous
location location
at a large at
circle at a large
distance, circle distance,
and Figure and
14 and Figure
shows 14
the 14 shows
collected the collected
eLoran signal. eLoran
Not
a mountainous location a large circle distance, Figure shows the collected eLoran
signal.
only wasNottheonly was the
skywave skywave interference
interference for this test for this but
strong, test the
strong,
SNR but
was the SNR
also verywas also
small,
signal. Not only was the skywave interference for this test strong, but the SNR was also
very small,
which made which
the made the groundwave
groundwave signal signal
severely severely distorted.
distorted.
very small, which made the groundwave signal severely distorted.
Normalized
Amplitude
Figure 12. eLoran signal collected in Test 1.

Figure 12. eLoran
Figure12. eLoran signal
signal collected
collectedin
inTest
Test1.1.
Normalized
Amplitude
Figure 13. eLoran signal collected in Test 2.

Figure13.
Figure 13. eLoran
eLoran signal
signal collected
collectedin
inTest
Test2.2.
Remote Sens.
Remote Sens. 2022,
2022, 14,
14, 250
x FOR PEER REVIEW 1616of
of 20
20
Figure
Figure 14.
14. eLoran
eLoran signal
signal collected
collected in
in Test 3.
Test 3.
The results
resultsafter
afterthe
thereceived
receivedsignals
signalsof of
thethe
three testtest
three sitessites
werewereprocessed
processed via the
viajoint
the
time–frequency
joint time–frequency domain cycle identification method are shown in Figures 15–17.num-
domain cycle identification method are shown in Figures 15–17. The The
ber of cycles
number identified
of cycles is 700.
identified is The
700. peak-to-peak
The peak-to-peak ratio ratio
errorserrors
of theofthree test points
the three were
test points
0.0124,
were 0.0676,
0.0124, and 0.0266,
0.0676, respectively,
and 0.0266, and the
respectively, anderrors werewere
the errors all less thanthan
all less 0.3, 0.3,
meeting the
meeting
peak-to-peak
the peak-to-peak ratioratio
evaluation
evaluationconditions. However,
conditions. as the distance
However, increased,
as the distance the disper-
increased, the
sion of the eLoran
dispersion signal signal
of the eLoran becamebecamesignificant in the propagation
significant process,
in the propagation leadingleading
process, to a larger
to a
standard deviation
larger standard of the calculated
deviation result of
of the calculated the peak-to-peak
result ratio. The
of the peak-to-peak standard
ratio. devia-
The standard
tions of theof
deviations peak-to-peak
the peak-to-peakratios ratios
were 0.0199, 0.0292,0.0292,
were 0.0199, and 0.0938, but these
and 0.0938, butresults
these had no
results
had no influence
influence on the evaluation
on the evaluation of the peak-to-peak
of the peak-to-peak ratio. Theratio.
cycle The cycle identification
identification results of
results
the of the
actual actual
signals signals
showed showed
very very identification
high cycle high cycle identification
success ratessuccess
of testsrates
1 andof2tests
and
1only
andone2 and only one identification
identification error at test pointerror1.atAlthough
test pointtest1. point
Although
3 wastestin apoint
far-off 3 was
moun- in
a far-offarea,
tainous mountainous
the success area,
ratethe success
of cycle rate of cycle
identification foridentification
this test reached for 72.86%.
this testCompar-
reached
72.86%. Comparing the
ing the experimental experimental
results results with
with the simulation the simulation
results, it can be seen results, it can
that the be iden-
cycle seen
that the cycle identification success rate of the actual measurement results
tification success rate of the actual measurement results is slightly worse. The main reason is slightly worse.
The main
is that reasoneLoran
the actual is thatsignal
the actual eLoran
received signal
contains received
pulse containscontinuous
interference, pulse interference,
wave in-
continuousand
terference wave interference
other and other
interference, whichinterference,
would cause which would of
the quality cause
the the quality
eLoran of the
signal to
eLoran signal
deteriorate andtoaffect
deteriorate and affect
the success the
rate of success
cycle rate of cycle
identification. In identification. In practical
practical applications, ac-
applications, accurate cyclecan
curate cycle identification identification
be achieved canthrough
be achieved through
multiple multiple identification
identification calculations.
calculations. Therefore, the experimental test fully verifies the practicality
Therefore, the experimental test fully verifies the practicality of the proposed joint of the proposed
time–
joint time–frequency domain eLoran cycle
frequency domain eLoran cycle identification method. identification method.
1.65
Average value: 1.5214, standard deviations: 0.0199
Cycle identification results (μs)
1.6
Peak-to-peak ratio
1.55
1.5
1.45
0 100 200 300 400 500 600 700
Cycle identification times

(a) (b)
Figure 15. Results of cycle identification for Test 1: (a) peak–peak ratio; (b) cycle identification result.
Remote
RemoteSens. 2022,14,
Sens.2022, 14,250
x FOR PEER REVIEW 17 of 20
17 20
1.8
1.8 Average value: 1.6014, standard deviations:0.0292
Average value: 1.6014, standard deviations:0.0292
1.75
(μs)
1.75
(μs)
1.7
results
1.7
results
ratio
ratio
1.65
1.65
identification
Peak-to-peak
identification
Peak-to-peak
1.6
1.6
1.55
1.55
1.5
Cycle
1.5
Cycle
1.45
1.45
1.4
1.4 0 100 200 300 400 500 600 700
0 100 200 300 400 times
Cycle identification 500 600 700
(a) (b)
(a) (b)
Results of
Figure 16. Results of cycle
identification for
for Test
Test 2: (a)
(a) peak–peak
peak–peak ratio;
ratio; (b)
(b) cycle
identification result.
result.
1.9
1.9 Average value: 1.5072, standard deviations:0.0938
Average value: 1.5072, standard deviations:0.0938
(μs)
(μs)
1.8
1.8
results
results
ratio
ratio
1.7
1.7
identification
Peak-to-peak
identification
Peak-to-peak
1.6
1.6
1.5
Cycle
1.5
Cycle
1.4
1.4
0 100 200 300 400 500 600 700

0 100 200 300 400 times
Cycle identification 500 600 700
(a) (b)
(a) (b)
4. Discussion
4. Discussion
As an important part of the PNT system, the eLoran system can form a powerful
As an important part of the PNT system, the eLoran system can form a powerful
supplement and backup to GNSS. Therefore, many countries are developing and improv-
supplement and and backup
backupto toGNSS.
GNSS.Therefore,
Therefore,many manycountries
countriesare developing
are developing and andimproving
improv-
ing eLoran systems to improve the service capabilities of the national PNT systems. Com-
eLoran systems to improve the service capabilities of the national
ing eLoran systems to improve the service capabilities of the national PNT systems. Com- PNT systems. Compared
pared with GNSS, the accuracy and reliability of the eLoran system needs to be improved.
with
paredGNSS, the accuracy
with GNSS, and reliability
the accuracy of the eLoran
and reliability systemsystem
of the eLoran needs needs
to be improved.
to be improved.Cycle
Cycle identification is a key factor in determining the positioning and timing accuracy of
identification
Cycle is a keyisfactor
identification a key in determining
factor the positioning
in determining and timing
the positioning and accuracy of eLoran
timing accuracy of
eLoran receivers.
receivers. However, However,
due to noise due to noise in the the space and the SWI caused by the signal
eLoran receivers. However, dueintothe space
noise in andthe space SWI andcaused
the SWIby the signalby
caused propagation
the signal
propagation
channel, channel,
the cycle the cycle identification
identification success rate in success
the eLoranrate in the eLoran receiver is affected.
propagation channel, the cycle identification success rate inreceiver is affected.
the eLoran receiverThe existing
is affected.
The
methodsexisting
are methods
poor in are poor
reliability in
and reliability and
anti-interference; anti-interference;
thus, this thus,
paper this paper
proposes pro-
a joint
The existing methods are poor in reliability and anti-interference; thus, this paper pro-
poses a joint
time–frequency time–frequency
domain cycledomain domain cycle
identification identification
method. After method.
the cycleAfter the cycle
identification identifi-
range
poses a joint time–frequency cycle identification method. After the cycle identifi-
cation
was range wasbydetermined
determined the spectrum by division
the spectrum
method, division
the method, theratio
peak-to-peak peak-to-peak
and the ratio and
waveform
cation range was determined by the spectrum division method, the peak-to-peak ratio and
the waveform
matching method matching
were used method were used
to determine theto toaccurate
determine SZC, the accurate SZC, which can im-
the waveform matching method were used determine thewhich can SZC,
accurate improve whichthe can
success
im-
prove
rate of the
cycle success rate
identification. of cycle identification.
prove the success rate of cycle identification.
Firstly,based
Firstly, based onthe the analysis of of the
the propagation
propagation channel channel and and interference
interference character-
character-
Firstly, based on on the analysis
analysis of the propagation channel and interference character-
istics of
istics of the
of the eLoran
the eLoran signal,
eLoransignal,
signal,the the principle
theprinciple of
principleofofthe the joint
thejoint time–frequency
jointtime–frequency
time–frequencydomain domain
domaincyclecycle
cycle identifi-
identi-
istics identifi-
cation method
fication method was
was given.
given. TheTheeffectiveness
effectiveness of the
of method
the method waswas verified by simulation,
verified by simulation,and
cation method was given. The effectiveness of the method was verified by simulation, and
the the
and anti-noise
anti-noise andand SWI SWIperformance
performance ofofthethe method
method was
was analyzed.
analyzed. Simulation results
Simulation results
the anti-noise and SWI performance of the method was analyzed. Simulation results
showedthat
showed that thismethod
method had a high success rate of cycle identification. When SNR
≥ 0 and≥0
showed that this
this methodhad hada high
a high success
success raterateof cycle identification.
of cycle identification.When SNRSNR
When ≥0
and
SGRSGR SGR ≤
≤ 23≤dB, 23 dB, the success rate of cycle identification was 100%; when
SNR ≥SNR SNR
−10≥and≥ −10 and
and 23 the
dB, success
the successrate ofratecycle identification
of cycle identificationwas 100%; when when
was 100%; −10 SGR
and
≤SGR ≤ 23the
23 dB, dB,success
the success
rate of rate
cycleof cycle identification
identification was greater
was greater than when
than 75%; 75%; whenSNR ≥ SNR
−13≥
SGR ≤ 23 dB, the success rate of cycle identification was greater than 75%; when SNR ≥
Remote Sens. 2022, 14, 250 18 of 20
and SGR ≤ 23 dB, the success rate of cycle identification was greater than 55%. When the
success rate of the current cycle identification method reached 100%, the delay-locked loop
method required SNR ≥ 20 dB, the waveform matching method required SNR ≥ 23 dB,
and the delayed-addition method required SNR ≥ 24 dB. Although the performance of
these methods can be improved by 10–18 dB after filtering, these methods can only meet
the requirements under the condition of SGR ≤ 6 dB. Obviously, the cycle identification
method proposed in this paper was better than the single cycle identification method for a
combination of noise and skywave interference.
Finally, we applied the cycle identification method to the developed eLoran receiver
and receive actual signals within 1000 km for testing. The test results show that under the
condition of high SNR, the cycle identification success rate can reach 100%. Even in the
862.5-km mountain test environment, the method can achieve a 72.86% cycle recognition
success rate. In practical applications, higher identification success rates can also be
achieved by counting the results of multiple cycle identifications. Thus, the method has
high engineering application value, and it is optimized for the design of the new eLoran
navigation and timing receivers. It will be able to improve the accuracy of the application
of modern eLoran systems, making it a more reliable backup for GNSS.
5. Conclusions
In this paper, a cycle identification method based on a joint time–frequency domain
was proposed for the standard zero crossing detection of groundwave signals in eLo-
ran positioning and timing receivers. We proved that this method has a good ability of
anti-skywave and anti-noise detection and can address the limitation of poor detection
performance of the current cycle identification method. This method also has a greater
application potential, because it can be used for period identification as well as to obtain the
time delay and amplitude of the skywave in the received signal, which is important for the
receiver. In the future, we will use this method to separate the skywave and groundwave,
in order to reduce the impact of skywave instability on the groundwave measurement.
This can improve the TOA measurement accuracy of the eLoran signal and the information
demodulation ability of the signal.
Author Contributions: Conceptualization, Y.H. and S.L.; methodology, W.Y. and M.D.; software, W.Y.
and J.Y.; validation, W.Y., C.Y. and Z.H.; formal analysis, M.D.; investigation, M.D. and J.Y.; resources,
S.L.; data curation, W.Y. and C.Y.; writing—original draft preparation, W.Y.; writing—review and
editing, W.Y. and Y.H.; visualization, Z.H. and C.Y.; supervision, Y.H. and S.L.; project administration,
W.Y. and C.Y.; funding acquisition, J.Y. All authors have read and agreed to the published version of
the manuscript.
Funding: This work was supported by Chinese Academy of Sciences “Light of West China” Program
(Grant No. XAB2018B21) and National Time Service Center “Youth Innovative Talent Project” (Grant
No. Y917SC1S08).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Restrictions apply to the availability of these data. The ownership
of data belongs to the National Time Service Center (NTSC). These data can be available from the
corresponding author with the permission of NTSC.
Acknowledgments: The authors would like to thank their colleagues for testing of the data provided
in this manuscript. We are also very grateful to our reviewers who provided insight and expertise
that greatly assisted the research.
Conflicts of Interest: The authors declare no conflict of interest.
Remote Sens. 2022, 14, 250 19 of 20
References
1. Xu, J.N. Analysis on underwater PNT system and key technologies. Navig. Position. Timing 2017, 4, 1–3. [CrossRef]
2. Zhang, H.Y.; Xu, B.; Li, H.B.; Liao, D.Y.; Chen, H.Q. Prospects for development of loran-C navigation system for building up
national comprehensive PNT system. J. Astronaut. Metrol. Meas. 2020, 40, 6–7. [CrossRef]
3. Gao, Z.Z.; Ge, M.R.; Li, Y.; Shen, W.B.; Zhang, H.P.; Schuh, H. Railay irregularity measuring using rauch-tung-striebel smoothed
multi-sensors fusion system: Quad-GNSS PPP, IMU, odometer, and track gauge. GPS Solut. 2018, 22, 22–36. [CrossRef]
4. Di, Q.; Dan, B.; Sherman, L.; Per, E. Reliable location-based services from radio navigation systems. Sensors 2010, 10, 11369–11385.
[CrossRef]
5. Specht, M. Determination of navigation system positioning accuracy using the reliability method based on real measurements.
Remote Sens. 2021, 13, 4424. [CrossRef]
6. Krasuski, K.; Savchuk, S. Determination of the precise coordinates of the GPS reference station in of a GBAS system in the air
transport. Commun. Sci. Lett. Univ. Zilina 2020, 22, 11–18. [CrossRef]
7. Li, Y.; Hua, Y.; Yan, B.R.; Guo, W. Experimental study on a modified method for propagation delay of long wave signal. IEEE
Antennas Wirel. Propag. Lett. 2019, 18, 1719. [CrossRef]
8. Specht, C.; Weintrit, A.; Specht, M. A History of maritime radio-navigation positioning systems used in Poland. J. Navig. 2016, 69,
468–480. [CrossRef]
9. Offermans, G.; Bartlett, S.; Schue, C. Providing a resilient timing and UTC service using eLoran in the United States. Navig. J. Inst.
Navig. 2017, 64, 339–345. [CrossRef]
10. Ren, S.H.; Lou, Y.L.; Yang, N.; Zheng, X.X.; Kang, S.L. Navigation warfare and its countermeasures. J. Navig. Pos. 2020, 8, 100–104.
[CrossRef]
11. Ge, Y.T.; Xu, L.L. Analysis on the disadvantage of American navigation and timing warfare. Aerodyn. Missile J. 2020, 6, 6–11.
[CrossRef]
12. Wang, X.Y.; Zhang, S.F.; Sun, X.W. The additional secondary phase correction system for AIS signals. Sensors 2017, 17, 736.
[CrossRef]
13. Strategy for the Department of Defense Positioning, Navigation and Timing (PNT) Enterprise—Ensuring a US Military PNT
Advantage. Available online: https://rntfnd.org/wp-content/uploads/DoD-PNT-Strategy.pdf (accessed on 15 October 2021).
14. Yan, W.H.; Zhao, K.J.; Li, S.F.; Wang, X.H.; Hua, Y. Precise loran-C signal acquisition based on envelope delay correlation method.
Sensors 2020, 20, 2329. [CrossRef]
15. About the Main Directions (Plan) of the Development of Radio Navigation CIS Member States for 2019–2024. Available online:
https://rntfnd.org/wp-content/uploads/CIS-Russia-Radionav-Plan-2019-2024.pdf (accessed on 18 October 2021).
16. Son, P.W.; Park, S.G.; Han, Y.; Seo, K. eLoran: Resilient positioning, navigation, and timing infrastructure in maritime areas. IEEE
Access 2020, 8, 193708–193710. [CrossRef]
17. Rhee, J.H.; Kim, S.; Son, P.W.; Seo, J. eLoran: Enhanced accuracy simulator for a future Korean nationwide eLoran system. IEEE
Access 2021, 9, 115042–115045. [CrossRef]
18. Zhou, L.L.; Yan, J.J.; Mu, Z.L.; Pu, Y.R.; Wang, Q.Q.; He, L.F. Field-strength variations of LF one-hop sky waves propagation in the
earth-ionosphere waveguide at short ranges. IEEE Trans. Antennas Propag. 2021, 69, 3443–3445. [CrossRef]
19. Last, J.D.; Farnworth, R.G.; Searle, M.D. Effect of skywave interference on the coverage of loran-C. IEE Proc. F-Radar Signal Process.
1992, 139, 306–312. [CrossRef]
20. Zou, D.C.; Bian, Y.J.; Wu, H.T.; Su, J.F.; Li, Y. Digital algorithm realized with SOPC for cycle identification of Loran-C. J. Harbin
Inst. Technol. 2005, 37, 1644–1646. [CrossRef]
21. Fisher, A.J. Loran-C cycle identification in hard-limiting receivers. IEEE Trans. Aerosp. Electron. Syst. 2000, 36, 290–297. [CrossRef]
22. Yan, Y.J.; Guo, Q. A novel signal search algorithm for loran-C receiver. Electron. Technol. 2010, 47, 20–21. [CrossRef]
23. Yan, W.H.; Hua, Y.; Yuan, J.B.; Zhao, K.J.; Li, S.F. A ioint detection method of cycle-identification for loran-C signal. In
Proceedings of the 2017 13th IEEE International Conference on Electronic Measurement & Instruments (ICEMI), Yangzhou, China,
20–22 October 2017. [CrossRef]
24. Lin, H.Z.; Luo, B.F.; Hu, D.L.; Yang, Z.J. Algorithm of matching wave for cycle identification of LORAN-C. Ship Electron. Eng.
2008, 167, 81–83. [CrossRef]
25. Sui, J.P.; Jin, X.Q.; Shu, D.L.; Ma, R.S. Improved Loran-C joint identification algorithm for sky-ground wave and cycle based on
optimized envelope. Electron. Meas. Technol. 2020, 43, 115–119. [CrossRef]
26. Li, S.F.; Gao, Y.Y.; Hua, Y. Joint cycle-identification method for Loran-C signal. J. Jiangsu Univ. 2014, 35, 547–551. [CrossRef]
27. Bian, Y.; Last, J.D. Loran-C skywave delay estimation using eigen-decomposition techniques. eLectronics Lett. 1995, 31, 133–134.
[CrossRef]
28. Mohammed, A.; Last, D. Loran-C skywave delay estimation using the AR algorithm. eLectronics Lett. 1998, 34, 2217–2219.
[CrossRef]
29. Zhu, Y.B.; Xu, J.N.; Wang, H.X.; Cao, K.J.; Hu, D.L. A New Loran C Sky-wave and ground-wave identification algorithm based on
IFFT spectral division. J. Electron. Inf. Technol. 2009, 31, 1153–1156. [CrossRef]
30. Hu, D.L.; Yang, Y.C.; Fan, J.H. Loran-C receiver sky-wave detection based on MUSIC algorithms. J. Nav. Aeronaut. Astronaut.
Univ. 2006, 18, 7–9. [CrossRef]
Remote Sens. 2022, 14, 250 20 of 20
31. U.S. Coast Guard and the U.S. Coast Guard Auxiliary. Loran-C User Handbook. Available online: https://www.loran.org/
otherarchives/-1992%20-Loran-C%20User%20Handbook%20-%20USCG.pdf (accessed on 13 October 2021).
32. Lo, S.C.; Peterson, B.B.; Enge, P.K. Loran data modulation: A primer. IEEE Aerosp. Electron. Syst. Mag. 2007, 22, 31–38. [CrossRef]
33. Li, S.F.; Wang, Y.L.; Hua, Y.; Xu, Y.L. Research of Loran-C data demodulation and decoding technology. Chin. J. Sci. Instrum. 2012,
33, 1407–1413. [CrossRef]
34. Zhou, L.L.; Yan, J.J.; Mu, Z.L.; Wang, Q.Q.; Liu, C.L.; He, L.F. Study on time delay characteristics of low frequency one-hop sky
waves in the isotropic ionosphere. J. Electron. Inf. Technol. 2020, 42, 1606–1610. [CrossRef]
35. Liatos, P.; Hussein, A.M. Characterization of 100-kHz noise in the lightning current derivative signals measured at the CN tower.
IEEE Trans. Electromagn. Compat. 2005, 47, 986–997. [CrossRef]
36. Minimum Performance Standards for Marine eLORAN Receiving Equipment. Radio Technical Commission for Maritime
Services. Available online: https://rtcm.myshopify.com/collections/maritime-navigation-equipment-standards/products/
copy-of-differential-gnss-package-both-of-the-current-standards-10402-3-and-10403-3 (accessed on 11 October 2021).
37. Yang, S.H.; Lee, C.B.; Lee, Y.K.; Lee, J.K.; Kim, Y.J.; Lee, S.J. Accuracy Improvement technique for timing application of LORAN-C
signal. IEEE Trans. Instrum. Measurement 2011, 60, 2648–2654. [CrossRef]
38. Wu, H.R.; Liu, R.Z. A new algorithm for sky-wave and ground-wave detection of loran C based on FFT/IFFT technology. J. Nav.
Aeronaut. Astronaut. Univ. 2009, 24, 317–319. [CrossRef]
39. Li, C.; Zhen, J.; Chang, K.; Xu, A.; Zhu, H.; Wu, J. An indoor positioning and tracking algorithm based on angle-of-arrival using a
dual-channel array antenna. Remote Sens. 2021, 13, 4301. [CrossRef]
40. Wu, M.; Xu, J.N.; Chen, Z.Y.; Wang, G.C. Skywave delay estimation of Loran C based on modern signal processing methods.
Hydrogr. Surv. Charting 2011, 31, 1–13. [CrossRef]

Remotesensing 14 00250

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Remotesensing 14 00250

Uploaded by

Copyright:

Available Formats

remote sensing

Received: 27 November 2021

Remote Sens. 2022, 14, 250. https://doi.org/10.3390/rs14020250 https://www.mdpi.com/journal/remotesensing

2. Materials and Methods

2.1. eLoran Signal Format

where t is time (µs), A is a normalization constant related to the peak amplitude, f is

= 20 log( ) = 20log( ), (3)

Ditital Bandpass Filter Fast Fourier Window Peak-to-peak Ratio

Figure 5. Principle of time–frequency domain joint cycle identification method.

2.3.2. Spectrum Division in Frequency Domain

F −1 ( Z ( f )) = δ(t − T1 ) + rSGR δ(t − T2 ) + F −1 [ H ( f )W ( f )], (12)

where F −1 represents the IFFT.

2.3.3. Peak-to-Peak Ratio and Waveform Matching in Time Domain

s(t + 2.5) 2 −10/65

Table 2. Theoretical calculation results of peak-to-peak ratio at different zero-crossing points.

∆h(n) = |hn (t) − 1.5338|, (14)

Remote Sens. 2022, 14, 250 10 of 20

Cycle identification error rate (%)

Pulse group signal

signal amplification Noise and interference Cycle identificationin Data collection

Digital signal processing （FPGA）

Data collection Computer

Figure 12. eLoran signal collected in Test 1.

Figure 13. eLoran signal collected in Test 2.

Cycle identification times

0 100 200 300 400 500 600 700

You might also like