HF source geolocation using an operational TDoA

Receiver network: experimental results

Ankit Jain, Pascal Pagani, Rolland Fleury, Michel Ney, Patrice Pajusco

To cite this version:

Ankit Jain, Pascal Pagani, Rolland Fleury, Michel Ney, Patrice Pajusco. HF source geolocation
using an operational TDoA Receiver network: experimental results. IEEE Antennas and Wireless
Propagation Letters, Institute of Electrical and Electronics Engineers, 2018, 17 (9), pp.1643-1647.
�10.1109/LAWP.2018.2860459�. �hal-01863584�

HAL Id: hal-01863584

https://hal.archives-ouvertes.fr/hal-01863584
Submitted on 21 Feb 2022

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 HF Source Geolocation using an Operational TDoA
Receiver Network: Experimental Results
Ankit Jain, Pascal Pagani, Rolland Fleury, Michel M. Ney and Patrice Pajusco

Abstract Determining the geographic location of high To further improve the accuracy of measured angle of arrival
frequency (HF) transmitters is of prime importance in both (AoA), several high resolution DoA estimation algorithms
defense and civil operations. In this letter, we present experimental such as MUSIC [3] and ESPRIT [4] can be used to resolve the
results of distant HF transmitter geolocation performed using the uncorrelated signals. These algorithms face problems when
time difference of arrival (TDoA) method. The HF signals
uncorrelated and coherent signals exist together. An efficient
considered for our study are transmitted from different locations
across Europe. Multiple receivers capable of capturing HF radio DoA method is detailed in [5] which estimates the DoA for
signals synchronously are built and a nationwide receiver network both uncorrelated and coherent signals separately. HF radio
with centralized control is deployed in France to assess the DF using MUSIC algorithm is experimentally validated by
feasibility of our approach. HF radio signals are captured deploying a receiver system based on a heterogeneous circular
simultaneously by all receivers in the network from different array [6]. There is a slight bias between the estimated azimuth
European HF broadcast transmitters. From the captured data, the and elevation angles when compared with the expected
propagation duration differences between the signals received by values.
all receivers are obtained using the cross-channel sounding Another HF geolocation technique, known as single site
method. Finally, the transmitter location is estimated using the
location (SSL) estimates the DoA in terms of azimuth and
geolocation algorithm based on the TDoA method. Experimental
results demonstrate that it is possible to locate the HF transmitters elevation angle of the incoming signal [7]. Moreover using
under favorable conditions with a relative geolocation error the SSL method, it is possible to locate a HF transmitter
ranging from about 0.1 to 10% of the actual ground distance, for located about 2000 km away from the receiving system with
most of the experimental scenarios. an error of about 10% [8]. Both SSL method and traditional
Index Terms Geolocation, HF propagation, Ionosphere, DF systems require the deployment of large antenna arrays at
TDoA single or multiple sites which can be quite expensive and
bulky. It is also mandatory to have information about the
I. INTRODUCTION ionospheric propagation path to estimate the ionospheric
reflection height.
Radio signals from the HF band cover large geographical A study for passive location systems using TDoA method
areas and can be used for long distance communication across for Line-of-Sight (LOS) transmissions is explained briefly in
intercontinental distances using light and portable devices. [9]. In the case of HF geolocation, signals are received
Thus, HF communication is extremely useful for defense and through non-line-of-sight (NLOS) mode. HF geolocation is
civil operations in remote areas. In this framework, the possible using both time of arrival (ToA) and time difference
location of HF transmitters can be passively retrieved by of arrival (TDoA). In our earlier work, it is shown that the
capturing and analyzing the HF communication signals. HF location of HF transmitters can be found using the ToA
radio signals propagate through the ionospheric channel using method without any information about the ionospheric profile
skywaves; reflected once or multiple times between the Earth along the propagation path [10]. Furthermore, single antenna
and the ionosphere. The ionosphere is composed of neutral portable receiver systems can be used at multiple sites which
atoms the makes it more cost-effective compared to a SSL or DF
. This results in free electrons which are system. When the transmission time of the signal is known,
responsible for the refraction of HF radio signals. The HF geolocation can be performed using the ToA principle
ionosphere is further split into different layers with varying whereas when the transmission time of the signal is unknown;
electron density and the characteristics of these layers are HF geolocation can be performed using the TDoA principle.
highly variable depending on the solar activity, geographic In the context of passive geolocation, the transmit time of
location and season [1]. the signal is not known. Thus, we use the TDoA method to
Traditionally, HF geolocation was performed by measuring evaluate the geographic location of the HF transmitter. This
the Direction of Arrival (DoA) of the incoming signals. letter presents for the first time experimental results of HF
Different forms of HF direction finding (DF) systems capable geolocation using the TDoA method, obtained by collecting
of measuring the DoA are explained in [2]. data from a country wide receiver network capable of
Manuscript received March 14, 2018. This work was supported in part by capturing HF radio signals synchronously. In Section II, HF
the Carnot Institute and the Brittany region in France. The authors thank the propagation and the geolocation method based on TDoA are
Atmospheric Optics Laboratory in University of Lille 1 and the Electrical described. Section III introduces the HF geolocation system
Engineering Department in IUT1-University Grenoble Alps for hosting our
receiver setup.
allowing the capture of HF radio signals. In Section IV,
The authors are with the Microwave Department, IMT Atlantique, Lab- experimental geolocation results of different HF transmitters
STICC UMR CNRS 6285, Brest, France (Corresponding author e-mail: are presented and discussed. In Section V, conclusions are
ankit.jain@imt-atlantique.fr). drawn and future work is described.
 II. HF PROPAGATION AND GEOLOCATION PRINCIPLE Nauen-Germany, Santa Maria di Galeria (SMDG)-Italy,
HF radio signals propagating using skywaves reach the Cerrik-Albania, Galbeni-Romania and Moosbrunn-Austria.
destination via a single or multiple reflections from any The geographic location of all these transmitters and the
ionospheric layer. Signals which are reflected once from the deployed receivers is presented in Fig. 2. Table 1 provides the
ionosphere are referred to as one-hop propagation mode. ground distances between different HF broadcast transmitters
Using one-hop mode, communication is possible over and receivers. Radio signals transmitted using HF
distances in the range of 500-3000 km. broadcasters are amplitude modulated and transmitted over a
The geometry of HF propagation paths, assuming a single 10-kHz band.
reflection from a uniform ionosphere, can be seen in Fig. 1.
Moreover, the Earth is assumed to be flat by neglecting its
curvature. The transmitting site is located at Tx while the
receiving sites are located at Rx1 and Rx2. Due to the
gradient-index in the ionosphere, the HF radio signals travel

the group path between Rx1-Tx indicated by the red curve is

equal to the triangular path Rx1-I1-Tx in vacuum; the same
applies for Rx2-Tx. Under the assumption of uniform
ionosphere, the actual ionospheric reflection height for both
propagation paths is the same and denoted by hm. This
reflection height is usually obtained from some ionospheric
modelling software or an ionosonde. Similarly, the reflection
heights for both triangular paths are also equivalent and Fig. 2 GPS coordinates of HF receivers and transmitter sites
denoted by hv.
h=2hv Tx' The automatically captured data files are post-processed and
data is retrieved in the form of complex samples (IQ data). To
remove other interfering signals which are captured along
I1 I2
with the signal of interest, the captured data are filtered in the
hv
hm frequency domain. Over the available 200 kHz band, only 5
kHz of signal on both sides of the carrier frequency of the
transmitter is passed and the rest is filtered out. The signal is
transformed back to the time domain and the
h=0
Rx1 Rx2 Tx
average power is calculated. The transformed signal is
demodulated to obtain the message signal.
Fig. 1 Geometry of HF geolocation in TDoA [11] The geolocation algorithm used to find the Tx location is
dependent on the reference receiver. Thus, the receiver with
From the electromagnetic image theory, the triangular path the highest signal average power is considered as the
length Rx1-I1-Tx is equal to the path length Rx1-Tx' and the reference receiver. The message signal from the reference
same applies for the second triangular path. Using this receiver is cross-correlated with the obtained message signals
geometry, the location of the HF transmitter can be obtained from the other receivers to obtain the propagation duration
by finding the coordinates of point Tx' without any differences. The TDoA ( ) corresponds to the maximum
knowledge of the actual ionospheric reflection height. The value of the cross-correlation function and is given as follows:
group path range differences are obtained by multiplying the
propagation duration differences with the speed of light from (1)
at least three different receiver pairs. From the known receiver where is the cross-correlation between the message
locations and the group path range differences, the point Tx' is signal received on the reference receiver denoted by
found by multilateration. In our previous work we have
and other receivers denoted by , respectively. Finally, the
presented the mathematical approach and the formulae for
Tx location is estimated using the obtained TDoA values by
obtaining the location of Tx' from the TDoA method [11].
solving a system based on quadratic equations [11].
III. HF GEOLOCATION SYSTEM
Table 1 Great circle distance between the receivers and HF transmitters
Four receivers capable of capturing HF signals Tx- Rx1 Tx- Rx2 Tx- Rx3 Tx- Rx4
Tx
synchronously were built using software defined radio (SDR) (km) (km) (km) (km)
in order to perform HF geolocation using TDoA method. We Nauen 1320 1313 976 710
installed these receivers in 4 cities in France namely, Rx1 in SMDG 1493 1085 632 1182
Brest, Rx2 in Bordeaux, Rx3 in Grenoble, and Rx4 in Lille. Cerrik 2049 1680 1197 1631
The HF receiver prototype is explained in detail in [12]. HF Galbeni 2376 2168 1676 1805
radio signals are captured synchronously by four receivers Moosbrunn 1553 1349 872 1004
from five different HF broadcast transmitters located in
 IV. EXPERIMENTAL RESULTS B. Measurement Example: Moderate Geolocation Error
In the measurement campaign, all HF radio signals were This example provides the analysis of a HF signal captured
captured for 5-second duration with a one-minute interval from the HF transmitter located in SMDG by all receivers in
between start times of two captures from the same transmitter. the network on 18th July at 16:11 UTC. Based on the received
The sampling rate was set equal to 200 kHz. Using azimuthal signal strengths at different receivers, the signal received in
equidistant projection system, the GPS coordinates of the HF Grenoble was considered as the reference signal. Further, the
transmitters and all receivers in the network were converted received signals were post processed as explained earlier and
into a rectangular coordinate system. Coordinates of the three TDoA estimates were obtained between the signal
selected reference receiver were used as the origin of the received in Grenoble-Brest, Grenoble-Bordeaux and
coordinate system. Details of the captures from different HF Grenoble-Lille, respectively. Using the geolocation algorithm,
transmitters over several days are summarized in Table 2. the Tx coordinates were estimated and the geolocation error
Some of these captured data from different HF transmitters was about 82 km. The geolocation error seems moderate and
are analyzed and discussed in the following section. is equivalent to a relative error of about 5.5%. The actual and
estimated transmitter location can be seen in Fig. 5.
Table 2 Details of capture schedule from different HF transmitters
Transmitter Capture Capture No. of C. Measurement Example: Geometric Dilution of Precision
Date
site freq (MHz) time (UTC) captures In this example, data captured from Moosbrunn on 27th July
17-07-17 Galbeni 15.130 13:15-13:18 4 at 20:11 UTC are analyzed. The signal received in Grenoble
18-07-17 SMDG 15.570 16:11-16:20 10 is considered as the reference signal. Using the known
21-07-17 Nauen 11.790 19:11-19:20 10 receiver positions and the TDoA estimates, the geographic
25-07-17 Cerrik 13.665 11:21-11:25 5 location of the transmitter is estimated as illustrated in Fig. 5.
27-07-17 Moosbrunn 11.880 20:11-20:15 5 The geolocation error is large, about 303 km which
corresponds to a relative error of 19.5%. Geolocation based
A. Measurement Example: Favorable Propagation Environnent on TDoA estimations suffers from an error caused by long
A HF radio signal which was captured synchronously by the baseline. The baseline refers to the
four receivers on 21st July at 19:17 UTC from Nauen is distance between receivers. TDoA based systems locate the
studied. The received signals on all receivers were target by using multilateration principle at intersection of
demodulated and the message signal was extracted as hyperbolas. For geolocation of the Tx in Moosbrunn and
explained in the earlier section. The signal received in Galbeni using our receiver network, it is found that the
Grenoble was considered as the reference signal on the basis hyperbolic line of positions (LOPs) are nearly parallel when
of the received signal power. An example of propagation exact TDoA values are known. A small error in TDoA due to
duration difference between the HF signal received in noise results in large geolocation error and this is referred as
Grenoble and Brest is presented in Fig. 3, calculated by using Geometric Dilution of Precision (GDOP). The GDOP gets
cross-correlation technique. As seen in Fig. 3, there is a single farther away from receiver baselines
cross-correlation peak which corresponds to one-hop [13], which can be seen from the geolocation errors obtained
propagation mode for signals received by both receivers. for the Tx in Galbeni. This example explains a case where a
Using obtained TDoA estimates, the HF transmitter site was ads to a large
located with an error as small as 1.1 km. This error geolocation error.
corresponds to a relative error of about 0.1% calculated with
respect to the maximum ground range among the four D. Measurement Example: Effect of Multipath Propagation
possible HF links. Fig. 5 presents the actual and estimated This example investigates data captured from the HF
geographic location of the HF Tx in Nauen. transmitter located in Cerrik on 25th July at 11:25 UTC. The
signal received in Lille was considered as the reference signal
considering the received signal intensity on different
receivers. By cross-correlating the reference signal with the
signal received by
obtained. Fig. 4 presents the cross-correlation output for the
signals captured in Lille and Grenoble. As seen in Fig. 4,
there are two peaks separated by 220 sec in cross-correlation
output with different amplitudes. This implies that the signal
was received at one of the receivers through multi-path
channels. We have discussed the effect of peak selection on
the geolocation error in [12]. In the present work, the time
delay corresponding to the maximum power was selected as
the TDoA estimate. Finally, Tx coordinates were estimated
using the geolocation algorithm. The geolocation error was
Fig. 3 TDoA estimate for HF signal captured in Grenoble and Brest from HF about 162 km corresponding to a relative error of about 8%.
transmitter located in Nauen, Germany on 21st July, 2017 at 19:17 UTC
The geographic location of estimated Tx can be seen in Fig 5.
 signals follow one-hop propagation mode. In order to assess
the feasibility of HF geolocation using the TDoA method,
four single antenna receivers were deployed in four different
cities in France. We performed a first series of experimental
measurements with different HF stations located around
Europe on different days. The TDoA estimates are obtained
using the cross-correlation method. From the known receiver

using the geolocation algorithm. Geolocation performance is

presented and discussed for HF signals captured from
different transmitters at different times of different days.
In favorable conditions, the minimum geolocation error for

Considering the coverage area for one-hop HF propagation

Fig. 4 TDoA estimate for HF signal captured in Lille and Grenoble from mode in km, these errors are small and acceptable. Obtained
Cerrik, Albania on 25th July at 11:25 UTC geolocation errors can be further reduced by capturing data on
a larger number of receivers [11]. From the analysis of
geolocation results, it can be concluded that HF geolocation is
possible using the TDoA method. In our future work, we plan
to evaluate the confidence degree of an estimated location
based on the output ionospheric height obtained from the
geolocation algorithm or the level of received power at
different receiving sites. Different super-resolution algorithm
needs to be studied in order to reduce the quantization error of
the TDoA estimates, which can further improve the
geolocation accuracy.

REFERENCES
[1] IEE Electromagnetic Waves Series 31,
Fig. 5 Estimated and actual geographic location of HF transmitters for pp. 124-152, n° ISBN: 0 86341 186 X, 1996.
[2] Proceedings of the Institution
different measurement examples of Electrical Engineers, vol. 113, no. 1, pp. 49 61, 1966.
[3] R. O. Schmidt, "Multiple emitter location and signal parameter
Table 3 presents the summary of geolocation error for all estimation", IEEE Trans. Antennas Propag., vol. AP-34, no. 3, pp. 276-
the HF radio signals captured from different transmitters 280, Mar 1986.
[4] R. Roy, T. Kailath, "ESPRIT Estimation of signal parameters via
around Europe. Out of the five captures from Cerrik, only two rotational invariance techniques", IEEE Trans. Acoust. Speech Signal
captures were exploitable. For the other three captures, the Processing, vol. 37, no. 7, pp. 984-995, Jul 1989.
data captured in Brest was too noisy thereby making them [5] Y. Zhang and Z. Ye, "Efficient Method of DOA Estimation for
unusable. The error in terms of distance for the Tx located in Uncorrelated and Coherent Signals," in IEEE Antennas and Wireless
Propagation Letters, vol. 7, pp. 799-802, 2008.
Moosbrunn and Galbeni were found to be large mainly due to [6]
GDOP but the algorithm output in terms of direction of finding operating on a heterogeneous array: principles and experimental
localization was approximately correct. Moreover, the Radio-Science, vol 39, n°1, pp. 1003:1-14, Jan-Feb 2004.
geolocation algorithm assumes one-hop propagation mode [7] -driven approach to HF
Signal Processing, vol. 93, no. 12, pp. 3487 3503, Dec.
and a uniform ionosphere which might not be the case for all 2013.
captured signals. For all HF signals captured from SMDG and [8] Site
Nauen, the geolocation error was always less than 7%. IEEE Military Communications Conference, 2007.
MILCOM 2007, 2007, pp. 1 5.
[9] IEEE
Table 3 Geolocation error for data captured from different HF transmitters Trans. on Aerospace and Electronic Systems, vol. AES-20, no. 2, pp.
Tx (Captures Error min. Error max. Error median 183 198, Mar. 1984.
considered) (km, %) (km, %) (km, %) [10]
Galbeni (3) 39.96, 1.68 845.88, 35.57 504.19, 21.20 transmitter geolocation through single-
The 8th European Conference on Antennas and Propagation (EuCAP),
SMDG (10) 44.02, 2.95 96.69, 6.48 82.28, 5.51 2014, pp. 2689 2693.
Nauen (10) 1.16, 0.08 71.94, 5.44 30.77, 2.33 [11] A. Jain, P. Pagani, R. Fleury, M.
Cerrik (2) 162.13, 7.90 283.14, 13.81 - The
Moosbrunn (5) 138.79, 8.93 464.66, 29.90 303.15, 19.51 11th European Conference on Antennas and Propagation (EuCAP),
2017, pp. 1852 1856.
[12] -
V. CONCLUSION Channel Sounding for HF Geolocation: Concepts and Experimental
The 12th European Conference on Antennas and Propagation
In this letter, we presented first results of geolocation of HF (EuCAP), 2018, accepted for publication.
sources which are located in the range of about 500-2500 km [13] R. A. Poisel, Elect
by using the TDoA method. It is explicitly assumed that Edition, Artech House, pp. 226-252, n° ISBN 13: 978-1-60807-523-2.