Forward-scatter Doppler-only Distributed Passive Covert Radar

Abstract

Forward-scatter geometry places the radar target in diffraction region, resulting in enhanced radar cross section. Passive covert radar using forward-scatter geometry with Doppler-only geolocation can provide medium to long range early warning capability with very moderate cost. In addition to direct exploitation of target data in command and control systems, the early warning capability can be used for allocation and cueing of other air surveillance resources. Pattern-recognition and inverse-problem approaches have given a computationally light track-before-detect algorithm in time-Doppler domain. Given time series of Doppler observations from two or more transmitter-receiver baselines that observe overlapping volumes of space, various methods can be used for constructing kinematic information. The transmitter-receiver baselines can utilise different or same transmitters as long as the fields of view of the baselines have enough overlap. We present a fault-tolerant network structure of low-cost passive radar stations. Detection and classification results from a proof-of-concept version of such network are presented. The ideas are extensible to correlation receivers and other deployment geometries.

Full text

STO-MP-SET-231 PAPER NBR - 1
NATO UNCLASSIFIED
NATO UNCLASSIFIED
Forward-scatter Doppler-only Distributed Passive Covert Radar
Ari J. Joki
Finnish Defence Forces' Logistics Command, Air Force Systems Division
P.O. Box 14; FI-41161 Tikkakoski; Finland
FINLAND
ari.joki@mil.fi
Piotr Ptak, Juha Hartikka, Mauno Ritola, Tuomo Kauranne
Lappeenranta University of Technology.
Lappeenranta, Finland
piotr.pawel.ptak@gmail.com
Azra Tayebi (until 31.5.2016), Adam Ludvig
Arbonaut Ltd.
Joensuu, Finland
ABSTRACT
Forward-scatter geometry places the radar target in diffraction region, resulting in enhanced radar cross
section. Passive covert radar using forward-scatter geometry with Doppler-only geolocation can provide
medium to long range early warning capability with very moderate cost. In addition to direct exploitation
of target data in command and control systems, the early warning capability can be used for allocation and
cueing of other air surveillance resources.
Pattern-recognition and inverse-problem approaches have given a computationally light track-before-
detect algorithm in time-Doppler domain. Given time series of Doppler observations from two or more
transmitter-receiver baselines that observe overlapping volumes of space, various methods can be used for
constructing kinematic information. The transmitter-receiver baselines can utilise different or same
transmitters as long as the fields of view of the baselines have enough overlap.
We present a fault-tolerant network structure of low-cost passive radar stations. Detection and
classification results from a proof-of-concept version of such network are presented. The ideas are
extensible to correlation receivers and other deployment geometries.
1.0 MOTIVATION
Air traffic monitoring over sparsely populated or ocean areas has similar challenges to homeland security
long-distance early warning. In both cases the need is to detect and track relatively far away airborne
objects. High measurement accuracy is not very important. Poor detections are better than none. The more
conventional microwave radar bands suffer from radio horizon limitations, whereas the frequencies where
propagation follows earth’s curvature a little better suffer from congestion of various users. A natural idea
is to see if it is possible to take advantage of these other users, leading to re-ignition of interest in passive
multistatic radar technologies.
One motivation for this work has been to find moderate-cost solutions to extending air surveillance

Forward-scatter Doppler-only Distributed Passive Covert Radar
PAPER NBR - 2 STO-MP-SET-231
NATO UNCLASSIFIED
NATO UNCLASSIFIED
coverage. Such need arises both in civilian and military applications, and may be of use even for non-
governmental organizations.
For air traffic safety augmentation one possibility is to take advantage of existing radio amateur and dx-
listener resources. During the periods where the proprietor is not using the receiving equipment for their
hobby, perhaps the receiver could be utilized as part of an air traffic monitoring network. An otherwise
obsolete laptop has proved sufficient for much of the needs of a station as described in this paper and its
references. The proprietor of the receiving station can be further motivated to donate resources by having
the station software display on a feedback channel the centrally compiled air traffic picture.
In rapid deployment of national or international security activities it may not be easy to take along a
sufficient amount of active radars to ensure at least surveillance of the airspace of the theatre of operations.
The solution in this paper can be constructed from moderately low-cost software defined radios and very
moderate computers.
The existence and operation of passive radar is by definition undetectable. The antennas of the proof of
concept stations described in this paper and its references are indistinguishable from consumer radio and
television antennas. It is no longer practically feasible and in most nations politically unthinkable to forbid,
prevent, or sanction radio reception activities as long as communication confidentiality is not breached.
National security organizations need to be aware that even if they are not deploying passive radar assets,
someone else may do so. Data from such installations can become openly accessible on the internet, or it
may remain the possession of the groups that operate such installations.
2.0 TECHNICAL BASIS
2.1 Forward-Scatter Geometry
When electromagnetic field propagating as a plane wave encounters an opaque object whose dimensions
are “large” compared to the wavelength, physical diffraction considerations permit us to express the
resulting forward-scatter radar cross section, no matter if the object is conducting or absorbing. [5], Section
25.7.
(1)
where
•σF Forward-scattering radar cross section;
•A Geometric cross-section of the object normal to the line of sight;
•f Frequency of the illuminating field;
•c Einstein’s constant - speed of light.
For the air targets considered, this forward-scatter gain is valid within 10 dB over a cone of at least ±20°,
often significantly more.
2.2 Illuminator Selection
For true long-range monitoring of airspace activities some amount of over-the-horizon propagation is
necessary. The number of high-power broadcast transmitters in the shortwave frequency range of 14 MHz
through 30 MHz has unfortunately diminished greatly since the widespread availability of internet
communications. Lower VHF frequencies still have sufficient refraction and diffraction into the shadow
zone to give appreciable secondary scattering from air targets.

Forward-scatter Doppler-only Distributed Passive Covert Radar
STO-MP-SET-231 PAPER NBR - 3
NATO UNCLASSIFIED
NATO UNCLASSIFIED
Figure 1 Behaviour of radio horizon at 100 MHz versus 3000 MHz. VHF signal has appreciable
intensity beyond optical horizon, and the rate of attenuation is much slower than in microwave
range. Graphics from AREPS 3.4.
The radio propagation in the shortwave region is strongly variable based on a number of geophysical
variables, time of day, and time of year. The Saudi Arabia to Finland baseline of Figure 2 can be observed
only for few months per year, and even those months not always 24 hours per day. Ensuring even roughly
consistent coverage would require a number of available illuminating transmitters in wide geographic area,
a situation which currently does not obtain. According to World Radio TV Handbook [7] and some
additional databases, where in 1991 between 15 and 26.1 MHz there were about 3000 transmitters of note,
in 2016 the same bands have about 600 transmitters, less by a factor of five.
Figure 2 Aircraft Doppler Scatter on a HF AM transmission. Transmitter Radio Saudi - ARS (Ara)
at 21,505.314 kHz, receiver near Joensuu, Eastern Finland, roughly 4000 km. Full width of
spectrum c 200 Hz, resolution bandwidth 0.24 Hz. Waterfall time duration c 5 minutes. Picture:
Mauno Ritola
For constant availability, OIRT analog TV transmitters have been selected. Aircraft Doppler histories have
been captured using transmitters at Archangelsk, Cherepovets, Segezha, and St Petersburg. Lower-power
transmitter of Nyandoma and very distant Moscow transmitter have also shown some promise.

Forward-scatter Doppler-only Distributed Passive Covert Radar
PAPER NBR - 4 STO-MP-SET-231
NATO UNCLASSIFIED
NATO UNCLASSIFIED
3.0 DOPPLER TIME SERIES
In the frequency band of interest for this work there are still analogue television transmissions in operation.
Ease of implementation was an important consideration for the proof of concept and therefore narrowband
reception of just 100 to 200 Hz slice centred at the carrier was selected. Moreover, a number of transmitters
operate at the same nominal frequency but with sufficient offsets to distinguish the transmitters and their
associated Doppler scatterer histories from each other. Thus, in a favourable case a number of baselines can
be observed within a single receiver narrow bandwidth and discriminated in the baseband frequency
domain.
At the frequencies considered here, the Doppler shift caused by the objects traversing the atmosphere
remains rather small, much of the time considerably below 50 Hz. Happily, moderately prised hobbyist
communication receivers and software defined radios easily provide necessary selectivity and
discrimination capabilities, as evinced in Figure 2 and Figure 3.
Figure 3 Two examples at 50 MHz band of single baseline Doppler observations. At 0 Hz the
recognized reference carrier, offset from it lower-power or more distant co-channel transmitters.
Electrostatic interference, high-altitude sporadic electric discharge scatterers, and auroral
phenomena can be seen in addition to aircraft scatter histories. Baselines roughly 300 and 270
km. Time ticks at 5-minute intervals. Picture: Juha Hartikka.

Forward-scatter Doppler-only Distributed Passive Covert Radar
STO-MP-SET-231 PAPER NBR - 5
NATO UNCLASSIFIED
NATO UNCLASSIFIED
3.1 Batch Doppler Extraction
Batch approach is useable for some air traffic safety applications and very long range surveillance, where
timeliness of tens of minutes is acceptable. The first approach developed was based on image-processing
and pattern-recognition approaches under the assumption of non-manoeuvring airliners. This assumption
holds reasonably well for airliners in the cruise segment of their flight. This section is adapted from Ptak et
al. [1]
To begin with we have a spectrogram of the kind show in Figure 2 or one of the halves of Figure 3. The
frequency range is constrained to the maximum Doppler shifts expected for targets of interest. The
frequency of the reference carrier is not necessarily stable, so this needs to be corrected. Any estimation
method may be applicable; for the proof of concept where the reference carrier was strong, taking in each
time slice the cell with the maximum amplitude in the vicinity of the nominal transmitter frequency as the
true transmitter frequency proved sufficient.
The spectrogram is subjected to binary thresholding using the Canny edge detector. Hough transform is
then used to segment the binary image into a collection of line segments. Hough transform will reject
possible image responses of the receiver as having angle of inclination that does not correspond to a
physically possible Doppler history. The family of Hough line segments is used to a) find a first estimate
for the zero-crossing time of the Doppler sequence; b) guide the generation of synthetic Doppler curves
later in the processing.
Figure 4. Result of Hough transform. Vertical line elements as well as elements in the first and third
quadrants from the zero-crossing time have been rejected. Grayscale - original spectrogram; cyan lines -
Hough line segments.

Forward-scatter Doppler-only Distributed Passive Covert Radar
PAPER NBR - 6 STO-MP-SET-231
NATO UNCLASSIFIED
NATO UNCLASSIFIED
The work has continued in real-time Doppler extraction with cell-averaging constant false alarm rate
detector and processing principles reminiscent of Track-Before-Detect -approaches. In [4] Ptak presents a
Probability Density Function of First-Order Derivative of Doppler shift and its use in modelling the
dependence between observable Doppler shift and target kinematics.
3.2 Geolocation
A coordinate description of bistatic measurement geometry that is approachable to the lay person is
presented in Figure 5.
Figure 5. Bistatic measurement geometry in operationally approachable form. Taken from [6]
Skolnik's Introduction, 1962, Figure 13.8.
(2)
While it is easy to determine the Doppler shift fdfrom the geometric and kinematic parameters, the inverse
problem is multiply ambiguous. Combining the observations from several baselines will help to resolve
some of these ambiguities, but inverting the function (2) in closed form is not feasible. In problems of this
kind, simplifying assumptions and an iterative approach often works.
As a first approach straight and level constant-velocity target trajectory may be considered. In the situation
where receiver and illuminator density is sufficient, assumption of straight-line target trajectories is not a

Forward-scatter Doppler-only Distributed Passive Covert Radar
STO-MP-SET-231 PAPER NBR - 7
NATO UNCLASSIFIED
NATO UNCLASSIFIED
severe problem; moderately manoeuvring trajectories can be sufficiently approximated by a collection of
straight-line segments. In the application and proof of concept discussed, an amount of time delay can be
accepted.
Under the straight trajectory assumption, the Doppler history will cross from positive, closing shifts to
negative, opening shifts at the point where the air target trajectory crosses the baseline between the
transmitter and the receiver.
Let us consider two baselines dI, I={ J, M}, which may exploit separate or common transmitters as long as
the area of coverage is somewhat overlapping. On each of these baselines a set of anchor points is
introduced. The initial location of the anchor point sets is dictated by the minimum and maximum distances
that an aircraft can traverse during the time span defined by the difference between Doppler zero crossing
times |tJ - tM|. The minimum and maximum distance values correspond to the lower and upper limits that
we set on the cruise speed of the aircraft Vc.As a result, we have two or four solutions, depending on the
time span and distances between the transmitter and receivers. The solution is then reduced to the pair with
the shortest spatial distance to the receivers.
Figure 6. Two baselines dJ and dM, here exploiting a common transmitter. Sets of trial anchor
points denoted by black dots, a late-stage candidate straight-line trajectory segment by red.
Let us denote the two sets of anchor points as Ua{I,n}, n =1 and refer to them as anchor points. The anchor
points are associated with the crossing time so that at the moment of the crossing time tI,x2, the aircraft is
intersecting the anchor points (one from each set). This assumption emerges directly from the equation for
fD{I}=0:
(3)
where
•fD{I} (t) Doppler history for baseline I as a function of time;
•ftFrequency of the illuminating field;
•cEinstein’s constant - speed of light;
•dSA(t) Distance from transmitter Source to Aircraft target;

Forward-scatter Doppler-only Distributed Passive Covert Radar
PAPER NBR - 8 STO-MP-SET-231
NATO UNCLASSIFIED
NATO UNCLASSIFIED
•dAI(t) Distance from Aircraft target to receiver I = {J, M};
•d()/dt Differentiation operator.
The Cartesian product of the two sets of anchor points Ua{J,n} and Ua{M,n}, gives a set of trial trajectories.
These trial trajectories are projected to the time-Doppler plane and compared to the extracted Doppler
histories. From the two closest neighbors to the best fitting trial trajectories a new set of anchor points
Ua{I,n} is then constructed. This process is iterated until the misfit is below a desired threshold of accuracy.
Finally, the generated trajectory segment is extended to the time limits of the observed Doppler history.
Figure 7. A pair of Doppler histories used to generate trajectory segments. Orange curves -
Doppler histories. Blue segments - trial trajectory segment after a few iterations, shown in
Figure 6. The spectrum history from receiver J shows image responses which the algorithm will
reject. The instability of the illuminating transmitter at 20:21:09 is compensated before the
exploitation of the Doppler history.
Resulting reconstructed trajectory is shown in Figure 8. The figure shows the final best-fit trajectory
segment between the two baselines, extension to the full extent of the straight-line Doppler history, and a
good continuation to available FlightRadar24 ADS-B reference data.

Forward-scatter Doppler-only Distributed Passive Covert Radar
STO-MP-SET-231 PAPER NBR - 9
NATO UNCLASSIFIED
NATO UNCLASSIFIED
Figure 8. Extended straight-line trajectory segment. Blue - generated trajectory from the
algorithm. Purple - reference data from FlightRadar24 service. J, M - receivers. S - transmitter. A
- starting point of data. Map background from Google Maps, data 2013 GIS Innovatsija, DATA+.
For real-time use an approach based on Multiple-Hypothesis Tracking is attractive. For maintenance of the
hypothesis tree an application of Viterbi algorithm is proposed in a later work.
4.0 NON-COOPERATIVE TARGET RECOGNITION
Civilian airliners move at regularly repeatable trajectories. Doppler time series are comparable from one
traverse to the next when the geometry remains sufficiently similar. It proved sufficient to discriminate
between the east to west and west to east traverses of Figure 9. Further discrimination between the route
bundles may increase the achieved recognition accuracy.

Forward-scatter Doppler-only Distributed Passive Covert Radar
PAPER NBR - 10 STO-MP-SET-231
NATO UNCLASSIFIED
NATO UNCLASSIFIED
Figure 9. Experimental setup for forward-scatter non-cooperative target recognition
experiments. T - Transmitter, R - Receiver. Solid lines - east to west airliner tracks; dashed lines
- west to east airliner tracks. Distance dTR Transmitter - Receiver ≈ 300 km. Airliners were
observed for approximately 330 km along their trajectory both sides of the line of sight dTR.
A smoothed or synthesised Doppler curve is used as an extraction gate over the Doppler spectrogram
history. For each time sample, the amplitude within a k-sigma -gate of the Doppler estimate is retrieved for
the amplitude time series. Despite the long segment of the airliner trajectory observed in the experiments
the properties of the propagation channel from transmitter to target remain reasonably stationary.
Normalization of the raw extracted amplitude by the transmitter - aircraft distance dTA, aircraft - receiver
distance dAR, and transmitter - receiver distance dTR is sufficient to render the amplitude histories
comparable.
Visual comparison of extracted scattering efficiency series was encouraging. A sufficient number of
scattering efficiency histories per airframe type is used as training data. Using a naïve correlation classifier
gives the results of Table 1. Discrimination between airframe categories even with this restricted data gives
workable results.

Forward-scatter Doppler-only Distributed Passive Covert Radar
STO-MP-SET-231 PAPER NBR - 11
NATO UNCLASSIFIED
NATO UNCLASSIFIED
cross-
range
T
x
R
x
Figure 10. Extraction of amplitude time series along the smoothed Doppler history. Left: Green
line down the centre - extracted reference carrier. Red curve - extracted Doppler history. Dashed
yellow curve - smoothed reference trajectory of associated FlightRadar24 ADS-B data. Yellow
dots - ADS-B outliers from FlightRadar24 service. Right: amplitude of Doppler samples along the
extracted Doppler history.
Figure 11. Comparison of two scattering efficiency profiles of same type of airliner. The
horizontal axis is the cross-range distance.

Forward-scatter Doppler-only Distributed Passive Covert Radar
PAPER NBR - 12 STO-MP-SET-231
NATO UNCLASSIFIED
NATO UNCLASSIFIED
Table 1 Confusion matrix between airliner categories. Ratios of number of occurrences interpreted as
frequentist probability.
Airliner categories
G1:
B788, A333
, A343
G2: B772, B77W, A346
G3: B744
G1
74
/101
=
0.73
34
/171
=
0.19
4
/59
=
0.06
G2
34
/171
=
0.19
162
/205
=
0.79
12
/75
=
0.16
G3
4
/59
=
0.06
12
/75
=
0.16
5
/
7
=
0.71
In future work the idea of comparison between fluctuation models and Doppler series has been proposed;
an ogive with a given maximum diameter and apex angle may match the body dimeter and nose cone angle
of the airliner. Thus it may be possible to reduce the recognition data to a small number of invariant
geometric parameters.
5.0 EARLY RESULTS
5.1 Achieved coverage
Comparison between Doppler curves synthetized from ADS-B data and observed Doppler time series has
been used to estimate the achieved coverage. Currently constantly reliable coverage can be achieved
approximately within the intersection of the radio horizons of the transmitter and receiver antennas. In
addition to the examples in Figure 12 the experimenters have pretty regular observations from up to 440
kilometres.
Figure 12. Examples of achieved coverage. Observed Doppler histories confirmed with FlightRadar24 data. Insert -
distribution of airframe types observed. Dashed circle - Radio horizon, circa 368 km. Left - St Petersburg 150 kW
transmitter, Joensuu West Dipole Array. Right - Nyandoma 46 kW transmitter, Joensuu East Uda-Yagi Array.

Forward-scatter Doppler-only Distributed Passive Covert Radar
STO-MP-SET-231 PAPER NBR - 13
NATO UNCLASSIFIED
NATO UNCLASSIFIED
5.2 Achieved accuracy
As an easy source of reference data, FlightRadar24 service of ADS-B data was used. The FR24 data
appears to have some problems, most likely in the handling of time stamps, the reported position of the
aircraft sometimes jumping backwards. These jumps show very well in the Doppler domain as seen in the
left panel of Figure 13. These problems are reflected in the uncertainty of the figures of merit of achieved
accuracy.
6.0 PROPOSED SYSTEM ARCHITECTURE
Encouraged by the tracking and classification results, an idea for easily deployable, low-cost network of
observation stations for supplementary air traffic monitoring is being developed. In this development work
dual-use considerations are always in the forefront. As a testing environment and as independent air safety
monitoring system, interested hobbyists will be recruited. The architecture and software developed in this
environment can then be utilised for purposes of homeland security and other safety organizations.
6.1 Station equipment
Antennas used at the proof-of-concept receiving sites have been of different kinds. Both Uda-Yagi and
uniform dipole arrays have showed good utility. In addition, on one site traveling-wave antenna of 250
meters (“long wire”) has shown that a simple structure that is easily deployed is very useable.
On receiving sites a variety of commercial hobbyist receivers have been used, as well as low-cost software-
defined radio modules. In situations where performance needs to be really optimised, a thorough evaluation
Figure 13. Comparison of estimated trajectory and data from FlightRadar24. Left - Doppler
domain - Right - Cartesian domain. Blue - Constructed from Doppler observations. Purple -
FlightRadar24 data. On the right picture only selected time-stamped points of the Doppler-
originating trajectory are shown connected to the FR24 data stamped for the same instant. The
difference is less than 1600 meters.

Forward-scatter Doppler-only Distributed Passive Covert Radar
PAPER NBR - 14 STO-MP-SET-231
NATO UNCLASSIFIED
NATO UNCLASSIFIED
of possible RF solutions is of course necessary. Even the communication receivers and Ettus units used so
far with simple antennas have shown themselves capable of discriminating also targets in 30-meter-class at
100-kilometer ranges, not to mention airliners at hundreds of kilometres. Even “thumb drive” receivers
have been used with promising results.
6.2 Clients and servers
We assume that there will at all times be a server or a collection of servers available. The servers could be
located by a service discovery mechanism. At the first stage, known URLs or IP addresses could be used.
As a station starts up, it will attempt to locate a suitable server. At the first stage, this will be by a cached
list of server locators or addresses. At installation there will be a default set of servers at the installed list;
periodically through operation the station will receive updates to the server list and store the updated list in
its local persistent storage.
In case of a widespread network, the concept of “suitable server” should include geographic proximity.
This geographic consideration will serve to promote clustering of stations that provide mutually
complementary data and helps minimize need for long-distance data transfer.
A receiver station will calculate a spectrum over the few seconds’ update and integration time. On this
spectrum a CFAR or other automatic detection procedure will be run. The resulting tuples of [timestamp,
stationID, N×(frequency, intensity)] will be transferred to the contacted server.
The server collates received Doppler-domain slices from its cluster of observing stations, and performs
target tracking on the detections that can be associated to existing tracks or between stations. Tracks leaving
the region of responsibility of a given server are handed over to another server where the connected stations
have better coverage of the projected trajectory. Similarly, detections that were not associated locally, will
be shared between servers for possible global track initiations.
To increase motivation of voluntary participants or to provide observer stations with up-to-date situation
picture, the server will disseminate the generated air tracks for display at the stations.
ACKNOWLEDGEMENTS
This paper is largely based on previously published works by Piotr Ptak et al. Figures and pictures not
otherwise credited are either from previous publications by Piotr Ptak et al. or made specifically for this
publication.
Part of this work continues with the support of Finnish Ministry of Defence via the Defence Board for
Science MATINE grant program.
[1] Piotr Ptak, Juha Hartikka, Mauno Ritola, Tuomo Kauranne. Long-Distance Multistatic Aircraft
Tracking With VHF Frequency Doppler Effect. IEEE Transactions on Aerospace and Electronic
Systems Vol. 50, No. 2 July 2014. DOI. No. 10.1109/TAES.2014.130246.
[2] Piotr Ptak, Juha Hartikka, Mauno Ritola, Tuomo Kauranne. Aircraft Classification Based on Radar
Cross Section of Long-Range Trajectories. IEEE Transactions on Aerospace and Electronic Systems
Vol. 51 No. 4 October 2015. DOI. No. 10.1109/TAES.2015.150139.
[3] Piotr Ptak, Juha Hartikka, Mauno Ritola, Tuomo Kauranne. Instantaneous Doppler Signature

Forward-scatter Doppler-only Distributed Passive Covert Radar
STO-MP-SET-231 PAPER NBR - 15
NATO UNCLASSIFIED
NATO UNCLASSIFIED
Extraction from within a Spectrogram Image of a VHF Band. IEEE Transactions on Aerospace and
Electronic Systems Vol. 52, No. 2 April 2016. DOI. No. 10.1109/TAES.2015.150077.
[4] Piotr Ptak. Aircraft Tracking and Classification with VHF Passive Bistatic Radar. Ph. D. Dissertation,
Lappeenranta University of Technology 2015. ISBN 978-952-265-815-9 (printed book), ISBN 978-
952-265-816-6 (PDF).
[5] Merril I. Skolnik (ed). Radar Handbook, 2nd edition. McGraw-Hill, 1990. LCC 89-35217, ISBN 0-07-
057913-X
[6] Merril I. Skolnik. Introduction to Radar Systems. McGraw-Hill, 1962. LCC 61-17675. Identifier-ark
ark:/13960/t42r4sf3d ; Identifier-access http://www.archive.org/details/IntroductionToRadarSystems
Visited on 2016-08-15.
[7] World Radio TV Handbook. WRTH Publications. Various editions. Edition 2016 ISBN: 978-0-
9555481-8-5