Received: 31 October 2022

DOI: 10.1049/rsn2.12399

ORIGINAL RESEARCH
- -Revised: 10 March 2023 Accepted: 15 March 2023

- IET Radar, Sonar & Navigation

Whole system radar modelling: Simulation and validation

Jithin Kannanthara1 | Darren Griffiths1 | Mohammed Jahangir2 |

Jonathan M. Jones1 | Chris J. Baker2 | Michail Antoniou2 | Colin J. Bell3 |
Henry White3 | Kai Bongs1 | Yeshpal Singh1

1
School of Physics and Astronomy, University of Abstract
Birmingham, Birmingham, UK
The ever‐expanding horizon of radar applications demands solutions with high‐end radar
2
School of Electronic, Electrical and Systems functionalities and technologies and is often limited by the available radar equipment, cost
Engineering, University of Birmingham,
Birmingham, UK and time. A practical method to tackle the situation is to rely on the modelling and
3 simulation of radar systems based on the user requirements. The comprehensive system‐
BAE Systems, Farnborough, UK
level modelling of a pulsed Doppler radar in MATLAB/Simulink consisting of all the
Correspondence fundamental blocks in the transmit chain, the environment, the receive chain, and
Jithin Kannanthara. the data processing chain is presented in this article. The first half of the article discusses
Email: [email protected] the high‐fidelity simulation of each building block in the radar model. In the second half
of the article, the range‐Doppler plot generated from the high‐fidelity radar model is
Funding information compared and validated using the range‐Doppler plot from a real radar trial. The radar
Horizon 2020 Framework Programme, Grant/
phase noise plays a crucial role in the detection of slowly moving, low radar cross‐section
Award Number: 820404; UK National Quantum
Technology Hub in Sensing and Timing, Grant/ targets in the presence of strong clutter. The article also briefly discusses the effects of
Award Number: EP/T001046/1; BAE systems radar oscillator phase noise in the range‐Doppler plot. The validated, fully flexible radar
model has the advantage of supporting the addition of further building blocks and
optimising the parameters based on user requirements.

KEYWORDS
Doppler radar, modelling, oscillators, phase noise, radar receivers, radar signal processing, radar transmitters,
thermal noise

1 | INTRODUCTION with real data was discussed in refs. [14, 15]. A relatively simple
method for precise simulation of the geometry of objects in
Radars are used for a wide variety of applications in the fields the environment in synthetic aperture radar was proposed in
of meteorology [1, 2], healthcare [3–5], mapping and astron- ref. [16]. The environment modelling in radar literature was
omy [6, 7], law enforcement [8], and defence [9–13]. It is mainly focussed on the radar performance in the coherent and
impossible to fully test a complex radar system across all non‐coherent radar sea clutter [17–20]. A hardware‐in‐the‐
hardware and environmental combinations in the real world. loop simulator for radar targets based on scattering model
One possible solution to test the radars would be to model the theory was discussed in ref. [21].
radar hardware and environment based on the user re- Even though there are different cases of radar modelling in
quirements and radar applications. At best, a subset of the the literature, each radar model focuses on a specific aspect of
conditions can be tested, and the results can be used to validate radar. The specific aspect could be modelling the radar envi-
more extensive radar modelling. ronment or modelling a limited number of building blocks in a
Over the years, there have been attempts to model the radar. In recent years, there has been very little literature on
different aspects of the radar. The development of phase noise radar hardware and radar system modelling. An incoherent
theory and a new design equation for the phase noise level in scatter radar simulation system using a modular design concept
frequency‐modulated continuous wave radar and its validation was discussed in ref. [22], and an introductory simulation of a

-
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is
properly cited.
© 2023 The Authors. IET Radar, Sonar & Navigation published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.

1050 IET Radar Sonar Navig. 2023;17:1050–1060. wileyonlinelibrary.com/journal/rsn2

 17518792, 2023, 6, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.12399 by Libya Hinari NPL, Wiley Online Library on [07/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
KANNANTHARA ET AL.
- 1051

pulsed Doppler radar system using MATLAB and Simulink the signal and the SNR at the output of each building block. In
was discussed in ref. [23]. A signal level simulator consisting of the results and validation section, we describe the generation of
algorithms for the simulation of raw radar return signals was range‐Doppler data from the simulated radar model. We then
discussed in ref. [24]. The literature lacks a high‐fidelity radar validate the simulated model and the range‐Doppler plot with
model with all the building blocks that could be used to the range‐Doppler data processed from real radar trials. Finally,
examine and predict each radar building block's performance in conclusion, we summarise the research presented in the
and fundamental limitations. paper.
The phase noise in radars is due to the random fluctuation
of the phase of the radar oscillator [25]. In range‐Doppler
plots, the phase noise floor emerges out of the thermal noise 2 | THEORY AND SIMULATION
floor for range bins with strong clutter power. The increase in
the noise floor due to the phase noise degrades the available Any generic radar design consists of a transmit chain to
signal‐to‐noise ratio (SNR) for target detection. Target detec- generate and transmit the signal, a receive chain to receive the
tion becomes more difficult for slowly moving low radar cross‐ reflected signal from the environment and a data processing
section (RCS) targets like drones. chain to process the received signal. The frequency of opera-
The novel radar model developed and discussed in this tion of the radar, the constituent blocks within the radar and
paper is a comprehensive model that consists of all the the parameters and power budgets depend on the radar
fundamental building blocks in the transmit chain, the receive application. The radar model discussed in the paper is based on
chain, the data processing chain and the environment of a Gamekeeper 16 U [26] staring radar with further simplifica-
pulsed Doppler radar. The radar model focuses on the fidelity tions. The model built in the MATLAB/Simulink framework
of every radar hardware block in the simulation. The high‐ can modify and optimise all the parameters at the component
fidelity radar model is also validated by comparing the range‐ level and include further building blocks. The schematic of the
Doppler plot generated from the model with the range‐ front end of the monostatic radar model is given in Figure 1.
Doppler plot from an actual radar trial. The effect of oscil- The radar model can be subdivided into four major sec-
lator phase noise in the range‐Doppler plot and in target tions: transmit chain, receive chain, the data processing chain,
detection is also briefly explored. The radar model discussed in and the environment. The building blocks within the transmit
this paper can integrate further building blocks and optimise chain, environment, and receive chain were represented as
every parameter in each building block to potentially replicate separate system‐level blocks in Simulink, with each Simulink
any radar design. block containing further sub‐blocks. All the building blocks
This paper is organised as follows. In the theory and were modelled in such a way that the output of one block was
simulation section, we discuss the theory behind each radar connected to the input of another block with test probes
building block and the realisation of each building block in the placed in between. The test probes were used to analyse the
simulation. We have also provided the mathematical model of input and output of each building block in the model. The data

F I G U R E 1 Schematic of the front end model of the monostatic radar in simulation. The front end consists of four main sections. The transmit chain
transmits the pulsed sinusoidal signal to the environment through the transmit antenna. The clutter and targets in the environment reflect the signal to the radar
receive antenna. In the receive chain, the received signal undergoes stages of amplification and down‐conversion to generate an intermediate frequency signal.
The intermediate frequency signal gets converted to a baseband signal in the data processing chain. The baseband signal undergoes further data processing stages
to enhance the target's signal‐to‐noise ratio. The building blocks in the transmit chain, the environment chain, and the receive chain are modelled as separate
blocks in Simulink. The building blocks in the data processing chain were generated using a separate code written in MATLAB.
 17518792, 2023, 6, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.12399 by Libya Hinari NPL, Wiley Online Library on [07/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1052
- KANNANTHARA ET AL.

at the output of the last block of the receive chain was taken to where SV(t) represents the pulsed signal at the output of the
the MATLAB workspace. A separate code was written in vector modulator and AV represents the amplitude of the
MATLAB for the data processing chain and the generation of signal at the output of the vector modulator.
the range‐Doppler plot. The amplifier was used to amplify the vector modulator
output signal. The amplifier block has the potential to amplify
the signal to any value of transmitted power, the power at the
2.1 | Transmit chain output of the transmit amplifier. The mathematical model of
the signal at the output of the amplifier is given by,
The first block in the transmit chain of the model is the
transmitter phase lock loop (transmit PLL) and is one of the SA ðtÞ ¼ AA cos½2πf T t þ θ�; ð3Þ
most crucial blocks in the radar simulation. Generally in radar
hardware, the radar oscillator signal at the reference frequency where SA(t) represents the signal at the output of the amplifier
is up‐converted to a radio frequency signal at the required and AA represents the amplitude of the signal at the output of
transmit frequency using a frequency up‐converter. The fre- the amplifier.
quency up‐converter could be a PLL or a frequency mixer. In In real radar, the transmit antenna is used to transmit the
the model, the transmit PLL was used to directly generate a real electromagnetic signal at the output of the amplifier to the
continuous sinusoidal wave at a specific L band transmit fre- environment. There are different transmit antenna architec-
quency. The transmit PLL in the model acts as a direct digital tures, and the transmit antenna generated in the model was a
synthesiser to generate the required transmit frequency. The simplified antenna with a single transmitter and a constant an-
model currently has a simplified version of the signal generator, tenna gain. In the model, the transmit antenna output was
and further blocks can be integrated to represent a separate connected to the input of the environment model. The math-
radar oscillator and PLL up‐converter. The transmit PLL has ematical model of the transmitted signal is given by [27, 28],
the further potential to integrate blocks to model square pulses
or frequency‐modulated waveforms at any transmit frequency ST ðtÞ ¼ AT cos½2πf T t þ θ�; ð4Þ
and amplitude, depending on the requirements of the radar
model. where ST(t) represents the transmitted signal, AT, fT, and θ
The phase noise is inherent to any radar oscillator. In the represents the amplitude, frequency, and phase of the trans-
model, the equivalent phase noise at the transmit signal can be mitted signal respectively. Even though the transmit signal is a
included. The phase noise of any oscillator at any frequency pulsed signal, for simplicity, the time dependency of the pulse
can also be simulated in the transmit PLL block by providing is not explicitly mentioned in Equation (4). Hence Equation (4)
the respective oscillator's frequency offset and the phase noise is a simplified model of the real‐valued transmitted signal. The
level. In the phase noise modelling discussed later in the paper Equations (1)–(4) are also connected to one another as SV(t) is
in Section 3.3, the equivalent oscillator phase noise at the a function of SPLL(t), SA(t) is a function of SV(t), and ST(t) is a
transmit frequency is added to the simulation. function of SA(t).
The mathematical model of the sinusoidal signal at the
output of the transmit PLL is given by,
2.2 | Environment
SPLL ðtÞ ¼ APLL cos½2πf T t þ θ�; ð1Þ
For a real radar, the environment consists of clutter that are
where SPLL(t) represents the sinusoidal signal at the output of stationary objects, including buildings, trees, and other
the transmit PLL, APLL represents the amplitude of the signal geographical features in the field of view. The environment
at the output of the PLL, θ represent the phase of the signal, also consists of moving targets, including drones, aircraft,
and fT represents the transmit frequency. ships, and birds. The transmitted signal gets reflected by the
The model's continuous sinusoidal wave generated at the objects in the environment and reaches the receive antenna of
transmit PLL was converted to a pulsed signal using a vector the radar. In radar theory, the radar range equation is used to
modulator. The vector modulator was a mixer that added the determine the power of the received signal and is given by,
amplitude envelope to the output of the transmit PLL. The
envelope generator provided the specific amplitude envelope
to the vector modulator. The envelope generator can generate P T GT GR λ2 σ
PR ¼ ; ð5Þ
a rectangular envelope or any shaped amplitude envelope 4π3 R4
depending on the requirements of the radar model. In the
model, the generated envelope was a shaped pulse with a pulse where PR is the received signal power, PT is the transmitted
width of about 1 μs and a pulse repetition interval (PRI) of signal power, GT and GR are the transmit antenna gain and the
about 136 μs. The mathematical model of the signal at the receive antenna gain respectively, λ is the wavelength of the
output of the vector modulator is given by, transmitted signal, σ and R are the RCS and the target range
respectively. The received signal power is the power of the
SV ðtÞ ¼ AV cos½2πf T t þ θ�; ð2Þ received signal at the output of the receive antenna.
 17518792, 2023, 6, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.12399 by Libya Hinari NPL, Wiley Online Library on [07/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
KANNANTHARA ET AL.
- 1053

In the model, a simple representation of the magnitude across all frequencies [29]. The model used the thermal noise
response of both the target and the clutter was realised. The block to generate and add complex thermal noise to the real
signal from the transmit chain was directed to object blocks valued received signal. The thermal noise consisted of both
representing clutter and targets. Algorithms following the radar positive and negative frequencies. The mathematical equation
range equation were performed to specify each object's range, for the thermal noise is given by,
the RCS, and the Doppler velocity. The objects were placed at
the radar bore‐sight and can potentially include additional al- N ¼ kB T B; ð7Þ
gorithms to place objects at a specific angle in the azimuth and
elevation. The reflected signal from the object blocks was where N, kB, T, and B represent the noise power, Boltzmann's
directed to the receive chain. constant, noise temperature, and noise bandwidth respectively.
The thermal noise block can model any value of temper-
ature. In the model, the temperature was kept at room tem-
2.3 | Receive chain perature of 290 K. The noise bandwidth in the thermal noise
block was equivalent to the sampling frequency used in the
In a real radar, the reflected signal from the objects in the radar simulation. The sampling frequency was kept at 5 GHz
environment is detected at the receive antenna. The receive and was at least two times higher than any of the frequencies
antenna can be of different configurations, either a single an- generated in the radar model.
tenna or a combination of many phased array antenna ele- The SNR of the received signal (SNRR) for a received
ments. In the model, the receive antenna was a single antenna signal power of PR from Equation (5) and a noise power of N
with a constant antenna gain which is used as a simplified from Equation (7) is,
representation even in the case where the model is compared
with a multiple receive array system. P T GT GR λ2 σ
Consider a radar that transmits a series of M pulses with SN RR ¼ : ð8Þ
4π3 R4 kT B
0 ≤ m ≤ M − 1, separated by a PRI of T. For a target with an
initial range of Ro and a velocity of ν, the range of the target The low noise amplifier (LNA) is a special kind of amplifier that
for the mth transmitted pulse is Ro − νmT. For the mth amplifies the weak received signal without drastically degrading
transmitted pulse, the signal that reaches the receiver antenna the SNR. The LNA noise figure (NF) is the parameter that
will have a time delay equivalent to 2c ðRo − νmT Þ. The quantifies the degradation of SNR by the LNA. The LNA was
mathematical model for the delayed received signal is given by, modelled as an amplifier with a specific NF. The LNA amplifies
both the signal and the noise. The noise power gets further
� � amplified by the NF of the LNA. The LNA model can specify
2
SR ðtÞ ¼ ST t − ðRo − νmT Þ ; any user‐required value for the amplification and the NF. The
c
SNR at the output of the LNA (SNRLNA) is given by,
� � � � ��
2ν 4πRo
¼ AR cos 2πf T t þ 2π mT þ θ − ; SNRLNA ¼ SNRR − NF: ð9Þ
λ λ
ð6Þ
The mathematical model for the received signal at the output
where 2νλ ¼ f d is the Doppler frequency of the target moving of LNA (SLNA) is given by,
towards the radar and θ − 4πR o 0
λ ¼ θ is the phase of the received
signal. AR represents the amplitude of the received signal and SLNA ðtÞ ¼ ALNA cos½2πf T t þ 2πf d mT þ θ0 � ð10Þ
follows the radar range equation.
Since the received signal power of the radar target is weak where ALNA represents the amplitude of the received signal
compared to both clutter and noise, the challenge is to bring after the LNA.
the signal power considerably above the noise floor. The SNR In a real radar, the output of the LNA is mixed down to a
describes the signal power over the noise floor. The SNR is convenient intermediate frequency (fIF) using a mixer. The
generally represented in dB. The noise provides the statistical mixer has two inputs, an RF and a local oscillator (LO) input.
basis for the target detection; hence, controlling noise within The RF input of the mixer takes the output of the LNA as
the radar receive chain is critical. The sources of noise in radars shown in Equation (10). The LO input of the mixer is generally
are both internal and external. One of the internal noise a local oscillator output at a frequency equivalent to fT + fIF.
sources in radars is thermal noise. Phase noise can also be a The mathematical model for the sinusoidal signal at the output
significant source of noise in some radar systems. of local oscillator is given by,
In the simulated radar model, the phase noise was inte-
grated into the transmit PLL block of the transmit chain. The SLO ðtÞ ¼ cos½2πðf T þ f IF Þt�: ð11Þ
thermal noise is due to the thermal agitation of the electrons in
the receive chain elements. The thermal noise is random, and The output of the mixer consists of two frequencies, the first
the power spectrum of the thermal noise is uniform and flat corresponding to the sum of the two input frequencies and the
 17518792, 2023, 6, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.12399 by Libya Hinari NPL, Wiley Online Library on [07/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1054
- KANNANTHARA ET AL.

second corresponding to the difference between the two input In the simulation, the fIF was sampled at a specific ADC
frequencies. The difference between the two frequencies is clock frequency (lower than the fIF) to generate aliased signals
equivalent to the fIF. The ratio between the input RF power at second intermediate frequency (fIF2). The ADC for aliasing
and the output IF power is known as the mixer conversion was realised using a signal‐to‐workspace block sampled at the
loss. For an ideal mixer with a LO input amplitude of unity and clock frequency. The output of the ADC was a sinusoidal
the RF input amplitude of A, the conversion loss is 6 dB signal at fIF2. The ADC has the potential to add further blocks
[30, 31]. The conversion loss affects both the signal and the to generate quantisation noise. The ADC in the simulation
noise. The effective SNR gain due to the conversion loss in a does not alter the SNR, and hence the SNR at the output of
mixer is zero. the ADC becomes:
The NF of the mixer (NFMXR) is another parameter that
affects the SNR. There is also an undesired frequency band SNRADC ¼ SNRBPF : ð15Þ
known as the image frequency for a mixer that down‐converts
the RF to IF using the LO, along with the desired frequency The mathematical model of the signal at the output of the
band. The noise will be present in both the desired and the ADC is given by,
undesired band for a real‐valued signal accompanied by a
complex‐valued thermal noise. The down‐conversion folds SADC ½t� ¼ AADC cos½2πf IF2 t þ 2πf d mT þ θ0 �; ð16Þ
back the noise in the image band on top of the noise in the
desirable band. The SNR at the output of the mixer SNRMXR is
where AADC is the amplitude of the signal after the ADC and
given by,
is equivalent to the signal amplitude before ADC. The signal‐
to‐workspace block carried the Simulink data to the MATLAB
SNRMXR ¼ SNRLNA − NFMXR : ð12Þ
workspace. The further processing of the data was performed
in MATLAB.
A narrow bandwidth band‐pass filter (BPF) is placed after the
mixer to transmit the difference frequency centred at fIF. The
signal power at the output of the BPF is equivalent to the 2.4 | Data processing chain
signal power at the input of the BPF. The noise power at the
output of the BPF is reduced by a factor of B/BBPF, where B is The first and foremost step in data processing was the down‐
the noise bandwidth before the BPF and BBPF is the bandwidth conversion to baseband and matched filtering. There are
of the BPF. The SNR at the output of the BPF (SNRBPF) is several methods for down‐conversion. In the model, we use
given by, the Hilbert transform. Theoretically, the Hilbert transform
� � generates the complex baseband signal containing the in‐phase
B
SNRBPF ¼ SNRMXR þ 10log10 : ð13Þ and the quadrature components. The simplified mathematical
BBPF model of the baseband complex signal at the output of the
Hilbert transform is given by,
The mathematical model for the signal at the output of the
band‐pass filter (SBPF) is given by, SHT ½t� ¼ AHT exp½2πf d mT þ θ0 �; ð17Þ

ALNA where AHT is the signal's amplitude after the Hilbert trans-
SBPF ðtÞ ¼ cos½2πf IF t þ 2πf d mT þ θ0 �; ð14Þ
2 form. The complex baseband radar signal is used to measure
the phase and amplitude independently [28].
where ALNA
2 ¼ ABPF is the amplitude of the signal after the BPF. The matched filter is a linear filter based on the principle of
In the simulation, the mixer block performs the mixing correlation in the time domain between the noisy received
operation between the noisy received signal at the output of signal and the reference signal to improve the SNR. For an
LNA and the receive PLL output signal. The receive PLL ideal matched filter, maximum SNR is obtained when the
generated a continuous sinusoidal wave at a specific L band reference signal is a time‐delayed mirror image of the received
frequency. The receive PLL can be modified to generate signals signal [32]. For a matched filter that provides an SNR gain of
at any required frequency. The output of the mixer consisted of GMF, the SNR at the output of the matched filter (SNRMF) is
both the sum and the difference frequency. The BPF was given by,
placed after the mixer to transmit the difference frequency
centred at the fIF. SNRMF ¼ SNRADC þ GMF : ð18Þ
An analogue‐to‐digital converter (ADC) is generally used
to convert the analogue signal to a digital signal. The ADC can The mathematical model of the signal at the output of the
also perform signal down‐conversion through frequency ali- matched filter is given by,
asing. Since the model was already in the digital domain, the
ADC block in the model was used to generate a down‐
converted frequency signal. SMF ½t� ¼ AMF exp½2πf d mT þ θ0 �; ð19Þ
 17518792, 2023, 6, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.12399 by Libya Hinari NPL, Wiley Online Library on [07/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
KANNANTHARA ET AL.
- 1055

where AMF is the amplitude of the signal at the output of the ranges and RCS. The targets and clutter are modelled as point
matched filter. scatters in the simulation. A range‐Doppler plot for a full
In the simulation, the real‐valued signal at fIF2 was con- coherent processing interval (CPI) was simulated to validate
verted to a complex signal at the baseband. The matched the model. The CPI consisted of 2048 pulses, each with a PRI
filtering was performed to improve the SNR. The matched of about 136 μs. The range‐Doppler plot was generated by
filter coefficient was derived from the simulated transmit performing FFT along the slow time axis. The slow time axis
signal. consists of all the PRIs within a CPI. The simulated range‐
The Fast Fourier transform (FFT) is used to convert the Doppler plot was compared with a real range‐Doppler plot
signal in the time domain to the frequency domain, thereby from a radar trial. The real range‐Doppler plot consists of a
providing the signal's frequency information. For a measured number of range bins and 2048 Doppler bins.
signal, the discontinuities due to the non‐integer number of A three‐level validation was performed. In the first level of
periods result in frequency components leaking into adjacent validation, the system‐level signal powers and the SNR values
frequency bins. The leakage causes the fine spectral lines in at the output of the different blocks of the simulation were
FFT to spread and is known as spectral leakage [33]. The compared with the expected values. In the second level of
spectral leakage is minimised by a technique known as win- validation, the basic parameters from the simulated range‐
dowing. The windowing multiplies the measured time domain Doppler plot were compared with the real radar values. In
discontinuous signal with an amplitude envelope that ap- the third level of validation, the CNR, the SNR, the thermal
proaches zeros at both ends [34]. The SNR at the output of the noise floor, and the phase noise floor from the range‐Doppler
windowing is degraded by a factor known as the loss factor plot for both the simulation and real data were compared.
(LF). The loss factor is defined as the ratio of the maximum
achievable SNR (SNR without windowing) to the actual SNR
(SNR with windowing) [35]. The SNR at the output of win- 3.1 | Comparison of system level signal
dowing (SNRW) is given by, power and SNR

SNRW ¼ SNRMF − LF: ð20Þ In the first level of validation, the signal power and the SNR at
different stages within the transmit chain, the receive chain,
In real radar, the FFT extracts the Doppler frequency/Doppler and the data processing chain were considered. In the first level
velocity information of both the targets and clutter. The data at of validation, a single stationary clutter was kept at a range of
the input of the FFT also undergo coherent pulse integration. 600 m and an arbitrary value of RCS. The simulation for a full
For a coherent integration of N pulses, the power in the in- CPI was performed. The test probes at the output of each
tegrated signal component is increased by N2 and the inte- building block of the radar model were used to evaluate the
grated noise power is increased by N. Hence, the effective signal power, the noise power, and the SNR values. The first
increase in the SNR is N in dB. The SNR at the output of the stage of validation is conveniently divided into two parts. The
FFT is given by, first part compares the signal powers, and the second part
compares the SNR values.
SNRFFT ¼ SNRW þ N: ð21Þ In the first part, the simulated signal power at the output of
the transmit amplifier, the transmit antenna, and the receive
Blackman‐Harris windowing and a 2048‐point FFT were antenna is compared with the expected values. The expected
applied to the matched filter output in the simulation. The value of the signal power at the output of the transmit amplifier
processed signal in the frequency space is generally represented and the transmit antenna was taken from the Gamekeeper
as range‐Doppler plots, with the vertical axis representing the 16 U power budget. The expected value of the signal power at
range of the radar and the horizontal axis representing the the output of the receive antenna was calculated from the radar
target's frequency (radial velocity). The data after the FFT was range equation (Equation 5). Table 1 compares the simulated
represented in the simulation as a range‐Doppler plot. and the expected value of signal powers. We can clearly see that
the simulated and the expected signal power are in very close
agreement. The thermal noise was added to the received signal
3 | VALIDATION OF RADAR MODEL at the output of the receive antenna. The thermal noise power

Validating the radar model is as crucial as developing the T A B L E 1 Simulated and expected values of signal powers at the
comprehensive radar model. The best method to compare and output of different radar building blocks.
validate a simulated radar model with a real radar is by
Simulated Expected Difference
comparing the range‐Doppler plots emphasising the crucial
Radar block (dB) (dB) (dB)
figures of merit. For a range‐Doppler plot, either real or
simulation, the significant figures of merit are the SNR and the Transmit amplifier 33.0 33.0 0.0
clutter to noise ratio (CNR). Transmit antenna 45.4 45.5 0.1
The comprehensive radar model presented in the paper has
Receive antenna −74.5 −74.4 0.1
the advantage of simulating clutter and targets at user‐specified
 17518792, 2023, 6, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.12399 by Libya Hinari NPL, Wiley Online Library on [07/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1056
- KANNANTHARA ET AL.

generated by the thermal noise block was expected and ob- were also measured from the simulated plot. The second level
tained at −107 dB. of validation was performed by comparing the basic parame-
In the second part, the simulated SNR at the output of ters, including the range bin, Doppler bin, range resolution,
each building block in the receive chain and the data processing and Doppler resolution for both the simulated radar and the
chain was compared with the expected values. The signal po- real radar. The range resolution for the simulated and real radar
wer and the thermal noise power at the output of each building was in very good agreement. The comparison of the remaining
block were used to calculate the simulated SNR values. The parameters is summarised in the Table 3.
expected SNR values were calculated using a combination of The expected range resolution and Doppler resolution
radar theory and power budget reference values. The SNR were taken from a prototype staring radar. The expected range
values at the outputs of different blocks are summarised and bin and Doppler frequency bin for both the clutter and the
compared in the Table 2. Table 2 clearly shows a good parity target were calculated from the range and Doppler velocity
between the simulated and the expected SNR values at the provided in the simulation, using the range resolution and
output of every building block. The Tables 1 and 2 show the Doppler resolution reference values. The simulated values were
fidelity of every building block in the simulated whole system the values measured from the simulated range‐Doppler plot.
radar model. Table 3 clearly shows that the basic parameters from the
simulation are very close to the expected parameters from the
real radar. The radar model discussed in the paper can be tuned
3.2 | Comparison of range and Doppler to simulate all the basic parameters, including the range reso-
basic parameters lution, Doppler resolution, CPI, and PRF, to any required
configuration.
In the second level of validation, the basic parameters from the
simulated range‐Doppler plots were compared with the pa- (a) Simulated range‐Doppler plot without the oscillator phase
rameters from the real radar. At first, a range‐Doppler plot noise. The plot consists of the stationary clutter and a
with thermal noise was simulated with a single stationary single simulated target in the uniform thermal noise floor.
clutter and a single moving target, placed at a distance of 600 (b) Simulated range‐Doppler plot with the oscillator phase
and 1500 m respectively. The range and RCS values were noise. The plot consists of the stationary clutter and a
arbitrary and did not follow a specific reference. The target's single simulated target in the overall noise floor. The
velocity was kept at 251 Hz, equivalent to 30 m/s for the overall noise floor consists of phase noise and thermal
specific L band transmit frequency.
The simulated range‐Doppler plot consisted of a number
of range bins. The PRF of the simulated range‐Doppler plot
was equivalent to the real radar range‐Doppler plot. Figure 2
shows the simulated range‐Doppler plot with the range bin on
the vertical axis and the Doppler frequency on the horizontal
axis. Figure 2 clearly shows a stationary clutter (highlighted in
green) with a zero Doppler and a target (highlighted in red)
with a Doppler frequency of around 251 Hz.
The range bin and the Doppler bin for both the target and
the clutter were measured from the simulated range‐Doppler
plot in Figure 2. The radar range and Doppler resolution
F I G U R E 2 Simulated range‐Doppler plot with range bin in the
T A B L E 2 Simulated and expected values of SNR at different stages vertical axis and Doppler frequency in the horizontal axis. The spectrum is
in receive chain and data processing chain building blocks. shown for −700 to +700 Hz. The single stationary clutter and the single
Doppler target are highlighted within a green and red box respectively. The
Simulated Expected Difference range‐Doppler plot is normalised to the clutter.
Radar block (dB) (dB) (dB)
Before LNA 32.5 32.6 0.1
TABLE 3 Simulated and expected values of range and Doppler basic
After LNA 28.0 28.1 0.1 parameters.

After mixer 25.0 25.1 0.1 Parameter Simulated Expected

After BPF 45.0 45.1 0.1 Clutter range bin 5 5
After ADC 45.0 44.8 −0.2 Target range bin 17 17
After matched filter 59 57.8 −1.2 Clutter Doppler bin 0 0
After windowing 55.5 54.7 −0.8 Target Doppler bin 70 70
After FFT 88.6 87.8 −0.8 Doppler resolution (Hz) 3.58 3.59
 17518792, 2023, 6, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.12399 by Libya Hinari NPL, Wiley Online Library on [07/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
KANNANTHARA ET AL.
- 1057

noise. The phase noise floor emerging out of the thermal The second step added the target information to the
noise floor for range bins with higher clutter is clearly simulated range‐Doppler plot. In the simulation, the target
visible. drone was realised as an object at a distance equivalent to the
(c) Real range‐Doppler plot from the radar trial at an airfield. range of the target. The Doppler frequency of the target was
The phase noise floor can be seen emerging out of the kept at 69.8 Hz, equivalent to the velocity of 8.3 m/s. The RCS
thermal noise floor for range bins with higher clutter. The of the drone was modelled such that the SNR of the simulation
plot also consists of other unwanted targets as seen by the is comparable with the SNR of the real data [36].
radar. For any real radar, phase noise plays a significant role in
detecting targets. The radar system's phase noise causes the
range‐Doppler plot's phase noise floor to emerge out of the
3.3 | Comparison of range‐Doppler plots thermal noise floor in range bins with stronger clutter power.
The increase in the overall noise floor makes the detection of
In the third level of validation, the CNR, the SNR, the thermal targets difficult. One great advantage of the sophisticated radar
noise floor, and the phase noise floor from the simulated model discussed in the paper is the ability to add oscillator
range‐Doppler plot are compared and validated with a real phase noise in the simulation. The simulation included a phase
range‐Doppler plot. For the third level of validation, the range‐ noise profile equivalent to the measured phase noise for the
Doppler plot of a real radar trial was taken as a reference. The radar transmit signal.
real radar trial was performed using the Gamekeeper 16 U in The range‐Doppler plot generated in the simulation is
an airfield with a controlled drone (target of interest) flying on compared with the real range‐Doppler plot and is given in
a predefined path. The range‐Doppler plot consisted of a Figure 3. All the range‐Doppler plots in Figure 3 are nor-
single frame equivalent to a full CPI from the radar trial with malised to the strongest clutter, the clutter present in the
the drone. For the selected frame, the drone flew at a specific second range bin. The first two figures, Figure 3a,b, shows the
distance and at a radial velocity of 8.3 m/s towards the radar. simulated range‐Doppler plots. Figure 3a is the simulated plot
Even though the radar trial was performed in a rural envi- without the oscillator phase noise, and Figure 3b is the simu-
ronment, the range‐Doppler plot consists of other unwanted lated plot with the oscillator phase noise. The real range‐
targets in the radar's field of view. Doppler plot is given in Figure 3c.
The third level of validation was performed by replicating Comparing Figure 3a,b, the increase in the overall noise
the real range‐Doppler plot in the simulation. As a first step, a floor and the emergence of the phase noise floor over the
range‐Doppler plot was simulated with the CNR of the thermal noise floor is clearly visible at ranges with extremely
simulation in parity with the CNR from the real range‐Doppler high clutter. The clutter‐induced phase noise due to large
plot. The clutter power from an individual range bin leaks to clutter backscatter is common to any radar system. The
the adjacent range bins making it impossible to generate a increased overall noise floor makes detecting slowing moving
replica of the real CNR for every range bin in the simulation. and low RCS targets difficult.
In the first step, the thermal noise was considered to calculate Figure 3b shows the simulated replica of the range‐
the CNR for both the real and simulated range‐Doppler plots. Doppler plot in Figure 3c. Qualitatively, the simulated plot
Table 4 compares the CNR from the simulated range‐ looks very similar to the real range‐Doppler plot. Since the
Doppler plot with the CNR from the real range‐Doppler CNR for every range bin cannot be exactly replicated in the
plot for different range bins. Table 4 shows a good agree- simulation, the overall noise floor for every range bin sepa-
ment between the simulated and real CNR. Since the exact rately does not match.
replica of the real CNR cannot be generated in the simulation, The target of interest is highlighted inside a red box for all
a difference of a few dB between the simulated and real CNR is the range‐Doppler plots in Figure 3. In all three cases, the
expected. target appears at the 19th range bin and 19th Doppler bin. The
SNR of the target was calculated from simulated and real
range‐Doppler plots and is given in the Table 5. Table 5 shows
T A B L E 4 Clutter to Noise Ratios from simulated and real an excellent agreement between the simulated and the real
range‐Doppler plots for different range bins. SNR.
The peak clutter power, average thermal noise power and
Simulated CNR Expected CNR Difference
average overall noise power for each range bin were taken from
Range bin (dB) (dB) (dB)
the Figure 3b,c to generate a range‐Doppler statistics graph.
A 103.43 105.72 2.29 The range‐Doppler statistics graph in Figure 4 compares the
B 94.96 90.13 −4.33 simulated and the real range‐Doppler plot from the Figure 3b,c
respectively.
C 93.89 91.92 −1.97
Figure 4 clearly shows the overlapping thermal noise floor
D 96.85 96.78 −0.07 for both the real and simulated range‐Doppler plots. The
E 65.24 64.79 −0.45 statistics graph shows a good agreement between the simulated
and real clutter power. The clutter power for the range bin 21
F 73.19 72.79 −0.4
from the real range‐Doppler plot was an exception which
 17518792, 2023, 6, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.12399 by Libya Hinari NPL, Wiley Online Library on [07/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1058
- KANNANTHARA ET AL.

T A B L E 5 The calculated SNR from the simulated and the real

range‐Doppler plots.

Parameter Simulated Real

Signal to noise ratio (dB) 33.17 32.34

F I G U R E 4 The range‐Doppler statistics graph shows the clutter

power, average thermal noise power, and average overall noise power for
every range bin. The statistics graph consists of data from the real and
simulated range‐Doppler plots with phase noise. The clutter power and
overall noise power for both the simulation and real data overlap to a good
extent. The thermal noise floor in the simulation fully overlaps with the real
data and hence is not differentiable. At the range bin with the highest
clutter power, the phase noise floor is raised at least 25 dB above the
thermal noise floor. The clutter‐induced phase noise due to large clutter
backscatter is a generic radar issue.

dB. The statistics graph shows a considerable increase in the

overall noise power due to the radar oscillator phase noise. For
the second range bin with the highest clutter power, the noise
power is increased by at least 25 dB. Figures 3b,c, and 4 show
the confidence and the fidelity of the comprehensive radar
model discussed in the paper and validate the preliminary
results.

4 | CONCLUSION
We have presented the modelling and simulation of a whole
system radar model. The high‐fidelity radar model consisted of
all the fundamental building blocks within the transmit chain,
the environment, the receive chain and the data processing
chain of a pulsed Doppler radar. A simplified mathematical
signal model and a very detailed system‐level SNR model were
F I G U R E 3 Comparison of the simulated range‐Doppler plots with the
real range‐Doppler plot. All the range‐Doppler plots are normalised to the
also presented. The radar model was validated at three levels.
strongest clutter. The target is highlighted in the red box. The target is In the first level, the signal power and SNR values at the output
present across all three plots in the same range and Doppler bin. The effect of each radar building block in the simulator were compared
of radar phase noise can be seen in the range‐Doppler plots with the phase with the expected values. In the second level, the range bin,
noise. The simulated range‐Doppler plot with phase noise and the real Doppler bin, range resolution, and the Doppler resolution
range‐Doppler plot show a good level of similarity.
obtained from the simulated range‐Doppler plot were
compared with the real radar data. In the third level, the
thermal noise floor, the overall noise power consisting of both
could not be replicated in the simulation. The statistics graph the thermal noise and the phase noise, the clutter power, the
also shows a close correspondence between the phase noise CNR and the SNR from the simulated range‐Doppler plot was
power for the real and simulated data within an offset of a few compared with the range‐Doppler plot from a real radar trial.
 17518792, 2023, 6, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.12399 by Libya Hinari NPL, Wiley Online Library on [07/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
KANNANTHARA ET AL.
- 1059

In all three levels of validation, the simulated values were in Additionally, the authors would also like to thank Stephen
good parity with the expected values. Harman (Aveillant) and Aveillant Ltd.
Comparing the simulated range‐Doppler plot with only the
thermal noise and with both the thermal noise and the phase C O N F L I C T O F I N TE R E S T S TA TE M E N T
noise, we can clearly see that the phase noise causes the overall The authors declare that there is no conflict of interest.
noise floor to rise above the thermal noise floor. The clutter‐
induced phase noise is common to all radar systems. For the DA TA AVA I L A B I L I T Y S TA TE M E N T
range bins close to the radar, the overall noise floor sits at least Data available on request from the authors.
25 dB above the thermal noise floor. We lose a dB of margin
for efficient target detection for every dB increase in the phase O R CI D
noise floor above the thermal noise floor. Improving the phase Jithin Kannanthara https://orcid.org/0000-0002-6156-
noise characteristic of the radar oscillator could ideally reduce 1181
the range‐Doppler plot phase noise floor and thus improve the
SNR available for target detection. An oscillator with lower R E F ER E N CE S
phase noise could be really helpful in detecting slow‐moving 1. Doviak, R.J., Zrnic, D.S., Sirmans, D.S.: Doppler weather radar. Proc.
targets with very low RCS. IEEE 67(11), 1522–1553 (1979). https://doi.org/10.1109/proc.1979.
The radar model discussed in the paper has the advantage 11511
of accommodating further building blocks and optimising 2. Liu, B., et al.: Characteristics and performance of wind profiles as
every parameter within each building block to represent any observed by the radar wind profiler network of China. Atmos. Meas.
Tech. 13(8), 4589–4600 (2020). https://doi.org/10.5194/amt‐13‐4589‐
radar in great detail. The validated model can also be used to 2020
predict the performance of the basic radar and explore the 3. Mercuri, M., et al.: Digital linear discrete FMCW radar for healthcare
benefits and limitations of (modelled) hardware changes and applications. In: 2019 IEEE MTT‐S International Microwave Sympo-
processing chain optimisations. The radar model has the po- sium (IMS), pp. 144–147 (2019)
tential to be developed as a digital twin to test any hypothetical 4. Gu, C., Li, C.: From tumor targeting to speech monitoring: accurate
respiratory monitoring using medical continuous‐wave radar sensors.
real radar scenarios within the simulation. One of the inter- IEEE Microw. Mag. 15(4), 66–76 (2014)
esting aspects would be to explore the limitation of the classical 5. Li, C., et al.: A review on recent advances in Doppler radar sensors
radar oscillator phase noise in target detection and the po- for noncontact healthcare monitoring. IEEE Trans. Microw. Theor.
tential of using a quantum oscillator with better phase noise Tech. 61(5), 2046–2060 (2013). https://doi.org/10.1109/tmtt.2013.22
56924
performance as a suitable alternative [37].
6. Carter, L.M., Campbell, D.B., Campbell, B.A.: Geologic studies of
planetary surfaces using radar polarimetric imaging. Proc. IEEE 99(5),
AUTHO R CO NT R I B U TI O N S 770–782 (2011). https://doi.org/10.1109/jproc.2010.2099090
Jithin Kannanthara: Conceptualisation, Data curation, 7. Moreira, A., et al.: A tutorial on synthetic aperture radar. IEEE Geosci.
Formal analysis, Investigation, Methodology, Software, Vali- Rem. Sens. Mag. 1(1), 6–43 (2013). https://doi.org/10.1109/mgrs.2013.
dation, Visualisation, Writing—original draft, Writing—review 2248301
8. Adnan, M.A., et al.: Vehicle speed measurement technique using various
& editing. Darren Griffiths: Conceptualisation, Data curation, speed detection instrumentation. In: 2013 IEEE Business Engineering
Formal analysis, Investigation, Validation, Writing—review & and Industrial Applications Colloquium (BEIAC), pp. 668–672 (2013)
editing. Mohammed Jahangir: Conceptualisation, Data 9. Ghadaki, H., Dizaji, R.: Target track classification for airport surveillance
curation, Supervision, Writing—review & editing. Jonathan radar (ASR). In: 2006 IEEE Conference on Radar, pp. 4 (2006)
10. Yang, Y., et al.: Effects of k distributed sea clutter and multipath on radar
M. Jones: Project administration, Supervision, Writing—re-
detection of low altitude sea surface targets. IET Radar Sonar Navig. 8(7),
view & editing. Christ J. Baker: Conceptualisation, Funding 757–766 (2014). https://doi.org/10.1049/iet‐rsn.2013.0285
acquisition, Methodology, Supervision, Writing—review & 11. Zhang, S., et al.: Target detection for multistatic radar in the presence of
editing. Michail Antoniou: Funding acquisition, Project deception jamming. IEEE Sensor. J. 21(6), 8130–8141 (2021). https://
administration, Supervision. Colin J. Bell: Formal analysis, doi.org/10.1109/jsen.2021.3050008
Funding acquisition, Methodology, Supervision, Validation, 12. Björklund, S.: Target detection and classification of small drones by
boosting on radar micro‐Doppler. In: 2018 15th European Radar Con-
Writing—review & editing. Henry White. Funding acquisi- ference (EuRAD), pp. 182–185 (2018)
tion, Supervision, Writing—review & editing. Kai Bongs. 13. Jahangir, M., Baker, C.: Persistence surveillance of difficult to detect
Funding acquisition, Supervision. Yeshpal Singh: Con- micro‐drones with l‐band 3‐d holographic radarTM. In: 2016 CIE In-
ceptualisation, Funding acquisition, Methodology, Supervision, ternational Conference on Radar (RADAR), pp. 1–5 (2016)
Writing—review & editing. 14. Siddiq, K., et al.: Phase noise in FMCW radar systems. IEEE Trans.
Aerosp. Electron. Syst. 55(1), 70–81 (2018). https://doi.org/10.1109/taes.
2018.2847999
ACKNO WL E DG EMEN T S 15. Siddiq, K., et al.: Phase noise analysis in FMCW radar systems. In: 2015
This work was partly supported by the UK National Quantum European Microwave Conference (EuMC), pp. 1523–1526 (2015)
Technology Hub in Sensing and Timing (Project EP/ 16. Brunner, D., et al.: Radar imaging simulation for urban structures.
Geosci. Rem. Sens. Lett. IEEE 8(1), 68–72 (2010). https://doi.org/10.
T001046/1), European Union's Horizon 2020 research and
1109/lgrs.2010.2051214
innovation programme (820404—IQ clock project), and BAE 17. Griffiths, H., et al.: Measurement and modelling of bistatic radar sea
Systems I‐CASE PhD. The authors would also like to thank clutter. IET Radar Sonar Navig. 4(2), 280–292 (2010). https://doi.org/
Bill Stafford and Caitlin Percy (ex‐members of BAE systems). 10.1049/iet‐rsn.2009.0124
 17518792, 2023, 6, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/rsn2.12399 by Libya Hinari NPL, Wiley Online Library on [07/01/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1060
- KANNANTHARA ET AL.

18. Grosdidier, S., Baussard, A., Khenchaf, A.: HFSW radar model: simulation 28. Richards, M.A., et al.: Principles of Modern Radar, vol. 1. Citeseer (2010)
and measurement. IEEE Trans. Geosci. Rem. Sens. 48(9), 3539–3549 29. Mazda, F.: Telecommunications Engineer’s Reference Book. Butterworth‐
(2010). https://doi.org/10.1109/tgrs.2010.2047022 Heinemann (2014)
19. Watts, S., Rosenberg, L.: A comparison of coherent and non‐coherent 30. Zumbahlen, H., et al.: Linear Circuit Design Handbook. Newnes (2011)
radar detection performance in radar sea clutter. In: International Con- 31. Li, C., et al.: Principles and Applications of RF/Microwave in Healthcare
ference on Radar Systems (Radar 2017), pp. 1–6 (2017) and Biosensing. Academic Press (2016)
20. Watts, S.: Modeling and simulation of coherent sea clutter. IEEE Trans. 32. Chen, W.K.: The Electrical Engineering Handbook. Elsevier (2004)
Aerosp. Electron. Syst. 48(4), 3303–3317 (2012). https://doi.org/10. 33. Lyon, D.A.: The discrete Fourier transform, part 4: spectral leakage. J.
1109/taes.2012.6324707 Object Technol. 8(7), 23 (2009). https://doi.org/10.5381/jot.2009.8.7.c2
21. Huang, H., Pan, M., Lu, Z.: Hardware‐in‐the‐loop simulation technology 34. Enggar, F.D., et al.: Performance comparison of various windowing on
of wide‐band radar targets based on scattering center model. Chin. J. Fmcw radar signal processing. In: 2016 International Symposium on
Aeronaut. 28(5), 1476–1484 (2015). https://doi.org/10.1016/j.cja.2015. Electronics and Smart Devices (ISESD), pp. 326–330 (2016)
07.006 35. Prabhu, K.M.: Window Functions and Their Applications in Signal
22. Yu, D., et al.: An incoherent scatter radar simulation system based on Processing. Taylor & Francis (2014)
MATLAB. Geosci. Rem. Sens. Lett. IEEE 17(9), 1513–1517 (2019). 36. Mohajerin, N., et al.: Feature extraction and radar track classification for
https://doi.org/10.1109/lgrs.2019.2950417 detecting uavs in civillian airspace. In: 2014 IEEE Radar Conference,
23. Al Zubaidy, M.A., Sayidmarie, K.H., Al Shamaa, S.S.: Radar system pp. 0674–0679 (2014)
simulator using pc and MATLAB Simulink. In: 2006 International Radar 37. Jahangir, M., et al.: Development of quantum enabled staring radar with
Symposium, pp. 1–4. IEEE (2006) low phase noise. In: 2021 18th European Radar Conference (EuRAD),
24. Brooker, M., Inggs, M.: A signal level simulator for multistatic and netted pp. 225–228. IEEE (2022)
radar systems. IEEE Trans. Aerosp. Electron. Syst. 47(1), 178–186
(2011). https://doi.org/10.1109/taes.2011.5705668
25. Robins, W.P.: Phase Noise in Signal Sources: Theory and Applications, How to cite this article: Kannanthara, J., et al.: Whole
vol. 9. IET (1984) system radar modelling: simulation and validation. IET
26. Gamekeeper Aveillant 16 U. (2020). https://www.aveillant.com/
Radar Sonar Navig. 17(6), 1050–1060 (2023). https://
products/gamekeeper/
27. Moran, B.: Mathematics of radar. In: Twentieth Century Harmonic doi.org/10.1049/rsn2.12399
Analysis—A Celebration, pp. 295–328. Springer (2001)