IET Microwaves, Antennas & Propagation

Research Article

ISSN 1751-8725
Efficient design of a radome for minimised Received on 15th January 2016
Revised on 10th May 2016
transmission loss Accepted on 5th June 2016
doi: 10.1049/iet-map.2016.0041
[Link]

Tamara Sheret 1 ✉, Clive Parini 1, Ben Allen 2

1
School of Electronic Engineering and Computer Science, Queen Mary University London, London, UK
2
Department of Engineering Science, University of Oxford, Oxford, UK
✉ E-mail: [Link]@[Link]

Abstract: A radome has to be carefully designed to ensure it has minimum impact on the performance of the antenna.
Traditional approaches involve equation approximations, complex ray tracing, iterative manufacturing process and or
electrically large and lengthy electromagnetic (EM) simulations. The novel approach described in this study involves a
two-dimensional ray tracing method simulation in MATLAB and an electrically small unit cell simulation in
high-frequency structure simulator. This new approach gives an optimal thickness result that is only 0.004 mm different
to that of a full EM simulation. The radome is manufactured and radio-frequency tested in an anechoic chamber, the
calculated thickness is shown to be the optimal thickness.

1 Introduction material, lo is the wavelength in free space of the EM radiation

(EMR), ɛr is the relative permittivity, n is an integer and αi is the
Radomes are structures that cover an antenna, providing angle of incidence of the EMR. Before the common use of EM
environmental protection whilst by design have limited effect on simulations to calculate the radome thickness, a radome would be
the performance of the antenna. The introduction of a radome to manufactured over-thick by several millimetre (mm),
an antenna system will increase transmission and receive losses, radio-frequency (RF) tested, the thickness reduced, RF tested and
distort the pattern and polarisation of the antenna pattern and iterated until the optimal thickness was found
cause errors in boresight [1, 2]. To reduce impact, radomes are
carefully tuned to be a matched thickness for the frequency of l0
electromagnetic radiation being used [3]. The matched thickness lm = √


(1)
1r
condition occurs when reflections are reduced and least power is
lost through the radome and there is least phase disturbance. The nlm
external profile, robustness and material of the radome vary greatly RMT = cos ai (2)
2
depending on its purpose.
Often, due to the operational requirements of the system, the Ray tracing methods calculate the path of the RF wave through the
radome cannot be a hemisphere and has to be a shape such as a radome system. It uses idealised narrow beams, rays and calculates
Von Kármán [4], meaning the refraction caused by the radome their propagation by known physical interactions at medium
changes as a function of scan angle of the antenna. In radar changes, in 2 or 3 dimensions. Snell’s law is used to define the
systems this apparent movement of the source, caused by the refraction of a wave at a medium boarder.
aberration effects of the radome, has to be compensated for during The radome under investigation in this paper is manufactured
measurement of the target [5]. Radomes can be made from a from BTCy-1, a pre-preg laminar composite material, with a ɛr =
single electrically thin dielectric layer [6] or a sandwich, which 3.6 and tan δ = 0.004 at a frequency of 15 GHz. Using (1) and (2),
has two dielectric layers with an alternative material between them the nominal thickness of this radome can be calculated to be 5.1
as a core at its simplest [7]. mm for normal incidences. As the centre frequency is low, n is set
This paper presents a novel and improved process to define the to 1 to keep the thickness of the radome to a minimum reducing
optimised thickness of a single walled radome. The antenna used transmission losses and maximising the antenna aperture.
for these simulations and measurements is a twist reflector
monopulse antenna (TRMAS) [8]. Traditional radome design
methods are time consuming and expensive often requiring
iterations of hardware manufacture and are based around complex 3 Combined approach for radome design
equation approximations [9], complex ray tracing [10] and or
electrically large EM simulations. The approach described in this This paper presents a novel and improved process to define the
paper uses a 2DRTM written in MATLAB and a unit cell optimised thickness of a radome. The approach described in this
high-frequency structure simulator (HFSS) simulation which paper is a specifically written two-dimensional ray tracing method
optimises the thickness of the radome without compromising on (2DRTM) MATLAB code which is based on the geometry of a
accuracy. Manufactured radomes are tested in an AC and the TRMAS. This outputs an average angle of refraction (αr) which is
thickness verified. used as the input to a unit cell HFSS simulation. The output of the
optimisation in the HFSS simulation is the optimal thickness of
the radome.
The radome shape to be optimised in this paper is a Von Kármán
2 Traditional radome design techniques with C = 0. This is a shape that has very good aerodynamical
properties, as it gives minimum drag for a given length and
The nominal radome matched thickness (RMT) for a radome can be diameter, but is not particularly good for RF beam quality due to
calculated from (1) and (2) where lm is the wavelength in the the sloped sides and pointed tip.

IET Microw. Antennas Propag., 2016, Vol. 10, Iss. 15, pp. 1662–1666
1662 & The Institution of Engineering and Technology 2016
 17518733, 2016, 15, Downloaded from [Link] Wiley Online Library on [02/04/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 1 2DRTM plot of the range of rays at 0.1 mm spacing and their path through the radome at a beam angle of 0°

3.1 2D ray tracing 3.2 EM unit cell model

The 2DRTM approach is valid for situations where diffraction is not To optimise the thickness of the radome in the RF window, a unit
a main effect. In summary, the 2DRTM creates the required cell HFSS simulation was constructed, see Fig. 2. The geometry is
geometry for the TRMAS and radome under investigation and a flat sheet of radome material of parameterised thickness
traces the path of a ray bundle from the twist plate and calculates vertically in the centre, with an air box above and below it that is
the refraction at both radome and air boundaries; and determines in excess of half the wavelength at 15 GHz. The radome material
the position where the ray travels to infinity. is BTCy-1 and losses are included. Floquet ports are used so the
Initially the 2DRTM programme sets up the external geometry of model becomes a repeated structure infinite in size, but it remains
the radome, a Von Kármán Ogive followed by a parallel section of electrically small so it remains quick to solve. αi of the radiation is
length 0.1 m, see Fig. 1. The internal geometry of the radome is the same as to the calculated αr in the ray tracing model, 76.5°.
drawn to give a constant wall thickness using the external In HFSS, the Quasi Newton optimisation tool [12] was used to
geometry as the defining surface. The thickness chosen is the determine the optimal radome thickness for a given incident
nominal thickness of 5.1 mm as calculated previously. The twist angle. For this investigation, a single frequency of 15 GHz is
plate radius is maximised in the available volume as 0.09 m and is used. The optimisation was set to optimise the thickness of the
set to an angle of 0°. radome at αi = 76.5° with a target S11 value of −32 dB and a
A TRMAS geometry can be simplified to a band of parallel rays convergence criteria of 1 e −6, (4), these values are derived from
emerging from the twist plate, see Fig. 1. The gap in the rays at experimentation and good engineering practice. Care must be
the centre of the twist plate is due to the blockage caused by the taken to assign these values correctly, or the final result of the
feed. This 0.1 mm spacing clearly shows how the ray tracing optimisation will be invalid. Results give a thickness of 6.136 mm
works for a ray bundle. As the required parameter from this
simulation is the refracted angle between the normal and the exit
ray, the ray spacing is reduced to 0.01 mm to give a high density
of rays.
At the intersection of the ray with the radome, refraction will occur
as the RF can travel through the radome. The refraction at the air to
radome boundary is calculated using a form of Snell’s law,
Heckbert’s method [11]. Heckbert’s method, (3), allows αi and αr
to be calculated by using I, the unit vector of the incident ray’s
direction and N the surface normal unit vector. N is calculated
from the gradient of the surface where the ray intersects

cos ai = −I · N






















  (3)
cos ar = 1 − n2 1 − cos a2i

As defined by (3), the ray then travels through the radome. At the
radome air boundary refraction occurs again, as defined by (3),
and the ray then travels out of the radome and away from the
antenna. Therefore, the thickness of the radome can be optimised
as a function of a single beam angle, as presented in this paper.
For this radome with look angle of 0°, αr ranges from 81.1° to 68.8°,
with the average angle being 76.5°. The symmetrical nature of this can
be clearly seen in Fig. 1. This value will be used as an input to the unit
cell model in HFSS to define the optimised radome thickness for
maximum transmission at the required beam angle. Fig. 2 HFSS simulation set up for a radome and TRMAS model

IET Microw. Antennas Propag., 2016, Vol. 10, Iss. 15, pp. 1662–1666
& The Institution of Engineering and Technology 2016 1663
 17518733, 2016, 15, Downloaded from [Link] Wiley Online Library on [02/04/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
enough to calculate the gain as a function of thickness, whilst
removing the full geometrical complexity. A single pyramidal horn
feed is used as the feed which is situated at the focus of a metal
paraboloid. An air box (vacuum) is constructed around radome
with a radiation boundary set up at edge of this vacuum. A wave
port is set up at the input end of the feed because it is inside the
air box. To look at the gain of the system, infinite radiation
spheres are set. Conductor and dielectric losses are included in the
simulation. The thickness of the radome is constructed as a
parameter, and the gain of the system is simulated as a function of
radome thickness. For a single frequency, beam angle and radome
thickness, this simulation takes 51:15 min to run on a distributive
computing high performance computing (HPC) with 96 cores and
576 GB total memory capacity. It was not possible to run this
simulation as an optimisation, therefore it is run multiple times at
difference thicknesses instead. The same simulation can be run on
a 48 GB stand-alone machine, in 26 h.
Fig. 4 shows the results for the gain as a function of radome
thickness for the full radome with antenna HFSS model in 0.1 mm
stages over the range of interest of 4.6–7 mm. The range of
thickness covered has to start below the calculated value of
thickness of 5.1 mm and continue until a definite peak is found at
6.1 mm.
Fig. 3 HFSS simulation set up for a radome and antenna model From the peak gain found at 6.1 mm, the next simulation range is
chosen as 6.127–6.142 mm thickness in 0.001 mm stages. The
maximum gain, i.e. the best tuned radome thickness, occurs at a
in 11 iterations thickness of 6.132 mm. The gain variation over this band width is
only 0.14 dB, this level is below measurement tolerance and
RMT = min (ai , RMT) (4) therefore is below measurement accuracy.
S11
The overall optimisation of the radome thickness has taken 44
simulations and 41 h of simulation time. If a mechanically
Now, αi has been calculated (2), can be used to calculate the accurate HFSS simulation had been run, each simulation would
thickness of the radome again. Using αi = 76.5° and n = 5 the have taken over 24 h without improving the results.
thickness of the radome is calculated as 6.0 mm. This is close to
the optimised value of 6.136 mm as calculated by the method
presented in this paper. 5 Comparison of approaches
Due to the difference between the starting thickness of the radome
used as an input to the 2DRTM and the end thickness it is valuable to It can be seen that both approaches presented in this paper, 2DRTM
determine if an iteration of the design process is required before with unit EM simulation and the full EM simulation, give a very
finalising the optimal thickness of the radome. After completing similar value for the best matched thickness, therefore both
this investigation at an angle of 0°, the simulations do not have to methods are suitable for use. The difference between the methods
be iterated. The matched thickness value of 6.136 mm will now be in obtaining an optimal thickness is 0.004 mm. This difference in
compared with the results of a full TRMAS and radome HFSS thickness leads to a drop in gain of 0.07 dB in the full HFSS
simulation. simulation, which is negligible in a real system. Indeed, in reality
a standard machining tolerance is ±0.1 mm, which means a
thickness of 6.13 mm would be chosen in both cases as the
4 Radome and antenna full HFSS model manufacturable thickness.
As the two approaches give the same answer, the real difference
A simplified TRMAS is to be simulated, see Fig. 3. A set of parallel between them is the time taken. The following analysis will not
rays are to be projected from the twist plate. In this simulation, this is include the set up time of the models and will only consider the
achieved by using a front fed parabola. This simulation is detailed run times. Table 1 shows that the approach described in this paper

Fig. 4 Full radome with TRMAS HFSS simulated gain results

IET Microw. Antennas Propag., 2016, Vol. 10, Iss. 15, pp. 1662–1666
1664 & The Institution of Engineering and Technology 2016
 17518733, 2016, 15, Downloaded from [Link] Wiley Online Library on [02/04/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table 1 Comparison of simulation times and CPU thus validating the design process. The two points on either side of
the peak are at ∼±0.1 mm and are <0.5 dB down on the peak gain.
Simulation Time taken, min Memory, GB Iterations
This shows that the machining tolerances of the unit are valid. The
Matlab ray tracing 0:12 4 ∼1
one outlier at −2 mm clearly shows a 2 dB drop in gain.
HFSS unit cell 2:44 4 ∼1
HFSS 51:15 576 44

7 Conclusions
This paper describes and validates a new method for the optimisation
sped up the simulation time by 984 times, and reduced the memory of the thickness of an antenna radome, 2DRTM and unit EM
required by 144 times. simulation. The presented method reduces the time to calculate the
The new method presented in this paper is able to adjust to new optimal thickness of a radome by three-orders of magnitude and
geometry shapes easily as it requires a change to the 2DRTM the required computer memory by two-orders of magnitude over
geometry and the incident angle in the HFSS unit cell model. The the full EM simulation without compromising accuracy. The
full HFSS antenna and radome simulation will take several days to presented approach is particularly useful as it removes the
construct and successfully simulate. Therefore, the approach requirement to produce and validate the complex full EM
reported in this paper will be quicker to set up the simulation for a simulation and it removes the need of a computer cluster to
new antenna and radome system. complete the design. As the timescales are reduced, it allows for
A radome of thickness calculated from the 2DRTM and unit EM fast design iterations and fault finding during manufacture. The
simulation will now be tested in hardware in an AC. calculated thickness of the radome is then validated by anechoic
chamber measurements of the full radome with antenna, showing
that in a real system the thickness has been optimised by the
simulation work.
6 Radome performance AC testing In this approach, the radome has been tuned specifically for a
single frequency and a single look angle. Whilst systems will have
If the radome under test is a large structure that is made in several a preferred frequency and look angle, it is likely that they will
pieces, it is possible to complete panel testing on each unit to have to be able to work successfully over a specified range of
measure the RF performance of the radome [13]. However, for angles as well as given bandwidth. To optimise over a range of
monolithic radomes, the interaction of the radome with the look angles, the resultant average angle of refraction from a range
antenna is greater and panel testing will therefore be useful for of beam angles in the 2DRTM would have to be averaged to
understanding the RF performance of the radome material [14]. calculate the required input to the unit cell EM simulation. To run
AC testing of the antenna and radome system should be carried over a frequency range, the HFSS unit cell simulation will be
out to characterise the radome. The positioner in the chamber altered to run over a range of frequencies and the results compared
should use axis transforms to correctly measure the antenna [15] over the different frequency range required. When optimising over
and the hardware set up have minimal phase disturbance to the scan angle and frequency range the time saving of the method
measurements during movement [16]. presented here is even more significant.
The thickness of the radome was determined as an average of 20 The approach presented here can be used to determine the
thickness measurement points across the RF window. Radomes of optimised thickness of a radome for any reflector antenna system
different thicknesses were tested with the same antenna at a beam with a simple alteration to the set up geometry in the MATLAB
angle of 0°, the gain as a function of radome thickness is shown in code. The alteration will require the band of parallel rays to be
Fig. 5, using the method described in [17] to create the sum gain. located at the correct position for the reflector antenna system. It is
Due to the cost of the units it was not possible to manufacture a also possible to extend the method to optimise a radome with a
large number of radomes that were known to be incorrect. variable thickness along its length. Instead of taking the average
However, one unit was manufactured 2.031 mm under thickness, angle of refraction over the whole of the aperture it is split into
and bottom, middle and top of tolerance units have been used to sections, corresponding to bands of equal height on the radome.
show the range in gain performance. For each band an average angle of refraction can be calculated, by
Fig. 5 shows that the radome thickness has an impact on the gain. making these bands small an optimised variable thickness radome
The highest gain is found from the calculated thickness of 6.13 mm can be constructed from the unit cell model.

Fig. 5 AC testing of radome gain as a function of radome thickness

IET Microw. Antennas Propag., 2016, Vol. 10, Iss. 15, pp. 1662–1666
& The Institution of Engineering and Technology 2016 1665
 17518733, 2016, 15, Downloaded from [Link] Wiley Online Library on [02/04/2025]. See the Terms and Conditions ([Link] on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 References 9 Li, L., Leong, M., Zhou, I., et al.: ‘Improved analysis of antenna radiation from a
circular aperture covered by a dielectric hemispherical radome shell’, IEE Proc.
Microw. Antennas Propag., 2000, 147, (2), pp. 144–150
1 Rudge, A.: ‘The handbook of antenna design’, ser. IEE electromagnetic waves
10 Nair, R., Suprava, M., Jha, R.: ‘Graded dielectric inhomogeneous streamlined
series. P. Peregrinus on behalf of the Institution of Electrical Engineers, 1983, no. v. 2
radome for airborne applications’, Electron. Lett., 2015, 51, (11), pp. 862–863
2 Renuka, A., Borkar, V.: ‘Computer-aided analysis for tangent Ogive airborne 11 Heckbert, P.S.: ‘Derivation of refraction formulas’ (Academic Press, 1989),
radome using physical optics method’. Proc. Asia-Pacific Microwave Conf. Proc. [Link]
2005, APMC 2005, December 2005, vol. 5, p. 4 12 Schoenberg, R.: ‘Optimization with the quasi-newton method’ (Aptech Systems,
3 Gibson, C., Bonebright, M., Weisbrod, S.: ‘Design of an 89 GHz radome’. Proc. Maple Valley, WA 2001)
1998 IEEE Radar Conf., 1998. RADARCON 98, May 1998, pp. 373–378 13 Qureshi, W., Hill, L., Scott, M., et al.: ‘Use of a gaussian beam range and
4 Fleeman, E.L.: ‘Technologies for future precision strike missile systems’. NATO reflectivity arch for characterisation of radome panels for a naval application’.
Lecture Series, 2001, vol. 221 Twelfth Int. Conf. on Antennas and Propagation, 2003, (ICAP 2003), March
5 Kuan-Kim, C., Po-Rong, C., Fang, H.: ‘Radome design by simulated annealing 2003, vol. 1, pp. 405–408, (Conf. Publ. No. 491)
technique’. IEEE Int. Symp. on Antennas and Propagation Society 1992. AP-S. 14 Audone, B., Delogu, A., Moriondo, P.: ‘Radome design and measurements’, IEEE
1992 Digest. Held in Conjuction with: URSI Radio Science Meeting and Trans. Instrum. Meas., 1988, 37, (2), pp. 292–295
Nuclear EMP Meeting., 1992, pp. 1401–1404, vol. 3 15 Sheret, T., Allen, B.: ‘Axis transform to characterise a monopulse twist reflector
6 Nair, R., Jha, R.: ‘Electromagnetic performance analysis of a novel monolithic antenna and radome in an anechoic chamber’. 2014 Loughborough Antennas
radome for airborne applications’, IEEE Trans. Antennas Propag., 2009, 57, and Propagation Conf. (LAPC), Loughborough, November 2014, pp. 612–616
(11), pp. 3664–3668 16 Sheret, T.L., Allen, B., Parini, C.G.: ‘Phase stable multi-channel antenna
7 Chang, K.: ‘Antenna angle error correction to radome curvature’. Antennas and measurements on a moving positioner’. European Conf. on Antenna and
Propagation Society Int. Symp., 1995. AP-S. Digest, June 1995, vol. 3, Propagation, 2016, pp. 1–4
pp. 1426–1429 17 Sheret, T., Parini, C., Allen, B.: ‘Monopulse sum and difference signals with
8 Naqvi, S.S.H., Gallagher, N.C.: ‘Analysis of a strip-grating twist reflector’, J. Opt. compensation for a failed feed element’, IET Microw. Antennas Propag., 2016,
Soc. Am., 1990, 7, (9), pp. 1723–1729 10, (6), pp. 645–650

IET Microw. Antennas Propag., 2016, Vol. 10, Iss. 15, pp. 1662–1666
1666 & The Institution of Engineering and Technology 2016