Analysis and optimal design of microstrip sensors

K.B. Khalid
T.S.M. Maclean
M. Razaz
P.W. Webb

Indexing terms : Microstrip and stripline, Permittivity

in fruit easily detached from a bunch as a measure of its

Abstract: The analysis and optimal design of a ripeness; however, this indicator does not relate to the
microstrip sensor for measuring the water content true oil quantity in the fruit [3]. There is, therefore, a
of rubber latex is described. The microstrip struc- need for an instrument to measure the water in liquid
ture consists of three layers: substrate, protective materials with a high moisture content (in the region of
layer and semi-infinite layer of wet medium. A 4&100%), especially for process materials such as hevea
functional relationship has been developed rubber latex, palm oil, milk, and sugar cane. Correspond-
between the attenuation and the water content of ingly a theoretical basis for such an instrument is also
the latex, and close agreement has been found required, and in this paper the assumption is made that
between the computed and experimental results. A all the water present contributes to the dielectric con-
computer program has also been developed which stant. In some materials, such as proteins [ 2 6 ] , this
optimises the sensitivity for a given water content. assumption would not be valid.
It has been pointed out by Kent [4] that stripline
sensors offer distinct advantages for these moisture-
1 Introduction content measurements. In such a method the material
under test is brought into contact with the free face of a
Since water is present in the majority of material encoun- stripline, and the insertion loss of the material is cali-
tered in solid or liquid form its determination and moni- brated against its moisture content. Various publications
toring are very important from an economic point of have appeared in the literature on the application of strip
view; such as when dealing with agricultural products, transmission lines to the measurement of moisture
food or coal. Low intensity microwaves have been content [ 5 , 6, 71 : most of these publications present
applied in microwave aquametry [4] for the nondestruc- experimental results relating geometry and the electrical
tive measurement of the moisture content of materials. properties of the microstrip sensor to the attenuation, or
In the natural rubber industry most of the rubber insertion loss, due to the water content. It is evident from
comes from hevea latex, obtained by tapping rubber our survey [25] that the information available on the
trees. Hevea latex is a biological product of complex analysis and optimum design of this type of microstrip
composition. The basis components of a freshly tapped sensor is inadequate. There is also a need for a quantitat-
natural rubber latex consist of 5G80% water, 1 8 4 5 % ive relationship between the geometrical and electrical
rubber and approximately 2-5% nonrubber constituents properties of a sensor and the optimum sensitivity of the
[l]. Latex collected in rubber plantations is sold to col- measuring instrument. This paper presents an approach
lectors, who pay according to the dry rubber content for the analysis and optimal design of microstrip sensors.
(DRC) or the water content. As the DRC of latex varies Using this approach a microstrip sensor was designed
according to the season, weather, soil conditions, clone, and constructed for measuring the water content of
stimulation progress, tapping system and adulteration, process materials, with specific reference to hevea rubber
the true DRC of the latex must be determined to ensure a latex. Good agreement between theory and experimental
fair price. results has been found.
Another potential application of microwave aqua-
metry is in palm oil industries. In palm trees, the oil
usually accumulates in the fruit within 40 days of ripe-
ness, and at the final stage the oil content drops drasti- 2 TEM analysis of double covered microstrip
cally from 80% to about 40% [2]. A preharvest is at w i t h semi-infinite layer
present carried out to give an indication of how ripe the In order to compute the characteristic impedance of the
crop is and so of the best time for it to be harvested. At configuration shown in Fig. 1 it is necessary to solve
present, no better ripeness indicator has been found.
Most palm oil researchers have used the percentage of oil
Er3 liquid
Paper 5991H (E12/S4/El) first received 30th July 1987, and in revised
form 21st January 1988 Er2 -
-W- : s protective layer
Dr. Khalid is with the University Pertanian Malaysia, Jabatan Fizik, Erl YI x ‘ h substrate
Fakulti Sains Dan, Pengajian. Alam Sekitar, Serdang, Selangor, / / / / / / /
Malaysia / ” / / ” \ground plane
Dr. Maclean, Dr. Razaz and Dr. Webb are with the Department of
Electronic and Electrical Engineering, University of Birmingham, PO Fig. 1 Cross-section of a semi-infinite double-covered microstrip line
Box 363, Birmingham B15 2TT, United Kingdom for measurement of moisture content

IEE PROCEEDINGS, Vol. 135, Pt. H , N o . 3, J U N E 1988 187

Laplace's equation subject to the appropriate boundary = 65.0, which corresponds to water for a frequency of
conditions. It has been shown by Yamashita et al. [lo] 9.0 GHz at a temperature of 20°C. These calculated
that the variational expression for the line capacitance results show that the effective permittivity E is drastically
per unit length C of double covered microstrip with a
semi-infinite lossless layer, is obtained from

where E , is the free space dielectric constant; Q the total

charge per unit length on the strip conductor; p(B) the
Fourier transform of the initially unknown charge
density distribution, and
1 ( ~ r 2cash BS + ~ r 3 )
(2)
(') =
[&,3(&,Z +
coth p s E , ~coth Bh) 20 -
+ +
E , ~ ( E , ~ erl coth Bs coth BO] 10-
where E , ~ ,E , ~and cr3 are the real relative permittivities of 01
the lossless substrate, protective layer and lossless semi- 01 W/h 10
infinite media, respectively. a
Different trial functions may be assumed for the 40
charge distribution p ( x ) . All such functions will have a
charge minimum along the central axis, rising with
varying degrees of steepness at the strip edge. Computa- 35:
tions of the resulting capacitances for a single microstrip
line, using such different distributions, show little change 25 -
in the resulting characteristic impedance for the different
distributions. Consequently the normalised trial function
of Yamashita [lo], given by
15-
W W
when - - < x <- (3)
2 2
=o
with Fourier transform
otherwise
5t
01
01 W/h 10
b
Fig. 2 Characteristic impedance and effective dielectric constant
against W l hfor a semi-infinite double-covered microstrip

sin -
-2-++
Bw
2
sin2 9j (4)
E , ~= 10.5; E , ~= 3.0; E , ) = 65.0

reduced, even for a small thickness of protective layer

and especially for small values of W / h where a greater
proportion of the field lines will lie above the strip. Con-

(9
sequently, it is desirable to keep the thickness of the pro-
tective layer as small as possible from an electrical point
Bw/2 of view, although mechanical considerations also apply.
has been chosen. Fig. 2 also indicates that for s/h = 0.1 the double-covered
When the distributed line capacitances per unit length microstrip is matched to a 50 R line when W / h = 0.5.
C and C , are evaluated for the actual and the air-spaced This corresponds to approximately 70 R for the micro-
dielectric layers, respectively, the effective dielectric con- strip line in air.
stant E, and the characteristic impedance 2, can be
obtained from 3 Effect of moisture content on characteristic
impedance and attenuation
C
E=-
The complex propagation constant of a wave travelling
CO through a homogeneous lossy medium with a complex
and relative permittivity E, = (E' - j ~ " is
) given by

Z,I! ' J cc,

c
y = (u + j B ) =jk(E' - J & " ) I ' Z
where o! is the attenuation constant, in this case due to
(7)

where c is the velocity of light in free space. the dielectric loss, B is the phase constant and k is the free
Fig. 2 shows the calculated variation of 2, and E, for space phase constant. Equating the real parts of eqn. 7
different normalised thicknesses s/h of the protective gives the attenuation constant as [ 111
layer expressed as a function of the substrate thickness.
Both E and 2, are shown as functions of the normalised (8)
width of the conducting strip W / h for a substrate erl =
10.5, a protective layer E , ~= 3.0 and a lossless wet media where tan 6 = E"/E',

188 IEE PROCEEDINGS, Vol. 135, Pt. H , N o . 3, J U N E 1988

 When tan 6 < 0.1 the inner square-root term may be (cited in Reference 14) as
expanded, by the binomial theorem, to give E,,, = BE: + (1 - 6 ) E ;
Elt
6,
CY = 4.343k 7dB/m
JE
(9) 1
- +-(1 - 6 )
Emin E: E:
This may also be written in terms of an equivalent con-
ductivity CJ,through the relation where 6, is the volume fraction of water and where E:
and E: are the complex permittivities of the water and
CJ = O E g El' dry inclusions, respectively. Because of the wide variation
between these limits as a function of moisture, an alterna-
so that eqn. 9 takes the form tive approximate single-valued complex permittivity has
been proposed by Kraszewski [151 as
a = 4.343k
WE,, JE' Eli2 = 6&:'2 + (1 - 6 ) E y (19)
Returning to the case of propagation along the double- Experimental results for clay-sand slurry [151 give agree-
covered microstrip with a semi-infinite layer, the effective ment with eqn. 19 for the real and quadrature parts of E
conductivity and permittivity will be written in terms of to within 2% and 5%, respectively. The volume fraction
the filling fraction 4 occupied by each dielectric as 6 is related to the moisture content by [161
+ 4 2 o2 + ( l - 41 - q 2 b 3 + LY!
Oeff = 4 l a 1

'eff

where C J ~o2
= q l E r l + 4 2 ' r 2 + (l - 41 - 4 2 ) & r 3

, and o3 are the conductivity of the substrate,

(12)
(11) 6 = M,/LM,
Yd
(1 - M,)

where M , is the wet basis moisture content, and yw and

1 (20)

protective layer and semi-infinite layer, respectively, and yd are the relative densities of the water and dry material,
q l , q2 are the dielectric filling fractions. These filling frac- respectively.
tions may be calculated by transforming the three layers Fig. 4 shows typical limit curves for hevea latex solu-
of the microstrip structure of Fig. 1 to the two layer tion as a function of water content, plotted for the equa-
structure of a single-covered microstrip line as shown in tions of Weiner and Kraszewski where the complex
Fig. 3. Both these structures are taken to have the same

€r23 -W-

Erl YI x 'h substrate

/ / / / / / / / / / / / / / / I
grou\;;d plane

Fig. 3 Cross-section of a semi-infinite microstrip line with as in

Fig. 1

effective dielectric constant E, obtained from eqn. 5, so

that for Fig. 3 using
'eff = q l E r l + (l - q l l E r 2 3 (13)
the effective dielectric constant of the upper layer of moisture content (wet basis),"/.
the two layer structure may be obtained by a numerical
root finding method. Knowing the value of the 40r a
We i ner 's
9
'
dielectric filling fraction q1 is obtained from eqn. 13: /'
upper b o u n d
30-
&eff - er23
41 =- (14)
&r1 - &r23 E"
Kraszewski's
Then, from eqn. 12, q2 follows as

Weiner's
lower bound
Substituting for oeffin eqn. 10 gives the dielectric loss in
the double-covered semi-infinite microstrip line as
k 0 20 40 60 80
a = 4.343 -[ q l ~ r tan
l 6, + q 2 e r 2 tan 6, moisture content (wet basis),"/.
b
JEeff
Fig. 4 Prediction of the complex permittivity of hevea latex at various
$- (1
tan 6 3 1 (16) - 41 - 42)&,3 moisture contents
This form of the attenuation constant is incomplete, in cHZ0= 60 - j31.6; = 2.2 - j0.02
that a relation between the moisture content of the latex
and its permittivity still has to be found. To achieve this permittivities for water and rubber have been taken [l5]
the latex will be considered as a biphase mixture, i.e. a as (60 - j37.6) and (2.2 - j0.02), respectively.
host medium with water and dry inclusions. For such a Fig. 5a, derived from eqn. 16, shows the variation of
biphase medium the upper and lower limits of the attenuation constant with water content for various
complex permittivity have been introduced by Weiner thicknesses of covered layer with = 2.3 and W / h =

IEE PROCEEDINGS, Vol. 135, i't. H , N o . 3, J U N E 1988 189

0.87, which is equivalent to 100 R in air. In this example 4 Factors affecting attenuation and sensitivity
the hevea latex solution is chosen, and the prediction of of the sensor
the complex permittivity for E from eqn. 19 varies with The factors affecting the dielectric loss and sensitivity of
the sensor according to eqn. 16 are described in the fol-
lowing Sections.

Or

!i/l
D

0
. 15
1
m
E 15-

3
b
C
L

motsture tontent (wet basls) ,% 0 10 20 30 LO 50 60 70 80 90 100

0 m o i s t u r e content (wet basis) ,Yo
a

"L s/h=l 0
50 _ _ _ _____

30 -
c
0

2ot
N 20-

0'
10
lot
o o k i r 7 r 2 r i o io $0 7 b io
moisture content (wet b a s i s ) , %
9b 160 0?
0 10 20 30 40 50 60 70 80 90 100
b moisture content ( w e t
b
Fig. 5 Variation of dielectric loss and characteristic impedance with
moisture content (hevea latex solution) for various thicknesses of covered Fig. 6 Variation of dielectric loss and impedance with moisture
content (hevea latex solution)for various thicknesses of covered layer
layer with E , ~ = 2.3 and Z , (air) = 100 R.
E,, = 10.5 -10.005; W / h = 0.95: Z,(air) = 50 R
yrubbcr= 0.92; E , . ~ ~ . , = 2.2 - j0.02; = 60 - j37.6; E , , = 2.3 - j0.005; E , ~= 3.0
- j0.005; Z, (air) = 100 n; W / h = 0.87; frequency = 9 GHz

Mixture model: Kraszewski 4.1 Operating frequency and temperature

The dielectric loss of the sensor is directly proportional
to the operating frequency, according to eqn. 16; this loss
is also affected by the frequency dependence of the polari-
water content in Kraszewski's model. Fig. 5a shows sation of water molecules. The variation of the permit-
that the sensitivity, defined as the slope of the attenuation tivity of free water with the frequency of the
curve for s/h = 0, is about 0.18 dB/cm/% m.c. and that it electromagnetic field is shown in Fig. 7. The experimental
drops to 0.1 dB/cm/% m.c. for s/h = 0.1. As s/h increases results are based on the data from Cook [lS] and Grant
the sensitivity is drastically reduced, and it varies accord- et al. [19], and the theoretical curves are described by the
ing to the range of water content. Fig. 5b shows that the Cole-Cole equation [24] :
variation of impedance in the range 0-100% water
content is about 60% for slh = 0 and 45% for s/h = 0.1.
The structure is almost matched to a 50 R line system for
s/h = 0.1, with the water content varying between 70% where E, is the static dielectric constant; E , is the dielec-
and 100%. Fig. 6a, derived from eqn. 16, and Fig. 6b, tric constant at very high frequency; w the angular fre-
derived from eqn. 6 show the attenuation constant and quency; z the relaxation time; CI the relaxation time
characteristic impedance as a function of water content spread parameter for z.
for various thicknesses s/h of the covering layer, for a The orientation of water molecules is also affected by
50 R line in air with E , ~= 10.5. As the thickness of the temperature. The variation of the permittivity of water
covering layer is increased the attenuation falls quickly, with temperature is also shown in Fig. 7.
while the characteristic impedance tends towards its It should be noted from Fig. 7 that at room tem-
value in air. The thickness of the covering layer is impor- perature (20"C), pure water has a dielectric constant
tant in determining the dynamic range and sensitivity of varying between 80 and 15, and a loss factor between 2
measurement, as well as in determining the matching. and 40, at frequencies between 580 MHz and 23.8 GHz.
tm IEE PROCEEDINGS, Vol. ISS, Pt. H , No. 3, J U N E 1988
Therefore different operating frequencies will give differ- 4.2 Choice of width of the conducting strip
ent relative permittivities of the semi-infinite layer, and The width of the conducting strip is inversely proportion-
hence affect the attenuation of the double-covered micro- al to the characteristic impedance of the line in the air.
strip. Fig. 9 shows the computed results for the attenuation in

45 r 9 1SGHz hevea latex

40 ~
s ~ r 2
12- hm,ErI
35 -

30 -

25 -
E"
20 -
74
15-

0 10 20 30 40 50 60 70 80 90 100
E'
Fig. 7 E" against E' representdon of microwave dielectric properties oj
water for different values of temptzrature and different frequencies
____ Cole-Cole equation
0 experimental points after CoJk [le] and Grant et a/. [I91
OA lb 2'0 io i o i o $0 7'0 io 90 Id0
molsture content ( wet basis 1 ,o/o
Fig. 8 shows a comparison between theoretical and
measured attenuation against moisture content (wet Fig. 9 Graph of attenuation in semi-infinite double-covered microstrip
basis) for sucrose solution at different frequencies. It is against moisture content in hevea latex solution for different characteristic
impedances
Mixture model: Weiner's upper bound E,, = 2.3 - jO.005; E , ~= 2.18 - j0.005;
sucrose s/h = 0.1; frequency = 9.0 GHz; temperature = 18°C
solution
a W / h = 0.89 (100 0)
S 'r2
b W / h = 1.66 (73.5 Q)
12-h Erl
c W / h = 3.09 (50 Q)
d W / h = 6.0(30.8 Q)

10-
the semi-infinite double-covered microstrip as a function
of water content for different strip widths, when the strip-
8- line is mounted on a polyguide substrate. The higher
attenuation seen for 50 R and 30.8 R lines in air is due to
the high reflective loss and dielectric loss. On the other
hand, the 100 R line in air is almost matched to the 50 R
1 8 "NIL line in this region of water content (see Fig. 5), and the
attenuation for this line is dominated by the dielectric
loss.
Two samples of different widths, with characteristic
impedances 73.5 R and 50 R, were tested experimentally
using the solutions from hevea latex, as shown in Figs. 10
and 11, with different values of the protective layer thick-
ness. The difference between experimental and theoretical
curves is almost 5%. The discrepancy increases as s/h
0 10 20 30 40 50 60 70 80 90 100 decreases, and this is due to the limitation of the TEM
moisture content (wet basis),"/. analysis for the small ratio of s/h. The sensitivity for both
Fig. 8 Graph of attenuation vtrsus moisture content (sucrose solutions) 73.5 R and 50 R lines is almost identical.
for different operating frequencies
mixture model: Kraszewski H,O (using Cole-Cole equation) E,, = 10.5 - jO.005
(RT-Duroid);F , =
~ 2.18 - j0.005 (polythene); temperature = 18°C 4.3 Effect of relative permittivity of the substrate
__ theoretical curve The effect of the relative permittivity of the substrate on
0 experimental point
the sensitivity can be seen from Figs. 11 and 12, where
attenuation has been plotted against percentage of water
observed that the attenuation, and hence the sensitivity, content in the latex solution for both a polyguide sub-
are the highest in the X-band region, which corresponds strate ( E , = 2.3 - j0.005) and an RT-Duroid substrate
to higher values of dielectric constant and loss. The ( E , ~= 10.5-jO.005). The results show that whereas the
average sensitivity for 5@-100% water content at 9 GHz sensitivity (in the range 5&100% water content) for RT-
is about 0.03 dB/% H,O/cm, whereas at 3 GHz this Duroid substrate is about the same as for polyguide its
sensitivity drops to 0.014 dB/% H,O/cm. level of attenuation is about 2 dB/cm better.
IEE PROCEEDINGS, Vol. 13.5, Pt. H , N o . 3, J U N E 1988 191
 4.4 Effect of thickness and relative permittivity ture content/cm at s/h = 0.07 to 0.015 dB/% moisture
of the protective layer content/cm at s/h = 0.52. For the case of the RT-Duroid
Fig. 11 shows that the attenuation and sensitivity of the substrate shown in Fig. 12 the sensitivity for s/h = 0.15 is
double covered microstrip are drastically affected by the
1 hevea latex

17 I hevea latex

'h *,Erl
ErZ

l2I Sh*Er'
Er2

/ ,0- - C
0
0
1 3 4-
a
l
L

0
0 10 20 30 40 50 60 70 80 90 100
moisture content ( w e t basis),'//.
0 10 20 30 40 50 60 70 80 90 100
m o i s t u r e content (wet basis ) , % Fig. 12 Attenuation against moisture content in hevea latex solution
Fig. 10 Graph of attenuation against moisture content in hevea latex on RT-Duroid substrate
solutions on polyguide substrate = 10.5 - j0.005; E , ~= 2.18 - j0.005; W / h = 0.95 (50 Q); frequency = 9 GHz;
temperature = 18-20°C
E , ~= 2.3 - j0.005; E , ~= 2.18 -10.005; W / h = 1.66 (73.5 Q); frequency = 9 GHz;
~ Weiner's upper bound
temperature = 18°C
- - - - Kraszewski's model
Weiners upper bound
~

_ - _ - Kraszewski 0 experimental points

0 experimental points
0.019 dB/% moisture contentlcm, and drops to 0.01 1 dB/
I hevea latex
YOmoisture content/cm at slh = 0.35.
This effect has been experimentally observed in earlier
work by Kent [4] and Steel [5] where it was found that
the attenuation and sensitivity decrease with increasing
thickness of the covering layer.
skimmed
milk
Er2

I er2 = 2 16-10 005

o L 8 # L m l i

0 10 20 30 AI
50 60 70 80 90 100
moisture :ontent ( w e t b a s i s ) , %

Fig. 11 Attenuation agaimt moisture content in hevea latex solution

on polyguide substrate
E,, =2.3 - j0.005; E,* = 2.8 - jO005; W / h = 3.09 (50 Q); frequency = 9.0 GHz;
temperature = 18-20°C
~ Weiner's upper bound 0 10 20 30 40 50 60 70 80 90 100
0 experimental points moisture content(wet b a s i s ) ,'L
Fig. 13 Attenuation against moisture content in skimmed milk solution
thickness of the protective layer: this effect was observed for different covering materials
in the discussion of the dielectric loss in Section 3. = 2.3 - j0.005; W / h = 3.09 (50 Q); s / h = 0.56; frequency = 9 GHz;
temperature = 18-20°C
Experimental results show in Fig. 11 that for the poly- ~ Weiner's upper bound
guide substrate the sensitivity falls from 0.04 dB/% mois- 0 experimental points

192 IEE PROCEEDINGS, Vol. 135, Pt. H , N o . 3, J U N E 1988

 In order to find the effect of the relative permittivity of 4.5 Effect of conducting strip length
the protective layer, two different protective materials In previous calculations the attenuation was given in
have been chosen i.e. RT-Duroid ( E , ~= 10.5 -j0.005) terms of dB per unit cm. In other words, we consider a
and polythene ( E , ~ = 2.18 -jO.OOS). linear relationship between attenuation due to dielectric
Fig. 13 shows that the covering material with a higher loss and the length of the stripline. The same effect has
dielectric constant decreases the level of attenuation to been experimentally observed by Kent [4] in stripline
about 50% when compared with the lower dielectric con- configurations. In the sensing head design this length is
stant. Both experimental and theoretical results show a properly chosen in order to keep the attenuation within
better sensitivity for the higher dielectric constant the dynamic range of most measuring instruments used
material. (typically N O dB).

T2 S'i 2

c d
Fig. 14 Test chamber and its simplification to a two-port network

IEE PROCEEDINGS, Vol. 135, t't. H , N o . 3, J U N E 1988 193

5 Experimental set-up for measurement of process, however, is to maximise the sensitivity of the
attenuation sensor with respect to the constraints in its geometrical
and electrical parameters.
The semi-infinite double-covered microstrip configu- There is a comprehensive review in the literature [21,
ration was constructed as in Fig. 14a where the semi- 221 of computer-oriented optimisation methods and their
infinite thickness of liquid is restricted by the perspex application to microwave circuit design. In our opti-
chamber. misation process the sensitivity S is the multivariable
This test chamber consists of three parts, the coupling nonlinear function
system representing the transition between coaxial and
microstrip line, the open microstrip line and the semi- s(x1, x2, ... x N ) (224
infinite double-covered microstrip. The test chamber can
Subject to the nonlinear inequality constraints
be represented by a two-port network as shown in
Fig. 14b. This can be simplified to Fig. 144 which can be < X , < H, k = 1, 2 ..., M
G, (224
used to calculate the insertion loss and attenuation. All
variables associated with the scattering parameters and where xl, x 2 , x3 . . ., x N are the explicit independent vari-
dissipation factor are available theoretically except for ables which represent crl, c r 2 , erg, frequency, length of
the reflection coefficients owing to the transition between conducting strip, etc. and XN+l, X N + 2 ..., X , are the
coaxial and microstrip, which have to be determined implicit variables and dependent functions of the explicit
experimentally. These parameters have been measured variables xl, x2 . . . , x N. In our case, the implicit variable
using a network analyser. The attenuation of the would be the attenuation of the line which covers the
chamber was determined experimentally by the substitut- dynamic range of the device.
ion method [20]. The optimisation of S subject to the constraints of eqn.
22b was carried out using the sequential search method
of Box [23] which has proved in practice to be efficient
6 Optimisation for nonlinear objective functions.
The close agreement between the computed and experi- The variation of sensitivity and total attenuation with
mental results obtained for the attenuation and sensi- respect to slh, Wlh and length 1 are shown in Figs. 15 and
tivity in the microstrip sensor provides further
hevea latex
encouragement to apply an optimisation process to its
W
design. The ultimate objective in the optimisation E r2

b
b
Fig. 16 Isometric configuration for the sensitivity of the microstrip
Fig. 15 Isometric configbration for maximum total attenuation of sensor in the hevea latex solution
microstrip sensor in the hevea latex solution mixture model: Kraszewski
shaded area is for attenuation 3 !O
: dB length = 2.5 cm;E,* = 2.15 j0.005; frequency = 9.0 GHz. temperature = 18°C;
~

mixture model: Kraszewski range of moisture = 70-100%

length = 2.5 cm; E , , = 2.3 - jO.005; E , ~= 2.18 - j0.005; frequency = 9.0 GHz; n E , ~= 2.3 j0.005
~

temperature = 18°C; range of moisture = 70-100% b E,, = 10.5 - jO.005; W / h = 0.92 (50 f2)

194 IEE PROCEEDINGS, Vol. I35, Pt. H , N o . 3, J U N E 1988

16. Figs. 15a and 16a show the variation with slh and 5 STEELE, D.J., and KENT, M.: ‘Microwave stripline techniques
Wlh, Figs. 15b and 16b show the variation with slh and 1. applied to moisture measurement in food materials’, 13th Micro-
wave Power Symposium, Ottawa, Canada, 1978
All the parameters and models involved in the calcu- 6 KENT, M.: ‘The use of stripline configurations in microwave mois-
lation of sensitivity and attenuation are indicated, along ture measurements 11’, J . Microwave Power, 1973,8, pp. 198-194
with the Figure concerned. Both Figures clearly show 7 KENT, M., and PRICE, T.E.: ‘Compact microstrip sensor for high
that sensitivity and total attenuation increase as s/h moisture content materials’, J . Microwave Power, 1979, 14, pp.
decreases. For fixed values of slh, sensitivity and attenu- 363-365
8 BAHL, I.J., and STUCHLY, S.S.: ‘Analysis of a microstrip covered
ation go up slowly as W / h and I increase. From Figs. 15a with a lossy dielectric’, IEEE Trans., 1980, MlT-28, pp. 104-109
and 15b we can determine approximately the optimised 9 YAMASHITA, E., and MITTRA, R.: ‘Variational method for the
values of slh and Wlh for a total attenuation 220 dB to analysis of microstrip lines’, IEEE Trans., 1968, MTT-16, pp. 251
be 0.13 and 1.O, respectively. 10 YAMASHITA, E.: ‘Variational method for the analysis of
microstrip-like transmission lines’, IEEE Trans., 1968, MTT-16, pp.
529-535
1 1 VON HIPPLE, A.R.: ‘Dielectric materials and applications’ (John
Wiley & Sons, New York, 1954), p. 20
7 Conclusions 12 GUPTA, K.C., GARG, R., and BAHL, I.J.: ‘Microstrip lines and
slotlines’ (Artech House, Inc., Dedlam, Massachusetts, USA, 1979).
In this paper a realistic optimisation-based design pp. 68-69
approach for microstrip sensors has been discussed. This 13 KRASZEWSKI, A.: ‘Microwave instrumentation for moisture
content measurement’, J. Microwave Power, 1973,8, pp. 323-325
approach is superior to previously published trial-and- 14 SURESH, N., CALLAGHAN, J.C., and CREELMAN, A.E.:
error methods in the literature. ‘Microwave measurement of the degree of binding of water
A reasonably close agreement has been found between absorbed in soils’, J . Microwave Power, 1967,24, pp. 129-137
the computed and experimental results of the attenuation 15 KRASZEWSKI, A,, KULINSKI, S., and MATUSZEWSKI, M.:
and sensitivity, with respect to the effects of the geometri- ‘Dielectric properties and a model of biphase water suspension at
9.4 GHz’, J . Appl. Phys., 1976,47, pp. 127S1277
cal and electrical parameters of the semi-infinite double- 16 KRASZEWSKI, A.: ‘Determination of the strength of water suspen-
covered microstrip. The results confirm that improved sions using a microwave bridge technique’, J. Microwave Power,
results for the sensor are obtained by selecting a higher 1974,9, pp. 295-302
operating frequency, a lower thickness of protective layer 17 KUHN, N.: ‘Simplified signal flow graph analysis’, Microwaoe J . ,
and a higher dielectric constant of the substrate and pro- 1963,6, pp. 59-66
18 COOK, H.F.: Brit. J. Applied Physics, 1952, 3, p. 249
tective layer. 19 GRANT, E.H., BUCHANAN, T.J., and COOK, H.F.: ‘Dielectric
A computer program has also been developed which behaviour of water at microwave frequencies’, J. Chem. Phys., 1957,
optimises the microstrip sensor sensitivity for a given 26, p. 156
water content. 20 WARNER, F.L.: ‘Microwave attenuation measurement’, IEEE
Monograph Series 19 (Peter Peregrinus Ltd., Herts, 1977), pp.
1 1 1-1 18
21 BANDLER, J.W.: ‘Computer optimisation of microwave circuits’,
Proceedings of European Microwave Conference, Stockholm, 1971,
8 References BSjS, pp. 1-8
22 BANDLER, J.W.: ‘Optimisation methods for computer aided
1 CHIN, H.C.: ‘Method of measuring the dry rubber content of field design’, IEEE Trans., 1969, MlT-17, pp. 533-552
latex’, RRIM Training Manual on Analytical Chemistry, Rubber 23 BOX, M.J.: ‘A new method of constrained optimisation and a com-
Research Institute of Malaysia, Kuala Lumpur, 1979, p. 63 parison with other methods’, Computer J., 1965,8, pp. 42-52
2 THOMAS, R.L., PHANG SEW,. MOK, C.K, CHAN, K.W., 24 HASTED, J.B.: ‘Aqueous dielectrics’ (Chapman and Hall Ltd.,
EASSAU, P.T. and NG, S.C.: ‘Fruit ripening in the oil-palm Elaeis London, 1973). pp. 23-25
Guineensis’, Ann. Botany, 1979, 35, pp. 1219-1225 25 KHALID, K.B::- ‘Analysis and design of microstrip and coplanar
3 CORLEY, R.H.V., HARDON, J.J. and WOOD, B.J.: ‘Oil palm stripline sensors’, Ph.D. thesis, University of Birmingham, England,
research-developments in crop science 1’ (Elsevier Scientific Publi- 1986
shing Company, Amsterdam, 1976), pp. 4 7 W 7 4 26 PETHIG, R., and KELL, D.B.: ‘The passive electrical properties of
4 KENT, M.: ‘The use of stripline configurations in microwave mois- biological systems: their significance in physiology, biophysics and
ture measurement’, J. Microwuve Power, 1973,7, 185-195 biotechnology’, Phys. Med. Biol., 1987,32, pp. 933-970

IEE PROCEEDINGS, Vol. 135, l ’ t . H , N o . 3, J U N E 1988 195