electronics

Article
Parameter Extraction for Equivalent Circuit Model of
RF Devices Based on a Hybrid Optimization Method
Zhimin Guan, Peng Zhao *, Qizhong Lao, Xianbing Wang and Gaofeng Wang *
Key Lab of RF Circuits and Systems of Ministry of Education, School of Electronics and Information,
Hangzhou Dianzi University, Hangzhou 310018, China; [email protected] (Z.G.);
[email protected] (Q.L.); [email protected] (X.W.)
* Correspondence: [email protected] (P.Z.); [email protected] (G.W.)




Received: 6 September 2019; Accepted: 28 September 2019; Published: 8 October 2019


Abstract: One major difficulty of parameter extraction for an equivalent circuit lies in applying an
optimization algorithm to efficiently approach the global minimum. Traditional genetic algorithm
(GA) is able to find the global minimum but it has very low convergence rate. On the other hand,
direct search method may be easily trapped in local minima. In this work, an improved parameter
extraction method which is capable of effectively performing the global minimum search is proposed.
This method combines GA and Powell’s method to efficiently determine circuit elements of an
equivalent circuit model. A spiral inductor and an interdigital capacitor are used as design examples
to illustrate the extraction procedure and validate the proposed hybrid method.

Keywords: parameter extraction; Powell’s method; genetic algorithm; equivalent circuit; spiral
inductor; interdigital capacitor

1. Introduction
Radio-frequency (RF) devices such as spiral inductors and microstrip interdigital capacitors
have been widely used in low noise amplifiers, power amplifiers and voltage-controlled oscillators;
and many other RF devices are used in silicon-based radio-frequency integrated circuits (RFICs) [1].
Accurate equivalent circuit models for these components are highly desirable for RFIC designs.
For a given device layout, an equivalent circuit can be constructed by a parasitic extraction tool.
In general, the parasitic extraction tool is based on quasi-static analysis, which makes the equivalent
circuit inaccurate at high frequencies. To improve the accuracy of the equivalent circuit model, it is
a common practice to adjust the circuit elements of the equivalent circuit model by using accurate
full-wave simulation results [2,3].
In practice, some RF devices can use simple T-, π- or ladder models as their equivalent circuits.
For relatively complex RF designs, however, one may build better but more complex equivalent
circuits based on their layouts with physical insights. For example, the equivalent circuits of the
aperture-coupled resonator frequency-selective surface (FSS) [4], spiral inductors [5,6], and interdigital
capacitors [7] have been constructed with physical insights. Once the equivalent circuit for a RF device
is determined, the element parameters of the equivalent circuit can be extracted based on the full-wave
results by optimization methods [8].
In order to build accurate equivalent circuit models, reliable and effective parameter extraction
methods are required. In general, the values of circuit element parameters can be determined by genetic
algorithm (GA) [9,10], which is an effective global and gradient-free optimization method. However,
for the optimization problem of continuous function (e.g., the parameter extraction), the convergence
rate of GA is slow, and particularly it is very ineffective for local optimization. More seriously, it is
highly reliant on initial population [11]. If the initial population is not set appropriately, GA may

Electronics 2019, 8, 1133; doi:10.3390/electronics8101133 www.mdpi.com/journal/electronics

Electronics 2019, 8, 1133 2 of 14

be unable to find appropriate results. If a small population is used, the solution set will have small
population diversity, and it may tend to cause the premature convergence problem [12]. As a result,
the optimization solution may be inaccurate. Although some methods have been proposed to handle
the premature convergence problem [13], the convergence rate of GA is hardly improved. In addition,
if a large population is used, the optimization results may be more accurate, but the convergence rate
becomes very slow. On the other hand, a conjugate gradient (CG) optimization algorithm, such as
Powell’s method [14] combined with a one-dimensional search method (e.g., golden section search [15]),
has strong local optimization ability and fast convergence rate [16], but the drawback of trapping in
local minima restricts its applications.
In this work, a parameter exaction method based on a hybrid GA–Powell’s method is proposed.
The solution can be close to the global minimum after applying GA. Then, the best individual of GA
is set as the start point of the local optimization process of Powell’s method. Determining when to
start the local optimization process is important. If the local optimization process starts too early, the
hybrid optimization method may easily trap at a local minimum. On the other hand, if it starts too
late the hybrid optimization method may waste some unnecessary time on the global search process.
To save the time on the global search process, the iteration number of the global search process is
brutally reduced from 500 to 100 in a handbook [17]. However, it cannot be guaranteed that the result
from 100 iterations of GA is an appropriate starting point of the local optimization process since the
convergence rate differs from case to case. To effectively determine the appropriate starting point of the
local optimization process, the variance of the fitness of population is used as a criterion. Using this
hybrid optimization algorithm, an accurate solution of the global minimum is achievable. This method
combines the merits of the global search ability of GA and the local search capability of Powell’s
method with good accuracy and rapid convergence. Hence, this hybrid method is very effective in the
parameter extraction of equivalent circuits.
To efficiently obtain the full-wave simulation results for a given device layout, a highly-efficient
full-wave EM simulation tool, called UltraEM® [18], is employed to generate the Y-parameters of the
device layout. These full-wave Y-parameters are set as the optimization objectives for the proposed
hybrid GA–Powell’s method to adjust values of the circuit elements so that the difference between the
full-wave Y-parameters and the Y-parameters of the equivalent circuit (which can be calculated by a
circuit simulation tool [19]) is globally minimized.
The proposed method may have some limitations: Firstly, although this hybrid optimization
method accelerates the parameter extraction of RF devices, it still cannot solve the problem of parameter
extraction in a very wide frequency band, since RF devices may have different properties in different
frequency bands [20]. That is, one specific equivalent circuit model may not be able to match the
physical model of a certain RF device in its entire frequency band. Secondly, the efficiency of the
hybrid optimization method is affected by its initial population of GA. If the initial population is
inappropriately set, the proposed method may not be able to find the appropriate global minimum.
The rest of this paper is organized as follows. In Section 2, the hybrid optimization method
based on GA and Powell’s method is introduced. The efficiency of this hybrid method is then tested
and verified by modeling a planar spiral inductor and a microstrip interdigital capacitor in Section 3.
A brief conclusion is given in Section 4.

2. Hybrid Optimization Method

The proposed hybrid optimization method can be applied for modelling components of RFICs.
For example, as shown in Figure 1, the layout of a planar spiral inductor is synthesized using technique
files, constrains files and so on. The EM simulation results (Figure 1c) of the planar spiral inductor is
used as the target of the equivalent circuit. The value of every equivalent circuit (Figure 1d) element is
able to be extracted from the proposed method effectively, and the model of the equivalent circuit can
be applied for circuit design.
Electronics 2019, 8, 1133 3 of 14

The process of the proposed hybrid GA–Powell’s method can be divided into two parts: global
search process and local optimization process.

Figure 1. Parameter extraction application of hybrid optimization method.

2.1. Global Search Process

GA is able to search in multiple regions among a solution space simultaneously. Thus, it reduces
the possibility of the algorithm falling into local minima, and makes the GA have better global
search ability.
Roulette wheel selection [21] is herein used as the selection method of GA. Assuming the
population size is N and the fitness of the individual xi is f (xi ), the probability of selection of individual
xi is defined as:
f (xi )
P(xi ) = PN , (1)
j = 1 f ( xi )

and the cumulative probability is calculated by:

Xi
qi = P(xi ). (2)
j=1

Then, a random number r with a uniform distribution of 0–1 is generated. If qi−1 ≤ r ≤ qi ,

the individual xi is selected to enter a new group of the next generation. This process will be repeated
until the number of individuals in the new group is equal to the size of the parent group.
The crossover process [22] is based on the following two formulas:
 t+1
 xA = αxB + (1 − α)xA
 t t

(3)
 xt+1 = αxt + (1 − α)xt


B A B

where xtA+1 and xtB+1 are individuals after the crossover process, xtA and xtB are two individuals randomly
selected, and α is a constant of the crossover with a value between 0 and 1 (α is herein set as 0.8).

1
Electronics 2019, 8, 1133 4 of 14

The mutation process [23] is based on:

 xA + k(xmax − xA )r, pick > 0.5

 t t

xtA+1

= (4)
 xt − k(xt − xmin )r, pick ≤ 0.5

A A

where xtA+1 is the individual after the mutation, xtA is the individual before the mutation, k is a constant
of the variation with a value between 0 and 1 (herein, k is set as 0.1 to maintain the diversity of
population), xmax is the upper limit of the individual, xmax is the lower limit of the individual, and r is a
random number. Note that pick is a random number with a uniform distribution in the range of 0 to 1.
When pick > 0.5, choose the first formula; when pick ≤ 0.5, choose the second formula [24].
With the iterations of GA, the solution can gradually converge to the global minimum. To accelerate
the convergence, a local optimization process is applied when the convergence rate of GA slows down.
To determine when to start the local optimization process, the variance of the fitness of population is
defined as:
1 Xn
var( f it) = ( f iti − f it) (5)
n−1 i=1

When the optimization converges to an optimum solution, the variance of the fitness for a
population is small. Otherwise, the population is scattered in the solution space, and the variance
of the fitness is large. When the variance of the fitness is small, the convergence rate consequently
slows down, in which it may take many iterations for the mutation process of GA to generate a new
individual that has a better fitness than the previous generations. Hence, the local optimization is
applied, once the variance of the fitness defined by Equation (5) is small. The CG search method is the
most effective way to solve a local optimization problem.

2.2. Local Optimization Process

After global search of GA, the individual with best fitness is chosen as starting point of Powell’s
(1)
method, which is defined as X0 .

(1)
(1) Set an optional initial point X(0) = X0 , and set the convergence thresholds as ε1 and ε2 (herein,
ε1 and ε2 are both set as 1 × 10−4 ). Use the initial basic direction group as the unit coordinate
vector, which can be defined as:
(k )
Si = ei (i = 1, 2, . . . , n) (6)

where ei is the i-th unit coordinate vector, and Si represent the i-th search direction, and k denotes
the k-th iteration round.
(k ) (k )
(2) Implement one-dimensional search along Si (i = 1, 2, . . . , n), and generate Xi (i = 1, 2, . . . , n)
(k ) (k )
from Xi−1 (i = 1, 2, . . . , n) at each direction search. The final point Xn will make the function of
(k )
f (Xn ) have the minimum value. Construct a new search direction:

(k ) (k )
S(k) = Xn −X0 . (7)

(k )
Then, perform a one-dimensional search in direction S(k) to generate a minimum point Xn+1 .
(3) If
(k ) (k )
Xn + 1 − X0 < ε1 , (8)

or
(k ) (k ) (k )
f (Xn+1 ) − f (X0 ) < ε2 | f (Xn+1 )|, (9)
(k ) (k )
stop the iteration and export the optimal solution Xn+1 and f (Xn+1 ). Otherwise, go to step 4.
Electronics 2019, 8, 1133 5 of 14

(4) Calculate
(k )
f (Xi ) (i = 1, 2, . . . , n), (10)
(k ) (k ) (k )
∆m = max{ f (Xi−1 ) − f (Xi )}, (i = 1, 2, . . . , n), (11)
(k ) (k )
where m denotes the corresponding index at which f (Xm−1 ) − f (Xm ) is the maximum one
(k ) (k )
among { f (Xi−1 ) − f (Xi )} (i = 1, 2, . . . , n). Then, the direction that contributes the most in the
k-th iteration round is:
(k ) (k ) (k )
Sm = Xm − Xm−1 . (12)

(5) To ensure that the search direction group for each iteration is linearly independent, it is needed to
determine whether the original search direction group can be directly used as the search direction
group in the next iteration. Firstly, determine mapping points:

(k ) (k )
X(k) = 2Xn − X0 , (13)

then calculate f (X(k) ) and set:

(k ) (k )
f1 = f (X0 ), f2 = f (Xn ), f3 = f (X(k) ). (14)

If f3 < f1 and
(k ) 2 (k )
( f1 − 2 f2 + f3 )( f1 − f2 − ∆m ) < 0.5∆m ( f1 − f3 )2 , (15)

go to step 6, and replace the most contributing search direction from Equation (12) with the new
search direction. Otherwise, go to step 7.
(6) Set the starting point and the search direction group of the (k + 1)-th iteration as follows:

(k +1) (k )
X0 = Xn + 1 , (16)
n o
Si (k+1) (i = 1, 2, . . . , n) = S1 (k) , . . . , Sm−1 (k) , S(k) , Sm+1 (k) , . . . , Sn (k) , (17)
(k )
and replace the most contributing search direction Sm with the new search direction S(k) . Then,
set k = k + 1 and go to step 2.
(7) Set the starting point and the search direction group of the (k + 1)-th iteration as follows: If f2 < f3 ,

(k +1) (k )
X0 = Xn . (18)

Otherwise
(k +1)
X0 = X (k ) , (19)

Si (k+1) = Si (k) (i = 1, 2, . . . , n). (20)

Set k = k + 1, and go to step 2 [14].

After the local optimization process, the solution is able to converge to an appropriate and accurate
global minimum. This hybrid method combines the advantages of the global search ability of GA and
the high accuracy and fast convergence of Powell’s method in the local search.

2.3. RF Device Modeling

The modeling process is shown in Figure 2. The full-wave simulation result of the layout of an RF
component is obtained by UltraEM® , and the corresponding Y-parameters are labeled as YEM . By using
Electronics 2019, 8, 1133 6 of 14

a Spice simulator, the Y-parameters of the equivalent circuit can be obtained from the calculated port
voltage and current, viz:
I = YV, (21)

where Y is an N × N matrix, and V and I are the port voltage vector and the port current vector,
respectively. The entry of the Y-matrix is defined as:

In
Ynm = | f or k,m . (22)
Vm Vk =0

If the voltage of the excited port is set as 1 volt, the Y-parameters are equal to the port currents, viz:

Electronics 2019, 8, x FOR PEER REVIEW Ynm = In . 6 (23)

of 14

Labelthe
Label theY-parameters
Y-parameterssimulated
simulatedfrom
fromthe
theequivalent
equivalent circuit
circuit asas 𝑌 . In. In
Yspice order
order to the
to fit fit the result
result of
of the
the equivalent
equivalent circuit
circuit withwith
the the accurate
accurate EMEM simulation,
simulation, the the objective
objective function
function tooptimized
to be be optimized cancan
be
be defined
defined as: as: X X feX N N
ob j(Yspice , YEM ) = (YEMmn )i − (Yspicemn )i , (24)
𝑜𝑏𝑗 𝑌 ,𝑌 = ∑
i = f s ∑ m =∑
1 n(𝑌
=1 ) − (𝑌 ) , (24)

where NN isis the

where the port
port number
number of
of the device; fsfs and
the device; and fefe represent
represent the
the start
start and
and the
the end
end frequencies
frequencies ofof
simulation,
simulation, respectively. To verify the effect of the method proposed in Section 2, a planar
respectively. To verify the effect of the method proposed in Section 2, a planar spiral
spiral
inductor
inductorand
andaamicrostrip
microstripinterdigital
interdigitalcapacitor
capacitorwereweretested.
tested.

RF device
Randomly set the value of
components of the equivalent
circuit to initialize chromosome
Layout model Equivalent circuit
simulated by UltraEM netlist format
Optimized by GA to find Hybrid
solution set that close to global optimization
Y-parameters minimum method
Y-parameters in
calculated from
certain frequency label
equivalent circuit label
as YEM Using the best individual of GA
as Yspice
as the starting point of Powell
method to find appropriate result
Set objective function

Figure 2. Flow chart model of the hybrid GA–Powell’s method.

Figure 2. Flow chart model of the hybrid GA–Powell’s method.
3. Results and Discussions
3. Results and Discussions
In order to illustrate the effectiveness of this hybrid method, a planar spiral inductor and a
In order
microstrip to illustrate
interdigital the effectiveness
capacitor are used as of this hybrid
examples. Themethod,
interested a planar spiral
frequency inductor
range and a
is assumed
microstrip interdigital capacitor are used as examples. The interested frequency
to be from 0.1 GHz to 2.3 GHz. Both examples are modeled by (1) pure GA with 500 individuals; range is assumed to
be pure
(2) fromGA 0.1 with
GHz1000to 2.3 GHz. Bothand
individuals; examples are modeled
(3) Hybrid GA–Powell’sby (1) pure GA
method with 500
(including 500individuals;
individuals (2)of
pure GA with 1000 individuals; and (3) Hybrid GA–Powell’s method (including 500
GA). In order to ensure accuracy and efficiency of the global search process in the hybrid optimization individuals of
GA). In order to ensure accuracy and efficiency of the global search process in the hybrid
method, the initial population of GA in the hybrid optimization is set as 500. In order to validate optimization
method, and
accuracy the efficiency
initial population of GA
of the hybrid in the hybrid
optimization optimization
method, the initialispopulation
set as 500.ofInthe
orderpuretoGAvalidate
is set
accuracy and efficiency of the hybrid optimization method, the initial population of the
as 500 and 1000 for comparison. For all these three cases, the local optimization process is started after pure GA is
set(as
var f it500
) < vand 1000herein.
(0.001) for comparison. For all these three cases, the local optimization process is started
after 𝑣𝑎𝑟(𝑓𝑖𝑡) < v (0.001) herein.

3.1. Planar Spiral Inductor

The planar spiral inductor is fabricated on a substrate with dielectric constant of 11.9 and
thickness of 750 microns. The thickness and conductivity of the metal are 2 microns and 2.8 × 107 S/m
respectively. The layout of the planar spiral inductor is shown in Figure 3, where the geometrical
parameters are shown in Table 1.
Electronics 2019, 8, 1133 7 of 14

3.1. Planar Spiral Inductor

The planar spiral inductor is fabricated on a substrate with dielectric constant of 11.9 and thickness
of 750 microns. The thickness and conductivity of the metal are 2 microns and 2.8 × 107 S/m respectively.
The layout
Electronics of8,the
2019, planar
x FOR PEER spiral
REVIEW inductor is shown in Figure 3, where the geometrical parameters 7 ofare
14
Electronics 2019, 8, x FOR PEER REVIEW 7 of 14
shown in Table 1.

Figure 3. Geometrical layout of the planar spiral inductor.

Figure 3. Geometrical layout of the planar spiral inductor.
Figure 3. Geometrical layout of the planar spiral inductor.
Table 1. Parameters of planar spiral inductor.
Table 1. Parameters of planar spiral inductor.
Table 1. Parameters of planar spiral inductor.
Name Values Units
Name
Name Values
Valueswidth Units
W Conductor 13 um Units
W
W Conductor
Conductor width
width 13
13 um
um
S Conductor spacing 7 um
S
S Conductor
Conductor spacing
spacing 773 um
N Number of turns integer um
N
N L1 Number
Number
Length of turns
of turns
of second outermost segment 33
300 integer
um integer
L1
L1 Length
L2 of second
Length outermost
of outermostsegment
segment
Length of second outermost segment 300
300
300 um um
um
L2
L2 Length of
Length of outermost
outermost segment
segment 300
300 um
um
The lumped
The lumped equivalent
equivalentcircuit
circuitofofthe
the planar
planar spiral
spiral inductor
inductor [6]be
[6] to
to tomodelled
be modelled by hybrid
by hybrid
hybrid GA–
The lumped equivalent circuit of the planar spiral inductor [6] be modelled by GA–
GA–Powell’s method
Powell’s method
method is is
is shownshown
shown in in
in FigureFigure
Figure 4.4. 4.
Powell’s
Cs
Cs

Ls
Ls Rs
Rs

Cox1 Cox2
Cox2
Cox1

Rsi1 Csi1 Rsi2

Csi1 Rsi2 Csi2
Csi2
Rsi1

Figure 4. Equivalent circuit of the planar spiral inductor.

Figure 4. Equivalent circuit of the planar spiral inductor.
Figure 4. Equivalent circuit of the planar spiral inductor.
With the three aforementioned settings, the errors of objective function for the planar spiral
With the
With the three
three aforementioned
aforementioned settings,
settings, the
the errors
errors ofof objective
objective function
function for the the planar spiralspiral
inductor are plotted in Figure 5, whereas the final errors and time costs are listed for
in Tableplanar
2. The initial
inductor are
inductor are plotted
plotted in in Figure
Figure 5,5, whereas
whereas the the final
final errors
errors and
and time
time costs
costs are listed
listed in Table
Table 2. 2. The
The
population of the global search process in the hybrid GA–Powell’s method isarethe samein as the pure GA
initial
initial population
population of
of the
the global
global search
search process
process in
in the
the hybrid
hybrid GA–Powell’s
GA–Powell’s method
method is
is the
the same
same as
as the
the
method with 500 individuals. From Figure 5, one can see that the convergence rate of the pure GA
pure GA
pure GA method
method withwith 500 individuals.
individuals. From FigureFigure 5, one one can
can see
see that
that the
the convergence
convergence rate rate of thethe
optimization gradually 500 slows down afterFrom 22 iterations.5,For the hybrid GA–Powell’s method, theoflocal
pure
pure GA optimization
GA optimization gradually slows down after 22 iterations. For the hybrid GA–Powell’s method,
optimization is appliedgradually slows down
after the global search.after 22 iterations. For the hybrid GA–Powell’s method,
the local
the local optimization
optimization is is applied
applied after
after the
the global
global search.
search.
In Table 2, it can be observed that for the hybrid GA–Powell’s method with seven iterations of the
In Table
Table 2,
2, it
it can
can be be observed that that for
for the
the hybrid
hybrid GA–Powell’s
GA–Powell’s method
method with
with seven
seven iterations of of
localIn optimization processobserved
after the global search, the error finally drops to 1.4157 × 10−2 .iterations
In contrast,
the local optimization process after the global search, the error finally drops
the local optimization process after the global search, the error finally drops to 1.4157 × 10−2. In to 1.4157 × 10 −2. In

contrast, when
contrast, when using
using 50 50 iterations
iterations ofof pure
pure GAGA with
with 500500 and
and 1000
1000 individuals,
individuals, the
the errors
errors are
are still
still 1.7347
1.7347
×× 10
10−1 and
and 2.03798
2.03798 ×× 1010−2,, respectively.
respectively. Moreover,
Moreover, the the total
total times
times costed
costed by
by the
the aforementioned
aforementioned process process
−1 −2

for the three settings are 574 s, 1197 s and 378 s,

for the three settings are 574 s, 1197 s and 378 s, respectively. respectively.
To validate
To validate the
the accuracy
accuracy of of the
the proposed
proposed method,
method, the the S-parameters
S-parameters calculated
calculated by by the
the full-wave
full-wave
simulation of the layout using UltraEM®® and the circuit simulation of the equivalent circuit extracted
Electronics 2019, 8, 1133 8 of 14

when using 50 iterations of pure GA with 500 and 1000 individuals, the errors are still 1.7347 × 10−1
and 2.03798
Electronics × 8,10x −2
2019, FOR, respectively.
PEER REVIEW Moreover, the total times costed by the aforementioned process 8 of 14 for
the three settings are 574 s, 1197 s and 378 s, respectively.
shownTo validate
in Tablethe accuracy
3. The circuitof the proposed
simulation method,
results the S-parameters
using setting calculated
1 cannot match by theresults,
the full-wave full-wave
simulation of the layout using UltraEM ® and the circuit simulation of the equivalent circuit extracted
but the results using setting 3 agree very well with the full-wave results.
Electronics 2019, 8, x FOR PEER REVIEW 8 of 14
by the three optimization methods are compared in Figure 6, while the extracted parameters are shown
in Table 3. The circuit simulation results using setting 1 cannot match the full-wave results, but the
shown in Table 3. The circuit simulation results using setting 1 cannot match the full-wave results,
results using setting 3 agree very well with the full-wave results.
but the results using setting 3 agree very well with the full-wave results.

Figure 5. Error of objective function of the planar spiral inductor. Solid line represents the error of
Hybrid GA–Powell’s method. Dash-dot line represents error of pure GA with population of 500. Dash
line represents error of pure GA with population of 1000.

Table 5.
Figure
Figure 2.5.Error
Error of
ofobjective
Errorafter function
optimization
objective of
of the
and time
function planar
cost
the spiral inductor.
inductor.
of optimization
planar spiral Solid
algorithm
Solid linethe
with
line represents the
examplethe
represents of errorofof
planar
error
Hybrid
spiral GA–Powell’s
inductor. method. Dash-dot line represents error of pure GA with population
Hybrid GA–Powell’s method. Dash-dot line represents error of pure GA with population of 500. Dash of 500. Dash
line represents error of pure GA with population
line represents error of pure GA with population of 1000.of 1000.
Number of
Table 2. Error after optimization and time cost of optimization algorithm with the example of planar
Table 2. Error after optimization and time cost Error after
of optimization Iterations
algorithm with the example of planar
spiral inductor.
Optimization Algorithm Time Cost
spiral inductor.
Optimization Powell’s
Error after GA of Iterations
Number
Optimization Algorithm NumberMethodof Time Cost
Optimization
GA Powell’s Method
Pure GA with 500 population size 1.7347 10−1
Error×after 50 IterationsNA 574 s
Pure GAOptimization
with 500 population
Algorithmsize 1.7347 × 10−1 −2 50 NA Time Cost574 s
Pure GA with 1000 population size 2.03798 × 10
Optimization 50 NA
Powell’s 1197 s
Pure GA with 1000 population size 2.03798 × 10−2 50 NA 1197 s
GA
Hybrid GA–Powell’s
Hybrid GA–Powell’s method
method × 10−2
1.41571.4157 × 10−2 22 22 77
Method 378 s 378 s

Pure GA with 500 population size 1.7347 × 10−1 50 NA 574 s

Pure GA with 1000 population size 2.03798 × 10−2 50 NA 1197 s

Hybrid GA–Powell’s method 1.4157 × 10−2 22 7 378 s

(a) (b)
Figure6.6.Comparison
Figure Comparison ofof S-parameters of
of the
theplanar
planarspiral
spiralinductor
inductorbetween
between UltraEM
UltraEM ® simulation
® simulation andand
equivalentcircuit
equivalent circuitsimulation
simulation modeled
modeled bybypure
pureGAGAwith
withpopulation
populationofof
500, pure
500, pureGAGA
with population
with population
ofof1000
1000andandhybrid
hybridGA–Powell’s
GA–Powell’s method.
method. (a) (a) Amplitude
Amplitudecomparison
comparisonofofS-parameters.
S-parameters.(b) (b)
Phase
Phase
comparisonofofS-parameters.
comparison S-parameters.
(a) (b)
Figure 6. Comparison of S-parameters
Table of thespiral
3. Result of planar planar spiral inductor
inductor between
parameters UltraEM® simulation and
extraction.
equivalent circuit simulation modeled by pure GA with population of 500, pure GA with population
of 1000 and hybrid GA–Powell’s method. (a) Amplitude comparison of S-parameters. (b) Phase
comparison of S-parameters.
Electronics 2019, 8, x FOR PEER REVIEW 9 of 14
Electronics 2019, 8, 1133 9 of 14

Circuit Pure GA with Pure GA with

Hybrid Method Units
Population
Components Table 3. Resultofof500
planar Population of parameters
spiral inductor 1000 extraction.
Rs 6.4598 6.4770 6.4896 ohm
Circuit
Ls Pure GA with
7.7692 Pure GA with
7.7662 7.7860 nH
Hybrid Method Units
Components
Cs Population
5.0782 of 500 Population 12.7123of 1000 15.161 fF
Cox1Rs 0.1271
6.4598 0.0496
6.4770 0.0436
6.4896 pF
ohm
Ls
Csi1 7.7692
11.5294 7.7662
10.9931 7.7860
0.0503 nH
pF
Cs
Rsi1 5.0782
18.3767 12.7123
14.0226 15.161
353.17 fF
ohm
Cox1
Cox2 0.1271
0.4024 0.0496
0.0557 0.0436
0.0243 pF
pF
Csi1
Csi2 11.5294
15.6328 10.9931
6.3849 0.0503
0.000668 pF
pF
Rsi1
Rsi2 18.3767
22.0197 14.0226
22.8286 353.17
1104.1 ohm
ohm
Cox2 0.4024 0.0557 0.0243 pF
Csi2 15.6328 6.3849 0.000668 pF
Rsi2 22.0197 22.8286 1104.1 ohm

3.2. Microstrip Interdigital Capcitor

3.2. Microstrip Interdigital Capcitor
The microstrip interdigital capacitor is fabricated on a substrate with dielectric constant of 12.9,
The microstrip interdigital capacitor is fabricated on a substrate with dielectric constant of 12.9,
loss tangent of 0.001, and thickness of 100 microns. The conductivity of the metal is 2.8 ×107 S/m.
loss tangent of 0.001, and thickness of 100 microns. The conductivity of the metal is 2.8× 10 S/m. The
The layout of microstrip interdigital capacitor is shown in Figure 7, where the geometrical parameters
layout of microstrip interdigital capacitor is shown in Figure 7, where the geometrical parameters of
of the parameters are shown in Table 4.
the parameters are shown in Table 4.

Ge W G

L Wt

Figure
Figure 7. Geometry layout of the microstrip interdigital capacitor.
7. Geometry

Table 4. Parameters of microstrip interdigital capacitor.

Table 4. Parameters of microstrip interdigital capacitor.

Name Name Values

Values Units Units
W W FingerFinger
widthwidth 30 30 um um
G G Gap between
Gap between
fingersfingers 30 30 um um
Ge Ge Gap at Gap
end at
ofend of fingers
fingers 30 30 um um
L L
LengthLength of overlapped
of overlapped regionregion 600 600 um um
Np Number of finger pairs 3 integer
Np Number of finger pairs 3 integer
Wt Width of interconnect 30 um
Wt Width of interconnect 30 um

The lumped equivalent circuit of the microstrip interdigital capacitor

capacitor [25] to be modelled by
hybrid GA–Powell’s method is shown in Figure 8a,b. The equivalent
equivalent circuit in Figure 8a is used for
lower frequency applications,
applications, while
while the
the equivalent
equivalent circuit
circuit in
in Figure
Figure 8b
8b is
is used
used for
for higher
higherfrequency.
frequency.
L C R

Cs1 Cs2

(a)
 L Length of overlapped region 600 um
Np Number of finger pairs 3 integer
Wt Width of interconnect 30 um

The lumped equivalent circuit of the microstrip interdigital capacitor [25] to be modelled by
Electronics 2019, 8, 1133
hybrid GA–Powell’s method is shown in Figure 8a,b. The equivalent circuit in Figure 8a is 10 of 14
used for
lower frequency applications, while the equivalent circuit in Figure 8b is used for higher frequency.
L C R

Cs1 Cs2
Electronics 2019, 8, x FOR PEER REVIEW 10 of 14
Electronics 2019, 8, x FOR PEER REVIEW (a) 10 of 14

L1 C1 R1 L2 C2 R2
L1 C1 R1 L2 C2 R2

Cs1 Cs2 Cs3

Cs1 Cs2 Cs3

(b)
(b)
Figure
Figure 8. Lumped
8. Lumped equivalent
equivalent circuits
circuits forfor
thethe microstrip
microstrip interdigital
interdigital capacitor.
capacitor. (a)(a) Lower
Lower frequency;
frequency;
Figure 8. Lumped equivalent circuits for the microstrip interdigital capacitor. (a) Lower frequency;
(b) higher frequency.

TheThe interdigital
interdigital capacitor
capacitor is also
is also modelled
modelled using using the aforementioned
the aforementioned threethree settings.
settings. The errors
The errors of
The interdigital
of objective function capacitor
are shown is also modelled
in Figure using the aforementioned three settings. The errors of
objective function are shown in Figure 9. It is9.observed
It is observed that 31
that after after 31 iterations
iterations the convergence
the convergence of the
of the
objective
global functionbecomes
are shown in Figure 9.
theIt hybrid
is observed thatstarts
after the31 iterations the convergence of
global searchsearch
becomes slow, where slow, where
the hybrid method method
starts the local optimization
local optimization process process
and can and
thecan
global search becomes slow, where the hybrid method starts the local optimization process and
rapidlyrapidly
converge converge
to a small to aerror.
small error.
can rapidly
The converge
final errors toand
a smalltime error.
costs
The final errors and time costs areare shown
shown in in Table
Table 5. In
5. In thethe hybrid
hybrid GA–Powell’s
GA–Powell’s method,
method, eight
eight
The final
iterations of errors
the and
local time costs
optimization are shownare
process in Table 5. In theafter
implemented hybrid 30 GA–Powell’s
iterations of method,
the global eight
search
iterations of the local optimization process are implemented after 30 iterations of the global search
iterations
process, of
andthe local optimization process are implemented afterwith
30 iterations of the globalthe search
process, and thethe error
error finally
finally drops
drops to 1.45255
to 1.45255 × 10 × −4
10.−4Compared
. Compared with thethe pure
pure GA GA in in which
which errors
the errors
process,
using and
50 the error with
iterations finally 500drops
and to
10001.45255 × 10 . Compared
individuals
−4
are 8.324 with
×−310 −3the
and pure1.16 GA× in−3which
10
−3 , the errorsthe
respectively,
using 50 iterations with 500 and 1000 individuals are 8.324 × 10 −3 and 1.16 × 10 −3 , respectively, the
using 50 iterations
results with 500 and 1000 individuals are 8.324to× the
10 total
andtime 1.16cost,
× 10 because
, respectively, the
results of of
thethe proposed
proposed method
method areare
muchmuch better.
better. In In regard
regard to the total time cost, because thethe proposed
proposed
results
methodof theonlyproposed
spends method are much better. Initregard to the412
total time cost, because theGAproposed
method only spends 3838 iterations
iterations to to converge,
converge, costs
it costs onlyonly
412 s. In s. In contrast,
contrast, thethe
purepureGA methodsmethods
method
using only
50 spends
iterations 38
with iterations
500 and to converge,
1000 it
individuals costs only
spend 412
529 s s.
andIn contrast,
1079 s, the pure
respectively, GA yet methods
still have
using 50 iterations with 500 and 1000 individuals spend 529 s and 1079 s, respectively, yet still have
using 50
larger iterations
errors than with 500 and 1000 individuals spend 529 s and 1079 s, respectively, yet still have
larger errors than thethe proposed
proposed method.
method.
larger errors
The than the parameters
extracted proposed method. using these three methods
The extracted parameters using these three methods areare listed
listed in in Table
Table 6. The
6. The full-wave
full-wave results
results
andThetheextracted
circuit parameters
simulation usingare
results these three
both shown methods
in are listed
Figure 10. in Table
The circuit 6.simulation
The full-wave results
results using
and the circuit simulation results are both shown in Figure 10. The circuit simulation results using
and the
setting circuit
1 cannot simulation
match results are both shown in Figure 10. The circuit simulation results using
setting 1 cannot match thethe full-wave
full-wave results.
results. ForFor thethe results
results of of setting
setting 2, the
2, the amplitude
amplitude of Sof11Scannot
11 cannot
setting
match 1 the
cannot matchresults.
full-wave the full-wave
In sharp results.
contrast,Forthe theresults
resultsusing
of setting
setting 2, 3the amplitude
agree very wellofwith
S11 cannot
thethefull-
match the full-wave results. In sharp contrast, the results using setting 3 agree very well with
match
wave the full-wave
results. The results.
accuracy Inofsharp
the contrast,
hybrid the results using
GA–Powell’s method setting
is 3 agree very
validated once well
again.with the full-
full-wave results. The accuracy of the hybrid GA–Powell’s method is validated once again.
wave results. The accuracy of the hybrid GA–Powell’s method is validated once again.

Figure 9. Error of objective function of the microstrip interdigital capacitor. The logarithmic
Figure
Figure 9.9.Error
Error
coordinates
of objective
areofused.
function
objective of theof
function
Solid line
microstrip
represents
interdigital
thethemicrostrip capacitor. capacitor.
interdigital
error of Hybrid
The logarithmic
GA–Powell’s method.
coordinates
The logarithmic
Dash-dot line
are used. Solid line
coordinates represents the error of Hybrid GA–Powell’s method. Dash-dot line represents error
representsare used.
error of Solid
pure line
GA represents the error
with population of Hybrid
of 500. GA–Powell’s
Dash line representsmethod.
error ofDash-dot
pure GAlinewith
of pure GAerror
represents with population
of pure GA ofwith
500. population
Dash line represents
of 500. errorline
Dash of pure GA with
represents population
error of pureofGA
1000.
with
population of 1000.
population of 1000.
Table 5. Error after optimization, number of iterations and time cost of optimization algorithm with
Table 5. Error after
the example optimization,
of planar number of iterations and time cost of optimization algorithm with
spiral inductor.
the example of planar spiral inductor.
Number of Iterations
Error after Number of Iterations
Optimization Algorithm Error after Powell’s Time Cost
Optimization Algorithm Optimization GA Powell’s Time Cost
Electronics 2019, 8, 1133 11 of 14

Table 5. Error after optimization, number of iterations and time cost of optimization algorithm with
the example of planar spiral inductor.

Error after Number of Iterations

Optimization Algorithm Time Cost
Optimization
GA Powell’s Method
Pure 2019,
Electronics GA with population
8, x FOR of 500
PEER REVIEW 8.324 × 10−3 50 NA 529 s11 of 14
Pure GA with population of 1000 1.16 × 10−3 50 NA 1079 s
Hybrid GA–Powell’s method 1.45255 × 10−4 30 8 412 s

(a) (b)

(c) (d)
Figure 10. Comparison
Figure 10. Comparison of of S-parameters
S-parameters of of the
the microstrip
microstrip interdigital
interdigital capacitor
capacitor between UltraEM®
between UltraEM®
simulation and
andequivalent
equivalentcircuit
circuitsimulation
simulation modeled
modeled byby pure
pure GAGA with
with population
population of 500,
of 500, purepure GA
GA with
with population
population of 500ofand
500hybrid
and hybrid GA–Powell’s
GA–Powell’s method.
method. (a) Amplitude
(a) Amplitude comparison
comparison of S-parameters
of S-parameters S11;
(b) amplitude
S11; comparison
(b) amplitude of S-parameters
comparison S21; (c)
of S-parameters phase
S21; comparison
(c) phase of S-parameters
comparison S11; (d)
of S-parameters phase
S11; (d)
comparison
phase of S-parameters
comparison S21. S21.
of S-parameters

Table 6. Results of microstrip interdigital capacitor parameters extraction.

Table 6. Results of microstrip interdigital capacitor parameters extraction.

Circuit
Circuit Pure GAGA
Pure with
with PureGA
Pure GA with
with
Hybrid Method
Components Population of 500 Population of 1000 Hybrid MethodUnits
Units
Components Population of 500 Population of 1000
Cs1 Cs1 0.1929
0.1929 0.1831
0.1831 0.1829
0.1829 pF pF
L L 11.5672
11.5672 1.2317
1.2317 0.5412
0.5412 nH nH
C C 0.1030
0.1030 0.1289
0.1289 0.1304
0.1304 pF pF
R R 8.3133
8.3133 5.3290
5.3290 1.4035
1.4035 ohmohm
Cs2 0.1952 0.1821 0.1831 pF
Cs2 0.1952 0.1821 0.1831 pF

3.3. More Discussions

To determine the appropriate starting point of the local optimization process, the variance of the
fitness of population is used. The error of objective function and the variance of the fitness of
population in each generation are recorded. It is clearly shown in Figure 11, when the variance of the
fitness of population becomes small, the convergence rate of GA slows down. After the variance of
the fitness of the population drops to 0.001, the error reduction of the objective function becomes very
Electronics 2019, 8, 1133 12 of 14

3.3. More Discussions

To determine the appropriate starting point of the local optimization process, the variance of
the fitness of population is used. The error of objective function and the variance of the fitness of
population in each generation are recorded. It is clearly shown in Figure 11, when the variance of the
fitness of population becomes small, the convergence rate of GA slows down. After the variance of the
fitness of the population drops to 0.001, the error reduction of the objective function becomes very
slow. Hence, the variance of the fitness of the population can be employed to effectively determine the
starting point of the local optimization process. As is shown in Figure 12, after a full 50 iterations of
GA, the local optimization process successful finds the appropriate global minimum for both examples.
Electronics
Electronics 2019,
However, 2019, 8, xx FOR
8,
if only 10PEER
FOR PEER REVIEW
REVIEW
iterations of GA are used, the local optimization process is trapped at 12
12 of
of 14
14
a local
minimum. For the proposed method, the local optimization process starts when var( f it) < v (0.001).
a local minimum. For the proposed method, the local optimization process starts when 𝑣𝑎𝑟 𝑓𝑖𝑡 < v
By so doing, this proposed method can successfully achieve the global minimum (i.e., without being
(0.001). By so doing, this proposed method can successfully achieve the global minimum (i.e., without
trapped in a local minimum) and yet save much time from unnecessary iterations of the global search
being trapped in a local minimum) and yet save much time from unnecessary iterations of the
process. The errors of the objective function by the three different hybrid optimization methods stated
global search process. The errors of the objec
above are shown in Table 7.

(a) (b)
Figure 11.
Figure 11. Error versus variance
Error versus variance of
of fitness:
fitness: (a) planar spiral
(a) planar spiral inductor
inductor and
and (b)
(b) microstrip
microstrip interdigital
microstrip interdigital
capacitor modeled by pure GA with population
with population
capacitor modeled by pure GA with population ofof 500.
of 500.
500.

(a) (b)
Figure 12. Error
Error versus variance of fitness: (a)
(a) planar spiral inductor and (b) microstrip
microstrip interdigital
capacitor modeled by hybrid optimization
capacitor modeled by hybrid optimization methodmethod
method with three different starting points
with three different starting points of
of local
local
optimization
optimization process.
process.

Table
Table 7.
7. Error
Error of
of the
the objective
objective function
function by
by three
three different
different hybrid
hybrid optimization
optimization methods.
methods.

Error after optimization

Hybrid optimization method Planar spiral Microstrip
inductor interdigital capacitor
Local optimization process
1.27055 × 10−2
−2 2.01342 × 10−4
−4
started when var(fit) < 0.001
Local optimization process
1.42699 × 10−2
−2 1.66245 × 10−4
−4
Electronics 2019, 8, 1133 13 of 14

Table 7. Error of the objective function by three different hybrid optimization methods.

Error after Optimization

Hybrid Optimization Method
Planar Spiral Inductor Microstrip Interdigital Capacitor
Local optimization process started
1.27055 × 10−2 2.01342 × 10−4
when var(fit) < 0.001
Local optimization process started
1.42699 × 10−2 1.66245 × 10−4
after 50 iterations of GA
Local optimization process started
0.22868 × 10−1 1.24562 × 10−2
after 10 iterations of GA

4. Conclusions
In order to efficiently determine the circuit elements of the equivalent circuit model, a hybrid
GA–Powell’s method has been proposed. This method combines the merits of the global search ability
of GA and the local search capability of Powell’s method with good accuracy and rapid convergence.
The variance of the fitness of population has been introduced as a key criterion to effectively determine
the starting point of the local optimization process. This hybrid method is able to effectively search
the global minimum of a parameter extraction problem. The planar spiral inductor and microstrip
interdigital capacitor have been used as design examples to validate the accuracy and efficiency of
this method. Some problems, such as very wide-band parameter extraction, may not be tackled by
the proposed method. In our ongoing research, the equivalent circuit topology could be adjustable
adaptively (e.g., increasing the order of ladder model) so as to maintain the modeling accuracy in a
very wide frequency band, which will be reported sometime in the near future.

Author Contributions: Conceptualization, Z.G., P.Z. and G.W.; Data curation, Z.G.; Funding acquisition,
G.W.; Investigation, Z.G., Q.L. and X.W.; Methodology, Z.G.; Project administration, G.W.; Supervision, G.W.;
Writing—original draft, Z.G.; Writing review & editing, Z.G., P.Z. and G.W.
Funding: This work was supported in part by the Natural Science Foundation of China under Grant 61601163,
in part by the Zhejiang Provincial Key Research & Development Project under Grant 2019C04003.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Burghartz, J.N.; Edelstein, D.C.; Soyuer, M.; Ainspan, H.A.; Jenkins, K.A. RF circuit design aspects of spiral
inductors on silicon. IEEE J. Solid State Circuits 1998, 3, 2028–2034. [CrossRef]
2. Karisan, Y.; Caglayan, C.; Trichopoulos, G.C.; Sertel, K. Lumped-element equivalent-circuit modeling of
millimeter-wave HEMT parasitics through full-wave electromagnetic analysis. IEEE Trans. Microw. Theory
Tech. 2016, 64, 1419–1430. [CrossRef]
3. Issakov, V.; Kehl-Waas, S.; Breun, S. Analytical equivalent circuit extraction procedure for broadband scalable
modeling of three-port center-trapped symmetric on-chip inductor. IEEE Trans. Circuits Syst. 2019, 66,
3557–3570. [CrossRef]
4. Zhao, P.; Chan, C.H. Design and analysis of a high-selectivity frequency-selective surface at 60 GHz.
IEEE Trans. Microw. Theory Tech. 2016, 64, 1694–1703.
5. Cao, Y.; Wang, G. A wideband and scalable model of spiral inductor using space-mapping neural network.
IEEE Trans. Microw. Theory Tech. 2007, 55, 2473–2480. [CrossRef]
6. Cao, Y.; Groves, R.A.; Huang, X.; Zamdmer, N.D.; Plouchart, J.O.; Wachnik, R.A.; Hu, C.
Frequency-independent equivalent circuit model for on-chip spiral inductors. IEEE J. Solid State Circuits
2003, 38, 419–426. [CrossRef]
7. Alley, G.D. Interdigital capacitors and their application on lumped-element microwave integrated circuits.
IEEE Trans. Microw. Theory Tech. 1970, 18, 1028–1033. [CrossRef]
Electronics 2019, 8, 1133 14 of 14

8. Ghayekhloo, A.; Afsahi, M.; Orouji, A.A. An optimized checkerboard structure for cross-section reduction:
Producing a coating surface for bistatic radar using the equivalent electric circuit model. IEEE Antennas
Propag. Mag. 2018, 60, 78–85. [CrossRef]
9. Watts, J.; Bittner, C.; Heaberlin, D.; Hoffmann, J. Extraction of compact model parameters for ULSI MOSFET
using a genetic algorithm. In Proceedings of the 2nd International Conference on Modeling and Simulation
of Microsystems, Cambridge, MA, USA, 19 April 1999; pp. 176–179.
10. Yun, I.; May, G.S. Passive circuit model parameter extraction using genetic algorithms. In Proceedings of the
49th Electronic Components and Technology Conference, San Diego, CA, USA, 1–4 June 1999.
11. Michalewicz, Z.; Janikow, C.Z.; Krawczyk, J.B. A modified genetic algorithm for optimal control problems.
Comput. Math. Appl. 1992, 23, 83–94. [CrossRef]
12. Diaz-Gomez, P.A.; Hougen, D.F. Initial population for genetic algorithm: A Metric Approach. In Proceedings
of the GEM 2007, Las Vegas, NV, USA, 25–28 June 2007.
13. Srinivas, M.; Patnaik, L.M. Adaptive probabilities of crossover mutation in genetic algorithms. IEEE Trans.
Syst. Man Cybern. 1994, 24, 656–667. [CrossRef]
14. Powell, M.J.D. An efficient method for finding the minimum of a function of several variables without
calculating derivatives. Comput. J. 1964, 7, 155–162. [CrossRef]
15. Keifer, J. Sequential minimax search for a maximum. Proc. Am. Math. Soc. 1953, 4, 502–506. [CrossRef]
16. Fletcher, R.; Powell, M.J.D. A rapidly convergent descent method for minimization. Comput. J. 1963, 2,
163–168. [CrossRef]
17. Wang, J.; Kusiak, A. Computational Intelligence in Manufacturing Handbook; CRC Press: Boca Raton, FL, USA,
2000.
18. Faraday Dynamics, Ltd. Available online: http://www.faradynamics.com (accessed on 30 September 2018).
19. Hansen, M.C.; Yalcin, H.; Hayes, J.P. Unveiling the ISCAS-85 benchmarks: A case study in reverse engineering.
IEEE Des. Test 1999, 16, 72–80. [CrossRef]
20. Van Nechel, E.; Ferranti, F.; Rolain, Y.; Latatire, J. A wide-band equivalent circuit model for single slot
defected ground structures. In Proceedings of the 2019 IEEE 23rd Workshop on Signal and Power Integrity
(SPI), Chambéry, France, 18–21 June 2019.
21. Holland, J. Adaptation in Natural and Artificial Systems; University of Michigan Press: Ann Arbor, MI, USA,
1975.
22. Larranaga, P.; Kuijpers, C.M.; Murga, R.H.; Yurramendi, Y. Learning Bayesian Network Structures by
searching for the best ordering with genetic algorithms. IEEE Trans. Syst. Man Cybern. 1994, 26, 487–493.
[CrossRef]
23. Patil, S.; Bhende, M. Comparison and Analysis of different mutation strategies to improve the performance
of genetic algorithm. Int. J. Comput. Sci. Inf. Technol. 2014, 5, 4669–4673.
24. Bodenhofer, U. Real-Coded Gas. In Genetic Algorithms: Theory and Application, Lecture Note Third
Edition—Winter 2003/2004; Bodenhofer, U., Ed.; SCCH: Hagenberg, Austria, 2004; Chapter 5; pp. 59–61.
25. Beeresha, R.S.; Khan, A.M.; Manjunath-Reddy, H.V. Design and optimization of interdigital capacitor. Int. J.
Res. Eng. Technol. 2016, 5, 73–78.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).