Professional Documents
Culture Documents
Abstract: Any order analog fractance circuit based on a method of constructing the rational approximation function in
the form of factorization is designed. The circuit is made up of level-N passive RC parallel network that makes the circuit
easy integration in which the size of N is equal to the order of rational function. Since the rational approximation function
in the form of factorization consists of several first order damp elements and first derivative elements, it will greatly
simplify the design process of fractance circuit. It shows that the designed fractance circuit based on a method of
constructing the rational approximation function is reasonable through three kinds of different performance simulations.
Key Words: rational approximation function; fractance circuit; fractional calculus operator; passive RC network
978-1-4799-7016-2/15/$31.00 2015
c IEEE 6011
that the fractance circuit can be achieved by RC networks. 2.1 Calculation of Transition Frequency
The difference between integral operator and differential
Assume that the transfer function L1 (s)= k1 / s which
operator is only the exchange of numerator and
denominator. So the paper only researches a method to corresponds to the parallel line L1 . The transfer function
approximate the integral operator. If the approximated
L2 (s)= k 2 / s which corresponds to the parallel line L2 .
operator C(s)= 1 / s , we want to adopt G(s) to approximate From fig.1, we can get
C(s). Assume that
20 Lgk1 -20 Lga1 - =20 Lg1 -20 Lga1 (3)
A(s/b1 + 1)(s/b2 + 1)...(s/bi + 1)...(s/bn + 1)
G(s) = (2) 20 Lgk 2 -20 Lga1 + =20 Lg1 -20 Lga1 (4)
(s/a1 + 1)(s/a2 + 1)...(s/ai + 1)...(s/an + 1)
From the equations (3) and (4),
where A , ai , bi are all positive numbers and solved / 20
k1 = 10 (5)
parameters. Because ai and bi are positive numbers, G(s) k 2 = 10 / 20
(6)
belongs to the minimum phase system. The amplitude- According to fig.1,
frequency characteristic and phase-frequency characteristic 20 Lga1 - Lgb0 =2 (7)
are in relationship of one-to-one correspondence, which
means that if the amplitude-frequency characteristic has a 20 Lgb1 - Lga1 +2 =20 Lgb1 - Lga1 (8)
good approximation effect, so is the phase-frequency From (7), we can get
characteristic. It can be seen from the expression (2) that the b0 = a1 10 /(10 ) (9)
rational function consists of N first-order damp elements From (8), we can get
and N first-derivative elements. The schematic diagram of b1 = a1 10 /(10(1 )) (10)
G(s) approximating C(s) is showed in fig.1.
The rest can be done in the same manner,
bi = ai 10 /(10 (1 )) i=1,2,... (11)
/(10 )
ai +1 = bi 10 i=0,1,2,... (12)
a1 = 10 2 , a2 = 10 10 / 9 , a3 = 10 2 / 9 , a4 = 106 / 9 ,
A(s / b1 +1)(s / b2 +1)...(s / bi +1)...(s / bn +1)
A2 = a5 = 1014 / 9 , b1 = 10 14 / 9 , b2 = 10 6 / 9 , b3 = 10 2 / 9 ,
(s / a1 +1)(s / a3 +1)...(s / ai +1)...(s / an +1) s=a
b4 = 1010 / 9 , b5 = 10 2 , A = 1010 / 9 =12.9155, A0 = 0.0744
2
0
The approximation error of amplitude-frequency
-50
characteristic by proposed method fluctuates within [-0.33,
0.33] and the maximum error is only 0.33dB which exists at
-100
the edge of approximation region endpoint in the region
[ 10 1 , 101 ]. It can be seen from the error curve comparison
0
Oustaloup
-45
Chareff
method is the best among the three. So it is reasonable to
proposed propose the method to design the fractance circuit.
-90
In fact, both of Oustaloups method and Chareffs
10
-4
10
-2
10
0 2
10
4
10 10
6
method adopt a form of factorization to approximate the
Frequency (rad/sec)
fractional calculus operator. Now lets think about a new
Fig.5 Performance comparison of proposed fractance circuit with
Oustaloups and Chareffs fractance circuit question. Why the proposed method is better than the two
former method? In order to explain it, we introduce the two
From the simulation diagram, it is seen that when the
method first.
highest order of transfer function is fifth order, in order to
Reference[16] gives the Oustaloups rational function used
approximate integral operator 1 / s 0.5 , the fractance circuit
to approximate the fractional calculus operator s . It
based on the method of rational approximation function in
assumes that
the form of factorization has the widest approximation
N 1 + s / k
region among the three. The approximation region H ous ( s ) = C 30
k = N 1 + s / k
'
proposed in this paper extends a lot, ranging from 10 1.9 to
102.1 , while the approximation region of oustaloups where
fractance circuit is [ 10 2 , 10 2 ], and the approximation k' +1 k +1
0' = 0.5 0 = 0.5 = = > 1
region of Chareffs fractance circuit is [ 10 2 , 101.3 ] . Within k' k
the limit of approximation region [ 102 , 10 2 ], the k' +1 log( N / 0 ) ;
= > 0 k' = > 0 N =
amplitude-frequency characteristic approximation effect of k k log( )
analog fractance circuit based on the above three methods is log ; = ;
great, basically the same and has little difference. From the = h b
log( )
phase-frequency characteristic diagram, it is seen that the
fractance circuit based on the proposed method is the best h , b are respectively the endpoints of approximation
among the three. It can be seen from the boundary region.
approximation frequency 10 2 and 10 2 that the phase based Reference[16] also gives the Chareffs rational function
on proposed method is closer to 45. used to approximate the fractional calculus operator s . It
Fig.6 shows the error curve between three kinds of assumes that
n 1
fractance circuit and integral operator 1 / s 0.5 in frequency (1 + s / zi )
H cha ( s ) = i =0 31
domain. n
(1 + s / pi )
Bode Diagram i =0
2
where
0 a = 10 y / 10(1 ) b = 10 y /10 ab = 10 y /10 (1 )
Magnitude (dB)
0
and phase-frequency characteristic are in relationship of
-20
one-to-one correspondence, which means that if the
-40 amplitude-frequency characteristic has a good
-60 approximation effect, so is the phase-frequency
characteristic. So is closer to 1, the whole approximating
0
0
Bode Diagram
40
-20
20
-40
Magnitude (dB)
0
-60
0
-20
-40
Phase (deg)
-60 -30
0
0.5 order ideal curve
ideal approximation curve
Phase (deg)
N=2 -60
-4 -2 0 2 4 6
-30 N=3 10 10 10 10 10 10
Frequency (rad/sec)
N=4
N=5 0.5
N=6
Fig.9(a) Bode diagram of fractance approximating 1 / s at level 2
-60
-4 -2 0 2 4 6
10 10 10 10 10 10
Frequency (rad/sec)
0
Journal of Process Control, Vol.20, No.7, 823-831, 2010.
-50
[5] Tavakoli-Kakhki M, Haeri M, Tavazoei M S. Study on
control input energy efficiency of fractional order control
systems. IEEE Journal on Emerging and Selected in Circuits
-100
0 and Systems, Vol.3, No.3, 475-482, 2013.
[6] Tripathy M C, Mondal D, Biswas K, et al. Experimental
-30
studies on realization of fractional inductors and fractional
Phase (deg)