Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

The triple resonance network with sinusoidal

excitation
Antonio Carlos M. de Queiroz
COPPE/EP - Electrical Engineering Program, Federal University of Rio de Janeiro
Rio de Janeiro, RJ, Brazil

Abstract—In a “triple resonance network” operating with zero- In [6], a different mode of operation was described, where
state response, the network is composed of a transformer instead of a charged primary capacitor, the source of energy
powered through a primary capacitor from a voltage source for the system is a sinusoidal voltage source connected in
that produces a burst of sine cycles, that has at its output side a series with C1 (Fig 2) (a square wave can also be used, with
shunt capacitor and a third coil, whose other end connects to a
just a small error). The output voltage has three (for Fig. 2a)
capacitive load. The objective is, at the end of a number of
cycles, to have all the energy in the system concentrated at the or four (for Fig. 2b) frequency components, one coming
load capacitance. The operation with the zero-state response from the input and the others from the natural frequencies of
allows the charging of the output from a much smaller input the network. The examples in [6] were all for the double
voltage than what would be required in a capacitor-discharge resonance case. In this paper, the triple resonance network
system, operating with the zero-input response. In the with sinusoidal excitation is discussed.
comparison with a similar “double resonance” network, a
kab
fundamental limitation in the voltage gain is identified, and
alternative solutions are proposed. These networks find +
applications in high-voltage generators for pulsed power C1 La Lb C 2 vout
a)
applications, and their design procedure uses curious aspects –
of the circuit theory of lossless linear circuits. kab L3
+
I. INTRODUCTION C1 La Lb C2 C 3 vout
b)
Triple resonance networks (Fig. 1b) were introduced in –
[2][3][4], as a variation of the double resonance network, or Figure 1. Double resonance (a) and triple resonance (b) networks.
Tesla transformer (Fig. 1a), with the purpose of charging
rapidly a relatively small capacitive load to high voltage. C1 kab
The double resonance circuit is composed of a primary +
capacitor C1, charged to moderate high voltage, that is a)
v in La Lb vout
C2 –
connected to the input inductor La of a transformer with low
coupling coefficient kab by a switch. The system enters then C1 kab L3
a transient involving two sinusoidal oscillations, that with
proper design ideally transfers all the input capacitor energy +
v in La Lb vout b)
to the output load capacitor CL = C2. Energy conservation C2 C3

dictates that the maximum voltage gain is: Figure 2. The same networks operating with zero-state response.
vout ( peak ) C1
= (1) II. DESIGN OF THE TRIPLE RESONANCE NETWORK
vC1 ( peak ) CL
EXCITED BY A SINUSOIDAL VOLTAGE SOURCE
The triple resonance version introduces a third inductor L3 C1 L2 L3
connecting C2 to the load CL = C3. With this the high-
voltage end can be moved away from the transformer, and it + +
is possible to obtain faster energy transfer, by using a v in C 0 iin L0 L1 C2 C 3 vout
– – Z out
transformer with increased coupling coefficient, now
possible due to the reduced insulation requirements. The Figure 3. Transformerless triple resonance network excited by a low-
transient after the closing of the switch involves oscillations impedance LC tank L0C0.
at three frequencies. The structure is sometimes cited as The derivation follows the same ideas discussed in [6],
Tesla “magnifier”. The design procedure described in [1], departing from the transformerless 8th-order network shown
and in an improved version in [5], departs from the ratio of in Fig. 3, excited by an initial current iin in the first inductor
the three resonance frequencies of the network, k:l:m, and L0. Considering that the structure places three transmission
from some given component values. In any case, energy zeros at 0 between an input current Iin(s) = iin/s in parallel
conservation fixes the maximum voltage gain as (1). with L0 and the output vout, Vout(s) has the form:
αs 2 And so, the output impedance of the network can be
Vout ( s ) = =
(
s + k s + l s2 + m2 s2 + n2
2 2
)( 2 2
)( )( ) (2) obtained as:
Bk Bl Bm Bn Z out ( s ) =
(B
1 − εA1 )s
+
(B 2 − εA2 )s
+
(B
3 − εA3 )s
+
(B
4 − εA4 )s (8)
= 2 1 2 + 2 2 2 + 2 3 2 + 2 4 2
(
s +k s +l s +m ) (
s +n ) ( ) ( ) s2 + k 2 s2 + l 2 s2 + m2 s 2 + n2
where k, l, m, and n are factors that multiply a common There are four values of ε that cause one of the terms to
disappear. When this happens, the tank L0C0 becomes a
frequency ω0, here normalized to ω0 = 1. The ratio k:l:m:n
then defines the mode of operation of the network. If these zero-impedance LC tank, that acts as an ideal sinusoidal
factors are successive integers with differences that are voltage source driving the network. These values are:
B B B B (9)
doubles of odd integers, as 1:3:5:7, 3:5:7:9, 3:5:7:13, etc., ε = − 1 ;ε = 2 ;ε = − 3 ;ε = 4
1 2 3 4
all the four sinusoids that form the output voltage add A1 A2 A3 A4
constructively at t = π/2, when the output voltage is Using ε1, the driving signal is at the first frequency,
maximum. Changing the origin of time to this instant, vout normalized to ω = k. The output impedance has the
becomes a sum of cosinusoids. Considering that all the expansion:
energy in the circuit is concentrated at C3, vout can be Z out1 =
(B2 − ε1 A2 )s + (− B3 − ε1 A3 )s + (B4 − ε1 A4 )s
s2 + l 2 s2 + m2 s 2 + n2 (10)
generated by an impulsive current source in parallel with C3, l (m − n )(k − n )(k + m )
and Vout(s) is proportional to the output impedance of the B2 − ε1 A2 =
k (k − l + m − n )(l + m )(l − n )
network. With a proper normalization, Vout(s) = Zout(s): m(l + n )(k − n )(k − l )
− B3 − ε1 A3 =
k (k − l + m − n )(l + m )(m + n )
B1 s B2 s B3 s B4 s (3)
Vout ( s ) = Z out ( s) = + + + n (l − m )(k + m )(k − l )
(s 2
+k2 ) (s 2
+l2 ) (s 2
+ m2 ) (s 2
+ n2 ) B4 − ε1 A4 =
k (k − l + m − n )(l − n )(m + n )
The residues Bi can be found, except for an arbitrary Using ε2, the excitation is at ω = l:
multiplying factor, by expanding (2) with α = 1. For Z out 2 =
(− B1 − ε2 A1 )s + (− B3 − ε 2 A3 )s + (B4 − ε 2 A4 )s
s2 + k 2 s 2 + m2 s2 + n2 (11)
convenience, the Bi can be scaled so their absolute values k (m − n )(l + n )(l − m )
add to 1. The results are: − B1 − ε 2 A1 =
l (k − l + m − n )(k − m)(k + n )
k (m − n )(l + n)(l − m ) l (m − n)(k − n )(k + m ) (4) m(l + n )(k − n )(k − l )
B1 = − ;B = − B3 − ε2 A3 =
(k − l + m − n )(k + l )(k − m )(k + n ) 2 (k − l + m − n )(k + l )(l + m )(l − n ) l (k − l + m − n )(k − m )(m + n )
m(l + n )(k − n )(k − l ) n(l − m )(k + m )(k − l ) n(l − m )(k + m )(k − l )
B3 = − ;B = B4 − ε 2 A4 =
(k − l + m − n )(k − n )(l + m )(m + n ) 4 (k − l + m − n )(k + n )(l − n )(m + n ) l (k − l + m − n )(k + n )(m + n )

The expansion of (3) with the residues (4) in the structure in Using ε3, the excitation is at ω = m:
Z out 3 =
(− B1 − ε3 A1 )s + (B2 − ε3 A2 )s + (− B4 − ε3 A4 )s
Fig. 3 would result in a “quadruple resonance network”, s2 + k 2 s2 + l2 s2 + n2 (12)
with complete energy transfer from L0 to C3 in π/2 seconds. k (m − n )(l + n )(l − m )
− B1 − ε3 A1 = −
Considering now that when the output voltage is maximum m(k − l + m − n )(k + l )(k + n )
there is a voltage vin left at C0, a voltage generated by an l (m − n )(k − n )(k + m )
B2 − ε3 A2 =
impulsive current source in parallel with C0 is added to vout. m(k − l + m − n )(k + l )(l − n )
The contribution of vin in vout has the form: n(l − m )(k + m )(k − l )
B4 − ε3 A4 =
m(k − l + m − n )(k + n )(l − n )
α' s3
Vout ' ( s) = = And using ε4, the excitation is at ω = n:
(
s2 + k 2 s 2 + l 2 s 2 + m2 s 2 + n2 )( )( )( ) (5)
(− B1 − ε4 A1 )s + (B2 − ε4 A2 )s + (− B3 − ε4 A3 )s
Z out 4 =
As As As As s2 + k 2 s2 + l 2 s 2 + m2 (13)
= 2 1 2 + 2 2 2 + 2 3 2 + 2 4 2
(
s +k s +l ) (
s +m s +n ) ( ) ( ) − B1 − ε 4 A1 =
k (m − n )(l + n )(l − m )
n(k − l + m − n )(k + l )(k − m )
The residues Ai can be obtained by expanding (5) with α’ = B2 − ε 4 A2 = −
l (m − n )(k − n )(k + m )
1, and, for convenience, scaling the residues so their n (k − l + m − n )(k + l )(l + m )
absolute values add to 1. The results are: m(l + n )(k − n )(k − l )
− B3 − ε4 A3 =
n(k − l + m − n )(k − m )(l + m )
A1 = −
(
k 2 m 2 − n2 l 2 − m 2 )( )
l 2 m2 − n2 k 2 − n2
; A2 = 2
( )( ) It’s seen that only Zout3(s) and Zout4(s) are realizable, with all
(
2 k 2 − m2 k 2 m2 − l 2n2)( 2 l − n2 k 2m 2 − l 2 n2) ( )( ) (6)
the residues always positive if k<l<m<n. This is a general
A3 = −
(
m2 k 2 − n2 k 2 − l 2 )(
; A4 = 2
)
n2 l 2 − m2 k 2 − l 2 ( )( ) property of these networks. Solutions exist only when the
( )(
2 k 2 − m 2 k 2m 2 − l 2n 2 )
2 l − n2 k 2 m2 − l 2n2 ( )( ) excitation corresponds to one of the two highest frequencies
Introducing a factor ε to control how important is this of the output signal. The final network is obtained by the
contribution, Vout(s) is given by: expansion of Zout(s) with the structure of Fig 3, using
B1 s B2 s B3 s B4 s
Cauer’s first form, followed by the conversion of L1-L2 into
Vout ( s ) = + + + =
(s 2
+k2 ) (s 2
+l2 ) (s 2
+ m2 ) (s 2
+ n2 ) (7) a transformer, and adequate impedance and frequency
scaling [1][6]. Explicit formulas for the elements become
 A1s As As As  too complex to be listed here, but it’s easy to carry out the
= Z out ( s ) + ε + 2 2 2 + 2 3 2 + 2 4 2
 (s ) ( ) ( ) ( )

2
+ k2 s +l s +m s +n calculations numerically.
A. Example 1: A. Energy left at C1:
Consider the design of a network operating in mode Considerations similar to the ones that lead to (8) and (11),
5:7:9:11, with driving signal at the normalized frequency ω result in:
= 9 rad/s. Using the formulas (12), the result for Zout(s) and
Z out 2 ( s ) =
(B
− ε 2 A1 − γD1 )s ( B3 − ε 2 A3 − γD3 )s ( B4 − ε 2 A4 − γD4 )s
1
+ + (15)
the normalized element values are: s2 + k 2 s2 + m2 s 2 + n2
5 49 77 m −n
2 2
1 k −m
2 2
s s s D1 = ; D3 = − ; D4 =
96
Z out 3 ( s ) = 2 72
+ 2 288
+ 2 (14) (
2 k 2 − n2 ) 2 2 k 2 − n2 ( )
s + 25 s + 49 s + 121
C3 = 1.000000000 F; L3 = 0.014925373 H; C2 = 4.110805861 F Where the Di are the residues of the expansion in partial
L2 = 0.002963489 H; C1 = 126.654620404 F; L1 = 0.000292956 H fractions of the contribution of an impulsive current source
in parallel with C1 to the output voltage, already considering
Introducing a transformer and scaling the network so C1 = 5
the tank L0C0 reduced to a voltage source. The expansion is
nF, C3 = 10 pF, and L3 = 30 mH, the final element values
identical to the one used in the design of a capacitor-
and the excitation frequency become:
discharge triple resonance network [1]. The parameter γ can
C3 = 10.00 pF; L3 = 30.00 mH; C2 = 41.11 pF; Lb = 6545 µH be always chosen so all the three residues of the expansion
C1 = 5.000 nF; La = 149.2 µH; k ab = 0.2999; f in = 319.5 kHz of Zout2(s) are positive, if it is between the limits:
B4 − ε 2 A4 B − ε 2 A3 (16)
>γ> 3
D4 D3

B. Example 2:
Consider again mode 5:7:9:11, but with excitation at ω = 7
rad/s. Using (15), Zout2(s) results as:
 175781 5  243 1  11 7 
 −γ s  + γ s  − − γ s
 875000 24  280 2  160 24  (17)
Z out 2 ( s ) = + +
s 2 + 25 s 2 + 81 s 2 + 121
The limits (16) are then γ > −243/140 = −1.736 and γ <
−33/140 = −0.2357. At the limits, one term of the expansion
disappears, and the tank L1C1 too. A reasonable value is
between the limits, γ = −0.9, what results, approximately, in
the maximum output voltage, of 10.82 kV. The resulting
element values are:
Figure 4. Voltages and currents with complete energy transfer
C3 = 10.00 pF; L3 = 30.00 mH; C2 = 33.70 pF; Lb = 14.81 mH
Fig. 4 shows the results of a simulation considering a
sinusoid with 200 V of amplitude applied to the circuit. A C1 = 5.000 nF; La = 54.17 µH; k ab = 0.4279; f in = 248.5 kHz
problem with this design can be observed. The voltage gain Fig. 5 shows the resulting waveforms. The voltage gain is
is quite low, with the output reaching only 6.504 kV. substantially increased, but significant energy (48.21 %) is
Consider the same C1 and C3 used in a double resonance left at C1.
network (Fig. 2a, design formulas and voltage gains in [6]),
operating in mode 5:7:9, excited at the central frequency.
The output voltage reaches 15.00 kV with the same 200 V
of input peak voltage. The triple resonance design above is
more similar to the double resonance design with excitation
at the upper frequency (m), mode 5:7:9, that results in only
4.998 kV of maximum output voltage. The triple resonance
design with excitation at the upper frequency (n) is even
worse, resulting in just 4.081 kV. The capacitor-discharge
version would produce, from (1), 4.472 kV.

III. INCOMPLETE ENERGY TRANSFER TO INCREASE THE


VOLTAGE GAIN
A possibility that turns realizable the triple resonance
network with excitation at the second frequency is to design
for incomplete energy transfer, leaving some energy in the
capacitors C1 or C2 when the output voltage is maximum. Figure 5. Voltages and currents when some energy is left at C1.
C. Energy left at C2: efficiency is slightly better. 45.96 % of the energy remains
The effect of a voltage v2 at C2 in vout can be calculated by in C2. It is also possible to combine the two transformations,
placing an impulsive current source q = C2v2 in parallel with leaving energy in C1 and C2, by applying first the
C2, and shifting it to the output, as shown in fig. 6. transformation on Zout(s) that leaves energy in C1. There is
no apparent advantage on this, however.
C1 L2 L3

+
L1 Vout
C2 q C3

C1 L2 L3 sL3q
+
L1 Vout
C2 C3 q

C1 L2 L3
+
L1 Vout + sL3q
C2 C3 q + s L3C3q
2

Figure 6. Finding the effect of a charge in C2 at the output.

The output voltage can then be expressed in function of the


output impedance, and expanded in partial fractions. After
Figure 7. Voltages and currents when energy is left at C2.
some manipulation, and using the fact that E1+E3+E4 = 1/C3,
the result is:
IV. CONCLUSIONS
(
Vout = Z out q + s 2 L3C3q − sL3q = )
(18) The detailed design of the triple resonance network with
=

s 2
E1s
+ k 2
s
Es
+ m s
Es 
+n 
(
+ 2 3 2 + 2 4 2  q + s 2 L3C3q − sL3q = ) sinusoidal excitation, assuming no losses, was described. It
 was verified that the obtainable voltage gain for fixed input
=
( ) ( )
E1q 1 − k 2 s E3q 1 − m 2 s E4 q 1 − n 2 s
+ +
( ) and output capacitances and number of input cycles is
s2 + k 2 s 2 + m2 s 2 + n2 significantly smaller than in a similar double resonance
network. There is no solution for an arbitrarily small C2,
Applying (18) in (7), using ε = ε2, the output impedance can
what would reduce the network to the double resonance case
be found as:
with excitation between the resonances, and so produce a
E1s
Z out =
Es Es
+ 2 3 2+ 2 4 2 = 1
(
B − ε 2 A1 − E1q 1 − k 2 s
+
( )) similar high voltage gain. Two possible designs with
(19)
s +k 2 2
s +m s +n s2 + k 2 incomplete energy transfer were then described, that result
(
B − ε 2 A3 − E3 q 1 − m 2 s
+ 3
(
+ 4
)) (
B − ε 2 A4 − E4 q 1 − n 2 s ( )) in larger voltage gain. A computer program that does all the
s +m
2 2
s 2 + n2 calculations and plots is available [7].
And the residues Ei of the expansion of the output
impedance are then: REFERENCES
B − ε 2 A1 B − ε 2 A3 B − ε 2 A4 (20) [1] A. C. M. de Queiroz, “Multiple resonance networks”, IEEE
E = 1 ;E = 3 ;E = 4
1 + q (1 − k 2 ) 1 + q (1 − m 2 ) 1 + q (1 − n 2 )
1 3 4 Transactions on Circuits and Systems I, Vol. 49, No. 2, pp. 240-244,
February 2002.
To make E4 positive and leave E3 positive too, q must be [2] F. M. Bieniosek, “Triple resonance high voltage pulse transformer
between the limits: circuit,” Proc. 6th IEEE Pulsed Power Conference, pp. 700-702,
1 1 (21) 1987.
<q<
n2 − 1 m2 −1 [3] F. M. Bieniosek, “Triple resonance pulse transformer circuit,”
Review of Scientific Instruments 61 (6) pp. 1717-1719, June 1990.
D. Example 3:
[4] F. M. Bieniosek, “Triple resonance pulse transformer circuit,” U. S.
With the same parameters of example 2, the limits for q are patent No. 4833421, 23 May, 1989.
0.0125 > q > 0.00833. The maximum voltage gain is [5] A. C. M. de Queiroz, “A generalized approach to the design of
obtained with q = 0.0088, approximately, that results in multiple resonance networks”, IEEE Transactions on Circuits and
maximum output voltage of 9.68 kV: Systems I, Vol. 53, No. 4, April 2006, pp. 918-927.
[6] A. C. M. de Queiroz, “Multiple resonance networks with incomplete
C3 = 10.00 pF; L3 = 30.00 mH; C2 = 140.2 pF; Lb = 4151 µH energy transfer and operating with zero-state response”, 2005 IEEE
C1 = 5.000 nF; La = 138.6 µH; k ab = 0.6001; f in = 215.7 kHz International Symposium on Circuits and Systems, Kobe, Japan, May
de 2005, Vol. I, pp. 236-239.
The voltages and currents are shown in Fig. 7. The voltage [7] A. C. M. de Queiroz, Program “trsstcd”, available for download at
http://www.coe.ufrj.br/~acmq/programs.
gain is smaller than in the previous example, although the

You might also like