Docstoc

Research and Design of a Common Mode Hybrid EMI Filter For

Document Sample
Research and Design of a Common Mode Hybrid EMI Filter For Powered By Docstoc
					      Research and Design of a Common Mode Hybrid EMI Filter For
                       Switch-mode Power Supply
                                      Wang Ping        Tao Chenbin         Zhang Jinghai

                            School of Electrical Engineering and Automation, Tianjin University, China
                                                    E-mail: pingw@tju.edu.cn

Abstract–After the study of electromagnetic interference
(EMI) mechanisms in switching mode power system (SMPS)                                       Z1
and the soft switch, this paper proposes a novel hybrid filter
(HEF) to suppress commend mode EMI in SMPS. This paper
reviews the limits of pure passive and active EMI filter and
discusses the advantages of the hybrid EMI filter (HEF). By
                                                                                                  Z2             A
studying the insertion loss, impedance compatibilities, and
reliability of the hybrid EMI filter, a new HFF topology is
proposed in this paper. And an example of HEF for flyback                              Fig.1: the HEF topology model
converter is designed. And the topology has been simulated in
Saber. The simulation result verifies its advantages in better     1.Passive Filter Design
attenuation, lower cost, and higher efficiency.                    The equivalent circuit with a LC filter is shown in Fig. 2.

Keywords–CM interference, hybrid active/passive filter,                           Zg                L
matching network, switch mode power system.

I. INTRODUCTION
                                                                                   Ug                    C           ZL        U2
Today, as nonlinear devices were widely used, the SMPS
becomes a large interference source [1]. These fast
switching devices generate high-voltage dv/dt and high
commend-mode voltages, causing some serious problems                             Fig. 2: Equivalent circuit with a LC filter
such as premature winding failures, ground leakage
currents, shaft voltages and bearing currents, and                 In Fig.2, ZL is the equivalent impedance of load resistance
conducted or radiated EMI.                                         RL. Ug is noise source, and Zg is noise impedance. The
                                                                   natural frequency of the LC filter is
To suppress EMI, conventionally passive filters have been
widely used. However, passive filters have some                             1
                                                                   0                                                         (1)
disadvantages which limit SMPS to be smaller, such as                       LC
large size, fixed compensation, and subtle impedance
calculation [2]. Recently, some active filter topologies are       then the insertion loss of the LC filter is
developed, most of which are not used in large current
converters yet[3]. An alternative to above two filters is to
                                                                                  P          2       ( L)2 .
use a hybrid EMI filter.                                            ILR  10lg(    1
                                                                                     )  (1  2 ) 2                           (2)
                                                                                  P2         0         RL2

In this paper, mating network is used to LC passive filter
to ignoring the effects of impedance mismatching. And in           In different SMPS, the circuit structure, high impedance of
active part, a feedforward filter is adopted. Its                  the elements, semiconductors used and working frequency
characteristics are analyzed and the current sensor is             are different. Therefore, the source impedance Zg is
discussed. Simulation and experiment result are presented,         different and hard detected. And the load impedance ZL is
demonstrating that good noise attenuation of this HEF.             also different when SMPS connected different part of the
                                                                   system, and it is also changing in a large scale. So Zg and
II. HEF DESIGN                                                     ZL are not certain, and impossible to meet impedance
                                                                   mating condition ZL=Zg, which means the EMI filter
The HEF topology in this paper is shown in Fig.1 [4]. The          cannot works at its best status. So it is difficult to design
noise current is first attenuated by the passive filter. Then      an appropriate EMI filter suitable for most SMPS.
the noise current volume must be controlled in a range that
the active filter can accept. In this paper, a LC filter is        In this paper, mating network was adopted to design
chosen as passive filter, and a feedforword filer is chosen        passive filter part. And the matching network is formed by
as an active filter. The Z1 is high impedance to noise, Z2 is      some matching devices such as inductors, capacitors and
low impedance to noise, A is active filter.                        resistors in input and output of the passive filter. The
                                                                   matching network could cancel the bad effects because of
                                                                   ZL≠Zg. The matching network used in this paper is shown
                                                                   in Fig.3.
                                                               the insertion loss is

                                                                                  (1   p )(1   s )
                                                               ILs  K  K 0                                             (9)
                                                                                          p s
                                 Fig.3: mating network
                                                               To realize that IL is nearly 20dB in passive part, after a
In Fig.3, Cp and Rp formed ‘P’ network, and Cs and Rs          series of calculation, the topology is shown in Fig.4.
formed ‘S’ network. The capacitive reactance of Cp is
much larger than Rp at working frequency, while its
capacitive reactance is much smaller than Rp above cutoff
frequency fc. And at work frequency, impedance loss of Rs
is decreased by low impedance choke Ls, for the
impedance Ls is much smaller, shorting Rs. And above
cutoff fc, the impedance is much higher than Rs, which
prevents resonance and decrease impedance mismatching.
                                                                          Fig.4: Passive filter with matching network
To simplified calculation, voltage attenuation factor K is
used to calculate insertion loss IL. And K0 is defined as
voltage attenuation factor when no matching network            2. Active Filter Design
exists. Then, the voltage attenuation factor K is
                                                               2.1.Current Sensor
                                                               The measurement of EMI noise currents makes necessary
       Ug        Ug                       U2
K                           K0                       (3)   the use of current transformer (CT) with a wide frequency
       Um          U1                     Um                   bandwidth and without distortion. An elementary structure
                                                               based on a toroidal current transformer, a winding, and a
Then the matching effects of P network and S network are       resistor load is used in this paper. The equivalent circuit is
defined as                                                     shown in Fig. 5.
                                                                                                  1:n
       K min | U g
p 
                 U1    min                               (4)                     Lμ                         R
         K c | f  fc
       K min | U
                   2


s 
               Um                                                              Fig. 5: Equivalent circuit of CT
                       min                               (5)
        K c | f  fc
                                                               Since the frequency of the EMI noise is as high as
Then constrains of Rp and Rs are                               30MHz,it is necessary to discuss its high frequency
                                                               performance. Fig.6 shows the high frequency equivalent
                                                               circuit of current transformer [5].
R p  X 1o (L)   p  1
                    2

                                                         (6)
       X 2 s (L)
Rs 
          s2  1

X10(ωL) presents the open-circuit impedance at f=f1
viewed from input, and X2S(ωL)presents short-circuit
impedance at f=f1 viewed from output.

P network is connected at input port of LC filter, and then              Fig.6: High frequency equivalent circuit of CT
the equivalent source impedance Zpg is formed by P
network impedance Zp in parallel with Zg.                      Ll1 and Ll2 are the primary and secondary leakage
Simultaneously, the equivalent load impedance Zsm equals       inductance, C1 and C2 are the magnetizing transformer
the impedance that S network impedance in series with Zs.      inductance, and    Cps1 and Cps2 are the primary and
Furthermore, Zpg and Zsm changes with Zg, Zm and               secondary inter winding capacitance, Lu is excitation
frequency. Here, the effect on input impedance by              inductance of CT. To analysis the circuit, the simplified
impedance matching is presented by ε, which is the ratio       model of CT is shown in Fig.7.
of input impedance viewed to source and load by filter
with and without the matching network. Then
 Z pg  Z g // Z p  2 p Z1i min                      (7)
                              Z 2i m in
Z sm  Z m  Z s                                        (8)
                               2 s

Consider the worst condition, ε is chosen to be 0.1. Then                       Fig.7: Simplified model of CT
Then the transfer function is                                                        III .SIMULATION EXPERIMENTAL RESULTS

I2
   
                                     1                                     (10)   The output filter was used in a full bridge DC/DC
I1 s 3C1C2' R ' ( L1  L'2 )  s 2C1 ( L1  L'2 )  sR ' (C1  C2' )  1          converter. And the filter is used to attenuate
                                                                                  common-mode noise. The simulating circuit is shown in
To guarantee small phase shift, it is necessary to make L u                       Fig.10. Fig.11 shows the noise currents with and without
large. Under certain n and R condition, to prevent                                active filter. i-g is noise current without active filter, i-c is
magnetic core saturating, the volume of core is                                   the active filter injection current, and i-q is the noise
                                                                                  current after compensation. Obviously, the active part
        c L I p
                2                                                                 produces compensate current that is same size of the noise
Vc         2
                                                                           (11)   current, but contrary phase.
           Bsat

μc is the permeability of the core, Ip is the peak value of
induced current, Bsat is the largest allowable flux density.
The current sensor and amplifier circuit is shown in Fig.8.




                                                                                         Fig.11: noise currents with and without active filter

                                                                                  The effectiveness of the HEF is shown in Fig. 12. From
             Fig.8: The current sensor and amplifier circuit                      the results, the HEF performs good attenuation. The
                                                                                  enlarging result is shown in Fig 13. After HEF the noise
2.2. current injection circuit                                                    current is attenuated 40dB, and its voltage peaked is
The current injection circuit is shown in Fig.9.                                  whittled.




                                                                                                    Fig.12: Effectiveness of HEF




                      Fig.9: The current injection circuit

The Op amp used in this paper is LM6361, and C3 prevents
direct current injecting to main circuit, and R4 is used to
reduce the dissipation of BJT.

3. Novel HEF topology
The novel topology of HEF is shown in Fig.10. The                                                      Fig.13: Enlarged Fig.12
nulling method we choose here is feedforward, of which
the input ig is noisy and the output iq is quiet.




                          Fig.10: The topology of HEF
 Fig.14: high frequency performance comparison between HEF
                            and PEF

Another passive EMI filter (PEF) with the same filtering
result was made in this paper. It also designed with
matching network. The devices shown in Fig.4 are
Rp=50Ω , Cp=25nF , L=1mH , C=250nF , Rs=91Ω ,
Ls=100μH. Fig.14 shows the frequency spectrum
comparison between hybrid filter and passive filter at high
frequency. By introducing passive devices to active filter,
the high frequency performance of HEF is much improved,
which is only a little worse than passive filter. However the
size of the HEF is only 20% of PEF. So the HEF has much
advantage in realization.

                   IV. CONCLUSION
To cancel the common-mode voltage generated by the
PWM inverter in the drive system, a novel structure of
active output EMI filter is proposed by combining it with a
passive filter to become a hybrid filter. The matching
network was added to traditional LC passive filter, which
is to ignoring input and output impedance of EMI filter. the
results show good attenuation to common-mode noise
and relatively smaller in size. The HEF has a good future
in realization.

                       REFERENCES
[1] Pairodamonchai, P., Suwankawin, S., Sangwongwanich, S.,
     “Design and Implementation of A Hybrid Output EMI Filter
     for High Frequency Common-Mode Voltage Compensation
     in PWM Inverters”, Power Conversion Conference - Nagoya,
     2007. PCC '07 2-5 April 2007, pp.1484 – 1491.
[2] Wenjie Chen, Xu Yang, Zhaoan Wang, “An Active EMI
     Filtering Technique for Improving Passive Filter
     Low-Frequency Performance”, IEEE Transactions on
     Electromagnetic                               Compatibility,
     2006,Vol.48,No.1,pp.172-177.
[3] Mingjuan Zhu, Perreault, D.J., Caliskan, V.,Neugebauer, T.C.,
     Guttowski, S., Kassakian, J.G., “Design and evaluation of
     Feedforward Active ripple filters,” Power Electronics, IEEE
     Transactions on Vol. 20, No.2, Mar 2005 ,pp.276 – 285.
[4] Wenjie Chen, Xu Yang, Zhaoan Wang, “Systematic evaluation
     of hybrid active EMI filter based on equivalent circuit
     model”, Power Electronics Specialists Conference,
     2006,18-22 June 2006, pp.1 – 7.
[5] Chen Wenjie, Yang Xu, Wang Zhaoan, “A study on design of
     an active emi filter for integrated power electronics
     modules”, Proceedings of the CSEE, Vol. 25, No. 24,
     2005,pp.51-55.

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:94
posted:12/10/2011
language:
pages:4