Read P-spice simulation for conducted EMI and over voltage investigations in a PWM induction motor drive system - Computers in Power Electronics, 2002. Proceedings. 2002 IEEE Workshop on text version

P-Spice simulation for Conducted EM1 and Over Voltage Investigations in a PVM Induction Motor Drive System

L. Arnedo and K. Venkatesan

Center for Power Electronics Systems Electrical and Computer Engineering Department Uiiiversity of Puerto Rico. Mayaguez PR 0068 I -9M2

'abslracL This Paper deals with the development 01 a P-spice model for a PWM induction motor drive rrstem that allows prdicrion of over voltage and conducted EM1 in the presence of long feeders. Motor has been represented by both d-q and High frequenq models whereas 3 lumped paranieters model has been used for the cable. A detailed model of the ICBT iiwerter has also been Included. Simulations have been csrried out for different cable lengths and inverter voltage rise times to investigate the effects of these on over-voltage a t moton terminals and common mode emissions in the system. RerulIs obtained from an experimental system are also presented.


An essential part of EM1 study is the modeling of device switching dynamic behavior. identification of different current paths and the extraction of parasitic inductances and capacitances. Fig.] shows the flow chart for a successful simulation of conducted emissions.

. . . .. .. .

.I. Introduction

Advances in power electronic switching devices such as Po~,erMOSFETs and IGBTs have enabled high frequency switching operations and improved the performance of PWM inverters for feeding induction motors. Howe\,er these new technologies have created new problems d a t e d to Electroma:metic interference (EMI) and over-voltages at the terminals of electric machines [I]. Prevntly EMC regulations are more stringent, imposing additional design objectives for power electronic systems. Some forms of filtering are required for the hput and output(s) lines of equipment. However, the optimum design approach is to minimize MI at the source of the emission. This reduces the size and volume ofthe filter and reduces the possibility EM1 being radiated internally to other sensitive components in the equipment. The over-voltage phenomenon has destructive effects upon both cable and machine insulation system due the energy contained in the transient overshoot caused by vollage wave reflection at the electric machine terminals. This pht:nomenon is also directly related with the conducted MI. Hence it is important that the EM1 and ovcr voltage characteristics of the system must be analyzed and predicted in the design stage. The simulation model of the system takmg into account the paths of current noise would be useful for POWH electronics circuits design and improve the MI reduction techniques. This research aims to develop a P-spice model of an electrical drive system that allows prediction of ovcr voltage at the terminal ofthe motor and conducted EM1 in presence of bog feeders. The developed model c3n be usoi to ~ u d y the e t k t ofowr-voltage mitigation techniques wed in PWM invertersupon conducted enlissions.

Fi: I Fhu.sLUfor Eh11 hldeliq

A. Idenrificatioir ofA'oise Cvrrenr Purlis.

In a PWM-VSI there are a lot of paths for the noise current and hence it is important to identify the most important ones such as: '?%e capacitance between the heat sink and IGBT ? W e capacitance between the bus bar and earth ?The cable capacitances ? W e stray capacitances of the motor ?The IGBT parasitic capacitances ?The IGBT parasitic inductances ?The diode Parasitic capacitances

B. Porasifir Errroction

Several techniques for extraction of pansitics have been developed but essentially they c m be divided in two categories:

0-7803-7554-8/02/$17.00 02002 IEEE


Authorized licensed use limited to: Rockwell Collins. Downloaded on June 16, 2009 at 18:23 from IEEE Xplore. Restrictions apply.

Mathematical modeling techniques: These use the threedimensional finite element analysis (FEA) aiid the partial element equivalent circuit (PEEC) method. Both methods are mathematical approaches and there are some limitations. Some tools such Maxwell and Inca use these inathematical methods. Measurement Based modeling techniques: Time domain reflectometry (TDR) has bern used extensively in the characterization of interconnects, and applied to extract the equivalent circuits consisting of lumped parameters [Z]. The Impedance analyzer is an equipment employed to measure the frrquency response of paver electronic devices, cables aiid electric machines.

C Circuit Simularion:





-* n.





Fig 3 High lieq~lcm~yt~xulcl link crpacilur ofLK

The simulation of the system can be perfomied in anyone ofthe two programs. Saber or P-spice. everything dependiog oil the preference of the designer. Maxwell and TDR measurements have subroutines that can be used with Pspice.



When the frequency increases the impedance of the capacitor decreases linearly ai rnte of -2OdBi'decade. The impedance of the inductor increase until it equals that of the capacitor at tho point of resonance. At this point the impedance is R,.'For higher frequencies the impedance ofthe inductor increases at rate of +2OdB/decade. The impedance of the DC link capacitor has a strong effect in the differential conducted emissions [4]. The panmeters of the inverter used are giveii in appendix

B. HF Cubic model

Tlie transient plienonlena o f cables have been explained in detail using Ranmission h e theory [SI and different cable mod4 configurations used in studies ofvoltage retlection we preseiited in [6].lo P-spice program although there are circuit simulation elements for transmission lines. the options are limited for multi-conductor cables. For obtaining an adequate model of the cable for high frequency studies sofiware Maxw*eIIZD extractor is used. The Maxwell ZD extractor uses finite element method to compute the circuit parqeter matrices such as inductance and capacitance for a w arbitrary multi-conductor transmission line. These circuit parameters depend upon the geometry structure and the charncteristics of the materials tlmt make up the structure. Once computed. these circuit parameters can be transformed into a P-spices sub circuit forming the lumped ? representation of the cable. In order to model appropriately the cable in a wide frequency range 64 lumped Sections have been used 171. Fig 4 shows the variation o f common mode impedance with frequency as calculated for a six meter SJ 4-14 A W G cable using (54 lumped sections. The experimentally detennined values closely agree with the calculated values as s h o w in the s i n e figure.

HF PWM Invener Model

For an accurate EMC model of the inverter it is necessary to take into account the HF parasitic palls. Fig. 2 shows the HF equivalent circuit ofone leg ofthe inverter [;I.


. c


Fig 2 Invmer model for EM1 stud&

The most important parasitic paths of this circuit are: the parasitic inductance of the emitter Le and the intemill parasitic capacitances of the IGBT. T3e value of L, is taken from device datasheet and parasitic capacitances ofIGBT are included i n the IGBT P-spice model. Stray inductaiices ofthe connecting wires (Li) have very small values and affect principally the differential conducted emissions. For this study this inductance has been neglected because the PWM iiiverter is enclosed in a package and the length of the connecting wire is very small. The Stray capacitaiice C, between the collector and grounded heatsiiik is measured with an impedance analyzer. I n the HF range the equivalent circuit of the DC: liiik capacitor consists of the series combination of capacitance, resistance and inductance, as shown in fig.;


Authorized licensed use limited to: Rockwell Collins. Downloaded on June 16, 2009 at 18:23 from IEEE Xplore. Restrictions apply.

be used for over roltlge and conducted EM1 studies. This can be implenlented in P-spice or Saber for anslpis of inverter fed induction motor drive systems.




111 older to compare the simulation and experimental results an experimental set up is implemented as shown i n fig. 6. The parameters of the iiiverier. cable and motor a15 given in appendix.

Fi 4 Gble " m n mode iwednnce

C. Indiicfion Mofor model

A motor model as shown in fy.5 suitable for low <and high frequencies is used [8]. The model b based on the experimental observation of frequency response and an approximation of the distributed HF motor model presented in [O] where it is possible to identify three dominant capacitances C, C;, C , at high frequencies. The model is diem in Fig5

* L

A . Efecf ~ / C Q / Jlength on over-volfoge u molor lcrminals. IP f

4 %

2" 2"

In these simulations different cable length are employed keeping inverter rise time constant. Figures 8. 9, 10 and I I illustrate that when the cable IengIh is increased the maximuni voltage overshoot at the motor terminals increases. However. bryond a cerIain length of the cable the voltage at motor terminals remain constant as shown in fig. I. The values of peak line voltage obtained from siniulations for different cable Ien,ds closely agree with experimental values. This demonctrates the accuracy of the developed Pspice models for the drive system There exists a different% in natural tiequency oscillation between simulated and experimental wave forms. This difference can be reduced by either using more lumped sections ar using distributed representation. That will demand more computational resources. Convergence problems may alw come up due to very small tinle-step.

Fig 5 Idwtion mmr model for uidc i q u w y nnp:

In above C C , Ci and Rrrepresenting phase or neutral IO , ground capacitance. phnse to phase capacitance, phase to neutral capacitance and eddy loss resistor are effective at high tiequencies. Series impedance elements consisting of h, 1 and C, are associated with phase to ground and neutral IO ground current paths at nledium frequencies.

The advantages of this model are that the paranutzrs can he determined by tiequency response tests and the model can


Authorized licensed use limited to: Rockwell Collins. Downloaded on June 16, 2009 at 18:23 from IEEE Xplore. Restrictions apply.

.........-.-...-.... . .

......... .........

Fig I Over-uolwp ='a?

B. E f f i l ofinvener. wlruge rise lime on motor terminal


The effect ofthe rise time upon the over voltage is studied for a given cable length of six meters The results are shown in fig. IO. From the fipres it seen that the rise time has a si&icant effect upon over-voltage. For a rise time of 110 11s the peak voltage at motor terminals is 460V whereas for a rise time of 25 ns it is 600 V. Experimental result for cable [email protected] of6 meters and rise time 110 ns is also MOV.


. ....








-&. *..& i

! I !

.A * :

' ' ? . e ' ,



. .

" l . W " **I . #V\ I 4 .I



.. ....




rig E ~ ~ ~ WMmn~muleC~W i I~ roc ms lengriL 12 ~ . v . cahk

Fi I O Rbtbm eIleF1 up)"


C: Coinmon nrode conducted emission Common mode conducted emissions is an important part of E 1 studies in a drive system [IO]. Simulations hdve been M carried out to study the effects of cable length and inverter voltage rise time on common mode currents.

Figures 13 to 16 show the simulated and measured total EM1 emissions for two different lengths, 6mts and 2Omts.

The siniulated EM1 specmm is obtained by applying FFT to tlie output of the LISN. The measured spectrum is obtained with 3 spectrum analyzer. The results indicate that the magnitude of the EM1 emissions is affected strongly with the increase of cable length due to increased stray capacitance to earth the over-voltage phenomenon and increased natural frequency. The simulated and experimental spectrum are i n good agreement.

Figs. 1 I , 12 slmw the simulated and experimental wave forms ofcommon mode current under steady state condition. Motor frame, gound conductor of cable and heat sink have been connected together to supply ground. The fi:mres shows the accuracy of the model to predict the peak value and the Wiavior of the common mode current


Authorized licensed use limited to: Rockwell Collins. Downloaded on June 16, 2009 at 18:23 from IEEE Xplore. Restrictions apply.

Figures 17 and 18 show the total EM1 emissions with inverter voltage rise time of 1 Ions and 52411s. The k n - ofthe cable ~ is kept constant at 6 meters. With increased rise times the total EM1 emission diminishes for frequencies in thi nnge of ZMhz to IO MHz . This is due to the decreased magnitude of high frequency harmonics for the pulws with large rise tinis. Then the magnitude ofthe high frequency c u m n t is less and this is reflected in the specbum ofthe total EM1 emissions.

.. -.












.. .

Fi:.?OTotalEMI Oimiriomarrkrfmlwmywna_.of IY20 Hr

Figures 19 and 20 show the spectrums of EM1 emissions with inverter carrier frequencies of 96OHz and 1920Hz respectively. The increnmt of twice the frequency modulation represents an increment of 5 dBuV in the total EM1 conducted emissions. The R M S \,slue of the comnwn mode current also increases from 35 mA to 55 nvl . These parasitic currents flow mainly through the insulation system of the cable and motor diminishing the life time of these devices


Authorized licensed use limited to: Rockwell Collins. Downloaded on June 16, 2009 at 18:23 from IEEE Xplore. Restrictions apply.


A P-spice model taking iiito account the h i i frequency

behavior of inverter. cable and induction motor has been developed for an induction motor drive system. This nmdel is used tu study over-voltages at motor terminals and conunon mode conducted EM1 in the system. The close ageement, between the results obtained experimentally and through simulation for diRerent cable lengths. inverter x,oitage rise times and frequency modulations validate the accuracy of modeling. In order to predict EM1 emissions and obtain a good accuracy in the magnitude of the spectrum at high frequencies, it is necessary to have time steps between 1 and 10 ns. The problem writli P-spice is that the time step is automatically adjusted by the program hised on the error estimation formulas. If the time step is large the range ofthe spectrum where the magnitude of the spectrum is well calculated will be short ( approximately I or 2 MHz). If the error tolerance is reduced to small values forcing the program to use snwll time steps. problems of convergence usually appear. The simulatioii time is another problem h a u s e with very small time stem a sinlulation for 2OOms could be as lonz as

This work was supported primarily by the ERC Program of the National Science Foundation under Award EEG9731 677.









13.5 '?




Authorized licensed use limited to: Rockwell Collins. Downloaded on June 16, 2009 at 18:23 from IEEE Xplore. Restrictions apply.


P-spice simulation for conducted EMI and over voltage investigations in a PWM induction motor drive system - Computers in Power Electronics, 2002. Proceedings. 2002 IEEE Workshop on

6 pages

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate


You might also be interested in

Catalog Cover 2008 Setup.qxp
BICC Industrial Cables Catalog
Anixter Wire & Cable Technical Information Handbook