Master Thesis
Converter and Cables in Propulsion Systems
Stephane Yannick Njiomouo
Stockholm, Sweden 2014
XR-EE-ETK 2014:012
drive system. In fact, power and audio frequency emissions could disturb track signaling and the control systems, while high frequency currents injected into cable screens could damage the cables. In order to ensure compatibility to conducted and radiated EMC requirements, and related infrastructure signaling specifica- tions, it is necessary to perform 3D modelling of the drive system to predict the EM emission during the design phase of the propulsion system. CST, an electro- magnetic analysis tool, is used to create the 3D model of the converter module and the cables. The model allows for the inclusion of the parasitic characteristics of the IGBTs, the bus-bars, and the motor cables. Influence of different grounding schemes is analyzed. The model predicts the EM field distribution at points inside the converter module and in the vicinity.
i
motor omvandlare och kablar i Framdrivnings Systems
Anv¨andningen av kraftomvandlare i j¨arnv¨agstraktionssystem introducerar h¨og- frekvens elektromagnetisk interferens (EMI) i framdrivningssystemet, vilket or- sakar elektromagnetiska kompatibilitetsproblem (EMC). Dessa h¨ogfrekvensfenomen orsakas av snabba variationer i str¨om och sp¨anning under omkopplingsopera- tioner i kraftomvandlare. H¨ogfrekvensstr¨ommarna alstrar elektromagnetiska (EM) st¨orningar, som kan p˚averka funktionaliteten hos det elektriska drivsystemet.
St¨orningar vid kraft- och ljudfrekvenser kan p˚averka signal- och kontrollsystemen, medan h¨ogfrekventa str¨ommar injiceras i kabelsk¨armar kan skada kablarna. F¨or att s¨akerst¨alla kompatibiliteten mellan EMC-kraven, vad g¨aller ledningsbundna och uts¨anda st¨orningar, och specifikationerna f¨or signalsystemets infrastruktur
¨
ar det n¨odv¨andigt att utf¨ora 3D-modellering av drivsystemet, f¨or att redan un- der designfasen av framdrivningssystemet kunna f¨oruts¨aga de elektromagnetiska st¨orningarna. CST, som ¨ar ett elektromagnetiskt analysverktyg, anv¨ands f¨or att skapa 3D-modellen av omriktarmodulen och kablarna. Modellen g¨or det m¨ojligt att ta med de parasitiska egenskaperna hos IGBT, ledningsmoduler och motork- ablar. Inverkan av olika jordningssystemen analyseras. Modellen f¨oruts¨ager det elektromagnetiska f¨altet vid olika punkter inuti omriktarmodulen och i dess n¨arhet.
ii
Ensa, Valerie, Giuseppe, Fabrice, TEK Boris and Nelly who assist me every time.
I could not forget my Family with all sacrifices they made for me and to tell you again THANK YOU and I love you.
iii
Abstract i
Acknowledgements iii
Contents iv
List of Figures vi
List of Tables ix
1 Introduction 1
1.1 Description . . . 1
1.2 Previous work . . . 2
1.3 Thesis Outline. . . 2
2 Cable modeling 4 2.1 Introduction . . . 4
2.2 Ground plane effects . . . 4
2.2.1 Surface currents . . . 5
2.2.2 The magnetic field . . . 5
2.2.3 S-parameter S11 . . . 7
2.3 Screen current estimation . . . 8
2.3.1 Background . . . 8
2.3.2 Assumptions. . . 10
2.3.3 Simulation results . . . 12
2.4 Power Cable . . . 14
2.4.1 Description . . . 14
2.4.2 Simulation results . . . 14
2.5 Conclusion . . . 17
3 Converter Models 18 3.1 Introduction . . . 18
3.2 Parasitic modeling of the IGBTs. . . 19
3.2.1 Parameter estimation . . . 20
3.2.2 Observations . . . 22 iv
4.3 3D field distribution . . . 39 4.4 EM field 2D plots . . . 40 4.5 Conclusion . . . 49
5 Discussion and Future Work 50
5.1 Discussion . . . 50 5.2 Future Work . . . 50 5.3 Conclusion . . . 51
2.1 The surface current distribution at 10 kHz with common ground . . 6
2.2 The surface current distribution at 10 kHz with separated ground. . 6
2.3 Zoom of the surface current close to port 1 . . . 6
2.4 The surface current distribution with common ground at 200 MHz . 7 2.5 The surface current distribution at 200 MHz with separated ground 7 2.6 Phase field distribution at 10 kHz with common ground . . . 8
2.7 Magnetic field distribution at 10 kHz with separated ground . . . . 8
2.8 Magnetic field distribution at 200 MHz with common ground . . . . 8
2.9 Magnetic field distribution at 200 MHz with separated ground . . . 8
2.10 Magnitude of S11 in the frequency domain: in blue when a common ground is used and in green when separated grounds are used . . . 9
2.11 Magnitude of S11 in the frequency domain: in blue when is used a common ground and in green when separated grounds are used. . . 9
2.12 Port impedance behaviour in the frequency domain: in blue when a common ground is used and in green when separated grounds are used . . . 10
2.13 Theoretical model used for building the equations [1] . . . 11
2.14 Screen current ratios computer by CST: Is/Icphase a in blue, Is/Ic phase b in green, Is/Ic phase c in red. . . 13
2.15 Screen current ratios measured. . . 13
2.16 Schematic of the reflection test in CST . . . 15
2.17 Port voltage in red, line close end voltage in green and far end voltage waveforms for ZL= 1M Ω . . . 16
2.18 Port voltage in red, line close end voltage in green and far end voltage waveforms for ZL= 1mΩ . . . 16
2.19 The port voltage signal (in blue) and the far end voltage signal (in green). . . 17
3.1 Simplification of the model . . . 19
3.2 DC link excitation . . . 19
3.3 Load setting in AC side . . . 19
3.4 Simultion setting . . . 19
3.5 Wire copper connecting terminals of IGBT . . . 20
3.6 Lumped element connecting the IGBT terminals . . . 20
3.7 Excitation signal pattern . . . 21
vi
3.17 Voltage responses using imported IGBTs . . . 29
3.18 Voltage responses using VCSW . . . 29
3.19 Current responses using imported IGBTs . . . 30
3.20 Current responses using VCSW . . . 30
3.21 Voltage responses using VCSW with rise time equals 0.4 µs . . . 31
4.1 schematic of the system . . . 32
4.2 Cable setting . . . 33
4.3 Cables coupled with converter . . . 34
4.4 IGBT locations . . . 35
4.5 Grounding and EMC filter setting . . . 35
4.6 schematic of the system . . . 36
4.7 load Voltages . . . 37
4.8 DC link current and voltage . . . 38
4.9 3D Electric field at 10 kHz . . . 39
4.10 3D Electric field at 500 MHz . . . 39
4.11 Cross section at the middle of the line at 10 kHz . . . 39
4.12 Cross section at the middle of the line at 500 MHz. . . 39
4.13 3D Magnetic field at 10 kHz . . . 40
4.14 3D Magnetic field at 500 MHz . . . 40
4.15 Probe location in the system . . . 40
4.16 Electric field in point A, close to the IGBT in V /m . . . 41
4.17 FFT Electric field in point A, close to the IGBT V /m . . . 41
4.18 Magnetic field in point A, close to the IGBT in A/m . . . 42
4.19 FFT Magnetic field in point A, close to the IGBT in A/m . . . 42
4.20 Electric field in point B, left to the MCM in V /m . . . 43
4.21 FFT Electric field in point B, left to the MCM in V /m . . . 43
4.22 Magnetic field in point B, left to the MCM in A/m . . . 44
4.23 FFT Magnetic in point B, field left to the MCM in A/m . . . 44
4.24 Electric field in point C, 1m from the cable middle in V /m . . . 45
4.25 FFT Electric field in point C, 1m from the cable middle V /m . . . 45
4.26 Magnetic field in point C, 1m from the cable middle A/m . . . 46
4.27 FFT Magnetic fieldin point C, 1m from the cable middle in A/m . 46 4.28 Electric field in point D, close to the cables in V /m . . . 47
4.29 FFT Electric field in point D, close to the cables in V /m . . . 47 4.30 Magnetic field in point D, close to the cables A/m . . . 48 4.31 FFT Magnetic in point D, close to the cables in A/m . . . 48
ix
Introduction
1.1 Description
Power converter and cables are very important components of the variable speed drives. The converter delivers switching waveforms of voltage and current. These fast voltage variations (dv/dt) inject into the cables current and voltage tran- sients which are sources of electromagnetic interferences (EMI). Depending of the path, the power propagations, on the motor side, can create over-voltage at the cable terminals which can be a critical issue for the motor insulation material.
The fast transients in the cables generate EM radiation which could disturb the smooth functionality of the control and other sensitive systems. In fact, current and voltage sensors installed on the converter module are usually victims of the EM disturbances from the converter and cable. These transients injected back to the grounding line could distort the the functionality of the track signaling systems. To prevent these issues and make sure that the drive system works un- der the EMI/EMC safety limits, it is necessary to have good and realistic models earlier in the design. CST EM modeling tool, is used to create 3D models of the converter module and the cable interconnections. From CAD model of the converter geometry is imported in CST microwave (MW) studio and the parasitic inductance and capacitance of the bus-bar interconnections are estimated. The IGBT parasitics are represented as lumped parameters. The cable connections are modeled as transmission lines in CST cable (CB) studio. The cable model is complete with the converter module to obtain a complete model of the drive system in CST design studio (DS). The model is excited with current or voltage
1
including the IGBT parasitic are modeled. It characterizes the IGBTs with lumped elements. It makes the excitation from the DC link bus, observes the results on the AC side.The EM field values at different locations are measured. The software tool CST (Computer Simulation Technology) was used. This thesis is a continuation of the work done in [2] and the new contributions are the following:
• The off-state capacitance chance to the Connector-Emitter capacitance;
• The excitation is set between the DC link bus-bars;
• The dynamic IGBT switching characteristic is introduced;
• Cables and ground plane are coupled with the converter module in order to complete the drive system.
1.3 Thesis Outline
The thesis is arranged as followed:
• Chapter 2deals with cable modeling including ground plane effects, screen current estimation, s-parameter estimation and voltage reflections at termi- nals;
• Chapter 3 deals with modeling of the converter module, including the rep- resented of the IGBTs, the input lumped filter elements and the converter modulation state;
• Chapter 4 presents the assembly of the converter module and the cable model with a given motor equivalent circuit representation. Predictions of
the EM fields of the EM field distribution within the converter module and along the cables are represented.
• Chapter 5 concludes the thesis with some discussions and suggestions for the future work.
In the motor drive system, power is mostly transmitted between components through cables. The performance of the cable can be affected by different factors, for example the ground system, the presence of other cables or the termination nature. The cable deals with problems like heating and breakdown voltage in the insulation. This chapter presents investigations which allow to understand the mechanism of these phenomena. The analysis will focus on the study of the ground plane effect in section 2.2, the screen current behaviour in the frequency domain in section 2.3 and the propagation of the voltage using lossy cable in section 2.4.
2.2 Ground plane effects
An issue when simulations are made is to make sure that the injected signal will follow the direction and the sense it is supposed to follow. For this aim, the nature and the setting of the ground plane must be studied very well. In this case for instance, a perfect electrical conductor is used as ground material. Two configurations have been considered:
• A common ground situation: the cable ends are both connected to the same ground through two ports with internal impedance equal to 0.35mΩ.
4
• A separated ground situation: the cable ends are connected to different grounds.
A 200 cm copper cable length is considered. After the excitations of both ports the surface currents and the magnetic fields are observed at 10 kHz and 200 MHz.
2.2.1 Surface currents
The figures 2.1 and 2.3 show the surface current distribution at 10 kHz with common ground while figure 2.2 shows the distribution for separated ground respectively. Figures 2.4 and 2.5 show the surface currents at 200 MHz.
Observations At 10 kHz
• At low frequency the surface current is reduced using separated grounds.
• The maximum current amplitudes for 10kHz are 1.026A/m and 0.1102A/m for the single ground and for the separated grounds respectively.
At 200 MHz
• At high frequency the surface currents are almost the same.
• The maximum fields amplitudes for 200M Hz are 0.07107A/m and 0.05855A/m for single ground and separated ground respectively.
2.2.2 The magnetic field
Using the settings presented in section 2.3 the EM field distribution could be observed. Figures 2.6 and 2.7 show the magnetic field distribution at 10 kHz with common and separated ground respectively . Figures 2.8 and 2.9 show the magnetic field distribution at 200 MHz.
Figure 2.1: The surface current distribution at 10 kHz with com-
mon ground
Figure 2.2: The surface current distribution at 10 kHz with sepa-
rated ground.
Figure 2.3: Zoom of the surface current close to port 1
Observations:
• At low frequency the magnetic field is reduced using separated ground. The maximum field amplitudes for 10 kHz are 2.489 A/m and 0.1102 A/m for single ground and separated ground respectively.
• At high frequency the magnetic fields are almost the same. The maximum field amplitudes for 200 MHz are 0.5551 A/m and 0.5515 A/m for single ground and separated ground respectively.
Thus the ground configuration affects the surface current and the magnetic field only in low frequency. In high frequency, currents have almost the same value.
Figure 2.4: The surface cur- rent distribution with common
ground at 200 MHz
Figure 2.5: The surface cur- rent distribution at 200 MHz
with separated ground
2.2.3 S-parameter S11
The scattering parameter S11 has been computed in the different ground configu- rations. The S11 parameter can be derived by the following formula:
Γ0 = S11= Zload− Zsource
Zload+ Zsource (2.1)
If Zload = Zsource(matchedload), S11 = 0. Figure 2.10 shows the phase of S11, figure 2.11 shows the magnitude of S11 and figure 2.12 shows the port impedance in the frequency domain. As can be observed in figure 2.10, in low frequency the phases overlap each other and we have the same resonance frequency. Then we have some deviation in high frequency. Figure 2.11 shows that in low frequency, the power goes first through the ground since it presents lower impedance than the cable and the port in the case of common ground. The ground impedance is zero, so the S11 = −1. In the case of distinct grounds, since the two ports have
Figure 2.6: Phase field distribu- tion at 10 kHz with common ground
Figure 2.7: Magnetic field distri- bution at 10 kHz with separated
ground
Figure 2.8: Magnetic field distri- bution at 200 MHz with common
ground
Figure 2.9: Magnetic field distri- bution at 200 MHz with separated
ground
the same impedance and the line impedance is very low, S11 will be equal to 0.
Figure 2.12 shows how much the equivalent port impedance is very high at the resonant frequency and is very low frequency (0.35 Ohm).
2.3 Screen current estimation
2.3.1 Background
High currents flowing in cable screens could cause heating issues and could even burn the cable in the worst case. In some applications there are requirements that limit the amount of the screen currents. Thus it is essential that the screen cables should be adequately well modelled in the design stage in order to prevent
Figure 2.10: Magnitude of S11 in the frequency domain: in blue when a common ground is used and in green when separated grounds are used
Figure 2.11: Magnitude of S11 in the frequency domain: in blue when is used a common ground and in green when separated grounds are used
Figure 2.12: Port impedance behaviour in the frequency domain: in blue when a common ground is used and in green when separated grounds are used
Elements Materials Diameter [mm] Thickness [mm]
Conductor Aluminum 12.8 -
Inner insulation PE - 3.5
Screen Copper - 0.1479
Outer insulation PVC - 2
Table 2.1: Cable parameters
high screen currents. The simulation of the three phase screen motor cables is presented. The characteristics of the cables are given in Table 2.1.
2.3.2 Assumptions
• The ground is a perfect conductor.
• The three-phase cables are modelled as shown in figure 2.13. The distance between conductors is 6 cm.
• A solid shield is considered for the calculations.
• Simulation range is 0 to 350 Hz to compare with the measurements performed in [3].
• Three symmetric current sources, with amplitude 1, are used to feed the phase center wires. Symmetric current phases are 0, 2π/3, −2π/3.
Figure 2.13 presents the screen cable model with the different current flowing in the system.
Figure 2.13: Theoretical model used for building the equations [1]
The following equations found in [1] link the phase currents and the screen currents.
0 = RsIa0 + jω(La0a0Ia0 + La0b0Ib0+ La0c0Ic0) + jω(La0aIa+ La0bIb+ La0cIc) (2.2) 0 = RsIb0 + jω(Lb0a0Ia0+ Lb0b0Ib0 + Lb0c0Ic0) + jω(Lb0aIa+ Lb0bIb+ Lb0cIc) (2.3) 0 = RsIc0 + jω(Lc0a0Ia0 + Lc0b0Ib0+ Lc0c0Ic0) + jω(Lc0aIa+ Lc0bIb+ Lc0cIc) (2.4) Where the parameters are:
results will be used.
2.3.3 Simulation results
According to the previous description, the model has been implemented in CST (Computer simulation technology) cable studio. The screen current over the phase current ratios has been computed for each phase and compared with the measured results. Figure 2.14presents the simulated results whereas figure 2.15presents the measured results. The measured results show some differences from the simulated ones. These differences might come out from the following reasons:
• Missing information about screen (braided shield characteristics);
• Screen is considered as solid shield in the model;
• The measured currents are generated by a converter which delivers the funda- mental frequency and the harmonics. The probes measure the current RMS values, which include high frequency current components. High frequency current components will increase the measured screen currents;
• The ground plane in the model is different from the actual ground plane in the project.
50 100 150 200 250 300 350 0
0.1 0.2 0.3 0.4 0.5 0.6
Frequency [Hz]
Figure 2.14: Screen current ratios computer by CST: Is/Ic phase a in blue, Is/Ic phase b in green, Is/Ic phase c in red.
Figure 2.15: Screen current ratios measured
insulation materials. As presented in [4] Two coefficients which allow to estimate the voltage at the terminal are:
• Reflection coefficient
ΓL= (ZL− Z0)/(ZL+ Z0) (2.5)
• Transmission coefficient
βL= 1 + ΓL (2.6)
Where ZL and Z0 are the load and line characteristic impedance respectively. The terminal voltage will be:
V = V++ V− = V++ ΓL∗ V+ = βL∗ V+ (2.7) Where V+ is the incident voltage and V− is the reflected voltage.
If ZL= ∞ (open load terminal) , V=1;
If ZL= 0 (short circuit load terminal), V=0;
2.4.2 Simulation results
A 10m length screen cable RG58 is used in this part. The port generates the impulse signal with unity amplitude and with rise time 11.3ns. The figure 2.16 shows the schematic of the system. The port impedance is equal to 50mΩ. Figure 2.16and figure 2.17show respectively the results when resistive load are1M Ω and 1mΩ. The following observations can be made:
• When the load is 1M Ω, over voltage phenomena are observed. In this case we almost reach two times the port voltage.
• When the load is 1mΩ, the reflected voltage is very small too.
The delay between the close end and the far end voltage depends on the cable length.
Figure 2.16: Schematic of the reflection test in CST
The travelling wave produces voltage oscillations at the ends of the line as can be seen in figure 2.18. This figure presents the cable far end voltage and the port voltage when the far end of the line is connected to 1 MOhm resistor. The port signal has the maximum voltage equal to 700 V with rise time equal to 0.2 micro seconds.
The damping of these oscillations depends on the cable terminal and the cable losses. The previous formula shows the relationship between the cable length and the frequency of the propagation wave. The frequency decreases when the cable length increases.
Figure 2.17: Port voltage in red, line close end voltage in green and far end voltage waveforms for ZL= 1M Ω
Figure 2.18: Port voltage in red, line close end voltage in green and far end voltage waveforms for ZL= 1mΩ
0 10 20 30 40 50 60 -600
-400 -200 0 200 400 600 800 1000 1200 1400
X: 10.6 Y: 1279
Time [ sec)]
Voltage [Volt]
Inverter output voltage Motor input voltage
Figure 2.19: The port voltage signal (in blue) and the far end voltage signal (in green).
2.5 Conclusion
The grounding method has been studied in section 2.2, the screen current esti- mation in section 2.3 and the termination effects of the power cables in section 2.4. From section 2.2, we could observe that the current can flow in the sys- tem without difficulties at any frequency when the common ground is used. The separated grounds do not facilitate the power flow at low frequency. The second section shows that the screen current magnitude depends on the frequency. The screen current can increase and reach more than 11 times the center wire cable until at resonant frequency. The last section shows that during the transient pe- riod, the terminal voltage amplitude depends on the load impedance. The over voltage occurs when the load impedance is large compared to the cable impedance.
Oscillating phenomena are the consequences of the propagation voltage wave and the frequency of these oscillations depends on the cable length.
This chapter will focus on the modeling of the internal components of a 1500V AC-DC motor converter module. The components include the IGBTs, The bus- bar interconnections and the DC link filters. Some simplifications have been made in order to allow good understanding of what happens. Figure 3.1 illustrates how the structure has been simplified for the study. The DC source is discrete port from the DC minus to the DC plus as shown in figure 3.2. Figure 3.3 shows the load in star connection here modeled by the discrete ports.
The high frequency responses of the IGBTs are modeled using the following two approaches:
• Parasitic model of the IGBT;
• IGBTs modeled as voltage controlled switch.
Only the high power conducting part of the converter module has been modeled.
This includes the IGBTs, The DC bus-bars, The bus-bars or the AC-side and the input filters. The low voltage components, for example, the gate drive unit and the drive control unit are not modeled. Figure 3.3 shows the setting of the load (connected to the AC side) and the DC signal source. The DC signal source is discrete port (figure 3.2) from the DC minus to the DC plus.
18
Figure 3.1: Simplification of the model
Figure 3.2: DC link excita- tion
Figure 3.3: Load setting in AC side
Figure 3.4: Simultion setting
3.2 Parasitic modeling of the IGBTs
This approach consists of replacing the IGBT by lumped elements. For instance, inductance can be used for the on-state and capacitance can be used for the off- state as presented in [2]. The main advantage of this model is that it does not take a big amount of memory. Then, it is fast, easy to check and it gives a very
Figure 3.5: Wire copper connecting terminals of IGBT
Figure 3.6: Lumped element connecting the IGBT terminals
good idea of the distribution of the electromagnetic field in the system in the frequency domain and the EMI characteristic of the system. Furthermore, the instantaneous load power is consistent when compared to simple lumped circuit converter models. The < 110 > state shown with the equivalent circuit given in figure 3.6, results shown in figure 3.7. It gives also an idea of the resonance behavior of the power converter module.
3.2.1 Parameter estimation
Table 3.1 shows the IGBTs parameter used in this simulation and figures 3.5 and 3.6 shows the connection setting in CST MW design studio on phase U. Figure 3.7 shows the excitation signal from the DC bus parameters with the following values:
• Rise time: 0.2 micro seconds;
• Fall time: 0.2 micro seconds;
Table 3.1: IGBTs model
IBGT states Physical connections Lumped element connections
On-state Copper wires 10nH inductance
Off-state No-connections 235 pF capacitance
Figure 3.7: Excitation signal pattern
Figure 3.8: Schematic of the circuit. The < 110 > state of the inverter is simulated with 100Ω in star connection
• Maximum voltage 1500 V.
The inverter will be in the configuration < 110 >. This means that in steady state the voltage will distribute itself according to equivalent circuit shown in figure 3.8. Since the system is almost lossless, the system can be analyzed as lossless transmission line. The are two IGBT cases considered in the < 110 >
state. This includes:
1. ”On” and ”Off” states represented by physical connections;
2. Lumped elements with values given in Table 3.1.
Figure 3.9: Voltage responses
3.2.2 Observations
The following observations were made:
• Figure 3.9 and 3.10 show small oscillations can be observed on the re- sponses at about 20 MHz. These oscillations are coming from the resonance behavior including from the IGBT parasitics, the bus parasitic inductances and distributed capacitances.
• The ripple amplitude is bigger on the DC link signals. That is because the situation is comparable with opening and closing a voltage source on the line terminals.
• There is increased damping in the case with lumped elements IGBT ripples as shown in figure 3.9 and 3.10.
Figure 3.10: Current responses
3.2.3 Results
The voltage and the current responses have been measured on the DC and the load side in both cases and respective results can be compared. Using the schematic in figure 3.8, the DC link is excited with 1500V dc. In the < 110 > state, the steady state voltages and currents are summarized in Table 3.3. Figure 3.9 and 3.10 show that the steady state model results are consistent. However, due to the parasitic capacitances and inductances, we can observe the presence of the ripples.
(charging resis- tor)
micro farads
The most invisible effect of the parasitic capacitances and inductances is the DC link. In both models, the ripple amplitudes are very large. The wire model presents higher damping path than the lumped elements’ model.
3.2.4 Improvement of the damping
The 20 MHz ripple amplitude could be reduce in different ways. For instance, by adding filters in the system or reducing the dv/dt. This section will present the first method. The filter parameter are presented in the table 3.2. Figure 3.11 shows the setting in 3D, while figure 3.12shows the lumped circuit. The voltage and current responses with the filters installed are shown in figures 3.13and 3.14 respectively.
Results
• Smaller oscillations can be observed on the responses at 20 MHz.
• The oscillation are damped faster than in the simplified model.
Table 3.3: Mean values of voltage and current response
Elements DC-link U V W
Voltage [V] 1500 500 500 1000
Current [A] 10 5 5 10
Figure 3.11: Setting of the system with the filter
Figure 3.12: Schematic of the filter
• The filter capacitors charged during the simulation.
• The DC link capacitor increases the supply current because it is empty at the beginning of the simulation as.
Figure 3.13: Voltage responses
3.3 IGBTs modeled as voltage controlled switches
3.3.1 Description
Connectors have been created in the IGBTs locations in the converter box module as shown in figure 3.15. A zoom of the IGBT connectors is shown in figure 3.16.
These terminals have been connected with the switches. Control voltage of the switches are delivered by the port excitation in CST design studio. The drawback of this approach is the increase in simulation time, and it requires more memory.
Figure 3.14: Current responses
However, it provides a means of controlling the IGBTs to simulate different mod- ulation patterns and observe the impact on the EM environment. IGBT models could be imported from different softwares or libraries. Regarding the DC link setting, a charged 4mF capacitor is used. It imposed 1500 Vdc between these converter terminals. It is large enough to maintain constant voltage during the simulation. Also, a connecting 30Ω resistor is used for limiting the current when the capacitor makes the contact with the structure. In this chapter, results ob- tained using Voltage controlled switches (VCSW) and imported IGBT from Pspice are compared. VCSW offers the advantage to control the rise time during the sim- ulation while imported IGBTs from Pspice have the fixe rise time.
Figure 3.15: Motor converter module with the IGBT ports
Figure 3.16: IGBT ports
3.3.2 Results
It could be observed that the system reacts at the DC link voltage connection.
Concerning the switching pattern, all the three phases are connected in the same time. The ringing strongly depends on the voltage excitation or/and the rise time.The riging strongly depends on the voltage excitation rise time. The voltage responses are presented in figures 3.17and 3.18. Figures 3.19and 3.20. Reducing the dv/dt, reduces the ringing. Figure figure 3.21 presents the voltage with the rise time increased to 0.4 µs. There is less ripple in figure 3.21compared to figures 3.17 and 3.18.
3.4 Conclusion
The resonance behaviour on a motor power converter has been studied in section 3.2 and IGBTs models have been proposed in section 3.3. From section 3.2,
Figure 3.17: Voltage responses using imported IGBTs
Figure 3.18: Voltage responses using VCSW
Figure 3.19: Current responses using imported IGBTs
Figure 3.20: Current responses using VCSW
Figure 3.21: Voltage responses using VCSW with rise time equals 0.4 µs
the parasitic resonance aspects of the converter could be predicted. The intro- duced ripples could be damped by using filters or reducing the speed the voltage variations (dv/dt) of the source. This approach is fast but it does not allow to change the switching configurations of the MCM without interrupting the simu- lation. Section 3.3 allows to switch the IGBTs while the simulation is running.
The dv/dt could be also modified in order to reduce the ripples after switching. It is more realistic but it takes more memory and uses different physics in the time.
This inverter could be coupled with cables and the 2D and 3D results could be obtained as done in chapter 4
This chapter focuses on the creating of the complete model of the motor drive system. The cable model presented in chapter 2 is coupled to the motor con- verter module (MCM) model presented in chapter 3. The motor load is simply represented as a Y-connected 100Ω resistors. The DC-link capacitor is charged through 30Ω resistor to charge the converter. Figure 4.1 presents a schematic of the complete drive system model. Unshielded cables are used to connect the load, just for study purposes. The cable configuration is presented in figure 4.2.
Figure 4.1: schematic of the system
32
Figure 4.2: Cable setting
Connections have been created in the switch locations in MW studio in order to connect the terminals with the voltage controlled switch as illustrated in figure 4.4. The simulation parameters are the following:
• DC link steady state voltage: 1500 V;
• The load: 3 resistors of 100 Ohms in star connection;
• Commutation from state 110 and 010;
The used cables are 5m unshielded cables with parameters given Table 2.1.The MCM is constructed in MW studio, while the cable model is constructed in CS studio. The MCM is coupled with cables model in MS design studio. Outside the MCM, the DC minus is grounded directly, while in the converter the DC minus is grounded through the EMI filters. Figure 4.5 illustrates the DC minus grounding through EMI filter.
The meshing should be fine enough for allowing the connection between cables and metallic structure. The possible ways of doing that is maintaining the frequency
Figure 4.3: Cables coupled with converter
and increase the wavelength ratio or maintaining the wavelength ratio and increase the frequency. By changing these parameters it is possible to find the solution which minimizes the time of simulation.
The schematic of the complete system is shown in Figure ??. The complete system in 3D view is shown in Figure 4.3, the 3D model is in-closed in the white box in the center. Only the external (hidden) lumped connections are visible on the schematic, for example the voltage control switches and the excitation ports.
4.2 Voltage and current responses
Figure 4.7 presents the load voltages while figure 4.8 dc link voltage and current and and the ground current. On the DC link, a first transient could be observed that corresponds to the contact between structure and dc link capacitor. When the current reaches zero amperes, the charging resistor is short-circuited and the voltage remains at 1500V. Between 0.5µs and 1µs, the converter switches to <
110 >, between 2µs and 3µs, it commutate from < 110 > to < 010 > and
Figure 4.4: IGBT locations
Figure 4.5: Grounding and EMC filter setting
Figure 4.6: schematic of the system
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 -1000
-500 0 500 1000
Phase U load voltage
time s
VoltageV
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
-500 0 500 1000 1500
Phase V load voltage
time s
VoltageV
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
-1500 -1000 -500 0 500
Phase W load voltage
time s
VoltageV
Figure 4.7: load Voltages
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 200
400 600 800 1000 1200
time s
VoltageVolt
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
-10 0 10 20 30 40 50
DC link current and ground curent
time s
CurrentinAmpere
DC link current Ground current
Figure 4.8: DC link current and voltage
Figure 4.9: 3D Electric field at 10 kHz
Figure 4.10: 3D Electric field at 500 MHz
Figure 4.11: Cross section at the middle of the line at 10 kHz
Figure 4.12: Cross section at the middle of the line at 500 MHz
at around 4µs the switches on the first leg are both opened. Figure 4.7 and figure 4.8 show the voltage response and the current response respectively.
4.3 3D field distribution
Once the current and voltage distribution in the model are obtained, the EM fields can be computed. Figure 4.9 shows the electric field at 10kHz, while figure 4.10 shows the E-field at 500M Hz. Figures 4.11and 4.12shows the E-fields in a plane traversing the cables. The magnetic fields are distributed at 10kHz is shown in figure 4.13 while the H-field at 500M Hz is shown in figure 4.14.
Figure 4.13: 3D Magnetic field at 10 kHz
Figure 4.14: 3D Magnetic field at 500 MHz
Figure 4.15: Probe location in the system
4.4 EM field 2D plots
Some probes have been put on different locations in the electric drive vicinity in order to measure the EM fields. Figure 4.15 shows the probe locations. The following probe results are shown for these locations:
A Probe measuring the E-field and The H-field to an IGBT inside the MCM;
B Probe measuring the E-field and The H-field 1 m left to the MCM;
C Probe measuring the E-field and The H-field 1 m above the mean length of the cables;
Figure 4.16: Electric field in point A, close to the IGBT in V /m
Figure 4.17: FFT Electric field in point A, close to the IGBT V /m
D Probe measuring the E-field and The H-field close to the mean length of the cables;
Figure 4.18: Magnetic field in point A, close to the IGBT in A/m
Figure 4.19: FFT Magnetic field in point A, close to the IGBT in A/m
Figure 4.20: Electric field in point B, left to the MCM in V /m
Figure 4.21: FFT Electric field in point B, left to the MCM in V /m
Figure 4.22: Magnetic field in point B, left to the MCM in A/m
Figure 4.23: FFT Magnetic in point B, field left to the MCM in A/m
Figure 4.24: Electric field in point C, 1m from the cable middle in V /m
Figure 4.25: FFT Electric field in point C, 1m from the cable middle V /m
Figure 4.26: Magnetic field in point C, 1m from the cable middle A/m
Figure 4.27: FFT Magnetic fieldin point C, 1m from the cable middle in A/m
Figure 4.28: Electric field in point D, close to the cables in V /m
Figure 4.29: FFT Electric field in point D, close to the cables in V /m
Figure 4.30: Magnetic field in point D, close to the cables A/m
Figure 4.31: FFT Magnetic in point D, close to the cables in A/m
4.5 Conclusion
Chapter4 describes the 3D assembly of motor converter module and unshielded cables. The steady state of the voltage and current responses present the same results that could be obtained from a 2D simulator and including ripples due to the parasitic resonant characteristic of the system. A resistive load has been connected to the system. This load allows high di/dt and dv/dt in the system. 3D results in section4.3 show how the electric field is perpendicular to the ground in low frequencies even though it is not the same in high frequencies. The magnetic field has an uniform distribution in low frequency unlike the high frequency field distribution. Section4.4shows the 2D measurements of the electromagnetic field in the vicinity of the electric drive system. The converter shields the electromagnetic radiations produced by the switches. Further investigations could be done in order to determine the more critical frequency components of the fields and thus prevent the unwanted interferences with the signalling system.
CST-computer simulation technology is a very suitable tool for the 3D modelling of the drive system. It allows for a suitable integral of the 3D models, the cable transmission line models and the lumped circuit models. Specific IGBT models can be imported. Different load cycles or modulation schemes can be simulated from the lumped circuit interface in CST MS and DS, to obtain most case field distributions. The the aim of the thesis, multiple physics have been used and this used leading to obtained results. However, some results cannot be used di- rectly from CST. For example, the frequency domain (Fourier transform) should be checked carefully. It is more prudent taking the time domain results and mak- ing the frequency domain analysis using a different tool like Matlab. Also, from [2], the impedance value (complex form) also have to be reviewed for avoiding mistakes in the conclusion.
5.2 Future Work
The thesis only investigate the converter and cable used in a propulsion system.
The use of resistive load allowing the model stability and robustness. Fast di/dt and dv/dt could study and they give an idea of the most critical case. Next step, could be the 3D motor coupling to be able to study the winding effect and completing the common current circuit. Furthermore, the achievement of the
50
control system (signal cables, IGBT drivers and other sensors) could ease the prediction of major issues in the drive system.
5.3 Conclusion
Electromagnetic phenomena are present in electrical drive systems. They can come out from different sources but this thesis work focused on converter feeding high frequency electrical components into the cables. In this thesis CST has been used to create 3D models of a motor drive system. The 3D models has been integrated with lumped circuit interface to simulate the different load cycles. The current and the voltage distribution is also obtained. The created model with some modifications could be used in projects to predict characteristics early in the design.
[4] Electrotechnical modeling and design. KTH, 2012.
52
