Modeling FrequencyDependent Effects in the Time Domain

Mark Christini, Mark Solveson Ansoft

Acknowledgments
Siemens for supplying measurements of coaxial cable used for verification of simulations and for supplying schematic for Maxwell 2D transient simulation.

Presentation Goals
Introduction of frequency-dependent modeling challenges. Discussion of three methods for creating frequency-dependent models. Calculate and verify R,L,C,G parasitics for a specific coaxial cable example. Creation of frequency-dependent model using Q3D. Simulation of cable model in Simplorer.

Introduction
The need to fully understand how devices operate through the desired frequency spectrum has increased due to:
the trend to operate devices at higher frequencies the increased use of inverter power supplies which cause non-sinusoidal voltages, currents, and extra losses

Products such as cabling, bus bars for power delivery, signal interconnects, and integrated power-electronic modules exhibit frequency-dependent behavior due to eddy current and skin effects. In these cases, it may not be sufficient to rely on resistance and inductance extracted at a single operating frequency.

Introduction
This presentation will demonstrate how Q3D and Simplorer can be used to easily simulate cable models in the time-domain. Three different methods used for cable modeling will be reviewed. Traveling waves, overvoltages, and resonant frequencies will be determined using a frequency-dependent cable model.

Q3D Overview
Q3D is a tool streamlined for quickly characterizing electrical parasitics of interconnect structures. Q3D includes:
Q3D Extractor: 2D Extractor: 3D quasi-static lumped RLC parameter extractor. Linear permeability = 1. 2D T-line RLGC parameter extractor. Linear permeability.
DC 1.5 GHz

15 GHz

???

Q3D L = 1cm 10 2D Extractor X

6

W = 5mm

HFSS

Three Cable Modeling Methods

1. Lumped, frequencyindependent, hierarchical structure model in Simplorer. 2. Lumped, frequencydependent model in Maxwell 2D transient using external circuits. 3. Lumped, frequencydependent model exported from Q3D into Simplorer.

VM_Linear_1MHz_Model ECELink4 E6

+
V

N_1

N_3

+
V

VM11
N_2 N_4 N_5
T2D

R11 1e3 Ohm

Method 1: Lumped, FrequencyIndependent, Hierarchical Structure

Cable Model Voltage Source Inverter Topology

A voltage source PWM IGBT inverter configuration is used to feed the motor. The cable model is frequency independent but lumped into 20 cells or segments. The electrical motor model is a surge impedance equivalent model obtained from a set of measurements on a 100 hp motor.
8

High Frequency Motor Model

Method 1: Lumped, FrequencyIndependent, Hierarchical Structure

R,L,G,C 5 Segment Model

20 Segment Model

Method 1: Lumped, FrequencyIndependent, Hierarchical Structure

Voltage results shown for varying cable length. Traveling wave over-voltages and resonant frequency vary with length. Considerations:
Model is manually constructed. Model is frequency independent.

10

Method 2: Lumped, FrequencyDependent Model using the Maxwell Transient Solver

Use Maxwell 2D transient with series of independent coaxial models in one FEA solution (assumes no coupling occurs). Connect in lumped network in 2D transient external circuit. Add in shunt capacitance from 2D electrostatic simulation.

Electr ost

atic A n

alysis

11

Method 2: Lumped, FrequencyDependent Model using the Maxwell Transient Solver

Considerations:
Voltage transients do not include high frequency harmonic components due to insufficient number of cells used. Circuit is manually created.
Simulation Data 200ft Cable
700 600 500 Volt 400 300 200 100 0 0.0E+00 5.0E-07 1.0E-06 1.5E-06 2.0E-06 2.5E-06 3.0E-06 3.5E-06 4.0E-06 Time
Voltage Step
Voltage at Cable End

Experimental Data 200ft Cable

800 700 600 500
Voltage Step
Voltage at Cable End

Simulation

Volt

400 300

100 0 0.0E+00 5.0E-07 1.0E-06 1.5E-06 2.0E-06 2.5E-06 3.0E-06 3.5E-06 4.0E-06 Time

Simulation Data 600ft Cable

800 700 600 500
800 700 600 500 Volt
Voltage Step
Voltage at Cable End

Experimental Data 600ft Cable

Volt

400 300 200 100 0 0.0E+00 2.0E-06 4.0E-06 6.0E-06 Time 8.0E-06 1.0E-05 1.2E-05

400 300 200 100 0 0.00E+00 2.00E-06 4.00E-06 6.00E-06 Time 8.00E-06 1.00E-05 1.20E-05

Voltage Step

Voltage at Cable End

12

Measured

200

Method 3: Frequency-Dependent, Model Using Q3D and Simplorer Preferred

Single phase conductor model for shielded coaxial cable. FEA results validated with analytical calculations for R,L,C,G. Frequency-dependent netlist is derived using Q3D (2D Extractor). Netlist imported into Simplorer for system simulation.
General Cable Uniblend
EPR/Copper Tape Shield/PVC, Medium-Voltage Power, Shielded 15kV, UL Type MV-105, #2AWG Copper shield Silicon Polyethylene Silicon Copper

13

Analytical Calculations for Coaxial Cable

Outer diameter of conductor (d1) = 6.858 mm Outer diameter material 1 (d2) = 7.874 mm Outer diameter material 2 (d3) = 19.177 mm Inner diameter shield (d4) = 25.396 mm Outer diameter shield (d5) = 25.65 mm

R DC =

1 1 1 1 L = + = + = 0.0021599 [/m] 7 2 2 7 2 7 2 2 7 A 5.8 * 10 * (r5 r4 ) 5.8 * 10 * r1 5.8 * 10 * (0.012825 0.012698 ) 5.8 * 10 * (0.00369392 )

L=

o d4 25.396 7 ln = 2 * 107 ln = 2.62 * 10 [H/m] 2 d1 6.858 2 o 2 * 2.25 * 8.85 * 1012 = = 1.29 * 10 10 [F/m] 1 d 2 1 d3 1 d4 1 7.874 1 19.177 1 25.396 ln + ln + ln ln ln ln + + r1 d1 r2 d2 r3 d3 11.9 6.858 2.25 7.874 11.9 19.177

C=

14

FEA Calculations for Coaxial Cable

Q3D and Maxwell 2D give identical results for R,L,C. Q3D also calculates R,L vs. frequency. Q3D exports frequency dependent R,L,G,C netlist. Frequency independent results for R,L,C,G
Analytical Capacitance (F/m) Inductance 1MHz (H/m) Resistance 1MHz (/m) Conductance (mhos/m) 1.29 10-10 2.62 10-7 --Q3D 1.2914 10-10 2.618 10-7 0.015369 1.5565 10-7 Maxwell 2D 1.2915 10-10 2.618 10-7 0.015391 --

15

FEA Calculations for Coaxial Cable

Frequency dependent results for R,L

Q3D

Identical results for R and L vs. frequency

Maxwell 2D
16

Q3D 2D Extractor Overview

Include Frequency Sweep

17

Q3D 2D Extractor Export

Exporting Netlist from Q3D
Format: Simplorer Choose number of lumped cells Specify length in meters Compute Cells . . . t_rise/t_delay > 20
For accurate representation of traveling waves.

Create Equivalent . . .
to export model

Check Frequency Dependent

18

Transmission Line Model with 1-cell

Frequency independent circuit

Lumped T-model

19

Transmission Line Model with 1-cell

Exported Netlist Lumped T-model
4 1

half shunt
3

half

Frequency dependent elements

Frequency independent elements

20

Q3D/Simplorer vs. Experimental Results

400V step input with risetime = 0.25usec. Lumped frequency-dependent model used with 50 cells. Over-voltage and resonant frequency vary with cable length.
Overvoltage Overvoltage vs.vs. time time Overvoltage vs. time 770 500 Input voltage 200ft cable 400ft cable 600ft cable

Voltage

-505 0
21

10.00u

20.00u

30.00u

40.00u 50.00u

Time (seconds)

Q3D/Simplorer vs. Experimental Results

800
600
Experimental Data 200ft Cable
800

200ft
V o l t

700 600 500 400 300

400

200

200 100

0
0
800

1u

2u

3u

0 0.0E+00 5.0E-07 1.0E-06 1.5E-06 2.0E-06 2.5E-06 3.0E-06 3.5E-06 4 Time

Experim ental Data 400ft Cable

800

600
400
200
0

400ft
V o lt

700 600 500 400 300 200 100 0 0.0E+0 1.0E-06 2.0E-06 3.0E-06 4.0E-06 5.0E-06 6.0E-06 7.0E-06 8.0E0 Time

0
800
600

1u

2u

3u

4u

5u

6u

7u
Experimental Data 600ft Cable
800 700 600 500
V o lt

600ft

400

400 300

200

200 100 0 0.00E+00 2.00E-06 4.00E-06 6.00E-06 Tim e 8.00E-06 1.00E-05 1.20E

0
0
1u
2u
3u
4u
5u
6u
7u
8u
9u 10u

Simulation
22

Experimental

Simplorer

Simplorer Transient Results

Simulation of three cases for coaxial cable:
Linear cable model exported at 1Hz (blue). Linear cable model exported at 1MHz (light blue). Frequency Dependent model (red).
Voltage versus Time Using Different 2D Extractor Models 2.0

VM_Linear_1MHz_Model
Voltage versus Time Using Different 2D Extractor Mode

ECELink4 E6

+
V

N_1
1.5 VM11

1.1

N_3

+
V

1.1

N_2
Voltage
1.1 Voltage

N_4 N_5
T2D

Damping

R11 1e3 Ohm

1.0

Phase
VM11.V [V] VM_Linear_1Hz_Model.V [V] VM_Linear_1MHz_Model.V [V][V] VM11.V VM_Frequency_Model.V [V] VM_Linear_1Hz_Model.V [V]
10.00u Time (Seconds)
18.00u 18.50u

1.0

500.0m
950.0m

0 0
900.0m 5.00u 876.5m 17.55u

15.00u
19.00u

VM_Linear_1MHz_Model.V [V] 20.00u VM_Frequency_Model.V [V]

19.50u 20.00u

Time (Seconds)

23

Simplorer Transient Results

E6

Simulation of Cable in with 100Hp:

ECELink4
N_1 N_3

Simulation of three cases for coaxial cable with load:

Linear cable model exported at 1Hz (blue). Linear cable model exported at 1MHz (light blue). Frequency Dependent model (red).

R7

R2

+
V

+
V

VM11
N_2 N_4
T2D

VM_Linear_1MHz_Model

C1

L1

N_5

R3

R4

R5

R6

C2

L2

C3

L3

Electrical motor model is a surge impedance equivalent model obtained from a set of measurements on a 100 hp motor.

Voltage versus Time Using Different 2D Extractor Models 2.0 1.8 1.5 1.3 Voltage 1.0 750.0m 500.0m 250.0m 0 0 5.00u 10.00u Time (Seconds)

VM11.V [V] VM_Linear_1MHz_Model.V [V] VM_Linear_1Hz_Model.V [V] VM_Frequency_Model.V [V]

15.00u 20.00u

24

3Phase Coaxial Cable Simulation

3Phase cable cross-section

cable 100Hp with 3 phase inverter:

ECELink1 N_1
U

N_4 N_5

+
Inverter
V

N_2

VM_ab

N_3

N_6

N_7

VM_ab2
V

+ +

T2D

VM_bc2
VM_ca2

+
V

VM_bc VM_ca

R5 R4 L4

R3

R7

R8

R9

Extracted Model

C4

C2

L2

C3

L3

Inverter and Motor Voltage vs Time

3Phase simulation results with a 100hp motor surge impedance load.

2.00k

Voltage

VM_ab2.V [V] VM_bc2.V [V] VM_ca2.V [V] VM_ab.V [V] VM_bc.V [V] VM_ca.V [V]

-2.00k 0 5.00u Time (Seconds) 10.00u

25

Summary
Q3D and Simplorer can be used to easily simulate cable models in the time-domain. Three unique methods for creating system models have been presented:
Lumped, frequency-independent, hierarchical using Simplorer. Lumped, frequency-dependent using Maxwell 2D transient with external circuit. Lumped, frequency-dependent model exported from Q3D into Simplorer.
VM_Linear_1MHz_Model ECELink4 E6

+
V

N_1

N_3

+
V

VM11
N_2 N_4
T2D

R11 1e3 Ohm

N_5

Voltage versus Time Using Different 2D Extractor Mode 1.1

1.1

Traveling waves, over-voltages, and resonant frequencies were shown in several examples.
26

1.1 Voltage

1.0

950.0m

900.0m 876.5m 17.55u 18.00u 18.50u Time (Seconds) 19.00u

VM11.V [V] VM_Linear_1Hz_Model.V [V] VM_Linear_1MHz_Model.V [V] VM_Frequency_Model.V [V]

19.50u 20.00u

Other Considerations: Q3D/Simplorer Cable Studies

Lumped cell study with different number of cells: Frequency dependent 200ft lumped with 1 cell. Frequency dependent 200ft lumped with 10 cells. Frequency dependent 200ft lumped with 50 cells.

27

Other Considerations: Q3D/Simplorer Cable Studies

Lumped cell study
Simulation of three cases for coaxial cable: number of cells = 1, 10, and 50. 200meter, lumped model, frequency dependent, 400V step with risetime = 0.25usec. Need 50 cells to get t_rise/t_delay > 20.
850.00

750.00 900.00

1 cell

800.00 500.00

1 cell 10 cells 50 cells

600.00

-290.00 0 50.00u
755.00

100.00u

400.00

500.00

50 cell model needed to get accurate voltages

10 cell

500.00

50 cell

200.00
0

-470.00

28
0 50.00u

0 0
100.00u

-420.00 0

10.00u

50.00u

100.00u

20.00u

Other Considerations: Q3D/Simplorer Cable Studies

Frequency Sweeps. Bode Plots can help aide in understanding the frequency response of a system.
Perform AC analysis in Simplorer. Observe gain and phase vs. frequency.

29

Conclusions
The Q3D Simplorer approach is the most robust and complete of the three methods presented. The Q3D (2D Extractor) models account for both frequency-dependency and traveling wave harmonics. The Q3D Simplorer approach is seamless and efficient. It automatically provides models for time domain simulations for cable modeling. This approach can help engineers model traveling waves, over-voltages, and frequency-dependent phenomenon that occur in all types of interconnect structures.

30

Future Work
Simulate and validate other cable designs. Include various machine inverter models to study effects of frequency dependency, line lengths, rise times, cable designs, and load configurations using Q3D and Simplorer. Include frequency-dependent exported models from Q3D of bus structures, connectors, IPEMs, or other electrical devices.

31

References
References:
A.C. Baisden, D. Boroyevich, J.D. van Wyk, Center for Power Electronics Systems (CPES) Virginia Polytechnic Institute and State University, Investigation of Conducted EMI in Converters with an RF Filter, IPEC, 2005. Introduction to Electromagnetic Compatibility by Clayton R. Paul, 1992 by John Wiley & Sons Electrical Transients in Power Systems by Allan Greenwood, 1971 by John Wiley & Sons Power System Analysis by Charles A. Gross, 1986 by John Wiley & Sons

32