Electromagnetic Field Simulation

Lecture 2: Circuit Simulation

Prof. Dr. Herbert De Gersem, Dr. Mané Harutyunyan, Prof. Dr. Sebastian Schöps
Computational Electromagnetics Group

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 1
Overview

1 Circuit Simulation
Kirchhoff’s Laws
Lumped Elements
Electric Circuits are Graphs
Modified Nodal Analysis
Remarks on the Implementation

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 2
Overview

1 Circuit Simulation
Kirchhoff’s Laws
Lumped Elements
Electric Circuits are Graphs
Modified Nodal Analysis
Remarks on the Implementation

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 3
Overview

1 Circuit Simulation
Kirchhoff’s Laws
Lumped Elements
Electric Circuits are Graphs
Modified Nodal Analysis
Remarks on the Implementation

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 4
 Kirchhoff’s Current Law (KCL)

Ampère’s law for statics

𝑖4
𝐴4
∫ ∇ ⋅ 𝐽 ⃗ d𝑉 = ∫ 𝐽 ⃗ ⋅ d𝐴 ⃗ = 0,
𝑉 𝜕𝑉 𝑖5 𝐴5
𝐴3
with volume 𝑉 surrounding each node 𝑖3

𝐴1 𝐴2
𝑖1
𝑖2

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 5
 Kirchhoff’s Current Law (KCL)

Ampère’s law for statics

𝑖4
𝐴4
∫ ∇ ⋅ 𝐽 ⃗ d𝑉 = ∫ 𝐽 ⃗ ⋅ d𝐴 ⃗ = 0,
𝑉 𝜕𝑉 𝑖5 𝐴5
𝐴3
with volume 𝑉 surrounding each node 𝑖3
boundary is union of surfaces 𝐴𝑘

𝜕𝑉 = 𝐴1 ∪̇ … ∪̇ 𝐴𝑛
𝐴1 𝐴2
𝑖1
𝑖2

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 5
 Kirchhoff’s Current Law (KCL)

Ampère’s law for statics

𝑖4
𝐴4
∫ ∇ ⋅ 𝐽 ⃗ d𝑉 = ∫ 𝐽 ⃗ ⋅ d𝐴 ⃗ = 0,
𝑉 𝜕𝑉 𝑖5 𝐴5
𝐴3
with volume 𝑉 surrounding each node 𝑖3
boundary is union of surfaces 𝐴𝑘

𝜕𝑉 = 𝐴1 ∪̇ … ∪̇ 𝐴𝑛
𝐴1 𝐴2
𝑖1
currents are integrals 𝑖𝑘 = ∫ 𝐽 ⃗ ⋅ d𝐴 ⃗ 𝑖2
𝐴𝑘

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 5
 Kirchhoff’s Current Law (KCL)

Ampère’s law for statics

𝑖4
𝐴4
∫ ∇ ⋅ 𝐽 ⃗ d𝑉 = ∫ 𝐽 ⃗ ⋅ d𝐴 ⃗ = 0,
𝑉 𝜕𝑉 𝑖5 𝐴5
𝐴3
with volume 𝑉 surrounding each node 𝑖3
boundary is union of surfaces 𝐴𝑘

𝜕𝑉 = 𝐴1 ∪̇ … ∪̇ 𝐴𝑛
𝐴1 𝐴2
𝑖1
currents are integrals 𝑖𝑘 = ∫ 𝐽 ⃗ ⋅ d𝐴 ⃗ 𝑖2
𝐴𝑘
𝑛 𝑛
Sum of all currents: ∑ ±𝑖𝑘 = ∑ ± ∫ 𝐽 ⃗ ⋅ d𝐴 ⃗ = ∫ 𝐽 ⃗ ⋅ d𝐴 ⃗ = 0
𝑘=1 𝑘=1 𝐴𝑘 𝜕𝑉

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 5
 Kirchhoff’s Voltage Law (KVL)

Faraday’s law in statics

∫ ∇ × 𝐸⃗ ⋅ d𝐴 ⃗ = ∫ 𝐸⃗ ⋅ d𝑠 ⃗ = 0, 𝑠3
𝐴 𝜕𝐴
𝑠4 𝑢3
with surface 𝐴 framed by conductors 𝑢4
𝑢2
𝐴 𝑠2
𝑢5 𝑢1
𝑠5
𝑠1

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 6
 Kirchhoff’s Voltage Law (KVL)

Faraday’s law in statics

∫ ∇ × 𝐸⃗ ⋅ d𝐴 ⃗ = ∫ 𝐸⃗ ⋅ d𝑠 ⃗ = 0, 𝑠3
𝐴 𝜕𝐴
𝑠4 𝑢3
with surface 𝐴 framed by conductors 𝑢4
𝑢2
Boundary is a union of curves 𝑠𝑘 𝐴 𝑠2
𝑢5 𝑢1
𝜕𝐴 = 𝑠1 ∪̇ … ∪̇ 𝑠𝑏 𝑠5
𝑠1

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 6
 Kirchhoff’s Voltage Law (KVL)

Faraday’s law in statics

∫ ∇ × 𝐸⃗ ⋅ d𝐴 ⃗ = ∫ 𝐸⃗ ⋅ d𝑠 ⃗ = 0, 𝑠3
𝐴 𝜕𝐴
𝑠4 𝑢3
with surface 𝐴 framed by conductors 𝑢4
𝑢2
Boundary is a union of curves 𝑠𝑘 𝐴 𝑠2
𝑢5 𝑢1
𝜕𝐴 = 𝑠1 ∪̇ … ∪̇ 𝑠𝑏 𝑠5
𝑠1
Voltage is the integral
𝑢 𝑘 = ∫ 𝐸 ⃗ ⋅ d𝑠 ⃗
𝑠𝑘

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 6
 Kirchhoff’s Voltage Law (KVL)

Faraday’s law in statics

∫ ∇ × 𝐸⃗ ⋅ d𝐴 ⃗ = ∫ 𝐸⃗ ⋅ d𝑠 ⃗ = 0, 𝑠3
𝐴 𝜕𝐴
𝑠4 𝑢3
with surface 𝐴 framed by conductors 𝑢4
𝑢2
Boundary is a union of curves 𝑠𝑘 𝐴 𝑠2
𝑢5 𝑢1
𝜕𝐴 = 𝑠1 ∪̇ … ∪̇ 𝑠𝑏 𝑠5
𝑠1
Voltage is the integral
𝑢 𝑘 = ∫ 𝐸 ⃗ ⋅ d𝑠 ⃗
𝑠𝑘
𝑛 𝑛
Sum of all voltages: ∑ ±𝑢𝑘 = ∑ ± ∫ 𝐸⃗ ⋅ d𝑠 ⃗ = ∫ 𝐸⃗ ⋅ d𝑠 ⃗ = 0
𝑘=1 𝑘=1 𝑠𝑘 𝜕𝐴

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 6
Overview

1 Circuit Simulation
Kirchhoff’s Laws
Lumped Elements
Electric Circuits are Graphs
Modified Nodal Analysis
Remarks on the Implementation

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 7
 Lumped element models

𝐶
𝑅 𝐿

(a) Resistor (b) Inductor (c) Capacitor

𝑢
𝑖

(d) Voltage source (e) Current source

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 8
 Resistor

An Ohmic resistor is constant and time invariant

𝑢(𝑡) = 𝑅 𝑖(𝑡) or 𝑖(𝑡) = 𝐺 𝑢(𝑡)

with resistance 𝑅 or conductance 𝐺 = 1/𝑅. 𝑅

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 9
 Resistor

An Ohmic resistor is constant and time invariant

𝑢(𝑡) = 𝑅 𝑖(𝑡) or 𝑖(𝑡) = 𝐺 𝑢(𝑡)

with resistance 𝑅 or conductance 𝐺 = 1/𝑅. 𝑅

A simple resistance is given by

1 Δℓ
𝑅=
𝜎 Δ𝐴
with specific conductivity 𝜎, length Δℓ and cross section Δ𝐴.

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 9
 Resistor

An Ohmic resistor is constant and time invariant

𝑢(𝑡) = 𝑅 𝑖(𝑡) or 𝑖(𝑡) = 𝐺 𝑢(𝑡)

with resistance 𝑅 or conductance 𝐺 = 1/𝑅. 𝑅

A simple resistance is given by

1 Δℓ
𝑅=
𝜎 Δ𝐴
with specific conductivity 𝜎, length Δℓ and cross section Δ𝐴.
The power loss and energy are
𝑡
𝑃Ohm = 𝑢 𝑖 = 𝑅 𝑖2 𝑊Ohm = ∫ 𝑅 𝑖2 (𝑠) d𝑠.
0

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 9
 Resistor

An Ohmic resistor is constant and time invariant

𝑢(𝑡) = 𝑅 𝑖(𝑡) or 𝑖(𝑡) = 𝐺 𝑢(𝑡)

with resistance 𝑅 or conductance 𝐺 = 1/𝑅. 𝑅

A simple resistance is given by

1 Δℓ
𝑅=
𝜎 Δ𝐴
with specific conductivity 𝜎, length Δℓ and cross section Δ𝐴.
The power loss and energy are
𝑡
𝑃Ohm = 𝑢 𝑖 = 𝑅 𝑖2 𝑊Ohm = ∫ 𝑅 𝑖2 (𝑠) d𝑠.
0

Open circuit (𝐺 = 0, 𝑅 = ∞) and short circuit (𝐺 = ∞, 𝑅 = 0)

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 9
 Inductor

A circuit element with two pins is called inductor if

ΦL (𝑖(𝑡), 𝑡) = ∫ 𝐵(⃗ 𝑟,⃗ 𝑡) ⋅ d𝐴 ⃗

𝐴

where the flux density 𝐵⃗ depends on the current 𝑖(𝑡). 𝐿

For the voltage holds

d
𝑢(𝑡) =
Φ (𝑖(𝑡), 𝑡) .
d𝑡 L
In the case of a linear time-invariant inductor:
d𝑖
ΦL (𝑖(𝑡), 𝑡) = 𝐿𝑖(𝑡) and 𝑢(𝑡) = 𝐿 (𝑡)
d𝑡
2
Simple coil: inductance 𝐿 = 𝜇0 𝜇𝑟 𝑁 ℓ 𝐴 with vacuum permeability 𝜇0 ,
relative permeability 𝜇𝑟 , number of turns 𝑁, length ℓ and area 𝐴.
June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 10
 Capacitor

A circuit element with two pins is called capacitor if

𝑞𝐶 (𝑢(𝑡), 𝑡) = ∫ 𝜚(𝑟,⃗ 𝑡) d𝑉 ,
𝑉 𝐶
where the charge depends on the applied voltage 𝑢(𝑡)
and for the current holds
d
𝑖(𝑡) =
𝑞 (𝑢(𝑡), 𝑡) .
d𝑡 𝐶
For linear time-invariant capacitors the relation is as follows
d𝑢
𝑞𝐶 (𝑢(𝑡), 𝑡) = 𝐶𝑢(𝑡) and 𝑖(𝑡) = 𝐶 (𝑡).
d𝑡
Simple plate capacitor: capacitance 𝐶 = 𝜖0 𝜖r 𝐴 𝑑 with vacuum permittivity
𝜖0 , relative permittivity 𝜖r , distance 𝑑 and area 𝐴.
June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 11
 Sources and Other Elements

Independent voltage sources Independent current sources

Hello 𝑉 Hello
Hello 𝐼

𝑢(𝑡) = 𝑢s (𝑡) 𝑖(𝑡) = 𝑖s (𝑡)

𝑢s ≡ 0, short circuit 𝑖s ≡ 0, open circuit

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 12
 Sources and Other Elements

Independent voltage sources Independent current sources

Hello 𝑉 Hello
Hello 𝐼

𝑢(𝑡) = 𝑢s (𝑡) 𝑖(𝑡) = 𝑖s (𝑡)

𝑢s ≡ 0, short circuit 𝑖s ≡ 0, open circuit
Other circuit elements
Controlled sources
𝑢(𝑡) = 𝑢(𝑢control , 𝑡) 𝑖(𝑡) = 𝑖(𝑖control , 𝑡) .
Diodes: e.g. simple Shockley model
𝑢(𝑡)
𝑖(𝑡) = 𝑖s (𝑒 𝑛𝑢T − 1)

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 12
Overview

1 Circuit Simulation
Kirchhoff’s Laws
Lumped Elements
Electric Circuits are Graphs
Modified Nodal Analysis
Remarks on the Implementation

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 13
Graph Definition

1
1 𝑑 5
𝑓 𝑐
𝑒 ℎ 𝑎
𝑖
𝑎 3 4 𝑐 𝑒
𝑓 2 3
𝑗 𝑔
𝑏
2 𝑏 6 𝑑

A graph 𝒢 = (𝒱, ℰ) consists of a set of nodes 𝒱 and branches ℰ.

The endings of branches are two nodes.
If the branches are oriented then 𝒢 is called a directed graph.
A graph is called connected if there is at least one path between each pair
of nodes of the graph.

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 14
Loop

A subgraph 𝐺𝑙 of a connected graph 𝐺 is called loop if

1. 𝐺𝑙 is connected.
2. Each node of 𝐺𝑙 belongs to exactly two branches of 𝐺𝑙

1 𝑑 5
Example:
𝑒 𝑖 ℎ
𝑎 3 4 𝑐
𝑓
𝑗 𝑔
2 𝑏 6

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 15
Loop

A subgraph 𝐺𝑙 of a connected graph 𝐺 is called loop if

1. 𝐺𝑙 is connected.
2. Each node of 𝐺𝑙 belongs to exactly two branches of 𝐺𝑙

1 𝑑 5
Example:
𝑒 𝑖 ℎ
𝑎 3 4 𝑐 {𝑎, 𝑏, 𝑐, 𝑑} is a loop, but
𝑓
𝑗 𝑔
2 𝑏 6

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 15
Loop

A subgraph 𝐺𝑙 of a connected graph 𝐺 is called loop if

1. 𝐺𝑙 is connected.
2. Each node of 𝐺𝑙 belongs to exactly two branches of 𝐺𝑙

1 𝑑 5
Example:
𝑒 𝑖 ℎ
𝑎 3 4 𝑐 {𝑎, 𝑏, 𝑐, 𝑑} is a loop, but
𝑓
𝑗 𝑔
{𝑎, 𝑒, 𝑖, 𝑗, 𝑓} is not a loop
2 𝑏 6

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 15
 Incidence Matrix

Let graph 𝒢 with 𝑛 nodes and 𝑏 branches be given. The 𝑛 × 𝑏 incidence

matrix is defined as Aa = (𝑎𝑖𝑗 ) with
⎧+1 if branch 𝑗 leaves node 𝑖
{
𝑎𝑖𝑗 = −1 if branch 𝑗 enters node 𝑖
⎨
{0 otherwise
⎩

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 16
 Incidence Matrix

Let graph 𝒢 with 𝑛 nodes and 𝑏 branches be given. The 𝑛 × 𝑏 incidence

matrix is defined as Aa = (𝑎𝑖𝑗 ) with
⎧+1 if branch 𝑗 leaves node 𝑖
{
𝑎𝑖𝑗 = −1 if branch 𝑗 enters node 𝑖
⎨
{0 otherwise
⎩
1
Example:
𝑓 𝑐
𝑎
1 0 1 0 0 −1 1
⎡−1 1 0 0 1 0⎤ 2 𝑒
Aa = ⎢ ⎥ 2 3
⎢ 0 0 −1 1 −1 0⎥ 3
⎣ 0 −1 0 −1 0 1⎦ 4 𝑏
𝑑
𝑎 𝑏 𝑐 𝑑 𝑒 𝑓
4
June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 16
 Incidence Matrix: Currents (i)

Let a circuit be represented by Aa and the vector of all branch currents is

i⊤ = [𝑖1 , 𝑖2 , ..., 𝑖𝑏 ].

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 17
 Incidence Matrix: Currents (i)

Let a circuit be represented by Aa and the vector of all branch currents is

i⊤ = [𝑖1 , 𝑖2 , ..., 𝑖𝑏 ].
Then, it holds
Aa i = 0

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 17
 Incidence Matrix: Currents (i)

Let a circuit be represented by Aa and the vector of all branch currents is

i⊤ = [𝑖1 , 𝑖2 , ..., 𝑖𝑏 ].
Then, it holds
Aa i = 0

because each row a𝑘∶ = [𝑎𝑘1 , 𝑎𝑘2 , ..., 𝑎𝑘𝑏 ] of the incidence matrix Aa sums
up the right currents
𝑏
a𝑘∶ i = ∑ 𝑎𝑘𝑙 𝑖𝑙
𝑙=1

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 17
 Incidence Matrix: Currents (i)

Let a circuit be represented by Aa and the vector of all branch currents is

i⊤ = [𝑖1 , 𝑖2 , ..., 𝑖𝑏 ].
Then, it holds
Aa i = 0

because each row a𝑘∶ = [𝑎𝑘1 , 𝑎𝑘2 , ..., 𝑎𝑘𝑏 ] of the incidence matrix Aa sums
up the right currents
𝑏
a𝑘∶ i = ∑ 𝑎𝑘𝑙 𝑖𝑙
𝑙=1

and the sum of all current entering/leaving a node must be zero due
Kirchhoff’s current law, i.e.
a𝑘∶ i = 0 ∀𝑘.

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 17
 Incidence Matrix: Currents (ii)

The rows in an incidence matrix of a connected graph are linearly

dependent, one row is superfluous, e.g.
1 0 1 0 0 −1
⎡−1 1 0 0 1 0⎤
Aa = ⎢ ⎥
⎢ 0 0 −1 1 −1 0⎥
⎣ 0 −1 0 −1 0 1⎦

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 18
 Incidence Matrix: Currents (ii)

The rows in an incidence matrix of a connected graph are linearly

dependent, one row is superfluous, e.g.
1 0 1 0 0 −1
⎡−1 1 0 0 1 0⎤
Aa = ⎢ ⎥
⎢ 0 0 −1 1 −1 0⎥
⎣ 0 −1 0 −1 0 1⎦
The reduced incidence matrix A ∈ {−1, 0, 1}(𝑛−1)×𝑏 is obtained from Aa
by removing one row (e.g. 𝑛).

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 18
 Incidence Matrix: Currents (ii)

The rows in an incidence matrix of a connected graph are linearly

dependent, one row is superfluous, e.g.
1 0 1 0 0 −1
⎡−1 1 0 0 1 0⎤
Aa = ⎢ ⎥
⎢ 0 0 −1 1 −1 0⎥
⎣ 0 −1 0 −1 0 1⎦
The reduced incidence matrix A ∈ {−1, 0, 1}(𝑛−1)×𝑏 is obtained from Aa
by removing one row (e.g. 𝑛).

This (𝑛-th) node is called mass or ground and is implicitly set to 0V.

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 18
 Incidence Matrix: Currents (ii)

The rows in an incidence matrix of a connected graph are linearly

dependent, one row is superfluous, e.g.
1 0 1 0 0 −1
⎡−1 1 0 0 1 0⎤
Aa = ⎢ ⎥
⎢ 0 0 −1 1 −1 0⎥
⎣ 0 −1 0 −1 0 1⎦
The reduced incidence matrix A ∈ {−1, 0, 1}(𝑛−1)×𝑏 is obtained from Aa
by removing one row (e.g. 𝑛).

This (𝑛-th) node is called mass or ground and is implicitly set to 0V.

Kirchhoff’s Current Law still holds for the other nodes, i.e.
Ai = 0.

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 18
 Incidence Matrix: Currents (iii)

Example: Aa i = 0
𝑖𝑎
1 0 1 0 0 −1 ⎡ 𝑖𝑏 ⎤ 0
⎡−1 ⎢ ⎥
1 0 0 1 0⎤ ⎢ 𝑖𝑐 ⎥ ⎡0⎤ 1
⎢ ⎥ =⎢ ⎥
⎢ 0 0 −1 1 −1 0⎥ ⎢𝑖𝑑 ⎥ ⎢0⎥
⎣ 0 −1 0 −1 0 1⎦ ⎢𝑖𝑒 ⎥ ⎣0⎦ 𝑓
𝑎
𝑐
⎣𝑖𝑓 ⎦ 𝑒
2 3

𝑏
𝑑

4
Node 4: ground

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 19
 Incidence Matrix: Currents (iii)

Example: Aa i = 0
𝑖𝑎
1 0 1 0 0 −1 ⎡ 𝑖𝑏 ⎤ 0
⎡−1 ⎢ ⎥
1 0 0 1 0⎤ ⎢ 𝑖𝑐 ⎥ ⎡0⎤ 1
⎢ ⎥ =⎢ ⎥
⎢ 0 0 −1 1 −1 0⎥ ⎢𝑖𝑑 ⎥ ⎢0⎥
⎣ 0 −1 0 −1 0 1⎦ ⎢𝑖𝑒 ⎥ ⎣0⎦ 𝑓
𝑎
𝑐
⎣𝑖𝑓 ⎦ 𝑒
2 3

and Ai = 0 𝑖𝑎 𝑏
𝑑
⎡𝑖 ⎤
1 0 1 0 0 −1 ⎢ 𝑏 ⎥ 0 4
⎡−1 𝑖𝑐 ⎥ ⎡ ⎤
⎢ 1 0 0 1 0⎤ ⎢
⎥ 𝑖 ⎥ ⎢0⎥
= Node 4: ground
⎣ 0 0 −1 1 −1 0⎦ ⎢ 𝑑
⎢𝑖𝑒 ⎥ ⎣0⎦
⎣𝑖𝑓 ⎦
June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 19
 Circuit Description: Voltages (i)

Let u denote the vector of branch voltages and 𝜑𝑖 the voltage between
the node 𝑖 and mass (called potentials)

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 20
 Circuit Description: Voltages (i)

Let u denote the vector of branch voltages and 𝜑𝑖 the voltage between
the node 𝑖 and mass (called potentials)
Vector of potentials: 𝜑⊤ = [𝜑1 , 𝜑2 , ..., 𝜑𝑛−1 ]

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 20
 Circuit Description: Voltages (i)

Let u denote the vector of branch voltages and 𝜑𝑖 the voltage between
the node 𝑖 and mass (called potentials)
Vector of potentials: 𝜑⊤ = [𝜑1 , 𝜑2 , ..., 𝜑𝑛−1 ]
If branch 𝑘 leaves node 𝑖 and enters node 𝑗 , then

𝑢𝑘 = 𝜑 𝑖 − 𝜑 𝑗 ,

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 20
 Circuit Description: Voltages (i)

Let u denote the vector of branch voltages and 𝜑𝑖 the voltage between
the node 𝑖 and mass (called potentials)
Vector of potentials: 𝜑⊤ = [𝜑1 , 𝜑2 , ..., 𝜑𝑛−1 ]
If branch 𝑘 leaves node 𝑖 and enters node 𝑗 , then

𝑢𝑘 = 𝜑 𝑖 − 𝜑 𝑗 ,

If node 𝑖 or 𝑗 is the mass node, then follows

𝑢𝑘 = −𝜑𝑗 or 𝑢𝑘 = 𝜑 𝑖 .

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 20
 Circuit Description: Voltages (i)

Let u denote the vector of branch voltages and 𝜑𝑖 the voltage between
the node 𝑖 and mass (called potentials)
Vector of potentials: 𝜑⊤ = [𝜑1 , 𝜑2 , ..., 𝜑𝑛−1 ]
If branch 𝑘 leaves node 𝑖 and enters node 𝑗 , then

𝑢𝑘 = 𝜑 𝑖 − 𝜑 𝑗 ,

If node 𝑖 or 𝑗 is the mass node, then follows

𝑢𝑘 = −𝜑𝑗 or 𝑢𝑘 = 𝜑 𝑖 .

In matrix-vector notation
u = A⊤ 𝜑.

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 20
 Circuit Description: Voltages (ii)

Example: u = A⊤
a𝜑
𝑢𝑎 1 −1 0 0
⎡𝑢 ⎤ ⎡ 0 1 0 −1 ⎤ 𝜑1
⎢ 𝑏⎥ ⎢ ⎥
𝑢
⎢ 𝑐⎥ = ⎢ 1 0 −1 0 ⎥ ⎡ 𝜑2 ⎤ 1
⎢ ⎥
⎢ 𝑢𝑑 ⎥ ⎢ 0 0 1 −1⎥ ⎢ 𝜑3 ⎥
⎢ 𝑢𝑒 ⎥ ⎢ 0 𝑓 𝑐
1 −1 0 ⎥ ⎣ 0⎦ 𝑎
⎣𝑢𝑓 ⎦ ⎣−1 0 0 1⎦ 𝑒
2 3

𝑏
𝑑

4
Node 4: ground

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 21
 Circuit Description: Voltages (ii)

Example: u = A⊤
a𝜑
𝑢𝑎 1 −1 0 0
⎡𝑢 ⎤ ⎡ 0 1 0 −1 ⎤ 𝜑1
⎢ 𝑏⎥ ⎢ ⎥
𝑢
⎢ 𝑐⎥ = ⎢ 1 0 −1 0 ⎥ ⎡ 𝜑2 ⎤ 1
⎢ ⎥
⎢ 𝑢𝑑 ⎥ ⎢ 0 0 1 −1⎥ ⎢ 𝜑3 ⎥
⎢ 𝑢𝑒 ⎥ ⎢ 0 𝑓 𝑐
1 −1 0 ⎥ ⎣ 0⎦ 𝑎
⎣𝑢𝑓 ⎦ ⎣−1 0 0 1⎦ 𝑒
2 3
and u = A⊤ 𝜑
𝑏
𝑢𝑎 1 −1 0 𝑑
⎡𝑢 ⎤ ⎡ 0 1 0⎤
⎢ 𝑏⎥ ⎢ ⎥ 𝜑 4
⎢ 𝑢𝑐 ⎥ = ⎢ 1 0 −1⎥ ⎡ 1 ⎤ Node 4: ground
𝜑
⎢𝑢𝑑 ⎥ ⎢ 0 0 1⎥ ⎢ 2 ⎥
𝜑
⎢ 𝑢𝑒 ⎥ ⎢ 0 1 −1⎥ ⎣ 3 ⎦
⎣𝑢𝑓 ⎦ ⎣−1 0 0⎦

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 21
Overview

1 Circuit Simulation
Kirchhoff’s Laws
Lumped Elements
Electric Circuits are Graphs
Modified Nodal Analysis
Remarks on the Implementation

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 22
 Sparse Tableau Analysis (i)

Sparse Tableau Analysis (STA) is based for each point in time

Ai(𝑡) = 0 and u(𝑡) = A⊤ 𝜑(𝑡)

as well as branch voltage/current relations for each element

d d
f( u(𝑡), i(𝑡), u(𝑡), i(𝑡)) = 0.
d𝑡 d𝑡

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 23
 Sparse Tableau Analysis (i)

Sparse Tableau Analysis (STA) is based for each point in time

Ai(𝑡) = 0 and u(𝑡) = A⊤ 𝜑(𝑡)

as well as branch voltage/current relations for each element

d d
f( u(𝑡), i(𝑡), u(𝑡), i(𝑡)) = 0.
d𝑡 d𝑡

This yields a differential/algebraic equation system with unknown

time-dependent variables i(𝑡), u(𝑡) and 𝜑(𝑡).
The resulting dimension of the system is 2𝑏 + 𝑛 − 1, where 𝑏 is the
number of branches and 𝑛 the number of nodes.

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 23
 Sparse Tableau Analysis (ii)

assumption: sort the branches by type so that

A = [AR AC AL AV AI ]

i⊤ = [i⊤
R i⊤
C i⊤
L i⊤
V i⊤
I ] and u⊤ = [u⊤
R u⊤
C u⊤
L u⊤
V u⊤
I ]

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 24
 Sparse Tableau Analysis (ii)

assumption: sort the branches by type so that

A = [AR AC AL AV AI ]

i⊤ = [i⊤
R i⊤
C i⊤
L i⊤
V i⊤
I ] and u⊤ = [u⊤
R u⊤
C u⊤
L u⊤
V u⊤
I ]

Resistors: conductance matrix G = diag(𝐺1 , … , 𝐺𝑏R ) and number 𝑏R

iR = G uR .

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 24
 Sparse Tableau Analysis (ii)

assumption: sort the branches by type so that

A = [AR AC AL AV AI ]

i⊤ = [i⊤
R i⊤
C i⊤
L i⊤
V i⊤
I ] and u⊤ = [u⊤
R u⊤
C u⊤
L u⊤
V u⊤
I ]

Resistors: conductance matrix G = diag(𝐺1 , … , 𝐺𝑏R ) and number 𝑏R

iR = G uR .
Capacitors: capacitance matrix C = diag(𝐶1 , … , 𝐶𝑏C ) and number 𝑏C
d
iC = C u ,
d𝑡 C

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 24
 Sparse Tableau Analysis (ii)

assumption: sort the branches by type so that

A = [AR AC AL AV AI ]

i⊤ = [i⊤
R i⊤
C i⊤
L i⊤
V i⊤
I ] and u⊤ = [u⊤
R u⊤
C u⊤
L u⊤
V u⊤
I ]

Resistors: conductance matrix G = diag(𝐺1 , … , 𝐺𝑏R ) and number 𝑏R

iR = G uR .
Capacitors: capacitance matrix C = diag(𝐶1 , … , 𝐶𝑏C ) and number 𝑏C
d
u , iC = C
d𝑡 C
Coils: inductance matrix L = diag(𝐿1 , … , 𝐿𝑏L ) and number 𝑏L
d
uL = L i .
d𝑡 L
June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 24
 Sparse Tableau Analysis (iii)

Findu, i, s.t. AR iR (𝑡) + AC iC (𝑡) + AL iL (𝑡) + AV iV (𝑡) + AI iI (𝑡) = 0 ,

iR (𝑡) = GuR (𝑡) ,
iC (𝑡) = Cu′C (𝑡) ,
uL (𝑡) = Li′L (𝑡) ,
uV (𝑡) = us (𝑡) ,
iI (𝑡) = is (𝑡) ,
uR (𝑡) = A⊤R 𝜑(𝑡) ,
uC (𝑡) = A⊤C 𝜑(𝑡) ,
uL (𝑡) = A⊤L 𝜑(𝑡) ,
uV (𝑡) = A⊤V 𝜑(𝑡) ,
uI (𝑡) = A⊤I 𝜑(𝑡) .

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 25
 Sparse Tableau Analysis (iii)

Findu, i, s.t. AR iR (𝑡) + AC iC (𝑡) + AL iL (𝑡) + AV iV (𝑡) + AI iI (𝑡) = 0 ,

iR (𝑡) = GuR (𝑡) ,
iC (𝑡) = Cu′C (𝑡) ,
uL (𝑡) = Li′L (𝑡) ,
uV (𝑡) = us (𝑡) ,
iI (𝑡) = is (𝑡) ,
uR (𝑡) = A⊤R 𝜑(𝑡) ,
uC (𝑡) = A⊤C 𝜑(𝑡) ,
uL (𝑡) = A⊤L 𝜑(𝑡) ,
uV (𝑡) = A⊤V 𝜑(𝑡) ,
uI (𝑡) = A⊤I 𝜑(𝑡) .

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 25
 Modified Nodal Analysis: Linear elements

Plugging the equations into each other and eliminating variables:

Find 𝝋, iL , iV , such that 𝑡 ∈ ℐ

d
AC CA⊤C 𝝋(𝑡) + AR GA⊤R 𝝋(𝑡) + AL iL (𝑡) + AV iV (𝑡) = −AI is (𝑡)
d𝑡
d
L iL (𝑡) − A⊤L 𝝋(𝑡) = 0
d𝑡
A⊤V 𝝋(𝑡) = us (𝑡) .

with initial values, e.g. 𝝋0 , iL,0 , iV,0 .

where „0” denotes the zero matrix or vector. Unnecessary rows or
columns (e.g. if there are no resistors in the network) are ignored.

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 26
 Modified Nodal Analysis: Matrix System

Matrix/vector-notation
AC CA⊤ 0 0 𝝋(𝑡) AR GA⊤R AL AV 𝝋(𝑡) −AI is (𝑡)
⎡ 0 C L 0⎤ d ⎡ i (𝑡) ⎤ + ⎡ −A⊤ 0⎤ ⎡ ⎤ ⎡ ⎤
⎢ ⎥ d𝑡 ⎢ L ⎥ ⎢ L 0 ⎥ ⎢ iL (𝑡) ⎥ = ⎢ 0 ⎥
⎣ ⏟0⏟⏟⏟
⏟⏟ 0⏟⏟⏟
0⎦ ⎣iV (𝑡)⎦ ⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟ ⎣ −AV
⊤ 0 0 ⎦ ⎣iV (𝑡)⎦ ⏟ ⎣⏟−u
⏟ s (𝑡)
⏟⏟ ⎦
=∶M =∶x(𝑡) =∶K =∶r(𝑡)

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 27
 Modified Nodal Analysis: Matrix System

Matrix/vector-notation
AC CA⊤ 0 0 𝝋(𝑡) AR GA⊤R AL AV 𝝋(𝑡) −AI is (𝑡)
⎡ 0 C L 0⎤ d ⎡ i (𝑡) ⎤ + ⎡ −A⊤ 0⎤ ⎡ ⎤ ⎡ ⎤
⎢ ⎥ d𝑡 ⎢ L ⎥ ⎢ L 0 ⎥ ⎢ iL (𝑡) ⎥ = ⎢ 0 ⎥
⎣ ⏟0⏟⏟⏟
⏟⏟ 0⏟⏟⏟
0⎦ ⎣iV (𝑡)⎦ ⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟ ⎣ −AV
⊤ 0 0 ⎦ ⎣iV (𝑡)⎦ ⏟ ⎣⏟−u
⏟ s (𝑡)
⏟⏟ ⎦
=∶M =∶x(𝑡) =∶K =∶r(𝑡)

It can be written abstractly as a kind of differential equation

d
M x(𝑡) + Kx(𝑡) = r(𝑡)
d𝑡

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 27
 Modified Nodal Analysis: Matrix System

Matrix/vector-notation
AC CA⊤ 0 0 𝝋(𝑡) AR GA⊤R AL AV 𝝋(𝑡) −AI is (𝑡)
⎡ 0 C L 0⎤ d ⎡ i (𝑡) ⎤ + ⎡ −A⊤ 0⎤ ⎡ ⎤ ⎡ ⎤
⎢ ⎥ d𝑡 ⎢ L ⎥ ⎢ L 0 ⎥ ⎢ iL (𝑡) ⎥ = ⎢ 0 ⎥
⎣ ⏟0⏟⏟⏟
⏟⏟ 0⏟⏟⏟
0⎦ ⎣iV (𝑡)⎦ ⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟ ⎣ −AV
⊤ 0 0 ⎦ ⎣iV (𝑡)⎦ ⏟ ⎣⏟−u
⏟ s (𝑡)
⏟⏟ ⎦
=∶M =∶x(𝑡) =∶K =∶r(𝑡)

It can be written abstractly as a kind of differential equation

d
M x(𝑡) + Kx(𝑡) = r(𝑡)
d𝑡
The unknown quantity is the time-dependent vector-valued function

x(𝑡) = [𝝋⊤ (𝑡), i⊤L (𝑡), i⊤V (𝑡)]⊤ ∈ ℝ(𝑛−1)+𝑏L +𝑏V

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 27
 Modified Nodal Analysis: Matrix System

Matrix/vector-notation
AC CA⊤ 0 0 𝝋(𝑡) AR GA⊤R AL AV 𝝋(𝑡) −AI is (𝑡)
⎡ 0 C L 0⎤ d ⎡ i (𝑡) ⎤ + ⎡ −A⊤ 0⎤ ⎡ ⎤ ⎡ ⎤
⎢ ⎥ d𝑡 ⎢ L ⎥ ⎢ L 0 ⎥ ⎢ iL (𝑡) ⎥ = ⎢ 0 ⎥
⎣ ⏟0⏟⏟⏟
⏟⏟ 0⏟⏟⏟
0⎦ ⎣iV (𝑡)⎦ ⏟⏟⏟⏟⏟⏟⏟⏟⏟
⏟ ⎣ −AV
⊤ 0 0 ⎦ ⎣iV (𝑡)⎦ ⏟ ⎣⏟−u
⏟ s (𝑡)
⏟⏟ ⎦
=∶M =∶x(𝑡) =∶K =∶r(𝑡)

It can be written abstractly as a kind of differential equation

d
M x(𝑡) + Kx(𝑡) = r(𝑡)
d𝑡
The unknown quantity is the time-dependent vector-valued function

x(𝑡) = [𝝋⊤ (𝑡), i⊤L (𝑡), i⊤V (𝑡)]⊤ ∈ ℝ(𝑛−1)+𝑏L +𝑏V

To determine the system state, initial value x(𝑡0 ) = x0 at time 𝑡0 is needed.

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 27
 Tunnel Diode Oscillator (i)

𝜑1 𝐺 𝜑2 𝐿 𝜑3 𝜑2
0.4 𝜑3

Potential 𝜑 in V
𝑢s 𝐶 𝑔(𝑢)
0.2

𝜑0 = 0 0
0 0.2 0.4 0.6 0.8 1
Zeit 𝑡 in s
⋅10−2

Parameters: inductance 𝐿 = 2 ⋅ 10−3 H, capacitance 𝐶 = 1 ⋅ 10−7 F,

conductance 𝐺 = 1S, const. voltage source 𝑢s = 0.25V and tunnel diode
(nonlinear resistor) 𝑔(𝑢) = 1.80048 − 8.766𝑢 + 10.8𝑢3 S.

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 28
 Tunnel Diode Oscillator (ii)

Incidence matrices
1 0 0 0 1
AR = ⎡
⎢−1 0⎤
⎥, AC = ⎡ ⎤
⎢0⎥ , AL = ⎡ ⎤
⎢ 1 ⎥, AV = ⎡
⎢0⎥
⎤ AI = ⎡
⎢⎥
⎤
⎣0 1⎦ ⎣1⎦ ⎣−1⎦ ⎣0⎦ ⎣⎦

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 29
 Tunnel Diode Oscillator (ii)

Incidence matrices
1 0 0 0 1
AR = ⎡
⎢−1 0⎤
⎥, AC = ⎡ ⎤
⎢0⎥ , AL = ⎡ ⎤
⎢ 1 ⎥, AV = ⎡
⎢0⎥
⎤ AI = ⎡
⎢⎥
⎤
⎣0 1⎦ ⎣1⎦ ⎣−1⎦ ⎣0⎦ ⎣⎦
Unknowns 𝜑⊤ = [𝜑1 , 𝜑2 , 𝜑3 ] and two currents iI = (𝑖L ) and iV = (𝑖V )

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 29
 Tunnel Diode Oscillator (ii)

Incidence matrices
1 0 0 0 1
AR = ⎡
⎢−1 0⎤
⎥, AC = ⎡ ⎤
⎢0⎥ , AL = ⎡ ⎤
⎢ 1 ⎥, AV = ⎡
⎢0⎥
⎤ AI = ⎡
⎢⎥
⎤
⎣0 1⎦ ⎣1⎦ ⎣−1⎦ ⎣0⎦ ⎣⎦
Unknowns 𝜑⊤ = [𝜑1 , 𝜑2 , 𝜑3 ] and two currents iI = (𝑖L ) and iV = (𝑖V )

Equations are given by

0 0 0 0 0 𝜑1 𝐺 −𝐺 0 0 1 𝜑1 0
⎡0 0 0 0 0⎤ ⎡𝜑 ⎤ ⎡−𝐺 𝐺 0 1 0⎤ ⎡𝜑2 ⎤ ⎡ 0 ⎤
⎢ ⎥ d ⎢ 2⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 0 𝐶 0 0⎥ ⎢𝜑3 ⎥ + ⎢ 0 0 𝑔(𝜑3 ) −1 0⎥ ⎢𝜑3 ⎥ = ⎢ 0 ⎥
⎢0 0 0 𝐿 0⎥ d𝑡 ⎢ 𝑖L ⎥ ⎢ 0 −1 1 0 0 ⎥ ⎢ 𝑖L ⎥ ⎢ 0 ⎥
⎣0 0 0 0 0⎦ ⎣ 𝑖V ⎦ ⎣ −1 0 0 0 0⎦ ⎣ 𝑖V ⎦ ⎣−𝑢s ⎦

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 29
 Tunnel Diode Oscillator (ii)

Incidence matrices
1 0 0 0 1
AR = ⎡
⎢−1 0⎤
⎥, AC = ⎡ ⎤
⎢0⎥ , AL = ⎡ ⎤
⎢ 1 ⎥, AV = ⎡
⎢0⎥
⎤ AI = ⎡
⎢⎥
⎤
⎣0 1⎦ ⎣1⎦ ⎣−1⎦ ⎣0⎦ ⎣⎦
Unknowns 𝜑⊤ = [𝜑1 , 𝜑2 , 𝜑3 ] and two currents iI = (𝑖L ) and iV = (𝑖V )

Equations are given by

0 0 0 0 0 𝜑1 𝐺 −𝐺 0 0 1 𝜑1 0
⎡0 0 0 0 0⎤ ⎡𝜑 ⎤ ⎡−𝐺 𝐺 0 1 0⎤ ⎡𝜑2 ⎤ ⎡ 0 ⎤
⎢ ⎥ d ⎢ 2⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 0 𝐶 0 0⎥ ⎢𝜑3 ⎥ + ⎢ 0 0 𝑔(𝜑3 ) −1 0⎥ ⎢𝜑3 ⎥ = ⎢ 0 ⎥
⎢0 0 0 𝐿 0⎥ d𝑡 ⎢ 𝑖L ⎥ ⎢ 0 −1 1 0 0 ⎥ ⎢ 𝑖L ⎥ ⎢ 0 ⎥
⎣0 0 0 0 0⎦ ⎣ 𝑖V ⎦ ⎣ −1 0 0 0 0⎦ ⎣ 𝑖V ⎦ ⎣−𝑢s ⎦

𝐺 −𝐺
The matrix contribution of each element is called stamp, e.g., [ ].
−𝐺 𝐺

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 29
 Tunnel Diode Oscillator (iii): Netlist

Netlist for SPICE (Simulation Program with Integrated Circuit Emphasis)

1 R N1 N2 1
2 C N3 0 0.1u
3 L N2 N3 2m
4 B N3 0 I=1.80048*v(N3)-8.766*v(N3)*v(N3)+10.8*v(N3)*v(N3)*v(N3)
5 .tran 0.01 uic
6 .backanno
7 .end

most lines define an element and its connectivity:

Type startnode (n+) endnode (n-) parameter

Command .tran defines time domain analysis (transient)

Command 0.01 defines the time interval [0, 0.01]s
Command uic enforces zero initial values

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 30
Overview

1 Circuit Simulation
Kirchhoff’s Laws
Lumped Elements
Electric Circuits are Graphs
Modified Nodal Analysis
Remarks on the Implementation

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 31
Summary Circuit Simulation

Kirchhoff’s laws are a special case of Maxwell’s equations

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 32
Summary Circuit Simulation

Kirchhoff’s laws are a special case of Maxwell’s equations

circuits can be interpreted as directed graphs

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 32
Summary Circuit Simulation

Kirchhoff’s laws are a special case of Maxwell’s equations

circuits can be interpreted as directed graphs
circuit topology is stored in netlists

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 32
Summary Circuit Simulation

Kirchhoff’s laws are a special case of Maxwell’s equations

circuits can be interpreted as directed graphs
circuit topology is stored in netlists
from netlists one can easily construct the incidence matrix A

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 32
Summary Circuit Simulation

Kirchhoff’s laws are a special case of Maxwell’s equations

circuits can be interpreted as directed graphs
circuit topology is stored in netlists
from netlists one can easily construct the incidence matrix A
matrix A allows to easily express Kirchhoff’s laws

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 32
Summary Circuit Simulation

Kirchhoff’s laws are a special case of Maxwell’s equations

circuits can be interpreted as directed graphs
circuit topology is stored in netlists
from netlists one can easily construct the incidence matrix A
matrix A allows to easily express Kirchhoff’s laws
modified nodal analysis is a very efficient formulation

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 32
Summary Circuit Simulation

Kirchhoff’s laws are a special case of Maxwell’s equations

circuits can be interpreted as directed graphs
circuit topology is stored in netlists
from netlists one can easily construct the incidence matrix A
matrix A allows to easily express Kirchhoff’s laws
modified nodal analysis is a very efficient formulation
the formulation consists of differential and algebraic equations

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 32
Summary Circuit Simulation

Kirchhoff’s laws are a special case of Maxwell’s equations

circuits can be interpreted as directed graphs
circuit topology is stored in netlists
from netlists one can easily construct the incidence matrix A
matrix A allows to easily express Kirchhoff’s laws
modified nodal analysis is a very efficient formulation
the formulation consists of differential and algebraic equations
time integrators (e.g. implicit Euler) are necessary for solution

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 32
Summary Circuit Simulation

Kirchhoff’s laws are a special case of Maxwell’s equations

circuits can be interpreted as directed graphs
circuit topology is stored in netlists
from netlists one can easily construct the incidence matrix A
matrix A allows to easily express Kirchhoff’s laws
modified nodal analysis is a very efficient formulation
the formulation consists of differential and algebraic equations
time integrators (e.g. implicit Euler) are necessary for solution
in each step a (non-)linear system of equations has to be solved

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 32
Pseudocode Circuit Simulation

1 read netlist
2 set x0
3 for i = 1 to #Timesteps
4 x1𝑖 :=x𝑖−1
5 for j = 1 to #Newton
6 for k = 1 to #Elements
7 get k-th stamp [a𝑗𝑘 ,b𝑗𝑘 ] from x𝑗𝑖
8 [A𝑗 ,b𝑗 ]:=[A𝑗 ,b𝑗 ]+[a𝑗𝑘 ,b𝑗𝑘 ]
9 end
10 solve A𝑗 x𝑗+1
𝑖 =b𝑗
11 end
12 store x𝑖 :=x𝑗+1𝑖
13 end

Nonlinear problems require additional steps, e.g. Newton’s method.

1 𝐶 −𝐶 1 1
Matrix via stamp 𝑎𝑘 = [ ] = [ ] 𝐶 [1 −1] .
ℎ −𝐶 𝐶 ℎ −1

June 9, 2022 | TU Darmstadt | Fachbereich 18 | Computational Electromagnetics Group | De Gersem, Harutyunyan, Schöps | 33