Download Generation of distributed electrical and electro

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Generation of distributed electrical and
electro-thermal networks with TRADICA
Yves Zimmermann
CEDIC, Technical University of Dresden
4th European HICUM Workshop
June 15/16, 2004
Bordeaux, France
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
OUTLINE
• Introduction
• Electro-thermal Characteristics
• Self-heating Networks
• Static Thermal Coupling
• Transient Thermal Coupling
• Distributed electrical networks
• Available Model Complexities
• Conclusion
• Outlook
Yves Zimmermann, CEDIC
1
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Introduction
• Shrinking device sizes lead to increased influence of parasitic effects on electrical
behavior.
• Deep trenches confine the heat flow resulting in higher operating temperatures.
• Multi-dimensional effects like avalanche breakdown are difficult to implement in
compact models.
• Need for accurate models for critical circuits and single devices
Yves Zimmermann, CEDIC
2
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Electro-thermal Characteristics
• Distinguish two different effects:
1. Self-heating (temperature rise due to power dissipation of a single heat source) and
2. Thermal coupling (temperature rise due to thermal interaction between several heat sources).
• Heat source definition:
Characteristic volume of integrated devices where heat is generated (e.g. BC-SCR of
BJTs)
Yves Zimmermann, CEDIC
3
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
PROBLEM-I
• Transient thermal characteristic is of complex exponential nature and varies with emitter contact configuration (CBEBC (solid), CBEBEBC (dashed), CBEBEBEC (dotted)).
BUT: Compact models use only simple thermal networks (parallel thermal
resistance Rth and capacitance Cth)!!!
Yves Zimmermann, CEDIC
4
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Self-heating Networks
• Common simple „built-in“ network (single RC-element):
-t
∆T(t) = P diss ⋅ R th  1 – exp  -------------------- 

 R ⋅ C 
th
Pdiss
th
Rth
Cth
∆T
• Extended network (multiple RC-elements):
Rth1
∆T(t) = P diss ⋅
n
∑i = 1
thi
Cth1
Pdiss
Yves Zimmermann, CEDIC
Rthn
..............
-t
R thi  1 – exp  ----------------------- 

 R ⋅ C 
thi
Rth2
Cth2
Cthn
∆T
5
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
PROBLEM-II
• Multi-emitter BJTs consist of several characteristic heat sources (one per emitter finger).
• Temperature distribution across the devices is NOT homogeneous due to thermal coupling
resulting in different operating temperatures at the fingers.
BUT: Compact models use only ONE ∆T for whole structure!!!
45%
Yves Zimmermann, CEDIC
6
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
PROBLEM-II
• Thermal coupling (intra- and inter-device) characteristic is NOT covered by compact models.
∆T 1
∆T 2
∆T 3
∆T 4
∆T 5
dT [K]
Heat sources for the multiple
internal base-collector junctions
10
9
8
7
6
5
4
3
2
1
14
12
10
0
1
8
2
3
x [µm]
Yves Zimmermann, CEDIC
6
4
5
4
6
7
y [µm]
2
8 0
7
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Static Thermal Coupling
• Total temperature rises (∆T1, ∆T2) of two interacting sources can be expressed as a sum of
both self-heating contributions (∆T11, ∆T22) using thermal coupling coefficients (c21, c12):
∆T 1 = ∆T 11 + c 21 ∆T 22
∆T 2 = ∆T 22 + c 12 ∆T 11
• Representation using voltage controlled voltage sources (VCVS) [D.J.Walkey]:
∆T1
self-heating
device 1
Pdiss1
Rth1
VCVS
Yves Zimmermann, CEDIC
∆T2
∆T11
c12∆T22
∆T22
c21∆T11
Rth2
self-heating
device 2
Pdiss2
VCVS
8
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Transient Thermal Coupling
• Time-dependence of coupling coefficient (solid line) is of exponential nature and can be
represented using one or multiple RC-elements (dashed line).
∆T 21 ( t )
-t
------------------c 21c ( t ) =
= c 21  1 – exp  ----------------------------------- 


∆T 11 ( t )
R thc21 ⋅ C thc21 
Yves Zimmermann, CEDIC
9
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Transient Thermal Coupling
• Voltage controlled current sources (VCCS) drive the delay networks for the coupling coefficients. VCVSs add the coupled temperatures to the respective self-heating contributions.
•
-t
∆T 12 ( t ) = ∆T 22 ( t ) ⋅ c  1 – exp  ----------------------------------- 
R

12 
thc12 ⋅ C thc12
-t
∆T 21 ( t ) = ∆T 11 ( t ) ⋅ c  1 – exp  ----------------------------------- 
R

21 
thc21 ⋅ C thc21
∆T1
Pdiss1
Rth1
∆T2
Cth1
∆T11
∆T22
Cth2
∆T12
VCVS1
VCVS2
c12∆T22
VCCS1
c21∆T11
Rth2
Pdiss2
∆T21
VCCS2
attenuation of heat wave
∆T12
Rthc12
Cthc12
Cthc21
Rthc21
∆T21
delay network
Yves Zimmermann, CEDIC
10
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Distributed electrical network
Finger-partitioning
• Every finger is represented by one transistor model.
• Distinguish three different types of transistor elements:
1. an outer element Tw containing the collector region
2. a center element Tc bounded by base contacts at each side (does not contain collector region)
3. an outer element Tn bounded by a base contact at one side (does not contain collector region)
Tn
Yves Zimmermann, CEDIC
Tc
Tw
11
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Distributed electrical network
Finger-partitioning
• External collector resistance is partitioned and pulled out of the transistor elements.
• Transistor model parameters are calculated from technology data using the respective
emitter area portion.
rsu
S
C’n
∆rCx
C’c1
∆rCx
S’
B
Tn
C’ck
∆rCx
S’
C’w
TcK
C
S’
S’
Tc1
Csu
rCx1
Tw
E
Yves Zimmermann, CEDIC
12
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Distributed electrical network
Finger-partitioning
• External access to the thermal nodes ∆T enables contacting thermal (coupling) networks
of arbitrary complexity.
∆Tn
Pn
∆Tc1
RTH,n
TnΣ
Yves Zimmermann, CEDIC
Pc1
Tw
TcK
Tc1
Tn
E
RTH,c1
Tc1Σ
∆Tw
∆TcK
PcK
RTH,cK Pw
TcKΣ
RTH,w
TwΣ
13
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Distributed electrical network
Emitter-partitioning
• The equivalent circuit of one emitter is split into external elements and an internal transistor
Qi, which is represented by a distributed network
rsu
S’
S
CjS
,
C BCx
CdS
ijSC
ijBCx
Csu
iTS
,,
C BCx
B*
CEox
C
C’
B
rBx
rCx
CjEp
ijBEp
Qi
E’
rE
E
Yves Zimmermann, CEDIC
14
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Distributed electrical network
Emitter-partitioning
• The finger (internal transistor Qi) is partitioned and represented by several transistor models.
• Distinguish three different types of transistor elements:
1. a generic element Tg bounded by another element in each direction
bE/k
bE/k
lE/n
Te
Tel
lE/n
Teb
Tg
Tg
lE/2
2. an edge element Te bounding to other elements on 2 sides
3. edge elements Tel and Teb bounding to other
elements on 3 sides
bE/k
lE/n
bE/2
Yves Zimmermann, CEDIC
15
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Distributed electrical network
Emitter-partitioning
• The internal base resistance is partitioned and pulled out of the transistor element model.
• Thermal (coupling) networks can be contacted through access to the thermal nodes ∆T.
r knz
r kny
r 2nz
C’
Tkn
r 2ny
r 1nz
T2n
r 1ny
C’
T1n
E’
E’
r k3z
B*
r k3y
Tk3
r 23y
r k2z
r k2y
C’
r 23z
r 22y
QjCi
QdC
QjEi
QdE
B’
r 22z
Tk2
ijBCi
T23
r 12z
T22
r 12y
C’
iAVL
iT
ijBEi
T12
E’
E’
C’
r k1z
r k1y
Tk1
r 31z
r 31y
r 21z
T31
r 21y
E’
C’
r 11z
T21
r 11y
T11
z
E’
E’
E’
equivalent circuit of a transistor element
(ideal 1-D Transistor model)
E’
y
Yves Zimmermann, CEDIC
16
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Available Model Complexities
electrical
representation
self-heating
(„built-in“)
self-heating
(improved external)
thermal coupling
(inter-device)
thermal coupling
(intra-device)
HICUM/L2
Cell/Device
1 TM per Device/ Cell
X
X
HICUM/L4
Cell
1 TM per Device
X
X
X
HICUM/L4
Finger
1 TM per Finger
X
X
X
X
HICUM /L4
Emitter
1 TM per Finger- subelement
X
X
X
X
(TM = transistor model)
Yves Zimmermann, CEDIC
17
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Conclusion
• Partitioning of emitter fingers enables better simulation of avalanche breakdown.
• Derived distributed models covering self-heating and thermal coupling effects.
• Implemented scaling equations and extraction routines for the respective parameters.
• The approach is based on „fast“ analytic expressions enabling efficient parameter
generation.
• Established automated network generation of different model complexities.
Yves Zimmermann, CEDIC
18
4th European HICUM Workshop, Bordeaux, 2004
Generation of distributed electrical and electro-thermal networks with TRADICA
Planned Activities - Outlook
• Verification of distributed electrical and electro-thermal networks.
• Development of scaling equations for substrate coupling.
• Automated generation of distributed models covering substrate coupling.
• Development of scaling equations for backend parasitics.
• Implementation into existing model generation procedures.
• Test and Verification.
Yves Zimmermann, CEDIC
19
Related documents