Wir schaffen Wissen heute fr morgen

Paul Scherrer Institut

12.5.2014
Ren Knzi
Thermal Design of Power Electronic Circuits
CERN Accelerator School 2!"# $a%en# Swit&erlan%
R. Knzi Thermal Design CAS 2014 12.05.2014 2
Motivation
Statement in a meeting:
The converter works as specified, there are just a
couple of tests missing to check its thermal behavior.
Statement in the follow-up meeting 2 weeks later:
We could only test up to 50% of the rated power, otherwise
the converter would have been burnt away.
Lesson learnt:
Take the losses and the heat dissipation issues into
account from the very beginning!
Heat Sources: Resistors
Resistors
Continuous load
Sufficient heat transfer to ambient
Temperature stable
Pulsed load
Absorb energy during pulses
sufficient active material
Temperature stable
Limited repetition rate
Cables
Use sufficient cross section additional investment vs. loss costs
Skin effect
Skin depth in Cu @ 1kHz: 2.1mm
@ 10kHz: 0.7mm
R. Knzi Thermal Design CAS 2014 12.05.2014 3
Heat sources: Magnetic Components
Core losses
Eddy current losses
Iron sheet cores f < 1kHz
Powder cores f < 10kHz
Ferrite cores f > 10kHz
Hysteresis losses
Proportional to the area of the hysteresis curve and the frequency
Core losses in transformers
Designed to have a minimized hysteresis curve area
Losses increase with frequency
Losses are also present at zero load
DC-chokes
Are built to store a lot of energy hysteresis curve spans a large area
Core losses depend on amplitude and frequency of the ripple current
R. Knzi Thermal Design CAS 2014 12.05.2014 4
Heat sources: Magnetic Components
R. Knzi Thermal Design CAS 2014 12.05.2014 5
Winding losses
Arise from the ohmic resistance of the winding
Dissipated power is proportional to I
2
Winding resistance increases approx. 0.4% per K (for Cu)
Keep losses and temperatures low because
Higher tempeatures cause even higher losses
Wasted energy
Costs of the wasted energy
Costs for recooling
Lower temperatures increase the life time
The investments will be returned
Heat Sources: Semiconductors
R. Knzi Thermal Design CAS 2014 12.05.2014 6
V
DC
= 250V
I
Out
= 200A @m = 0.8
f
s
= 20kHz
IGBT losses
Conduction losses
(Blocking losses)
Switching losses
(ON and OFF)
Diode losses
Conduction losses
(Blocking losses)
Switching losses
(Recovery losses)
FF600R06ME3 [1]
Example:
Heat Sources: IGBT losses
Conduction losses IGBT:
P
CI
= m I
0ut
I
CL
Note: V
CE
depends on the junction
temperature
P
CI
= u.8 2uuA 1.1I = 176w
R. Knzi Thermal Design CAS 2014 12.05.2014 7
Heat Sources: IGBT losses
Switching losses IGBT:
P
SI
= s E
on
+E
o]]

I
C
I
c]
K

K
1

with K
v
= 1.S .. . 1.4
and K
1
= 1 + u.uuSK
-1
I
]
-I
Rc]
[2]
In our example we assume T
J
90C
K
1
= 1 + u.uuSK
-1
9u -12S
K
1
= u.89S
P
SI
= 2u
i
uuus
-1
S + 8 1u
-3
[

2SuI
SuuI
1.35
u.89S = 182 w
R. Knzi Thermal Design CAS 2014 12.05.2014 8
Heat Sources: IGBT losses
A softer switching (higher gate
resistance) increases the switching
losses dramatically
R. Knzi Thermal Design CAS 2014 12.05.2014 9
Heat Sources: Diode losses
Conduction losses Diode
P
C
= (1 -m) I
0ut
I
P
Note: Negative temperature coefficient!
P
C
= (1 -u.8) 2uuA 1.1SI = 46w
R. Knzi Thermal Design CAS 2014 12.05.2014 10
Heat Sources: Diode losses
Switching losses Diode
P
S
=
s
E
Rcc

I
C
I
c]
K

K
1

with K
v
= u.6
and K
1
= 1 + u.uu6K
-1
I
]
-I
Rc]
[2]
In our example we assume T
J
90C
K
1
= 1 + u.uu6K
-1
9u -12S
K
1
= u.79
P
S
= 2u
i
uuus
-1
S 1u
-3
[
2SuI
SuuI
0.6
u.79
= 71 w
R. Knzi Thermal Design CAS 2014 12.05.2014 11
Heat Sources: Diode losses
In contrary to the IGBT, a softer
switching (higher gate resistance)
reduces the switching losses of the
freewheeling diode remarkably.
R. Knzi Thermal Design CAS 2014 12.05.2014 12
Summary losses
IGBT losses
P
I
= P
CI
+ P
SI
P
I
= 176W+ 182W = 358W
Diode losses
P

= P
C
+P
S
P

= 46W+71W = 117W
R. Knzi Thermal Design CAS 2014 12.05.2014 13
Total losses per module
P = P
I
+P

P = SS8W+ 117W = 47SW

Heat transfer: Packages
Isolated module package
Isolated Cu Base
1 heat sink for several devices
Heat sink on ground potential
Solder contacts
Cooling from one side only
Open circuit after failure
Parallel connection
Press pack devices
Cooling via power terminals
Individual heat sinks for all devices
Heat sink on high voltage
Presspack contacts
Cooling from both sides
Short circuit after failure
Series connection
R. Knzi Thermal Design CAS 2014 12.05.2014 14
Heat transfer: Temperatures
Maximum ambient temperature
Get thermal resistances from
module data sheet
Get thermal resistance from heat
sink supplier
IGBT- and diode-losses from
calculation
Calculate heatsink temperature
Calculate base plate temperature
Calculate diode and IGBT chip
temperatures
There should be a safety
margin of min. 25K!
R. Knzi Thermal Design CAS 2014 12.05.2014 15
35C
0.09K/W
0.15K/W
358W
117W
T=(358+117)W*0.1K/W=48K
T=(358+117)W*0.009K/W=4K
83C
87C
T=117W*0.15K/W=18K
T=358W*0.09K/W=32K
105C 119C
0.009K/W
0.1K/W
Transient thermal impedance
R. Knzi Thermal Design CAS 2014 12.05.2014 16
Continued fraction circuit (Cauer model)
Transient thermal impedance
R. Knzi Thermal Design CAS 2014 12.05.2014 17
Partial fraction circuit (Foster model)
Stacking of different materials
R. Knzi Thermal Design CAS 2014 12.05.2014 18
Temperature variations
Power Cycling
Temperature variations
of the silicon chips (fast)
Causes mechanical
stress to bond wires
Bond wires lift off
V
CE
increases
Losses increase
Larger temperature
variations
Device fails
Thermal Cycling
Temperature variations
of the base plate (slow)
Causes mechanical
stress to soldering joint
Aging of soldering joint
R
th
increases
Larger temperature
variations
Device fails
R. Knzi Thermal Design CAS 2014 12.05.2014 19
LESIT Project
R. Knzi Thermal Design CAS 2014 12.05.2014 20
Power Cycling of standard modules from various manufacturers (~1995)
Practical Experience (1)
R. Knzi Thermal Design CAS 2014 12.05.2014 21
SLS Booster Dipole Power Supply in Topup Mode
Junction temperature variations
IGBT failures after 3.5a of operation
Practical Experience (2)
R. Knzi Thermal Design CAS 2014 12.05.2014 22
3Hz Oscillations: 6882C T
j
= 14K, Tm = 75C
40 weeks/a, continuous 80*10
6
cy/a 6a to failure
40 weeks/a, 16cy every 100s 15*10
6
cy/a 33a to failure
Extreme extrapolation: Result valid ???
500*10
6
cy
Practical Experience (3)
R. Knzi Thermal Design CAS 2014 12.05.2014 23
100s Oscillations: 3086C T
j
= 56K, Tm = 58C
40 weeks/a, 1cy every 100s 0.25*10
6
cy/a 4a to failure
that is, what we have experienced!
1*10
6
cycles
Thermal Design of a PCB
R. Knzi Thermal Design CAS 2014 12.05.2014 24
10A H-bridge on a PCB 100mm 160mm
First prototype
R. Knzi Thermal Design CAS 2014 12.05.2014 25
Infrared image at full load
Replace DC/DC converter 24V/15V
R. Knzi Thermal Design CAS 2014 12.05.2014 26
New:
SMD DC/DC-
converter in a
metallic case
Protoype 1:
THT DC/DC-
converter in a
plastic case
R. Knzi Thermal Design CAS 2014 12.05.2014 27
New:
Switched mode
converter
15V / 5V
Prototype 1:
Linear converter
15V / 5V
Replace DC/DC converters 15V/5V
Replace filter chokes
R. Knzi Thermal Design CAS 2014 12.05.2014 28
Neu:
Four chokes of
3.9uH/12A each
Prototype 1:
Two chokes of
8.2uH/13A each
Replace snubber resistors
R. Knzi Thermal Design CAS 2014 12.05.2014 29
New:
2W snubber-
resistors
Prototype 1:
100mW snubber-
resistors
Improve heat flow
R. Knzi Thermal Design CAS 2014 12.05.2014 30
Thermal conductivity of different materials [W/K m]
Gold 318
Silver 429
Copper 401
Aluminum 237
Steel 50
Heat transfer foil 2
PCB core material (FR-4) 0.3
Air 0.025
Improve heat flow
R. Knzi Thermal Design CAS 2014 12.05.2014 31
W
K
cm W
K m mm
A
l
R
th
3 . 13
4 3 . 0
6 . 1
2
=

K W
W
K
P R T
th
133 10 3 . 13 = = =
Heat path through the PCB
Mounting area of semiconductors A = 2cm x 2cm = 4cm
2
Distance from top to bottom layer l = 1.6mm
Thermal conductivity PCB core (FR-4) = 0.3 W/K m
Total losses in the semiconductors P 10W
Thermal resistance:
Temp. difference
Top - Bottom
Improve heat flow
R. Knzi Thermal Design CAS 2014 12.05.2014 32
2 2 2 2 2
0255 . 0 ) ) 15 . 0 ( ) 175 . 0 (( ) ( mm mm mm ri ra A = = =
W
K
mm W
K m mm
A
l
R
th
8 . 156
0255 . 0 400
6 . 1
2
=

W K R
th
/ 45 . 2 =
K
K
W
K
W
W
R R
P T
Via th LP th
7 . 20
45 . 2
1
3 . 13
1
1
10
1 1
1
=

=
+
=

Additional heat path through vias
Outside radius ra = 0.175mm
Inside radius ri = 0.15 mm
Via heigth l = 1.6 mm
Thermal conductivity of Cu = 400 W/K m
On 4 cm
2
there is space for 8 x 8 = 64 vias
Thermal resistance of 64 vias:
Temperature
difference
Top - Bottom
Improve heat flow
R. Knzi Thermal Design CAS 2014 12.05.2014 33
Original design
Improved design
Heat transfer to ambient
R. Knzi Thermal Design CAS 2014 12.05.2014 34
Heat sink
Heat transfer foil
MOSFETs
PCB
Al bars
Cover
Comparison Prototype 1 and 2
R. Knzi Thermal Design CAS 2014 12.05.2014 35
Better heat
spreading
Lower
temperatures
Good corre-
lation with
calcualtions
R. Knzi Thermal Design CAS 2014 12.05.2014 36
References
[1] Infineon, Datasheet FF600R06ME3 Rev. 3.1
[2] Semikron, Applikationshandbuch Leistungs-
halbleiter, VSL Verlag 2010
[3] Infineon, Thermal equivalent circuit models,
AN2008-03
[4] Infineon, Technical Information IGBT modules,
Use of Power Cycling curves for IGBT4,
AN2010-02