Application Note
ShowerPower®
cooling concept
This application note gives an introduction to liquid cooling of power modules,
why cooling is needed, how it is done and a description of the Danfoss
ShowerPower® cooling concept. Then follow sections on ShowerPower® thermal
performance data, design guidelines, material choice recommendations and finally
a description on how to perform thermal tests on a ShowerPower® cooler.
siliconpower.danfoss.com
Content
and references
Cooling of power electronics ............................................................................................... 3
Why is cooling needed in power Electronics? 3
Why liquid cooling? 3
Liquid cooling – indirect vs. direct cooling 4
The ShowerPower® concept 4
ShowerPower® thermal performance 6
Simulations 6
Measurements 7
Design considerations 8
Corrosion issues 8
Water tightness 9
Material choices 9
Setting up the test 10
Laboratory equipment needed 10
Connecting the bathtub to the coolant supply 10
Assembly of the power module on the bathtub 11
Connecting the power module 11
References
1. K. Olesen et al., ”ShowerPower® New Cooling Concept”, PCIM Conference Proceedings 2004, Nuremberg.
2. F. Osterwald et al., “Innovative Kühltechnologie für Leistungsmodule”, ‘
Bauelemente der Leistungselektronik und ihre Anwendungen, 10.-11. Oktober 2006, Bad Nauheim.
3. R. Bredtmann et al., “”Power under the hood” Increasing power density of inverters with a novel 3D-approach”,
APE, March 25-26 2009, Paris.
4. K. Olesen et al. “Designing for reliability, liquid cooled power stack for the wind industry”,
EWEC Conference Proceedings 2012.
5. COG “O-ring basics”. http://www.cog.de/en/o-rings-products/o-ring-basics.html
DKSP.PB.301.A3.02
2
Cooling of
Power Electronics
Why is cooling needed in Power Electronics?
Every electronic circuit generates heat
during operation (excluding superconductivity); this is due to conductive and switching losses in active
devices as well as ohmic losses in
conductor tracks. And since every
new generation of power semiconductors becomes smaller than the
predecessor and the market expects
smaller and more compact solutions
the demands to the thermal design
engineer keeps growing. Suffi cient
cooling of Power Eectronics is crucial.
The dominant failure mechanisms in
power semiconductor components
Why liquid cooling?
are related not only to high absolute
temperatures but to changes in
temperature: temperature swings
produce thermo-mechanical induced
stresses and strains in the materialinterfaces of the components (mismatches in coeffi cients of thermal
expansion, CTE) which lead to fatigue
failures.
The most important failure mechanism in power modules is the bond
wire lifto where the aluminum wire
bonds (with a CTE of 24ppm/K) pop
o the silicon chip surface (CTE of Si is
2-3pmm/K) .
Bond wire liftoff after power cycling
Liquid cooling of power electronics
has been around for many years,
primarily due to the ever increasing
power densities and due to the
availability of liquids in certain
applications. Liquid cooling outperforms air cooling by having heat
transfer coeffi cients several orders of
magnitude higher thus enabling
much higher power densities and
more compact solutions.
The acceptance for liquid cooling
varies from business segment to
business segment. The automotive
industry for example has used liquid
cooling for cooling the combustion
engines for more than a hundred
years so the idea of cooling power
electronics in an automotive application is not frightening for the design
engineers. In other segments the idea
of having uids owing through
power electronic assemblies is most
disturbing.
Shower Power® Cooling Concept 3
Liquid cooling – indirect vs. direct cooling
The large amount of liquid cooling
solutions may be divided into two
groups: indirect and direct liquid
cooling.
Indirect cooling means that the
power module is assembled on a
closed cooler, e.g. a cold plate.
Cold plates may be realised e.g. by
gun drilling holes in aluminium plates
or by pressed-in copper tubes in
aluminium extrusions, an example of
which is shown here.
The rst picture shows a P3 module
being assembled onto a cold plate.
When dealing with cold plates it is
necessary to apply a layer of TIM
between the power module and the
cold plate.
Direct liquid cooling on the other
hand means that the coolant is in
direct contact with the surface to be
cooled. Here the cooling effi ciency is
improved by increasing the surface
area of the surface and this is commonly done by various pin n
designs. Below an example of how
the P3 module would look if it was
equipped with a baseplate with pin
n s .
Direct liquid cooling eliminates the
layer of TIM that is traditionally
needed between the backside of the
power module and the cold plate.
Because the TIM layer accounts for
30%-50% of the Rth, junction-coolant,
this TIM-elimination results in an
improved thermal environment for
the power module. Since the dominant failure mechanisms are temperature-driven, this will lead to higher
reliability.
A cold plate for the P3 module.
A baseplate with pin ns for the P3
module may look like this
The ShowerPower® concept
ShowerPower® is a concept for direct
liquid cooling developed by Danfoss.
The main motivation for the concept
was to solve the classical problem
associated with liquid cooling of
power modules namely:
Classical problems
with liquid cooling:
■
Inhomogeneous cooling due to
the calorimetric heating up of the
coolant
■
Thermal interface material (TIM)
related quality issues like pumpout and dry-out e ects
■
High cost
Key freatures are:
■
Ability to homogeneously cool
large at baseplate power modules, and systems of modules,
thereby eliminating temperature
gradients thus improving live and
facilitating paralleling of many
power chips
■
No TIM-related pump-out and
d r y - o u t e e c t s
■
Very low di erential pressure drop
■
Compact, low weight, high degree
of design freedom enabling 3D
designs
■
Low cost: metal-to-plastic conver-
sion into simple plastic parts.
The key element of the concept is the
ShowerPower® turbolator that guides
the coolant along the module
baseplate in cells that are supplied
with coolant in parallel thereby
securing uniform module temperatures. Actually the term turbolator is
misleading: under normal ow
conditions the liquid ow in the ow
channels is laminar; typical Reynolds
numbers range around 500 and the
transition into turbulence occurs at
Reynolds numbers around 2400.
Temperature gradient in standard cooler.
No gradient as in ShowerPower®
DKSP.PB.301.A3.024
The ShowerPower® concept – continued
The general ShowerPower® plastic
part having several cooling cells in the
X and Y directions needs a manifold
structure on the backside of the
plastic part; this ensures that each
cooling cell receives water with the
same temperature.
Since the P3 module is relative long
and narrow only one cell is necessary
across the module; this makes the
plastic part much simpler since the
manifold structure on the backside
becomes obsolete.
Shown below is a ShowerPower®
cooler assembly for a wind application featuring seven P3 IGBT modules,
turbulators, sealings and manifold.
The design ensures that all chips in all
modules are cooled equally well. The
concept enables tailored cooling if
hot spots need extra attention; this is
simply done by designing the cooling
channels individually. For further
information on the principles of
ShowerPower® please refer to 1, 2, 3,
4.
The general ShowerPower® turbolator
The ShowerPower® for the P3 module.
Shower Power® Cooling Concept 5
ShowerPower®
thermal performance
Numerous simulations and measurements have been done over the years on various
ShowerPower® designs. The two most investigated power module coolers are the E+
and the P3 module-cooler combinations.
E+ module and cooler. P3 module and cooler.
Simulations
Simulations, (thermal, uid, mechanical, stress, vibrational etc.) are always
important in any product development project; the obvious reason is to
reduce the number of time-consuming and costly tests.
There are basically three approaches
for doing simulations of uid ow
problems.
The best way to simulate a liquid cooling system is to use computational
uid dynamics, CFD. Here the uid
ow is solved numerically so that the
correct heat transfer rates and
pressure conditions are found and
thus the relevant temperatures, e.g.
semiconductor junction temperatures
are found.
In some cases though, e.g. in complicated transient situations, it makes
more sense to use nite element
analysis, FEM. Here a heat transfer
coeffi cient is applied as a boundary
condition to the wetted surface of the
power modules; this heat transfer
coeffi cient is found either from
measurements or from CFD analysis.
Typically a thermal step response
analysis is being made and the
resulting thermal impedance curve is
curve- tted to a sum of exponential
functions from which thermal
resistances and capacities can be
extracted and forming the basis for a
Foster network analysis. For analysing
complicated mission pro les this is
the only way; using a transient CFD
analysis may take years of CPU time!
Thirdly simple calculations using e.g.
Excel is often used to get a rst
glimpse of the thermal performance
and pressure drop. An Excel model
can also be used to analyse the
thermal stack layer by layer and assess
the thermal resistances and heat
capacities of each layer which can
then be used in a Cauer network for
transient analysis.
DKSP.PB.301.A3.026
Measurements
Below a collection of measurements
done on the P3 module is shown.
Note that there are two dierent
module types represented: Al2O3
DCBs on copper baseplate and AlN
DABs on AlSiC baseplate. All measurement sets but one are done with
glycol/water 50%/50%; the other set
was done using glycol/water
40%/60%. All the measurements were
done by turning on the IGBTs only
and using a DC current source for
generating the heat loss. Temperature
measurements were done using
thermography.
Based on the measurements done on
the P3 cooler for ShowerPower® two
empirical expressions have been
derived that predict the thermal
resistance (junction-water) and the
dierential pressure drop as function
of volume ow rate.
Thermal resistance, junction-water, for one single IGBT chip in the P3 module.
is the thermal resistance (junc-
R
thJA
tion – coolant) for a single IGBT chip
(SIGC186T170R3) and V is the volume
.
ow rate [l/min] for the whole
module.
The measurements from customer A.
Parameter Value
R
[K ⁄ W] 0.48 x V
thJA
∆P [mbar] 1.07 x V
.
–0.235
.
2
Shower Power® Cooling Concept
7
Design
considerations
When designing a liquid cooled
system several issues have to be
considered in order to ensure a
reliable solution that is capable of
delivering the performance needed
over the required lifetime of the
system.
When designing a liquid cooling
system for a power electronics
applications the issues in the table
need to be considered.
Additionally it is very important to
look into corrosion-, tightness-,
sedimentation- (including biogrowth) and anti-freezing issues is discussed in the following sections.
It is recommended to apply a lter of
100µm before the ShowerPower®
cooler.
It is also recommended to design a
independent cooling loop for the
ShowerPower® system for ensuring
the best environment that will yield
long life and high performance.
Parameter Unit Comment
Volume ow rate l/min What is the ow rate available
Coolant Glycol/water mixture
Absolute pressure in the
system
Dierential pressure drop
allowed
T
in
Power losses W
T
Jmax
∆T
max
Distribution of heat sources
Geometric and weight
constraints
Material compositions in the
cooling system
Particle sizes in coolant
bar
mbar
°C
°C Maximum allowed component temperatures
K
High pressure may need special power modules
(thicker baseplates) for avoiding excess mechanical
deformation of power modules
From inlet to outlet, determines the size of the pump
required
Inlet coolant temperature; high temperatures may
require special materials
All heat that need to be transported away by the
cooling system
Maximum allowed temperature variations over the
power module assembly
The best is to have the physical layout of the power
modules so the optimum cooling can be designed
Size and form factor; is the system to t into larger
assemblies
Tubing, bathtub, power module baseplate; determine
the chemistry of the coolant e.g. regarding the correct
anti-corrosive additives
Determines the minimum allowed geometries in the
cooling system for avoiding clogging of narrow
channels
Corrosion issues
Metals subjected to water are prone
to corrode. Crevice and galvanic
corrosion are the most important
processes that need to be controlled.
Crevice corrosion is related to small
gaps where the cooling liquid has
limited ow- access, i.e. the water is
moving very slowly. Therefore care
should be taken when designing the
groove for the sealing and for the
ShowerPower® plastic insert.
Galvanic corrosion occurs when
metals with dierent electrode potentials are immersed in an electrolyte,
like water.
DKSP.PB.301.A3.028
Typical metals in liquid cooled setups
are aluminum, copper and nickel. If
pure water is used as coolant galvanic
corrosion will quickly corrode away
the aluminum.
In order to avoid these corrosion
issues anti-corrosive chemicals are
added to the coolant. By far the most
commonly used materials are ethylene-glycol mixtures with suitable
anti-corrosives like the uids used in
every car in the world. Other substances used include ethylene-glycol
and propylene-glycol.
The mixture needs to contain more
than 30% glycol in order to avoid
bio-growth: lower glycol concentrations may actually act as “food” for
some microorganisms. Typical
mixtures have 40-50% glycol and
60-50% water. This also acts to avoid
freezing issues down to -35-40°C. The
type of glycol depends on the
material combination in the cooling
system (for combustion engines the
blue type is used for iron-cast motors
and the red type for aluminum cast
motors.
Water tightness
One of the biggest issues related to
direct liquid cooling principles is how
to design a reliable watertight
solution that will remain watertight
throughout the design life of the
product. The correct sealing concept
is key to success. Not only is a correct
groove design important but the
material choice and surface roughness also play important roles.
Sealing design
The design of a sealing concept
involves the design of the sealing
itself but also the groove into which
the sealing is placed. Basically two
types of sealing concepts are used:
■
Standard O-ring
The size of the O-ring depends on
the geometry and size of the
power module. For the P3 module
Ø3-Ø4mm would be the best
choice.
The recommended groove design for
an O-ring is seen in the gure to the
right.
■
Speci c Gaskets
Danfoss can provide a speci c
gasket for the P3 modules. It is a
rubber-plastic-compound component with a double sealing line and
drainage holes in between the two
lines. So if the inner sealing should
leak the coolant can be drained
outside the converter. The second
sealing line will then take over.
This gasket was successfully tested
in an 8.000 hours test at 105°C with
water glycol under operating
conditions (assembled and with
ow of coolant).
The surface quality of the gasket’s
contact surfaces (power module baseplate and gasket groove) needs to be
speci ed as Rz6.3 / Rmax10.
d b t r1 r2
2.00 2.85 1.45
2.50 3.55 1.90
2.65 3.80 2.00
3.00 4.20 2.30
3.55 5.00 2.75
3.70 5.15 2.90
4.00 5.55 3.20
Recommended groove geometry for
diff erent O-ring sizes, 5.
0.3±0.1
0.3±0.1
0.3±0.1
0.6±0.2
0.6±0.2
0.6±0.2
0.6±0.2
0.2
0.2
0.2
0.2
0.2
0.2
0.2
Material choices
Baseplate
The part of the power modules that is
in contact with the coolant, most
typically a copper baseplate, needs a
surface treatment. The Danfoss base
plate for Shower Power applications is
plated with speci c combinations of
Ni layers to assure good robustness
against corrosion.
Bathtub
The bathtub is most commonly made
of aluminum. Depending on the
manufacturing method there are
di erent aluminum alloys having
suffi cient corrosion robustness:
■
Extrusion, e.g. EN AW 6060
■
Machining from casted block
material, e.g. EN AW 5083
■
Die Cast, e.g. EN AC 44300
Sealing
There are some vital parameters for
selection of the right sealing material:
■
Expected lifetime at maximum
temperature level
■
Temperature range (operation and
storage)
■
Coolant type
■
Potential contaminations of the
coolant
■
Hardness of the material
■
Material of contact surfaces (power
module and bathtub)
■
Surface roughness of the sealing
area of Rz. 6.3.
The most commonly used material in
glycol / water loops is EPDM (ethylene-propylene-diene-monomer) 70
shore A.
It is generally recommended to clean
O-rings or gaskets in isopropanol
prior to assembly.
Coolant
Care must be taken choosing suitable
coolant chemistry. The coolant must
withstand the operating and storage
temperatures (anti-freeze) and
protect the materials included into
the coolant loop. Danfoss recommends:
■
Tyfocor® by Tyforop Chemie GmbH
■
Antifrogen N® by Clariant
Most often the bathtub neeeds to be
machined at the sealing area in order
to ensure a properly low surface
roughness not exceeding Rz. 6.3.
The application kit includes a double
gasket comprising a plastic carrier
upon which the EDPM rubber has
been vulcanized.
The double gasket for the P3 module
featuring drainage holes for optional
leakage detection.
Shower Power® Cooling Concept 9
Setting up
the test
The application kit comprises the following:
■
Power module, P3, not encapsulated
1
and painted black
infrared imaging
■
Bath tub
■
ShowerPower® insert
■
Double gasket
■
Screws for assembling
the module to the bathtub
■
A copy of this application note
1
Note that the black paint is only insulating up to 100V voltage levels;
this is not an issue under normal thermal test conditions with
DC-voltage levels of 3-4V.
for optimum
Laboratory equipment needed
Necessary equipment for making
measurements on a liquid cooling
system:
■
Liquid cooling circuit
– Closed loop system incl. pump
– Heat exchanger for re-cooling
the coolant. With typical power
dissipations of up to several kW
in the power module it is
recommended to connect a
heat exchanger to the cooling
circuit for re-cooling the water;
otherwise the cooling water will
heat up quickly making further
testing impossible
– Pressure transmitters, prefer-
ably as close as possible to the
bathtub for getting the most
precise assessment of di erential pressure drop of the power
module cooler
– Temperature transmitters,
preferably as close as possible to
the bathtub for getting the most
accurate measurements of the
calorimetric heating up of the
coolant.
– Flow meter
Connecting the bathtub to the coolant supply
– Air bleed, if there is air
entrapped into the cooling
system this could lead to hot
spots on the power module
baseplates so the system must
be degassed properly
■
Power supply for generating the
power losses in the power module
■
Voltage supply for turning the
gates of the IGBT chips on
■
Infrared camera for measuring
chip temperatures
If it is required to take out the bathtub
leak free from the cooling circuit
Danfoss recommends to use Stäubli
SPT connectors made of aluminum.
Please note that these connectors
10
DKSP.PB.301.A3.02
have a substantial inherent pressure
drop. Every time the cooling loop has
been opened the system needs
degassing properly after reassembly.
Assembly of the power module on the bathtub
Assembling a power module on a
heat sink, a cold plate or an open
cooler must be done carefully
otherwise the module may be
damaged. The prescribed mounting
sequence of the screws most be
followed for minimizing the stresses
inside the power module. The
recommended mounting sequence
for the P3 module is as follows.
The M5 bolts should be preset in a
rst step until the bolt heads touch
the baseplate top side. In a second
step the bolts must be torqued with 6
+/- 0,5Nm in the shown sequence.
The bolt strength class needs to be
8.8 minimum; the use of impact
wrenches is not permitted because of
the ceramic components inside the
power module. An electronic controlled screw driver with soft stop is
the best choice in a series production.
Recommended mounting sequence for the P3 module (In neon/Danfoss).
Connecting the
Test mode Connections Comments
power module
Active IGBTs, passive diodes
When the IGBTs are turned on the
voltage drop typically is 3V which
means that the power supply must be
capable to deliver currents up to
several hundred Amperes. This also
means that the cabling must be
suffi ciently dimensioned e.g. having
cross sectional areas of 70-120mm2.
Testing on an IGBT module, like the P3
module supplied in the application
kit, can be performed in a number of
di erent modes; and because the P3
module is a half bridge module is can
be decided whether the IGBTs and or
diodes are active during the test.
Active IGBTs, passive diodes
Passive IGBTs, active low side
diode
Passive IGBTs, active high side
and low side diodes
Low side IGBT is active and the
diodes are active, the IGBT is
turned on by a 15V gate
voltage
High side and low side IGBTs
are active, the IGBTs are
connected to the power
supply is in series and both
IGBTs are turned on by a 15V
gate voltage
Only the low side diode is
active; alternatively the high
side diode can become active
Both diodes are active by
connecting them in series
Active IGBTs, active diodes
High side diode and low side
IGBT are active; the IGBT needs
to be turned on by a 15V gate
voltage; alternatively the high
side IGBT and low side diode
can become active
Shower Power® Cooling Concept 11
Danfoss Silicon Power
Based in Flensburg, Germany, Danfoss Silicon Power is a leading developer
of customer specifi c IGBT and MOSFET modules and power stacks for power
intensive applications.
Our power modules and power stacks
are a preferred choice in demanding
automotive and wind power
applications and a wide variety of
industrial applications.
Our 35,000 m2 research, development
and production facility is certifi ed
according ISO 9001, ISO/TS 16949, ISO
14001, ISO 50001 and OHSAS 18001.
This enables us to quickly transfer
development projects to high volume
production that can be integrated
seamlessly into our customers’ supply
chain with full focus on quality.
Danfoss Silicon Power is a subsidiary
of the Danfoss Group, the largest
industrial company in Denmark. Danfoss
employs more than 24,000 people in
100 countries within development,
production, sales and support.
Danfoss Silicon Power GmbH, Husumer Strasse 251, 24941 Flensburg, Germany, Tel. +49 461 4301-40, Fax +49 461 4301-4310
www.siliconpower.danfoss.com, E-mail: dsp-sales@danfoss.com
DKSP.PM.301.A3.02 Produced for DSP 2016.04
Loading...