session 4 packaging techniques
TRANSCRIPT
SESSION 4
PACKAGING TECHNIQUES
Sponsored by: National Science Foundation D4.1
Power Cycling Tests for Passive Integrated Power Electronics Modules
Seung-Yo Lee, Ning Zhu, W.G. Odendaal and J.D. van Wyk20032003
q Power Cycling Tests for Passive Integrated ModulesØTemperature Chamber( Ambient Temperature:-2 °C )
Integrated ModuleIntegrated Module
Vc
Vds
iLC
ØCircuit Diagram for Power Cycling ØWaveforms in Power Cycling Circuit (for 2-Turn Module)
Current of the LC Module (5A/div): iLC
Drain-Source Voltage of the Top Switch (20V/div): VdsApplied Capacitor
Voltage (25V/div): Vc
Ø Power CycleØ Temperature Profile (for 2-Turn Module) Ø Temperature Profile (for 3-Turn Module)
Crack in Ceramic
Ceramic substrate
Copper
Epoxy
Delamination at the Interface between Copper and CeramicLayers Crack in Ceramic
Spring to hold the module
Temp(°C)
About 92°C
Heating Cooling
About 2 °C
Heating
Time(sec)48sec 48sec
q Power Cycling Tests ResultsØ Optical Microscopic Images after Power Cycling
for 2-Turn Module
ØElectrical Characteristics of The Integrated Modules before and after Power Cycling§ 2-Turn Module
q Passive Integrated Modules Used in Power Cycling Testsq ObjectiveØInvestigate Failure Mechanismof the Passive IPEMsØHelp to Estimate Lifetime of the
Passive IPEMs
• 2-Turn Module •3-Turn Module
- Relative Dielectric Constant of Ceramic Substrate: 1720~2690
q ConclusionØ Cracks and Delaminations Were
Found after Power Cycling TestsØ Capacitances Were Changed within
10,000 Cycles with about 90°C~97°CTemp. Variation and 96sec Power Cycle
• Electrical Characteristics of LC Modules
§ 3-Turn Module
A Comparison of Lifetime and Thermal Performance of Lead-free Solder Alloys for Bonding Power Devices in Flux
and Flux-less Reflow Processes
Sponsored by: NSF
D. Huff, D. Katsis, J.D. van Wyk, G-Q Lu
D4.2
20032003
Vacuum-reflowedPower Devices
Resulting Scanning Acoustic Micrograph of Solder Junction
Vacuum Reflow Furnace
Graphite Fixture
0 cycles
2000x
0 cycles
6000x
60 cycles
2000x60 cycles
6000x
0 cycles
2000x
0 cycles
6000x
60 cycles
2000x60 cycles
6000x
Progression of void growth, 0 to 500 temperature cycles
Coarsening of solder grain structure
New Device 1000 2000 3000 4000
5000 6000 7000 10000
Void Growth
10
20
30
40
50
0 2000 4000 6000 8000 10000Power Cycles
Void
%
Typical Large AreaSolder Application
Materials
•Substrate
•Solder
•Plating
Processing
•Equipment and Tooling
•Atmosphere
•Process Profile
Accelerated Life Testing
•Temperature Cycling
•Power Cycling
Project Scope: Die Attach of Silicon Power Devices
Flux-less Vacuum Solder ReflowUsing Lead-free Solder
Reliability Analysis
Analysis Tools
•Scanning Acoustic Microscopy (SAM)
•Scanning Electron Microscopy (SEM)
•Thermal Performance Measurement
•Modeling
MOSFETS
Chamber Shelf
MOSFETS
Chamber Shelf
-60
-40
-20
0
20
40
60
80
100
0 2 4 6 8 10time-minutes
tem
pera
ture
-C
Actual Profile Temp
Die Attach Process
Direct Bond Copper (DBC) Substrate with Au/Ni Plating
Temperature/Power Cycling Void Growth
SAM Analysis SEM Analysis
For details on Thermal Performance Measurementand Modeling please see poster #R52 in room 633C !
Reflow Equipment and Tooling Process Profile
Solder
Solder Types
•Sn100
•Sn95.5Ag4Cu0.5
•Sn95Sb5
•Sn96.5Ag3.5
•Sn97Cu2Sb0.8Ag0.2
•Sn63Pb37
•SnBi
Courtesy of Kelly Stinson-Bagby
Coining of America, LLCKester, Inc.
•<50 mtorr Vacuum
•Ar95H25 Forming Gas
•N295H25 Forming Gas
•N2 Gas
Atmosphere
Substrate and Plating
Calorimeter for Accurate Measurement of Power Losses in IPEMs
Sponsored by: NSF
Chucheng Xiao, Gang Chen, W.G. Odendaal
D4.3
20032003
Motivation: Power loss measurement in active IPEMs, passive IPEMs and integrated magnetics.
Heat Sink
DUT
RP2
TC
I2Rm/2
I2Rm/2
αΙTC
αΙTH
TH
1/Km
RP1
Rhs
Rfs
To
Rio
Roa
Tin
Ta Ta
RP2
TC
I2Rm/2
I2Rm/2
αΙTC
αΙTH
TH
1/Km
RP1
Rhs
Rfs
To
Rio
Roa
Tin
Ta Ta
TE Modules
Base Plate II
Heat Flux Sensor
Two covers
Leads
Base Plate I
Schematic of proposed calorimeter Heat transfer model
Temperature Controller for two covers(limits heat leakage)
TE modules as cooler to direct heat(controls temperature on DUT)
Heater
+15
Inner cover
Outer cover
AD595
AD595
DifferentialAmplifier
CMP
+15
-15+15
Driver
+15
-15
vPolished double covers
vInsulator: fiberglass
vAir-tight test chamber: feedthroughs, spring-loaded connectors, gasket, adjustable draw latches
Features:Ø Direct measurement of heat flux
Ø Error < 5%
Ø Control an arbitrary temperature on DUT
Ø Easy to setup
TEModules
Base Plate II AD595
AmplifierCMP
+15
-15+15
Driver
+15
-15
Vref
DUT
Outer coverOuter cover
Inner coverInner cover
InsulationInsulation
Base plateBase plate
Heat flux sensor
Heat flux sensor
Thermoelectric Modules
Thermoelectric Modules
Base coverBase cover
TemperatureControlCircuits
Power Supply
Power Supply
HeaterHeater
VDS
VGS
Control signals of TE moduleControl signals of TE moduleOutput of heat flux sensor for 4.26 W power
loss measurementOutput of heat flux sensor for 4.26 W power
loss measurement
20
30
40
50
60
70
80
90
100
0 10 20 30 40 50time-ms
tem
p-C
0 15% 40%55% 15% Avg 40% Avg55% Avg
A Finite Element Modeling of Dynamic Hot Spot Effects in MOSFET Dies
Sponsored by: National Science Foundation
D. C. Katsis and J.D. vanWyk
D4.4
20032003
=CapacitanceJ/ ºC=Resistanceθ, ºC/W=AmpsW=VoltsºC
RC Circuit Thermal Models
58ºC
55ºC
51ºC
48ºC
55% Voided Area
40% Voided Area
15% Voided Area
0% Voided Area
C, 10k cycles
A, 0 cycles
B, 7k cycles
15% Die-Attach Area Voided
40% Die-Attach Area Voided
55% Die-Attach Area Voided
SimulatedActual
FEA Simulation Development FEA Results vs. Experimental Observation
5ms 10ms 15ms 20ms
30ms 35ms 40ms
45ms 50ms
25ms
FEA Heating Simulation
21ms
30ms
22ms20ms
25ms23ms
FEA Cooling Simulation
RC Circuit Response With Encapsulation
Voids
CrossSection
Investigation of Gate-driver Selection based on Operating Conditions
Sponsored by: National Science Foundation / Raytheon
J.T. Strydom, J.D. van Wyk
D4.5
20032003
Bus
vol
tage
(V)
Gate capacitance
RV
VS
Nor
mal
ized
vol
tage
RN=1
RN=2RN=4
RN=10
1
0
0.2
0.4
0.6
0.8
RN=8RN=6
200 2 4 6 8 10 12 14 16 18
RN=1
RN=2
RN=4RN=6
200 2 4 6 8 10 12 14 16 18
1
0
0.2
0.4
0.6
0.8
Nor
mal
ized
cur
rent
Nor
mal
ized
cur
rent
IN=1
IN=1/2
IN=1/4IN=1/6
100 1 2 3 4 5 6 7 8 9
1
0
0.2
0.4
0.6
0.8
IN=1/8IN=0.1
Nor
mal
ized
vol
tage IN=1
IN=1/2IN=1/4
IN=0.1
100 1 2 3 4 5 6 7 8 9
1
0
0.2
0.4
0.6
0.8
IN=1/8IN=1/6
Gate capacitance
RC
RFIS
Gate capacitance
RRLR
RP Clamped resonance
0.6
0.4
0.2
00 2ππ 3π
0.8
Nor
mal
ized
Indu
ctor
Cur
rent
Q=5
Q=2.5
Q=0.5
Q=0.625
Q=1
Clamped resonance- Vs =Vg
1.5
1
0.5
00 2ππN
orm
aliz
ed C
apac
itor
Vol
tage Full resonance
Clamped resonance- Vs =Vg
Q=5Q=2.5
Q=0.5
Q=0.625
Q=1
0.8
RV (3), RR (15)
IS (6)
RF (9)
RR, Q=5
0.01
0.1
1
100
Z0 (14)
50 0.5 1 1.5 2 2.5 3 3.5 4 4.5
10
RRHQ (18)
Rgate=1Ω
Res
ista
nce
(Ω),
Cur
rent
(A
) Maximum clamped resonance frequencyMaximum full resonance frequency ~ Q = 5
Maximum frequency: Voltage source driver
Increasing Q
Maximum freewheeling resistance: Pc-
baseline=PV
Decreasing max. freq.Full resonance Q=5
Current driven: RF = 0
0.01
0.1
1
Full resonance Q=1/2
50 0.5 1 1.5 2 2.5 3 3.5 4 4.5
10Current driven: RF = 0.1RV
Increasing Q
Voltage driven
Increasing freewheeling losses current driven
Pow
er d
issip
ation
(W)
Max. crossover point for losses between voltage and current driven topologies
20mm
Semiconductor area
Passives area
Semiconductor area
Passives area
Semiconductor area
Passives area
Gate-driver Investigation Most neglected part of converter
design Influences converter performance Most pronounced at:
High frequency High voltage Low power
Different topologies possible Comparative study to evaluate
gate-driver topologies
3π
Opto-isolation
Galvanically isolated supply
Opto-isolation
SiC BootstrapSupply
LV level shift and Si Bootstrap supply
HV level shiftBootstrap supply
0.5 1 2 3 4 5Switching frequency (MHz)
200
0
400
600
800
1000
1200
1400
Representative voltage and frequency ranges of different commercial gate-driver systems
Representative gate voltage for base-line requirement of all gate-driver topologies
Simplified charge-interval circuit for the Voltage Source gate-driver topology
Simplified charge-interval circuit for the Current Source gate-driver topology
Normalized time
Normalized time
Normalized time
Normalized time Normalized time
Normalized time
Simplified charge-interval circuit for the Resonant gate-driver topology
Charge time ≤ 5% of period
Switching periodDuty cycle variation
Shape dependent on gate-driver ∆t
0.8VS
VS
Current and Voltage waveforms for different Q-values for the resonant topology
Current and Voltage waveforms for different current source values
Current and Voltage waveforms for different series resistance values
Frequency (MHz)Frequency (MHz)
Gate-driver comparison resultsUnder the presented assumptions: Below cross over point: Current
source has lowest losses Above cross-over: Current
source - highest losses, but only solution for required rise time
Resonant lowest losses, but most frequency limited
Maximum allowable resistances for the three gate-driver topologies as function of frequency
Maximum allowable resistances for the three gate-driver topologies as function of frequency
Same topology Same technology
Voltage driven - 200kHz, hybrid Voltage driven – 1MHz, SMT Current driven – 1MHz, SMT
Processing and Reliability Study of Pressure-assisted Low-Temperature
Sintered Die-attach
Sponsored by: National Science Foundation
Zach Zhang and Dr. Guo-Quan Lu
D4.6
20032003
Conclusion:
Increasing the temperature and pressure improved the electrical and thermal conductivity, and the shear strength.
Increasing the time beyond two minutes did not change the properties significantly.
I Processing parameters investigation
Thermal cycling
0 . 0 0 E + 0 0
5 . 0 0 E + 0 4
1 . 0 0 E + 0 5
1 . 5 0 E + 0 5
2 . 0 0 E + 0 5
2 . 5 0 E + 0 5
3 . 0 0 E + 0 5
3 . 5 0 E + 0 5
4 . 0 0 E + 0 5
4 . 5 0 E + 0 5
5 . 0 0 E + 0 5
1 7 0 O C 2 4 0 O C 3 0 0 O C 1 0 M P a 2 0 M P a 4 0 M P a 2 M in 5 M in 1 0 M in
s int e re d a t 4 0 M P a, 5 M ins int e re d a t 2 4 0 o C , 5 mins int e re d a t 4 0 M P a, 2 4 0 o C
0
2 0
4 0
6 0
8 0
1 0 0
1 2 0
1 4 0
170OC 240OC 300OC 10MP a 20MP a 40MPa 2Min 5Min 10Min
s int e re d a t 4 0 M P a, 5 M ins int e re d a t 2 4 0 o C , 5M ins int e re d a t 2 4 0 o C , 4 0 M P a
c
0
1 0
2 0
3 0
4 0
5 0
6 0
7 0
1 7 0 O C 2 4 0 O C 3 0 0 O C 1 0 M P a 2 0 M P a 4 0 M P a 2 M i n 5 M i n 1 0 M i n
s i n t e re d a t 4 0 MP a , 5 Mi ns i n t e re d a t 2 4 0 o C , 5 Mi ns i n t e re d a t 4 0 MP a , 2 4 0 o C
II Reliability investigation
Before cycling
Solder (Sn63Pb37) Reflow Die-attach layer
SAM
300 cycles
Silver sintering(240oC, 40MPa, 5Min)
Die-attach layerSAM
Appearance of New Voids
Growth of Void Area
Edge Cracking
Destructive adhesion test accomplished by chiseling the chip from the substrate
Conclusion:
The sintered silver die-attach layer is highly resistant to void-induced failure.
Inhomogeneous pores distribution, binder residue may cause local concentrations of stress during the cycling load because of CTE mismatch. These stress concentrations usually initiate and propagate cracks these sintered silver die-attach layer.
Initial voids
Silver sintering(240oC, 40MPa, 5Min)
Die-attach layerSEM
Largepores
Small pores
Largermagnification
Shea
r Stre
ngth
(MPa
)
Effect of Processing Parameters on Shear Strength
Processing Parameters
Effect of Processing Parameters on Thermal conductivity
Ther
mal
con
duct
ivity
(W/m
-K)
Processing Parameters
Effect of Processing Parameters on Electrical Conductivity
Elec
trica
l Con
duct
ivity
((Ohm
*cm
)-1)
Processing Parameters
Planar Litz - A method to reduce high frequency conduction losses in integrated
components
Sponsored by: CPES
M.A. de Rooij, J.T. Strydom, J.D. van Wyk
D4.7
20032003
Position
Wide planar conductor with flux lines (for a current into or out of the page)
Current density along the center line in the conductor I[A]
Substrate Lower l ayer of planar litz conductor
Upper layer of planar litz conductor
Layer inter -connects
Planar litz conductor direction
Planar litz lines
Inner layer strands
Outer layer strands
Strand angle
Vias wPLC
Inner layer Outer layer
1 1
1
2
1
3
3
4 4
5
5
litz blocks
27 strands 20mils x 20mils
57 strands 20mils x 20mils
57 strands 10mils x 10mils
Inner layer and core Outer layer and core Assembled planar litz conductor and core
12345 A
B
CR LR
LM
4 : 2
A
B
D
100V 24V
L-L-C-T module
IP
Outer layer
Via layer
Inner layer
Top layer Bottom layer
Outer layer
Via layer
Inner layer
Terminal blocks used to accommodate narrowing of the planar litz conductor
44 mm
91mm
12mm
Primary Winding
Secondary Winding
Alignment holes Vias
Copper sputtered onto substrate, and electroplated
Step 1: Cutting substrate Step 2: Application of copper to substrate
Step 3: Etching of inner layer Step 4: Masking of inner layer
Step 5: Applicat ion of copper for outer layer
Step 6: Etching of outer layer
Step 7: Masking of outer layer Step 8: Final cutting
Current density distribution and magnetic flux lines in an isolated wide planar conductor
( )maxmax20 torofWcond δ⋅>Definition of a wide planar conductor
where Wcond = Width of the planar conductor [m]d = Skin depth of the conductor [m]tmax = Maximum thickness of the conductor [m]
Definition for a planar litz conductor
Strand/s: Narrow conductor/s used to make up a planarlitz conductor.
Planar litz Conductor: A planar conductor comprised of many isolated strands that are weaved together to form a single high frequency, higher current carrying capacity conductor.
Planar litz conductor guidelines1. Any improvement over a single planar conductor is an improvement.
2. An extra layer is required. The additional thickness is small . Multi-layer structures are also possible by zigzagging additional layers as shown in the figure.
3. Minimum number of litz strands is 3.
4. The higher the number of strands the more efficient the effect.
5. All the litz strands must be of equal length and width so that the impedances of each strand will be the same.
6. The litz strand may zig-zag within the winding window as many times as is possible.
7. An optimal strand angle is 45º.
8. The width of the planar litz conductor will be greater than or equal to half the width of a single conductor.
9. Inter-connects are used to connect the layers of the planarlitz conductor together on the outer edges of the planar litz conductor only.
Simplified top views of the planar litz conductor building blocks
Converter Parameter Value Units Maximum Input voltage (Vi n) 200 V Maximum Output voltage (V o) 48 V Maximum Output power (Po) 500 W Resonant frequency (fo) 1.1 MHz Resonant capacitance (CR) 50 nF Resonant inductance (L R) 400 nH Magnetizing inductance (L M) 20 µH Primary winding (NP) 4 turns Secondary winding (N S1) = (NS2) 1 turn
Steps followed to construct a planar litzconductor for experimental
1. A corner block.
2. A square block. The following equation describes the line position change of a strand where a strands starts in line x and ends in line y and where:
3. A rectangle block.
4. A termination block.
5. A termination compensation block.
x nlines≤y n xlines= + −( )1
Different #2 blocks sizes for the same and different strand and spacing widths
Top views of the assembled planar litzconductor with a planar E-core
EXPERIMENTAL VERIFICATION OF THE PLANAR LITZ CONDUCTOR
Power circuit used to test the planar litzconcept
Experimental converter specifications
Primary side litz conductor patterns
Secondary side litz conductor pattern
Photo of an experimental winding made using planar litz for an integrated passive component
EXPERIMENTAL
Using a half grid to eliminate the gap in the corner blockWide planar conductors utilize the
conductor area poorly and a new method is sought to capitalize on the benefits of
litz wire for planar conductors
THE PLANAR LITZ CONDUCTOR
Wide planar conductors are widely used in many integrated power electronic
passive components and are usually made to be skin depth thick at the
designed operating frequency
An Overview of Electromagnetic Modeling of Resonant Integrated Spiral Planar Power
Passives (ISP3)
Sponsored by: National Science Foundation
J.T. Strydom, J.D. van Wyk
D4.8
20032003
LM
LR1CR
NP NS
LR2
CMSCMP
Ciw
Complete L-L-C-TStructure
Ferrite Core
Leakage Layer
Secondary Winding
Primary Winding
Main core
LR
NS
NP
LM
LMCR
Acore
ka⋅lcore
lcore
wwAC
ka⋅lcore /2
lT
l
wT
Bottom conductor Mth layer
Top conductor Mth layer
Bottom conductor mth layer
Top conductor mth layer
w
Bottom conductor 1st layer
Top conductor 1st layer
HNI/w mNI/w MNI/w
I/Mw
(m-1)NI/w
x = 0
x = l
Structure length
N conductors
Bottom conductor Mth layer
Top conductor Mth layer
Bottom conductor mth layer
Top conductor mth layer
w
Bottom conductor 1st layer
Top conductor 1st layer
HNI/w mNI/w MNI/w
I/Mw
(m-1)NI/w
x = 0
x = l
Structure length
N conductors
Prototype 2
Primary side
Secondary side
Heat sink
Gate drives
Prototype 3 - connected
Prototype 1
Thermo-couples
Input voltage measurement
Output voltage measurement
Input current measurement
DC-input
DC-output
0.001
1
0.01
0.1
0.2
0.5
0.02
0.05
0.002
0.005
Cor
e as
pect
rat
io k
a
Inherent leakage inductance constraint
Local-minimum507W.cm3
Volume-loss product constraint
600
1k
2k
12501.5k
1500 10050 200Flux density BT (mT) (a)
0.001
1
0.01
0.1
0.2
0.5
0.02
0.05
0.002
0.005
Cor
e as
pect
rat
io k
a
Inherent leakage inductance constraint
Local-minimum312W.cm3
Volume-loss product constraint
1k500
1.5k
350
400
1500 10050 200Flux density BT (mT) (b)
0.001
1
0.01
0.1
0.2
0.5
0.02
0.05
0.002
0.005
Cor
e as
pect
rat
io k
a
Inherent leakage inductance constraint
Local-minimum232W.cm3
Profile restriction
250
1k500
375
1.5k
1500 10050 200Flux density BT (mT) (c)
0.001
1
0.01
0.1
0.2
0.5
0.02
0.05
0.002
0.005
Cor
e as
pect
rat
io k
a
Local-minimum194W.cm3
Profile restriction
Inherent leakage inductance constraint
250
1k500 375
1500 10050 200Flux density BT (mT) (d)
Initial construction technology
Improved electromagnetic modeling and optimization
Initial electromagnetic modeling
1kW, 1MHz prototype (1999) – 130W/in3
Improved construction technology CPES
½kW, 1MHz prototype 1 (2001) – 240W/in3
½kW, 1MHz prototypes 2 and 3(2002) – 480W/in3
Process development
Core aspect ratio
Jmax
0
1
0
Bmax0,001
Current density Flux density
Design Parameter Prototype 1 Prototype 2 Prototype 3
Input Voltage (V) 200.0 ±0.7 200.0 ±0.7 200.1 ±0.7
Input Current (A) 2.89 ±0.02 2.87 ±0.02 2.83 ±0.02
Output Voltage (V) 49.9 ±0.06 50.3 ±0.07 50.3 ±0.07
Output Current (A) 10.15 ±0.02 10.17 ±0.02 10.00 ±0.02
Input Power (W) 577.7±6.5 574.7±6.4 564.9 ±6.4
Output Power (W) 506.3 ±1.9 511.7 ±1.9 503.3 ±1.9
Total Converter Losses (W) 71.4 ±8.4 63.0 ±8.4 61.6 ±8.3
Converter Efficiency (%) 87.7 ±1.3 89.0 ±1.3 89.1 ±1.3
Module power density [W/cm3 / W/in3] 14.1 / 230.5 29.8 / 487.6 28.6 / 468.6
Top core temperature 47.4 45.9 42.2
Middle inside temperature 57 51 48
Bottom inside temperature 51 44 43
Heatsink temperature 26.2 26.3 26.2
Design Parameter Prototype 1 Prototype 2 Prototype 3
Input Voltage (V) 200.0 ±0.7 200.0 ±0.7 200.1 ±0.7
Input Current (A) 2.89 ±0.02 2.87 ±0.02 2.83 ±0.02
Output Voltage (V) 49.9 ±0.06 50.3 ±0.07 50.3 ±0.07
Output Current (A) 10.15 ±0.02 10.17 ±0.02 10.00 ±0.02
Input Power (W) 577.7±6.5 574.7±6.4 564.9 ±6.4
Output Power (W) 506.3 ±1.9 511.7 ±1.9 503.3 ±1.9
Total Converter Losses (W) 71.4 ±8.4 63.0 ±8.4 61.6 ±8.3
Converter Efficiency (%) 87.7 ±1.3 89.0 ±1.3 89.1 ±1.3
Module power density [W/cm3 / W/in3] 14.1 / 230.5 29.8 / 487.6 28.6 / 468.6
Top core temperature 47.4 45.9 42.2
Middle inside temperature 57 51 48
Bottom inside temperature 51 44 43
Heatsink temperature 26.2 26.3 26.2
Design Parameter Prototype 1 Prototype 2 Prototype 3
Input Voltage (V) 200.0 ±0.7 200.0 ±0.7 200.1 ±0.7
Input Current (A) 2.89 ±0.02 2.87 ±0.02 2.83 ±0.02
Output Voltage (V) 49.9 ±0.06 50.3 ±0.07 50.3 ±0.07
Output Current (A) 10.15 ±0.02 10.17 ±0.02 10.00 ±0.02
Input Power (W) 577.7±6.5 574.7±6.4 564.9 ±6.4
Output Power (W) 506.3 ±1.9 511.7 ±1.9 503.3 ±1.9
Total Converter Losses (W) 71.4 ±8.4 63.0 ±8.4 61.6 ±8.3
Converter Efficiency (%) 87.7 ±1.3 89.0 ±1.3 89.1 ±1.3
Module power density [W/cm3 / W/in3] 14.1 / 230.5 29.8 / 487.6 28.6 / 468.6
Top core temperature 47.4 45.9 42.2
Middle inside temperature 57 51 48
Bottom inside temperature 51 44 43
Heatsink temperature 26.2 26.3 26.2
Design ParameterDesign Parameter Prototype 1Prototype 1 Prototype 2Prototype 2 Prototype 3Prototype 3
Input Voltage (V)Input Voltage (V) 200.0 ±0.7200.0 ±0.7 200.0 ±0.7200.0 ±0.7 200.1 ±0.7200.1 ±0.7
Input Current (A)Input Current (A) 2.89 ±0.022.89 ±0.02 2.87 ±0.022.87 ±0.02 2.83 ±0.022.83 ±0.02
Output Voltage (V)Output Voltage (V) 49.9 ±0.0649.9 ±0.06 50.3 ±0.0750.3 ±0.07 50.3 ±0.0750.3 ±0.07
Output Current (A)Output Current (A) 10.15 ±0.0210.15 ±0.02 10.17 ±0.0210.17 ±0.02 10.00 ±0.0210.00 ±0.02
Input Power (W)Input Power (W) 577.7±6.5577.7±6.5 574.7±6.4574.7±6.4 564.9 ±6.4564.9 ±6.4
Output Power (W)Output Power (W) 506.3 ±1.9506.3 ±1.9 511.7 ±1.9511.7 ±1.9 503.3 ±1.9503.3 ±1.9
Total Converter Losses (W)Total Converter Losses (W) 71.4 ±8.471.4 ±8.4 63.0 ±8.463.0 ±8.4 61.6 ±8.361.6 ±8.3
Converter Efficiency (%)Converter Efficiency (%) 87.7 ±1.387.7 ±1.3 89.0 ±1.389.0 ±1.3 89.1 ±1.389.1 ±1.3
Module power density [W/cm3 / W/in3]Module power density [W/cm3 / W/in3] 14.1 / 230.514.1 / 230.5 29.8 / 487.629.8 / 487.6 28.6 / 468.628.6 / 468.6
Top core temperatureTop core temperature 47.447.4 45.945.9 42.242.2
Middle inside temperatureMiddle inside temperature 5757 5151 4848
Bottom inside temperatureBottom inside temperature 5151 4444 4343
Heatsink temperatureHeatsink temperature 26.226.2 26.326.3 26.226.2
L-L-C-T Loss Component (W) Prototype 1 Prototype 2 Prototype 3
Ferrite Core Losses 4.8 4.3 3.7Dielectric Losses 5.5 5.6 5.0Primary Conductor Losses 14.1 11.6 14.8Secondary Conductor Losses 2.9 4.6 5.3Total Estimated L-L-C-T Losses 27.3 26.1 28.8Semiconductor Losses 31.0 31.1 30.8Estimated Converter Losses 58.3 57.2 59.6Experimental Losses 71.4 ±8.4 63.0 ±8.4 61.6 ±8.3Unaccounted Losses 13.1 ±8.4 5.8 ±8.4 2.0 ±8.3Calculated module efficiency [%] 94.6 ±1.3 94.9 ±1.3 94.3 ±1.3
L-L-C-T Loss Component (W) Prototype 1 Prototype 2 Prototype 3
Ferrite Core Losses 4.8 4.3 3.7Dielectric Losses 5.5 5.6 5.0Primary Conductor Losses 14.1 11.6 14.8Secondary Conductor Losses 2.9 4.6 5.3Total Estimated L-L-C-T Losses 27.3 26.1 28.8Semiconductor Losses 31.0 31.1 30.8Estimated Converter Losses 58.3 57.2 59.6Experimental Losses 71.4 ±8.4 63.0 ±8.4 61.6 ±8.3Unaccounted Losses 13.1 ±8.4 5.8 ±8.4 2.0 ±8.3Calculated module efficiency [%] 94.6 ±1.3 94.9 ±1.3 94.3 ±1.3
L-L-C-T Loss Component (W)L-L-C-T Loss Component (W) Prototype 1Prototype 1 Prototype 2Prototype 2 Prototype 3Prototype 3
Ferrite Core Losses Ferrite Core Losses 4.84.8 4.34.3 3.73.7Dielectric LossesDielectric Losses 5.55.5 5.65.6 5.05.0Primary Conductor LossesPrimary Conductor Losses 14.114.1 11.611.6 14.814.8Secondary Conductor LossesSecondary Conductor Losses 2.92.9 4.64.6 5.35.3Total Estimated L-L-C-T LossesTotal Estimated L-L-C-T Losses 27.327.3 26.126.1 28.828.8Semiconductor LossesSemiconductor Losses 31.031.0 31.131.1 30.830.8Estimated Converter LossesEstimated Converter Losses 58.358.3 57.257.2 59.659.6Experimental LossesExperimental Losses 71.4 ±8.471.4 ±8.4 63.0 ±8.463.0 ±8.4 61.6 ±8.361.6 ±8.3Unaccounted LossesUnaccounted Losses 13.1 ±8.413.1 ±8.4 5.8 ±8.45.8 ±8.4 2.0 ±8.32.0 ±8.3Calculated module efficiency [%]Calculated module efficiency [%] 94.6 ±1.394.6 ±1.3 94.9 ±1.394.9 ±1.3 94.3 ±1.394.3 ±1.3
Prototype 1
Prototype 2
Prototype 3
A
B
C
DE
A
B
C
DE
A
B
C
DE
Series resonant ISP3 L-L-C-T structure: (a) equivalent circuit and (b) exploded view
(a)
(b)
Structural dimensions for the L-L-C-T structure (cross-sectional view from top)Two-dimensional magnetic field intensity for the primary-
side winding window showing the spatial distribution
3D design space showing some idealized design constraints
Remaining designs meeting
design constraints
Efficiency constraint
Inherent inductance constraint
Volumetric constraint
0
1
Bmax0,001
Current density
Core aspect ratio
Flux density
Jmax
Partial design spaces showing volume-loss product for current densities of (a) 8A/mm2, (b) 12A/mm2, (c) 16A/mm2, (d) 20A/mm2 for the 500W, 1MHz prototype converter
Experimental setup showing integrated L-L-C-T prototypes
Future prototypes –1000W/in3 possible
Design space showing location of selected sections
Modeling results compared to experimental measurementsExperimental measurement results Prototypes
(a) Temperature recording and curve fitting
(b) Thermal equivalent circuit of four time constants
An Experiment for Accurate Characterization ofTransient and Steady-state Die Thermal Parameters
Sponsored by: Philips Research
Jian Yin, J. D. van Wyk and W. G. Odendaal
D4.9
20032003
(b) Gate-source voltage of the MOSF (c) Experimental waveform of Vgs
time
Achieve transient thermal parameters to improve thermal management compared to the existing approachesObjective
Setup
Experiment
Recording of cool-down curve and achievement of RC equivalent circuit
On-state resistance Measurement
• Accurate determination of thermal parameters of power device packaging
• Identification of the physical meaning
• A valuable method to investigate and evaluate die connecting approaches
• Further careful analysis provides a key to an improved or new double-sided cooling structure
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.520
22
24
26
28
30
32
34
Time (S)
Tem
pera
ture
(oC
)D. U. T
DBCSolder
Heat Sink
Insulating Materials
Thermo-electric CoolerHeat Spreader
Heat Spreader
Adjustable Fixture
Wire bond sample
0.1156 0.2184
0.0217 0.1224 1.6697
0.0655
To
0.3815
0.2250
Die HeatSpreader
(a) Temperature transient curve
Vgs
T
time
d
D
D.U.T
Heat Sink
Heating andMeasurement
TemperatureControl
Voltage
T
TE Module
(d) Control Process
• Self heating by measurement:0.00134°C (one pulse)
• MOSFET die is controlled tolinear region to heat and saturation region to measure
• Bottom temperature of the heatspreader is controlled
• Encapsulation and natural air flow
• Blue uneven line is temperature recording curve and red line iscurve fitting curve
Measurement and Calculation of Residual Stress between Substrate and Copper Deposition in
Integrated Passive Modules
Sponsored by: National Science Foundation
Ning Zhu, S.Y.Lee, J.D.van Wyk, W.G.Odendaal, J.N.Calata
Problem:Possible reasons for failures of passive modules: Intrinsic residual stress & thermal stress in
metallization layer. Objective:
Compare the contributions of thermal-mechanical stress and intrinsic residual stress to the mechanical failures of the moduleMethodology:• use smooth,flat substrate to show up the metallization stresses• use same metallization procedures as processing passive modules• measure the sample bending using Dektak surface profile measuring system • calculate the residual stress based on Stoney Equation• simulate the thermal-mechanical stress of the same films using IDEAS software
D4.10
20032003
16mmGlass
16 mm18mm
18mm
Copper
• Sample Construction: • Sample bending after electroplating:
-5000
0
5000
10000
15000
20000
25000
30000
35000
40000
• Measurement results: example• Measurement equipment:
0
5000
10000
15000
20000
25000
30000
35000
⋅
−+⋅
−⋅+⋅= 33
11)(61
ss
sf
f
f
fsff d
Ed
E
dddR υυσ
• Residual stress Calculations:
• Thermal stress simulation results: example
0
5000
10000
15000
20000
25000
30000
35000
40000
Measurement Length 5mm
Strain δ 3.4µm
Bottom side of the copper Top view of the sample
100MPa
Sputtered copper (400-500nm)
Glass substrate(150µm)
Electroplatedcopper (100µm)
• Cross-section of the sample:
Electroplated copper
Glass substrate
Sample processing steps: 1. Sandblast substrate(Grit #400) for adhesion; 2. Sputter Cu; 3. Electroplate Cu for 100 µm.
Electroplating conditions: 1. Room temperature: around 20°C; 2.Current density 300 A/m2 ; 3.Plating time: 4 hours
• Electroplating tank:
Conclusions: 1. Residual stresses in metallized films are measurable by substrate distortion. 2. By careful experimental investigation, consistent results can be obtained.3. A methodology to compare residual and thermal-mechanical stresses in metal film on
substrate has been developed.Future work: What’s the relations between metallization process parameters and the residual stresses?
3851.38Thermal Cond’ty(W/m-K)
11064Young’s Modulus(Gpa)
0.3430.19Poisson’s Ratio
16.44CTE(µm/m-°C)100150Thickness(µm)Cuglassmaterials
Material properties: Comparison between residual stress and thermal -mechanical
stress: example
1.5
100
1
10
100
residual thermal-mech’al
MPa
1.528 MPa1.894 m3.3µmMeasurement (c)
1.505 MPa1.923 m3.25µmMeasurement (b)
1.574 MPa1.838 m3.4µmMeasurement (a)Residual stressCurvature RStrain δ
A Planar Process Module for Active IPEM Manufacturing
Sponsored by: NSF
Z. Liang and J. D. van Wyk20032003
Base Substrate
CSP
Base Substrate
CSP CSP
Surface Mount
Double sided Cooling Capability
Single Chip Packaging
Dual Chip Packaging
IPEM Packaging
To Integrated Power Chips Stage
3-D Assembly
Gat
e D
river
High-density Assembly
Double MOSFET Module for Half-bridge Switching Configuration
Embedded Power
From Power MOSFET, IGBT, Diode Bare Dice
2 CoolMOSFETs+2 SiC Diodes for PFC Power Switching
DPS System Active IPEM: 2 CoollMOS+2 SiC Diodes +2 Si MOSFETs
Discrete MOSFET Device
S1 S2S1 S2
G1 G2
S1 S2S1 S2
G1 G2
D1 D2D1 D2
D1 D2D1 D2
G1 S1
D1
G2
S2
D2
Gat
e D
river
Gat
e D
river
Drain (Cr/V-Ni/Au)
Source (Al)Gate (Al)
Ceramic Chip Carrier Preparation
Multichip Mount
SiSi
Interlayer Dielectric Coat
Si Si
Metallization Pattern Deposition
Si Si
D4.11
Preliminary Electromagnetic Modeling of an Integrated RF-EMI Filter
Sponsored by: NSF
Lingyin Zhao and J. D. van Wyk
S8.1
20032003
The prototype
Preliminary calculation results.
1 MHz10 kHz 100 MHz~22 MHz
0 dB
0 °
-180 °
-50 dB
1 MHz10 kHz 100 MHz~22 MHz
0 dB
0 °
-180 °
-50 dB
1 MHz10 kHz 100 MHz~22 MHz
0 dB
0 °
-180 °
-50 dB
Slope: 40 dB/decade
103 104 105 106 107 108-60
-40
-20
0
20
Frequency (Hz)
Gai
n (d
B)
Rg=1m Ohms
103
104
105
106
107
108
-200
-100
0
100
200
Frequency (Hz)
Pha
se (
degr
ee)
Rg=1m Ohms
Measurement results of transfer gain at ZL=50Ω.
Cu
Al2O3Ni
NiBaTiO3
Al2O3
Cu
HF: High AttenuationLF: Conducting
+
_Integrated Power Electronic Module LOAD_
+SOURCE RF-EMI
Filter
+
_Integrated Power Electronic Module LOAD_
+_+
SOURCE RF-EMI Filter
• Proximity effect does influence the resonances and current distributions.• The filter behaves like a R-C network at very high frequency due to the dominance of resistances over inductances.• The measurement setup used can not measure the real transfer gain. • Accurate loss modeling – future work
• Proximity effect does influence the resonances and current distributions.• The filter behaves like a R-C network at very high frequency due to the dominance of resistances over inductances.• The measurement setup used can not measure the real transfer gain. • Accurate loss modeling – future work
C12∆x
L1∆x
L2∆x
G12∆x
R1∆x
R2∆x
I1(x,t)
I2(x,t)
I1(x +∆x,t)
I2(x +∆x,t)
C34∆x
L3∆x
L4∆x
G34∆x
R3∆x
R4∆x
I3(x,t)
I4(x,t)
I3(x +∆x,t)
I4(x +∆x,t)
C23∆x G23∆x
C12∆x
L1∆x
L2∆x
G12∆x
R1∆x
R2∆x
I1(x,t)
I2(x,t)
I1(x +∆x,t)
I2(x +∆x,t)
C34∆x
L3∆x
L4∆x
G34∆x
R3∆x
R4∆x
I3(x,t)
I4(x,t)
I3(x +∆x,t)
I4(x +∆x,t)
C23∆x G23∆x
C12∆x
L1∆x
L2∆xG12∆x
R1∆x
R2∆x
C34∆x
L3∆x
L4∆xG34∆x
R3∆x
R4∆x
C23∆x G23∆x
M3,12∆x
M3,12∆x
M4,12∆x
M4,12∆x
M2,34∆x
M2,34∆x
M1,34∆x
M1,34∆x
C12∆x
L1∆x
L2∆xG12∆x
R1∆x
R2∆x
C34∆x
L3∆x
L4∆xG34∆x
R3∆x
R4∆x
C23∆x G23∆x
M3,12∆x
M3,12∆x
M4,12∆x
M4,12∆x
M2,34∆x
M2,34∆x
M1,34∆x
M1,34∆x
(a) Model without considering proximity effect
(b) Model with proximity effect considerations
Proximity Effect Modeling
I
I
A
B
C
D
IZL
ZL VoutVin
1
23
4
0 l x
I
I
A
B
C
D
IZL
ZL VoutVin
1
23
4
0 l x
Multi-conductor Transmission Structure Model
Measurement setup.
Equivalent connections
Measurement Issues
Parameter Extraction
Zg IZg
I
I
A
B
C
D
IZL
ZL VoutVin
1
23
4
0 l x
Zg IZg
I
I
A
B
C
D
IZL
ZL VoutVin
1
23
4
0 l x
Difference made by common-ground measurement setup.
103 104 105 106 107 108-60
-40
-20
0
20
Frequency (Hz)
Gai
n (d
B)
Rg=1M OhmsRg=1m Ohms
103 104 105 106 107 108-200
-100
0
100
200
Frequency (Hz)
Pha
se (
degr
ee)
Rg=1M OhmsRg=1m Ohms
Objectives: To model the high frequency characteristics of an integrated RF-EMI filterTo facilitate the filter design
Objectives: To model the high frequency characteristics of an integrated RF-EMI filterTo facilitate the filter design
High Density Integrated Power Passive with Single Layer Symmetric Structure
Sponsored by: NSF
Wenduo Liu, J. D. van Wyk, W.G. Odendaal
S8.2
20032003
)(0 Hll
l
AL
r
geg
corecell
µ
µ−
+=
)(0 Fh
AC
d
cellrcell
εε=
Cr Lr1 Lr2
Cp LM
n:1
)1(
)1)(1(2
21
2
−
−−≈
pMr
rMpropen CLC
CLCLjZ
ωω
ωω)1'(
)1](1)'([
22
12
212
−
−−+≈
prr
prrrrshort CLC
CLCLLjZ
ωω
ωω
Open Circuit
7.67E+06
8.71E+05
5.19E+05
0.1
1
10
100
1000
1.00E+05 1.00E+06 1.00E+07 1.00E+08
f (Hz)
Impe
danc
e (o
hms)
-90
-70
-50
-30
-10
10
30
50
70
90
Phas
e
Impedance Phase
Short Circuit
7.76E+06
3.56E+06
2.40E+06
0.1
1
10
100
1000
1.00E+05 1.00E+06 1.00E+07 1.00E+08
f (Hz)
Impe
danc
e (o
hms)
-90
-70
-50
-30
-10
10
30
50
70
90
Phas
e
Impedance Phase
Internal Cell
Magnetic fluxCurrent flow
1st Order Approximation of a Single Cell
Single Layer Symmetric Structure
A
B
1234
5678
Interconnections are achieved on the vertical surface.
A
B
1 2 3 4
8 7 6 5
Schematic Diagram
Various interconnections provide flexibility to achieve various energy transfer functions. 1st Order Approximation of a Module
with Specific Interconnection
Schematic of One-dimensional Thermal Model
X
T (ºC)
Z
X
0 1 2 3 4 5 6 7 8 9 10 …13 14
A section in the ferrite layer
Ferrite
Insulation layer Copper Dielectric
Current Distribution and Conductor Loss Simulation Design Optimization Map
Thermal Constraint
Constructional Constraint
L-L-C-T Module
2nd Order Approximation
524kHz
871kHz
7.56MHz 2.39kHz
4.57kHz
7.94kHz
Calculation Results
Experimental Results
Z
T (ºC)
X
Modeling of Skin- and Proximity Effect Losses in Foils and in Planar Litz Windings
Sponsored by: National Science Foundation
Shen Wang, M.A. de Rooij, W.G. Odendaal, J.D. Van Wyk, D. Boroyevich
S8.4
20032003
MotivationsThe new trend in power converters is to design planar magnetic components, aiming at low profile. However, the increasing ac resistance at high operating frequency is an important constraint.
Assumption: End-effect can be ignored. The magnetic fields are always perpendicular to current flow and parallel to the flat surface of the conductor.
z
yx
hw
Infinite length
( w >> h )I
H
To account for the losses due to skin- and proximityeffect respectively, we define two factors andFΛ GΛ
dceGdcrmsF RHwRIP 222)1( Λ++Λ=
1coscoshsinsinh
2−
−+⋅=Λ
ααααα
F ααααα
coscoshsinsinh
+−⋅=Λ G
Single Foil Conductor
Planar Litz Conductor
3D FEM Modeling
A 3-dimensional finite element analysis method was used to calculate the effective ac resistance.
Litz winding can result in lower ac resistance than planar winding in a specific freq. range
Strand/s: Narrow conductor/s used to make up a planar litzconductorPlanar Litz Conductor: A planar conductor comprised of many isolated strands that are weaved together to form a single high frequency, higher current carrying capacity conductor.Strand angle: The angle the strand makes with the direction of the planar litz conductor.
Top Bottom
Copper layers
AB
AB
JJPlanar Winding and E Core
High eddy current due to proximity effect
Uniform current
each strand subjected to the magnetic field everywhere in the winding window
Litz Winding and E Core
I I
Top layer strand
Bottom layer strand
Vias
Planar conductor
Strand angle
Planar Litz conductor
High Density Integrated Power Passives with Vertical Interconnect and Stacked Structure
Sponsored by: (NSF)
Wenduo Liu, J. D. van Wyk, W.G. Odendaal
NP4.1
20032003
Magnetic flux
Current flow
The Stacked Structure
3D Interconnection
Planar Symmetric Structure Small Profile and Volume
Complete Usage of Space
Interconnections determine the energy transfer functions in the module.The challenge is to achieve interconnections for modules with many internal cells.
Interleaving Leads Stacked Structure Encapsulated ModuleCreating Interconnection
by Metal Deposition
Height
Generation 4
57.6MHz
1.93MHz
66.8MHz1.00
10.00
100.00
1000.00
10000.00
100000.00
1.00E+05 1.00E+06 1.00E+07 1.00E+08
Frequency (Hz)
Mag
nitu
de (o
hms)
-100.00
-80.00
-60.00
-40.00
-20.00
0.00
20.00
40.00
60.00
80.00
100.00
Phas
e (°
)
Magnitude Phase
Module: 2MHz Resonator
Current Distribution and Conductor Losses Simulation
One-Dimensional Thermal Calculation
Design Optimization Map
Height
Thermal Constraint
1st Order Approximation
Measurement Result