Integrated Design Needs from A Power Electronics Reliability Perspective
description
Transcript of Integrated Design Needs from A Power Electronics Reliability Perspective
Integrated Design Needs from A Power Electronics Reliability Perspective
Mike Shaw, Jun He, Vivek Mehrotra, Fred Morris, Bruce Beihoff, Rich Lukaszewski, Sriram Chandrasekeran and Qingda Yang
July 19, 2000
2
Relevant Electronic Package Examples
Power Modules, Chip-Scale Packages RF Power Packages Focal Plane Arraysdc-dc Converters
Interposer
PC Board
Semiconductor Chip (1-2 mm)
MetalBaseplate/Heatsink
CeramicIsolation
Layer
SiliconeEncapsulant Wirebonds
Semiconductor Chip(multiple @ 1 cm) Semiconductor
Chip (5-20 mm)
Ceramic Carrier
WirebondsMetalBaseplate/Heatsink
WirebondsOverfill
Solder Joints
> 1000 V < 5 V < 50 V < 20 V> 1000 A < 5 A < 20 A < 10 A~ kHz MHz GHz kHz - MHz- 40 / +150 C - 40 / +150 C - 40 / +150 C - 77 / +60 C
Today’s Focus
Semiconductor Chip (~20mm)
3
• Motor Drives
• Radar / Microwave Communications
• dc to dc Converters
• Power Supplies
• Electric Vehicle Drives
• Weapons Systems
Power Electronic Systems
Today’s Topics
4
Converts AC power (fixed frequency, voltage) to AC Power (variable frequency, current, and voltage)
Enables exact control of speed (RPM) and torque of motors
Motors become controlled electromechanical energy converters.
Rockwell Automation - Allen Bradley 1336 Force Drive
Rockwell AutomationReliance Electric AC Motor
Drive & Motor Automation System
Performance Metrics: • Power Density • Cost • Reliability
5
Generic Electronic Packaging Technology Parameters
Controlled Power Density (W/m3)High Power Requirements from DevicesHigh Packaging DensitiesWeight Requirements
Cost ($/Function)
Reliability (MTBF)
6
Lifetime Estimation of Critical Importance
1) Silicon Failure
2) Wirebond Failure
3) Solder/Attachment Failure
4) Encapsulant Failure
5) Substrate Failure
Most Failure Mechanisms are Thermally Activated or Enhanced
Primary Failure Modes in Si-IGBT Power Modules
Example in Power Electronics Communications: Friend-or-foe Interrogation by Microwave Transmission….
7
Typical Coupled Predictive/Experimental Reliability Approach
Assume:1) Tj measured through VCE or VGE
methods;2) Tc , Ta measured with a thermocouple
method;3) All Tj, Tc, Ta measurements recorded
automatically throughout the experiments4) Power is DC excitation only
Cas
e T
emp,
Tc
Time, t
Tc, maximum
Tc, minimum
Tc
300 sec
30 sec
Figure
Tj
Tj, maximum
• Detailed, 3D Numerical Analyses
• Contrasted with Experimental Power-HALT Analyses
• Are Either/Both Correct?
Copper Post
Silicon
Solder
Strain Concentration and Crack Initiation Point
Predicted Crack Propagation Path
8
E.g., Power Electronics Lifetime Governed by Temperature Swing During Operation
Number of cycles
10.000
100.000
1.000.000
10.000.000
30 40 50 60 70 80 90 100
delta Tj (°C)
heel crack
diode bonds
FZ1200R16, LotA, failed 2/6, Tj=60°C_100°C
FZ1200R16, LotA, failed 2/6, Tj=60°C_120°C
FZ1200R16, Lot B, failed 3/3, Tj=75°C_125°C
FZ1200R33, Lot C, failed 2/2, Tj=46°C_112°C
FZ1200R33, Lot D, failed 1/1, Tj=49°C_127°C
estimated trend <10% failure rate
estimated trend =50% failure rate
Ref: K. Sommer, eupec GmbH + Co, Trodheim, 1997.
KF4 Power Modules
9
Si, 2,6 10-6 K-1
Al, 23,8 10-6 K-1
crack
Ref.: E. Wolfgang, PCIM Conference 1999 (Nuremberg)
Tj
Nf
Wirebonds in Power Modules
10
Thermal Cycling of Sn - Pb Joints (Elastic/Plastic)
As Soldered 1 cycle 10 cycles 100 cycles 1000 cycles
Copper
Kovar
= 14.1 ppm
= 2.7 ppm
Cu or Kovar
SiSn-Pb Solder
Crack within Solder Layer
He et al Reliability at “Effects of Plasticity on Reliability in Multilayered Electronic Packages,” ITHERM Conference Proceedings, in press, Las Vegas,
5/00.
11
Motor / Load
Essential Motor Drive Components
ControlInterface
Power Supply(low power)
Gate DriverBoard
Control Board
Power Supply,High Power
PowerModule
Heatsink
Control Signals
Antenna
RF Power
PowerTransistors
(RF Power Analogy)
12
Power Range vs. Thermal Management Approach (Motor Drives) - 2000
HP (kW) Range of Motor Package Heat sink type
0.5 - 1 (0.4 - 0.75) Discrete None
1 - 5 (0.75 - 3.7) Discrete Forced/Natural Air cooled
5 - 40 (3.7 - 30) 6-Pack Module Forced Air cooled
40 - 150 (40 - 113) Parallel Discrete Modules Forced Air cooled
150 - 500+ (113 - 375+) Parallel Discrete Modules/Press-pack Forced Air or Water cooled
13
Experimental Measurements of Device Temperature Distributions by IR microscopy - Power Transistors in Motor Drives
The magnitude as well as gradient of Tj needs to be reduced
139 C
131 C
127 C117 C
124 C
119 C
122 C
137 C127 C
130 C
134 C
123 C117 C
122 C
133 C
130 C124 C
120 C116 C
118 C
132 C
129 C
126 C
121 C119 C
112 C
120 C
134 C137 C
132 C
127 C
131 C
120 C
117 C
124 C
138 C132 C
131 C
129 C
122 C
115 C
141 C
138 C
137 C
134 C123 C
125 C
135 C
124 C
130 C
129 C
126 C
119 C
120 C
J. He, V. Mehrotra and M.C. Shaw, “Thermal Design and Measurement of IGBT Power Modules: Transient
and Steady-State, IEEE IAS Conference Proc., 34th Annual Mtg.,
Phoenix, Az, 10/99.
14
Wide Range of Application Profiles Complicate Reliability Analyses
• Pumps• Fans• Radar Protocols………….
• What are the consequent thermal, mechanical loads over a 20-year life?
15
Decrease in System Volume Through Utilization Of Silicon Carbide (SiC) Electronics
Silicon PowerDensity = 106 W/m2
Baseplate Power Density ~ 105 W/m2
Heatsink Power Density ~ 103 W/m2
T fin = 55C
SiliconTj ~125-150 C
SiC PowerDensity = 106 W/m2
Baseplate Power Density ~ 105 W/m2
Heatsink Power Density ~ 104 W/m2
T fin >200C
Silicon CarbideTj ~300 - 350 C
Smaller, hotter
heatsink feasiblewith SiC
(Q=hAT)
16
Conclusions
• Advantages of new power electronics designs must be demonstrated at the system level.
Device Power Density (A/cm2 or W/cm2 )System Power Density (W/m3)Lifetime Assurance of Entire SystemSystem Cost Analysis Ultimately Required
•Highly localized heating around active cell regions leads to sharp temperature gradients.
•Complex interactions occur between electronics function, thermal loads and physical failure mechanisms
•Integrated Design Methodology Needs:
- Cellular design methodology yields ideal design process- Closed loop electrical, thermal, mechanical coupling essential- Well-established (calibrated), physically-based reliability models required- Statistical distributions of failures critical- Self-optimization schemes offer enormous potential