1. ABCs of Power Electronic SystemsBy
Dr. Doug Hopkins & Dr. Ron Wunderlich
DCHopkins & Associates
Denal Way, m/s 408
Vestal, New York

2. Our Professional Challenge
“The illiterate of the 21st century will not be
those who cannot read and write, but those
who cannot learn, unlearn and relearn.”
-- Alvin Toffler
Dr. Toffler, Ph.D., is one of the world's preeminent futurists. As co-author of War and Anti-War, he sketches the emerging economy of the 21st century, presenting a new theory of war and revealing how changes in today's military parallel those in business.

3. About the Course Authors?
Dr. Doug Hopkins
PhD. Virginia Tech, VA Power Electronics Center
GE-CR&D, Carrier Air Conditioning Company(UTC), University at Buffalo, and DCHopkins & Associates (President)
R&D for advanced power electronic systems
Dr. Ron Wunderlich
Ph.D. Binghamton University
IBM Power Systems, Celestica Power Systems, Transim Corp, and Innovative Design and Development (President)
Chief Engineer in design and development of power supplies for the computer and telecom industries.

4. Course Topics 1. Overview of Power Electronics Technology
1a. Introduction to the power electronics system
2. Knowing your specifications
2a. Design for safety
3. Choosing the correct topologies
3b. Knowing where disaster can strike
4. Characterizing power components
4a. A safe operating area
4b. The dual faces of MOSFETS
4c. The circuit is a component
5. Design approaches and tools
5a. Simulating reality
5b. Input filtering
6. Design approaches and tools
6a. Design case study

5. DCHopkins & Associates - Products
Products designed by our Associates, photographed by our Associates.
600W family of isolated DC/DC building blocks
Multi-output telecom power supply
Pictures courtesy of Celestica, Incorporated

6. DCHopkins & Associates - Products
30A high efficiency, high-transient isolated power supply
Isolated power supply for high-end micro-processor
Pictures courtesy of Celestica, Incorporated

7. Where to go for news (other than suppliers)?
News Sources
Where to go for news (other than suppliers)?
(PowerPulse Daily)
(see electricnet)
http// (see poweronline)
(see attached catalog)
Conferences:
(the IEEE Power Electronics Society)

8. Introduction to The System
Source
Load
Power
Processor

9. “Conversion” changes one energy form to another.
Conversion or Supply
“Conversion” changes one energy form to another.
Power
Processor
Load
Source
Electrical Source
Types of Loads
Motor drives
Linear
Rotational
Lighting
Fluorescent
HID
Halogen
Pulsed power
Ignition
Flash lamp
Pulsed propulsion
POWER CONVERSION

10. “Supply” changes only the attributes.
Conversion or Supply
“Supply” changes only the attributes.
Power
Processor
Load
Source
LOAD
Computer Applications
Desktops
Workstations
Servers
Mainframes
Circuits:
CPU
Memory
Bus Terminators
Logic
Graphics
Handheld Applications
PDA’s
Notebooks
Cell-phones
Circuits:
RF Amps
CPU/Logic
Memory
Display
Audio Amps
Telecom Applications
Routers
Tele. Switches
Circuits:
Optical Amps
CPU
Memory
Switch Cards
Logic
POWER SUPPLY

11. Uninterruptible Power Supply Systems
Electronic Circuits are
Any electronic equipment that requires clean, reliable AC utility
Computers, Telecom equipment, Home appliances
Sources are
DC such as a battery or solar cells
AC utility that is of poor quality
Power
Processor
Electronic
Circuit
Clean AC
Utility
Noisy AC

12. The System - Source Characteristics
Power
Processor
Load
Source
SOURCE

13. The System - Source Characteristics
Load
Power
Processor
THE POWER PROCESSOR
Converts an unregulated power source to a regulated output.
Like CPU’s processing information - Power Supplies process energy.
Linear Regulator
Absorbs the energy difference
Switch-mode Regulator
Chops and averages energy packets

14. Knowing Your Specifications and the User’s Requirements

15. Developing User Requirements
Responsible Design is from Cradle to Grave
Typically, User Requirements are derived through a polling process.
This brings forward the highest-priority requirements, but are limited to personal experiences.
A comprehensive approach uses a matrix of
Five Taxonomies
and
Three Characteristics

16. Grouping User Requirements
Characteristic
Unspoken Expectations
Articulated Needs
Unexpected Features
Taxonomy
Financial
Legal
Social
Environmental
Technical
MATRIXED

17. Taxonomies in User Requirements
Financial requirements: represent cost and is base metric for other matrix entries.
Legal requirements: include intellectual property as a source of revenue, strategic positioning or enticement.
Social requirements: represent the corporate culture and image, global perceptions, and ethical conduct.
Environmental requirements: represent government regulations and broader global concerns.
Technical requirements: science based metrics related to ‘energy forms’ and provide the “SPECIFICATIONS.”

18. Characteristics of User Requirements
Unspoken Expectations:
requirements for a product, process or service to be acceptable to all end users. Though labeled as unspoken, these may be new requirements that develop while a business has not been keeping up with the competition or market place, or basic requirements for entry into new markets.
Articulated Needs:
typical, open and printed “specifications. ” Discerns one user from another. There should be no question that these needs are requirements that must be met for each user.
Unexpected Features:
exciters that make the product, process or service unique and readily distinguishable from the competition. (This is what the sales force lives for.) Features are speculative requirements.

19. Example User Requirements
Unspoken Environmental Expectation: the product is not lethally hazardous to shippers
Articulated Technical Need: the products will operate from -40°C to +100°C.
Unexpected Legal Feature: the product can have exclusive patent protection.

20. Defining Specifications
Power electronic circuits condition and convert
many energy forms!
Power
Supply
Iin
Vin
Iout
Vout
Electric
Magnetic
Electromagnetic
Thermal
Mechanical
Chemical
Photonic
Technical User Requirements provide the
SPECIFICATIONS
for each Energy Form.

21. Framework leading to Specifications
Responsible Design is from Cradle to Grave.
Characteristics
Taxonomies
Technical Characteristics
Energy Forms
Conditions
Start-up
Shut-down
Normal operation
Fault operation

22. Electrical Specs Electrical Spec Input Output Controls Misc AC DC
Vout, Iout
PGood, On/Off
Efficiency
Vin, Iin
Vin, Iin

23. Specifying Vin depends on the source voltage range.
DC Input Spec
Specifying Vin depends on the source voltage range.
Typical DC sources:
Car Battery  typical 12 volts with 11 to 14 volts variation
Solar Cell  0.5 to 1 volt per cell depending on sunlight
Telecom Bus  typical 48 volts with 36 to 72 volts variation
PC Internal 5V Bus  5 volts, +/- 10%
Example: A Telecom bus has a Vin operating range of 36 to 72 volts
If the input voltage drops below 36V, typically, a PS will shut down.
If the input voltage exceeds 72V, typically, a PS will be damaged by the excessive high voltage.
A PS can be designed so it can handle short duration of high input voltage such as line transients due to lightning.
This is known as a surge rating.
For example, this PS may have a surge rating of 100V for 100usec.

24. DC Input Spec - Iin, Pout, Pin, 
Iin is the current drawn by the PS and derived by
Pout (output power) = Vout x Iout
Pin (input power) = Vin x Iin
 (efficiency of the PS) = Pout / Pin
Typically between 0.5 to 0.98
Substituting and solving for Iin
Iin = (Vout x Iout) / (Vin x )
Note: Worst case - Iin occurs at lowest value of Vin, e.g.
for telecom PS most current is at Vin=36 volts.

25. DC Input Spec
Iin will have ripple current, Irip, from the switching stage within the PS.
Specified as peak-to-peak.
Occurs at usually < 10Mhz
Typically, < 10% of max Iin
E.g., if Iin max is 10A, Irip p-p should < 1A
Irip
Iin
Time

26. Iin will have switching noise.
DC Input Spec
Iin will have switching noise.
Iin will have switching noise that occurs at >10Mhz.
The noise is due to the internal capacitive coupling parasitics
Typically, the peak-to-peak noise is less than 1% of max Iin

27. Iin will have a surge during start-up.
DC Input Spec
Iin will have a surge during start-up.
Surge current, Isurge, is due to charging of internal capacitors
Usually Isurge is less than 5 times max Iin
This can cause problems with fusing.

28. Specifying Vin depends on the source voltage range
AC Input Spec
Specifying Vin depends on the source voltage range
Typical AC sources for the home
Doorbell, heating systems  24Vrms +/- 30%
Household wiring  Typically 110Vrms with 90 to 130 range
Electric stoves  Typically 220Vrms with 180 to 260 range
Actually, 220Vrms with a center-tap is delivered to the home. 110Vrms is derived from the center-tap
Typical AC sources for business (single phase derived from three phase)
Office wiring  Typically 120Vrms with 90 to 140 range
Industrial/Computer  Typically 208Vrms with 180 to 260
Smaller businesses will use the household AC utility
Europe and some other countries are wired with either 208Vrms or 220Vrms

29. AC Input Spec Vin for typical products
Desktop PC sold in the US, 90Vrms – 140Vrms
Desktop PC sold Worldwide, 180Vrms – 260Vrms
High-end servers sold worldwide, 180Vrms – 260Vrms
Desktop PC with “universal” PS, 90Vrms – 260Vrms
Why not use a “universal” PS in all desktop PC’s ?
“Universal” PS are more expensive and difficult to design
Operating frequency for Vin is specified as
USA - 60Hz; Europe and other countries - 50Hz, range is +/-3Hz
A “universal” PS operates from 47Hz to 63Hz
This is not a cost or a design problem

30. Vin is from the wall outlet or a UPS for “Off-Line”converters
AC Input Spec - Vin-rms
Vin is from the wall outlet or a UPS
for “Off-Line”converters
AC sources are:
Single Phase
Three Phase (>5kW, not covered)
Vin is understood to be Vin-rms;
Vin-rms = Vpk / *
RMS makes calculations easier
For DC, Pin = Vin x Iin
For AC, Pin = Vin-rms x Iin-rms
For single frequency sine wave
Vpk
Power
Supply
Iin
Vin
Iout
Vout

31. AC Input Spec - sags, surges, and transients
AC voltage will have transients and surges
2000V spikes are not uncommon
Florida is the worst US state
Due to lightning, industrial equipment and solar flares
The “front-end” PS circuitry must be able to shunt this energy The PS cannot have direct connection between input and output. Hence, isolation is required. This is a safety requirement.
AC supply has brown outs, sags, or drop outs in power
This occurs when
The utility transformer in a sub-station goes bad
The grid becomes overloaded from air-conditioners, etc.
Solar flares induce too much voltage and “pop” the breakers
These occur quite often
More than 99% of the drop outs are less than 20ms in length

32. AC Input Spec - Hold-up Time
When AC momentarily is interrupted
For non-mission-critical devices
e.g., televisions, radios, VCRs
PS can shut down temporarily
For mission-critical devices
e.g., high-end servers
PS shall maintain operation for a loss of AC up to 20ms
After 20ms it can shut down This is known as hold-up time
This is accomplished by a large energy storage device such as a capacitor in the input (PFC).
Typical specifications for hold-up is 20ms.

33. AC Input Spec - Power Factor
Ideally, Iin should follow Vin emulating a resistor
A bridge rectifier with a large capacitance is usually at the PS input.
Iin, with respect to Vin, will be distorted.
Iin-rms is now significantly higher than for a resistor input to have the same usable energy flow.
The distortion adds frequency harmonics.

34. AC Input Spec - Power Factor (con’d)
Apparent power is Pa = Vin-rms x Iin-rms
Real power is the average Pr = Vin x Iin
Power Factor, PF PF = Real Power / Apparent Power
The lower the PF, the higher the Iin-rms for the given power
The problems with lower PF are
Wire sizes must be increased to handle the higher Iin-rms current
Power Loss increases by the square of current!
This is extra power for which the feeders and fuses must be size
Iin is rich in harmonics which adds noise and circulating currents in 3-phase systems

35. AC Input Spec - Inrush Current
Like DC, Iin has inrush issues with AC applications.
Usually, peak Iin is specified to be <5X the steady- state Iin-rms.
Another factor to consider is fusing and circuit breakers.
If the inrush current is too high or can occur throughout the day, fuses and circuit breakers can be weakened, damaged, or open up.

36. Total Harmonic Distortion, THD -
AC Input Spec - THD
Total Harmonic Distortion, THD -
the same for your stereo as for the power supply
Any waveform can be broken down into a sum of sine waves with different amplitudes
If there is any distortion, then
I = x [I1sin(2ft)+I2sin(4ft)+I2sin(6ft)+…]
I1 is the rms of the “fundamental” current waveform
I2 is the second-order harmonic, I3 is third, etc.
The Total Harmonic Distortion is then THD = {sqrt[ (I2)^2 + (I3)^2 + (I4)^2 + …] / (I1)} x 100%
A good value for THD < 5%

37. Conducted versus Radiated Noise
AC Input Spec - Noise
Conducted versus Radiated Noise
Conducted noise current is measured on the line cord.
The frequency is less than 30Mhz
A “LISN” box is connected to the cord to filter out the 50/60hz
A frequency-spectrum analyzer then displays the noise spectrum
Federal specifications must be met
If the frequency is > 30Mhz, this is known as radiated
This is measured with an antennae usually 10 meters away
At these frequencies, line cords and cables become very effective antennae
Federal specifications that must be met

38. Line, Load & Temperature
Vout Specs
Line, Load & Temperature
Static Regulation
Load Step
Dynamic Regulation
VOUT
Noise
Ripple & HF Noise
Drift
Long Term Stability

39. Static Regulation Line Regulation Load Regulation
% change in output voltage versus input voltage at a given load
Typically 1-2%
Load Regulation
% change in output voltage versus load at a given input voltage
Typically 0.1-3%
Vout Temperature Effect
% change in output voltage versus temperature for given input and load
Typically 0.2-1%

40. Static Regulation Cross-Regulation (multi-output converters)
Change in output voltage of channel 2 for a change in load on channel 1 at a given input voltage
Typically %

41. Dynamic Regulation
Change in output voltage is due to the dynamic behavior of the power supply
The output voltage initially changes because of the I step x ESR of the output cap (5A x 0.3ohms)
The second part is due to the loop response of the converter
The change in output voltage is measured from the nominal output voltage
5% for this example

42. Another effect shows up as L x (di/dt)
Dynamic Regulation
Another effect shows up as L x (di/dt)
This is due to inductance of
Output capacitor
Connector
Bus distribution
This is not always included in the spec.
Could typically be < 5%

43. Ripple and Noise Ripple High Frequency Noise Typically 0.2-3%
Triangular-shaped current at the switch frequency
Due to inductor current x ESR of output cap
High Frequency Noise
Noise > 10 x fSW
Either random or the excitation of high-frequency parasitics.
Typically 0.2-3%
Vout
Vrip
Time

44. Over time, a reference voltage can change.
Drift
Over time, a reference voltage can change.
Drift is due to
Aging
Soldering
Package compression
Typically < 0.2%

45. Question -
How can you improve the transient response of the converter without…
changing the components or
changing the switching frequency?

46. Answer Use adaptive control (positioning)
At no load, start at +X1% above nominal Vout
At full load, change Vout to be X2% below nominal Vout
In the previous example, dynamic regulation was 5%
This can be changed to 3% dynamic regulation by modifying VREF for the control loop scheme
Common in IC’s

47. Over current trip point
Iout Specs
Below is a typical Iout load behaviour
Iout
Time
di/dt rate
Minimum current
Maximum current
I step
Over current trip point

48. Question What happens to current in COUT if IOUT’s frequency
>> than the bandwidth of the converter ?

49. Answer
Normally, the ripple current in Cout is the same as the inductor current
If the load is switching faster than the bandwidth of the converter
the ripple current in Cout is due to Iout (load shift).
the converter will not respond to the load changes so the current it delivers will be the average of Iout
The ripple current in Cout due to Iout may be significantly higher than that due to the inductor current
This condition occurs with most modern micro-processors when executing certain software
Local decoupling caps help solve this problem

50. Design For Safety Standards, Certificates & Regulations

51. Standards, Certificates & Regulations
A power supply has many standards and regulations to meet
Only the major ones will be covered
Safety
Corporate
Standards
Pisces, Fielday IBM) are FEM based simulators which simulate at the holes/electrons level. Very limited external circuits can be added.
Ansoft is a FEM program for magnetics
SMS – Switchers Made Simple is a canned program that designs the components around a power supply chip.
EMC
Features
Robustness

52. Safety - http://www.i-spec.com
Many countries have their own safety agencies
Europe has Conformity European Mark
Canada has Canadian Safety Agency
US has Underwriters Laboratories
To sell a product and/or to be protected from liability, the product must be approved by a safety agency
Most countries follow standard IEC-60950
The Product Designer's “on-line guide”
to compliance with the
International Safety Standard for
Information Technology Equipment,
IEC 60950
i-Spec also covers national standards based on IEC 60950, including EN 60950, UL 1950/CSA C , AS/NZS 3260.

53. Safety
For example:
A product that will operate from 240VAC requires that the primary-secondary spacing be greater than 8mm
The FR4 Card must meet UL 94V-0 standard for flammability
There is even safety consideration for battery-operated equipment when the battery fails short
To obtain safety approval
The product must be taken to an agency for testing
Performed by a person within the company who has been certified by the safety agency
Approval by one safety agency will be accepted by others
To obtain CE and CSA approval for a power supply that has been approved by UL, only the test report need be shown
Many labs will do all the required testing and the paper work for a fee

54. Electro-Magnetic Compatibility (EMC)
Electro-Magnetic Compatibility - (Love / Hate)
EM Emission - EM Susceptibility
EM emission limits are required by law for products
For the US, FCC part 15
For Europe, CSIPR
Both are similar
Class A
typically for industrial equipment
Class B
typically for commercial / home equipment
Class B is 10dB more stringent

55. <30 MHz FREQUENCIES >30 MHz
Conducted or Emitted
<30 MHz FREQUENCIES >30 MHz
Noise is measured through a device called a LISN on the AC cord
LISN – Line Impedance Stabilizer Network is a set of filters that filters signals above 60Hz to a spectrum analyzer
The noise is measured with an antennae 10 meters away
All testing is done in a shielded chamber
Certifications must come from approved sites

56. Why 30 MHz? Question Answer
Why are measurements done through the line cord at <30Mhz and with an antennae at 10m for >30Mhz?
Answer
The speed of light, c, is 300 x 106m/s
At f = 30Mhz (30 x 106/s), the wavelength (=c/f) is 10m
At frequencies <30Mhz, the emitted noise is carried out in the wiring which is not an effective antennae
At frequencies >30Mhz, emitted noise is radiated from the line cord and circuit wiring since these now become effective antennas

57. Lower Susceptibility is increased Robustness
These standards help the user design a product that will last a reasonable time in every day environments.
There are no requirements to meet any of these standards. However, they contain a wealth of experience.

58. Susceptibility - Circuit Card Effects
The Institute for Interconnecting and Packaging Electronic Circuits developed standards for the packaging of products
For example
For connectors, FR4 cards and sheet metal
Spacing between primary to secondary wring on a FR4 card is well defined in safety guidelines
IPC defines the spacing between primary-to-primary and secondary-to-secondary wiring
If the primary-to-primary spacing is reduced below the IPC guidelines, arcing can occur
There is no facility to test against the IPC spec.
This is left up to the designer

59. Susceptibility - AC Utility Effects
The AC utility line has surges and transients
Surges are caused by abrupt load changes and “bank” switching
Transients are caused by lightning strikes and line faults.
IEC and IEC801-5 provide test procedures that ensure your product survives most cases
These tests can be performed by the designer with the right equipment or by outside labs

60. Susceptibility - Electro-static Discharge
Products must also be protected or withstand
electro-static discharges (ESD)
These occur when products are physically handled
IEC provide test procedures to ensure your product survives most cases
These tests can be performed by the designer with the right equipment or by outside labs

61. Susceptibility - ElectroMagnetic
EM Susceptibility tests how sensitive a product is
to EM emissions
The product should behave as expected with EM fields up to a certain strength
The standards for this are
IEC for radiated susceptibly
IEC for conducted susceptibly
Testing for this is usually performed in EM shield chambers, same place as for FCC approval

62. Corporate standards are policed within the company
Corporate Standards should be all encompassing
They can toughen existing requirements, such as IPC guidelines
They can be guidelines on how a product should be designed
Topology A is chosen over topology B
SMT vs. PTH
They can be guidelines on how a product looks
Placement of labels
Color of products
They can be guidelines on de-rating of components
Some product specs will cite MIL-217F or Bellcore

63. Corporate Standards - Features: ENERGY STAR
Features are specifications that make a product more valuable
Some features later become requirements
ENERGY STAR
A “feature” developed by the US-EPA
Products must reduce their power consumption significantly for a period of time or when not in use, known as sleep mode
These tests can be performed by the designer with the right equipment or by outside labs
The guideline for computers can be found at

64. Corporate Standards - Features: PFC & THD
Low Power Factor and Low THD apply to
AC off-line supplies
In US, still a feature
In Europe, this has become a requirement
This is an example of a feature that has become a requirement
The standard for this is IEC-555
This test can be performed by the designer with the right equipment or by outside labs

65. Choosing the Correct Topology

66. Linear Regulators Switch is used as programmable resistor
Fast dynamic response
Minimal filtering
Poor efficiency
Relatively large with heat sink
Load
(VOUT)
dc source
(VIN)
PL = (VIN - VOUT) * IOUT

67. Switchmode Regulators
Switch is used as a chopper
Dynamic response depends on switching frequency
Requires filtering
High efficiency
High density
PL : steady state + switching
dc source
(VIN)
chopper
(fT)
filter
(fF)
load
(VOUT)

68. Demystifying the Circuits - Duality
Using Simple principles of Duality
Duality
Current is voltage; Voltage is current
L is C; C is L
R is R is R
Series is parallel; Parallel is series
Transistor is diode; Diode is Transistor
Open is closed; Closed is open

69. Demystifying the Circuits – Non-isolated
load
Boost dc
source
Buck
load dc
source
DUALITY
CASCADE dc
source
load
Buck/Boost
Cuk
not covered
DUALITY

70. Demystifying the Circuits – Conversion Ratios
Buck Regulator
(Step-down converter)
Boost Regulator
(Step-up converter)
Buck/Boost
(Up/down converter)
VOUT
VIN
= D
VOUT
VIN 1
1-D =
D: duty cycle of switch
TON
TPERIOD
VOUT
VIN -D
1-D =
TON
TPERIOD
D =

71. Demystifying the Circuits – Transformer Isolateddc
source
• •
load
Flyback
Buck/Boost
Isolated dc
source
• •
load
Forward
Buck
Isolated

72. Demystifying the Circuits – Bridgedc
source
LOAD
Half Bridge
Buck derived topologies dc
source
LOAD
Full Bridge

73. Demystifying the Circuits – Resonant Bridgedc
source
LOAD
Series Resonant
LOAD dc
source
Parallel Loaded
Series Resonant

74. Partially Resonant Topologies
Discontinuous-Resonant topologies known as
Zero-Voltage Switched circuits
Zero-Current Switched circuits
Resonant Transition topologies
Zero-Voltage PWM topologies
Characteristics:
Uses internal parasitics for nearly lossless switching
Fairly involved design approach
Next level of sophistication
Beyond this course

75. Knowing Where Disaster Can Strike
Do you have the “knack?”

76. Disaster is Only Nanoseconds Away
Inductive Switching 101 or Understanding the Waveforms
You can be a rich power electronics designer too! It is all in battling Mother Nature.
She likes continuity and easy flow, e.g. Sinewaves, exponentials and Gaussians.
We give her
v=L*di/dt and i=C*dv/dt
Buck Load
Buck-Boost Load

77. Inductively Induced Voltage
Power Mosfets can switch 10A in 5ns
Internal lead inductance could be 5 nH each terminal
v=L*di/dt, or lead inductance creates a 20 V spike.
Lower the Mosfet rating, the faster the device
All parameters work against you
Thank you, Mother Nature

78. Inductive Switching - Ideal Circuit

79. Inductive Switching - Ideal Circuit, Real Switch

80. Inductive Switching - Diode Inductance

81. Inductive Switching - Diode Capacitance

82. Inductive Switching - Circuit Inductance

83. Inductive Switching - Slower Switch Transition
Vds is worse if Fet is slowed down. Suspect something with model. Everything else ok.

84. Inductive Switching - Snubbing transients

85. Characterizing Power Components

86. Semiconductors Zeners Diodes Varistors (MOVs) Transistors
typical operation
transient suppression
Diodes
Rectifiers
Fast recovery
Ultra-fast recovery
Reverse Recovery Charge
Forward turn-on delay
Package parasitics
Varistors (MOVs)
clamps (not crowbars)
should thermally fuse
Transistors
Power Mosfets
vertical structure
IGBTs
“TopSwitch”
modules
Bipolars
Triggered semiconductors
SCR’s
crowbar applications
Phase-controlled bridges
high power
Unijunctions

87. Do's and Don'ts of Using MOSFETs
Be Mindful of
Reverse blocking characteristics of the device
A vertically conducting device
Handling and testing power HEXFETs
Unexpected gate-to-source voltage spikes
Drain or collector voltage spikes induced by switching
Pay attention to circuit layout
Do not exceed the peak current rating
Stay within the thermal limits of the device
Be careful when using the integral body-drain diode
Be on your guard when comparing current ratings

88. MOSFET Gate Drive Characteristics
Gate drive -vs- base drive
Driving HEXFETs from linear circuits
TTL gate drive for a standard HEXFET?
The universal buffer
The most important factor in gate drive: The impedance of the gate drive circuit
Gate drive approaches
Simple and inexpensive isolated gate-drive supplies
Optocouplers, pulse transformers, choppers, photovoltaic generators
Bootstrap gate-drive supply
Maximum gate voltage and the use of Zeners
Driving in the MHz? Use resonant gate drivers
Power dissipation of the gate drive circuit is seldom a problem

89. Paralleling MOSFETs General Guidelines Steady State Sharing
The inherent positive temperature coefficient provides dc (steady state) sharing while in the on-state!
Dynamic Sharing at Turn-On
Requires close matching of gate-threshold voltages
Avoid gate resonance by using ferrite gate beads (few nH)
Must have matched inductive paths
Clamping MOSFETS are beneficial
Dynamic Sharing at Turn-Off
Requires some matching of gate-threshold voltages
Requires close matching of “Miller Capacitance” path

90. Diode Reverse Recovery
Recovery produces sharp current transients and EMI ta tb tn
IRR IF
Soft Recovery ta tb tn
IRR IF
Abrupt Recovery
Buck Load

91. Safe Operating Area - the holy grail
SOA combines transient and thermal limits ID
Transient thermal limit
Steady state (DC) limit Fusing current Thermal path limit
Switching speed dc -to- pulsed
Breakdown limit
MAXIMUM POWER AREA
VDS

92. Capacitors - Circuit Equivalent
Ceramic
high frequency
sensitive to thermal transient
Tantalum
polarized, also organic leads
high energy density
Electrolytic, also oscon
polarized
highest energy density C
ESL
ESR
leakage
Equivalent Circuit
R, L, C
limited internal temperature from “RMS heating,” i.e. current ripple
Staged for reducing ESR

93. Magnetics - Circuit Equivalent
Transformers
Leakage is loss of coupling from primary to secondary
Skin effect is determined by copper and core magnetic fields
litz wire and foil help in high-frequency designs
Thermal hot-spots of most concern:
from high flux densities in core
from eddy current losses in core and wires
potting can trap heat Rp Xp Xs Xl Cs
Approx.: Xl = 10 *Xp

94. The Dual Faces of Power MOSFETS
Getting the heat out with Synchronous Rectification

95. Synchronous Rectification - Output Drop
As voltage requirements from micro-processor’s and logic drop, efficiency becomes a problem
For output voltages < 3.3V, the best case efficiency can be approximated by
Vd is the voltage drop due to the output diodes
load
Boost dc
source

96. Synchronous Rectification - Efficiency
The best Schottky diode voltage is 0.25V and high current Schottky diodes are as high as 1V
For example, converter with 0.5V for Vd, can have an efficiency of 67% best case
For every 100W out, 50W is wasted as heat!
Other advantages for increasing efficiency
Greater utilization of AC feeder capacity
Reduced electrical bill for the customer
Increased reliability with less thermal issues
More “green” friendly

97. Synchronous Rectifiers
Solution is to use Synchronous Rectifiers
Replace or parallel the output diode with a low Rds-on Fet
For this to work, the Fet must turn on when the current is in the direction of the diode
I x Rds-on < Vd
Efficiency of 90% can be achieved with power supply! I

98. Synchronous Rectifiers - Notables
What to watch out for
If the current reverses and the Fet is on, you have a short-circuit condition across, usually, a transformer
Timing is critical
The MOSFET body diode may come on
Placing a Schottky diode in parallel with the body diode will not, in all cases, reduce power loss – Ramp down effect
Very low Rds-on Fets require a large amount of gate drive energy
For example, a converter, 2% efficiency loss to gate drives is not uncommon
This ramp down effect is because the inductances of the Fet and diode are in separate packages. Even though the inductances is low so is the voltage across them. So it takes time for the current to ramp down from Schottky to Fet. Times of 700ns are typical. At 1Mhz, sync. Rect. Is useless with a parallel Schottky. Just use the body diode.

99. Synchronous Rectifiers - Parallel Modules
Some PS can source and sink current
At light loads, this could happen with parallel modules
Circulating current can be as high as several hundred amps
Solution is to shut sync rect off at light loads

100. The Circuit is a Component
Insights into Power Packaging

101. Electrical v. Physical Circuits
Power electronic circuits [PHYSICAL CIRCUITS] condition and convert many energy forms!
Electric
Magnetic
Electromagnetic
Thermal
Mechanical
Chemical
Photonic +
We do not do ONLY electrical designs

102. Typical Electrical Structure
Lead Inductance
Finite resistance
Skin Effect
Inter-Conductor Capacitance
Coupled Capacitance

103. Conductor Resistance -Sheet Resistancel
R = r l / (t × w)
let l / w = 1 = “one square”
Rsheet = r / t [ W / sq. ]
A corner is squares l t w

104. “1 oz. copper” is weight for one square foot
Conductor Thickness
“1 oz. copper” is weight for one square foot
Thickness and Resistance from Common Conductors

105. DC Power Supply Example - Output Conductor Resistance
Terminal
Calculate the voltage drop and power loss of the output leads for a 5V, 100A supply. Consider 1oz., 2oz. and 3oz. copper conductors. ~1
~.22
Cap
No. of squares for both sides is:
Squares = =
For 2oz. copper
Rtotal =
Vleads =
Pleads = ?
Terminal

106. DC Power Supply Example - Output Conductor Resistance
Terminal
Calculate the voltage drop and power loss of the output leads for a 5V, 100A supply. Consider 1oz., 2oz. and 3oz. copper conductors. ~1
0.56
0.56
Cap
~.22
No. of squares for both sides is:
Squares = 2( )
= 4.68 sq.
For 2oz. copper
Rtotal =(0.252 mW/sq) (4.68 sq)
=1.18 mW
Vleads = (1.18 mW) (100 A) = 118mV
or 2.8%
Pleads = (118 mW) (100 A) 2 = 12 W
0.56 ~1
0.56
Terminal

107. Output Conductor Resistance

108. Coupled Capacitance Substrate Coupling Example:
Conductor #1: 100mils x 1 inch
Conductor #2: 400mils x 1 inch
Substrate: ceramic loaded polymer, 3 mils thick, er = 6.4
Find Capacitance:
C =

109. Coupled Capacitance Substrate Coupling Example:
Conductor #1: 100mils x 1 inch
Conductor #2: 400mils x 1 inch
Substrate: ceramic loaded polymer, 3 mils thick, er = 6.4
Find Capacitance:
C1 = 47.9 pF, C2 = 192 pF
C = C1 series with C2 = 38.3 pF

110. Ground Coupling Example: Switching current
coupled into header from FET drain.
FET: 400mils2, tf = 20 ns
(+20 mil conductor periphery)
(+100 mils2 drain bond pad)
(+200 mils x 400 mils drain lead)
Substrate: Al2O3
25 mils thick, er = 9.4
Voltage source: 425 Vdc
continued Vd

111. Ground Coupling (continued)
Bond Pad
100mils2
Find Capacitance:
Find switching current:
i = C (dV/dt )
i = ?
20mils
400mils2
Drain Lead
100x200mils

112. Ground Coupling (continued)
Bond Pad
100mils2
20mils
Find Capacitance:
A = in2 = 183 mm2
d = 25 mils = mm
Then: C = 24 pF (d-s Cap)
Find switching current:
i = C (dV/dt )
= 24 pF (425/20ns)
i = 0.51 A
400mils2
Drain Lead
100x200mils
For ceramic loaded polymer
C = 136 pF and i = 2.9 A

113. Inductive Effects Non-Transmission Line Mode
Self Inductance of Conductors
Minimum is non-coupled in free space
Xe ( W / sq ) = Ld Gl
Ld = Rd = ( 2 s d )-1
Gl = (sinh n - sin n ) / ( cosh n - cos n )
n = t / d
t is the thickness (m)
d = ( p f m s )-1/ 2, skin depth
s is conductivity in (s/m)
f is frequency
m is permeability ( m0 = 4 p x H/m)

114. Inductive Effects Example - High Frequency Lead Inductance
Calculate the per-square self-inductance of a 1oz, 2oz and 3oz
copper lead needing to conduct a 1MHz signal.
For 1oz copper:
d = 66.0 mm, n = t / d = 0.516
Ld = mW / sq, Gl = 0.172
Xe = 22.4 mW / sq, or Le = 3.57 pF / sq
Note: max self-inductance =
( 4p fs / m ) -1 / 2
Self-Inductance, 1MHz

115. Inductive Loops Non-ferrous headers Aluminum Copper Si C Al Si C
Ferrous headers / substrates
Invar ( 64% iron, 36% nickel )
Kovar ( 54% iron, 29% nickel, 16% cobalt)
Ferrite (substrates)
Porcelainized steel (substrate)

116. Temperature as the Culprit
Junction Life Statistics
150, 0.2
Failure Rate ( /105 runs)
100, 0.05
50, 0.005
Junction Temp (C)

117. Thermal Issues Factors affecting DT Convection/conduction in medium
Chip size
Chip attach
Heat spreader
Conductor type and thickness
Substrate type and thickness
Substrate attach
Heatsink

118. Rule of Areas (Hoppy’s Rule)
Power Supply
h = P0 / Pi , Pl = P0 ( 1 - h ) / h
Load
P0, zero % efficient electrically
Load
PwrSupply
heat
heat
For first-level type packaging (e.g.. chip and wire) the thermal area densities are equal: P0 Pi
h, Pl PL
Pl / Aps = PL / AL

119. Rule of Areas (continued)
For thermal enhancements (e.g. thermal vias) a Thermal Density ratio, TDr , is defined
TDr = ke, l / ke, ps
where ke is an equivalent thermal conductivity for that area.
Aps/AL
TDr=1
Then TDr ( Pl / Aps ) = PL / AL
Aps / AL = TDr ( h ) h

120. Thermal Resistance Model - 1D
k A
Chip
Rq =
Solder
Spreader
1 l
s A
R =
Conductor
Substrate i q
Attach
R [W] = v[V] / i [A]
R [oC/W] = T [oC] / q [W]
Baseplate Dv DT
Attach R Rq
Heatsink

121. Comparative Thermal Resistances (°C/kW cm2)
Thermal Typical Thickness DT(°C) Material Conductivity Rq/cm (mils) IGBT
(W/m °C) (°C/kW cm2) @0.2kW/cm2
Silicon (Si)
Solder (95Pb-5Sn)
Molybdenum (Mo)
Alumina (Al2O3)
Aluminum
Nitride (AlN)
Beryllia (BeO)
Aluminum Silicon
Carbide (AlSiC)
Aluminum (Al)
Copper (Cu)
Polymer Ceramic
Glass Epoxy (FR-4)
Thermal Grease 84 63
146
20-26
170
240
393
3.2
1.1 42 16 17
244 37 26 -
2.6
476
3000
924 14 4 10 25 -
4 (3oz) 6 20
8.4 8
3.4 49 -
5.2
0.52 95
600
185

122. Example Structure width (mm) depth (mm) thick (mm) Material k (W/m °C)Si
width
(mm)
depth
(mm)
thick
(mm) k
(W/m °C)
Material
DBC Si
Solder Cu
Al2O3 Al
AlSiC
10.2
12.7
15.2
10.2
12.7
15.2
360
102
204
635 51
1.27* 84 63
393 26
240
170
Al2O3 Al
AlSiC
* in mm

123. One-Dimensional Model -Using Bulk Dimensions-
Rq = (t / A) / k Si
t = thickness, A= width x depth
DBC
Rq(°C/W)
layer
t (mm)
w(mm)
w’(mm)
D(mm)
D’(mm)
Ae(mm2) Si
Solder Cu
Al2O3 Al
AlSiC
360
102
203
635 51
1.27*
10.2
10.4
11.0
12.3
12.7
15.2
103
107
121
122
152
0.041
0.016
0.003
0.105
0.001
0.002
Al2O3
10.2
12.7
15.2 Al
AlSiC
*in mm
Rq, total = °C/W

124. - ” 45° ” Spreading Angle - Hence: A = (Wu + t)(Du + t)
Assumption : For an isotropic material, heat flows laterally at the same rate it flows vertically.
Hence: A = (Wu + t)(Du + t)
Rq = (t / A) / k
width
(mm)
depth
(mm)
thick
(mm) Si k
(W/m °C)
Material
DBC Si
Solder Cu
Al2O3 Al
AlSiC
10.2
12.7
15.2
10.2
12.7
15.2
360
102
204
635 51
1.27* 84 63
393 26
240
170
Al2O3 Al
* in mm
Rq = 0.232° C/W
AlSiC

125. - Adjustable Spreading Angle -
Thermal interaction of layers changes the thermal spreading angle, a
an = tan-1(kn / kn+1)
A’n = [W’n + 2tn tan (an)]
Rq,n = (tn / A’n) Kn
DW > tn tan (an)
Asi
Acu
Example Spreading Angles
Composite
Material
Spreading
Angle in * a2 a2
Acer
DBC* on Al2O3
DBC* on BeO
Cu* on Fr-4
AlSiC* on Al
85°
57°
89.8°
30°

126. Adjusted Spreading for Structure
Layer
Angle (°)
W’(mm)
D’(mm)
A’(mm2)
Rq (°C/W) Si Si
Solder Cu
Al2O3 Al
AlSiC 86
6.2 55 30
10.2
16.0
(12.7)
12.8
13.0
14.5
10.2
16.0
(12.7)
12.8
13.0
14.5
104
---
107
165
169
209
0.041
0.016
---
0.003
0.148
0.001
0.036
DBC
Al2O3 Al
Rq = 0.290° C/W
AlSiC

127. Presentation Goes Off-Line
We break to another topic.
See supplemental material.
Review of packaging paraphernalia

128. Design Approaches and Tools

129. The Good the Bad and the Ugly
In the Best of Designs...
Compliments of Celestica, Inc.
The Good the Bad and the Ugly

130. Presentation Goes Off-Line
We break to another topic.
See supplemental material.
Review of physical hardware
Compliments of Celestica, Inc.

131. Simulating Reality Our best guess at Mother Nature

132. Overview on Design Tools
Webench, SMS, SwitcherCAD
Design Programs for Power Supplies
Cost $$$
Pspice, AWB, SIMetrix, Simplis
SPICE based or State-Space Simulator
Ease of Use
Pisces, Fielday, Ansoft
FEM Based Simulators
Pisces, Fielday IBM) are FEM based simulators which simulate at the holes/electrons level. Very limited external circuits can be added.
Ansoft is a FEM program for magnetics
SMS – Switchers Made Simple is a canned program that designs the components around a power supply chip.
Device
Physics
Component
Modeling
Circuit
Simulation
Specific
Circuits
Physical Level

133. FEM Design Tools Expensive and require significant learning
Pisces and Fielday, IBM tools, simulate semiconductor devices at the electron level
Ansoft simulator models electro-magnetic devices with FEM
On the right is a gapped ferrite core showing the flux lines
See additional Ansoft foils

134. 1=Cadence, 2=Simetrix inc
SPICE Design Tools
Easy to use but requires circuit design experience and $$$
Pspice1, AWB1 and SIMetrix2 use time differentials for solving circuits.
Good for modeling electrical circuits
Transistor and op-amps are modeled as equivalent circuits
On the right is a simple circuit and waveform from Pspice
1=Cadence, 2=Simetrix inc

135. SPICE Design Tools - Limitations
When simulating switchmode supplies, SPICE has limitation
Need to simulate long times to look at control loop behavior in milliseconds, yet ...
SPICE will calculate in nanoseconds because of the time domain calculations
One solution is to use “Average Models,” where the switching waveform is averaged out.
Models require mathematical definitions and a good understanding of the subject

136. State-Space Design Tools
Another solution is to use a state-space simulator such as Simplis1
Simplis calculates based on the topology and only at the switching points
Simulation speed for switchmode power supplies is improved up to 100X
You can enter the circuit as is
Purple is the output voltage with a transient response
Yellow is the primary switch current
Red is the error voltage
1=Transim Corp

137. Webench Design Tool - www.webench.com
Webench is a design tool from National Semi. in conjunction with Transim Corp.
Webench helps you pick the IC, simulate and build.
Within Webench is Websim which uses Simplis as the simulation engine
Webench is a Web based tool
Very easy to use and free but not flexible
Purple is the output voltage with a transient response
Yellow is the primary switch current
Red is the error voltage

138. SMS Design Tool
Switcher Made Simple (SMS) is a PC program from National Semi., in conjunction with Transim Corp
The program
helps the user select the appropriate controller IC
designs and selects components
easy to use but not flexible
free
Great for the novice that needs a quick power supply design
Purple is the output voltage with a transient response
Yellow is the primary switch current
Red is the error voltage

139. Input Filtering (Not selective hearing)

140. Input Filter Why is an input filter needed ?
Reduce ripple current from the PS
Prevent filter oscillation
Reduce the di/dt of the load reflected back to the input

141. Why is an input filter needed ?
Reduce ripple current from the PS
Prevent filter oscillation
Reduce the di/dt of the load reflected back to the input
Irip
Iin
Time
Ideally, Iin should be a clean DC current
There will be the ripple current, Irip, from the PS switching stage
To reduce the input ripple, use an L-C network on the front-end of the power supply
The resonant frequency << Fsw

142. Input Filter
If the resonance of L1,C1 is around Fsw of the PS, a large amount of current can oscillate between L1 and C1
The amount of current depends on the Q of L1 and C1
Very common if L1 is just the board trace between the PS and the Vin source
This oscillation can depend on the length of board trace!
Adding an inductor will lower the resonance and make this parameter controllable
If the resonance of L1 and C1 still a problem, dampen it with an R-C across L1 or use lossy core material for L1

143. Input Filter Another problem arises if L1 and C1 have a large Q
Even if the resonance is less than Fsw, this peaking effect can cause problems with the control loop
This resonant frequency can show up on the output of the power supply
Again, solutions are either an R-C across L1 or use a lossy core material for L1

144. Input Filter
Another characteristic is reduction of input di/dt during load transients
Problems caused in the Vin bus
Ringing on the board traces
Vin not able to respond to load change
Solution: absorb the load energy
How?
Large cap on Vin bus – PTH parts on SMT board? No
Adding more output caps to absorb the energy? Expensive - No
Add second stage filter? Inexpensive SMT parts - Yes
First filter L1,C1 filters the high-frequency switching components. Second filter L2,C2 is a low-pass filter to smooth out the reflected load transient

145. Input Filter
Shown is a two-stage filter with input current and load current
Beware of inter-stage oscillations

146. A Different Approach to a DESIGN
Optimally Selecting Packaging Technologies
and Circuit Partitions Based on Cost and Performance
APEC’ 2000 Conference
John B. Jacobsen and Douglas C. Hopkins

147. Full-Cost Model Other OH Overhead Production Cost Depreciation Wages
Packaging Materials
& Production Costs
(controllable)
Wages
Packaging materials
Comp. packaging
Standard unit cost
Materials
Cost
Materials
Cost
Minimum packaged
components

148. Centers of Cost Materials cost* Production cost* Partitioning cost
*Full Cost
Partitioning cost
Product business cost (return on investment for development of one product)
Company business cost (return on investment for cross products)

149. Centers of Cost (con’d)
Materials cost represent direct costs of packaging materials.
Production cost includes factors for wages and product volume, but are independent of material costs.
Partitioning cost is incurred for each technology used.
Full cost combines material costs and production costs.
Product business cost, i.e. return on investment for development of one product, is an investment in future payback. The total cash flow from development until end of production determines the business costs for a product.
continued

150. Centers of Cost (con’d)
Company business cost, i.e. return on investment for cross-product usage, reflects the cost of sub-optimization within one single product.
Reusing the same packaging technologies, designs (diagrams) and even physical circuits (building blocks) across different products should be measured at the company level. The value of building blocks becomes obvious through savings in repetitive development costs and maintenance of function

151. Production Cost Dependency by Volumeyr
700%
600%
500%
Other overhead costs
400%
Production Cost
300%
200%
Depreciation
100%
Wages 0%
10k
32k
100k
320k
1000k
Products/Year

152. Cost Variation Within a Technology
Packaging &
Packaging Performance:
Production Costs
(electrical, thermal, mechanical 1
0.8
TF module
Changing technology
& leadframe
110 SMDs
to change density
14 leadet
0.6
Relative Cost
FR4
0.4
Functional
0.2
integration within
70 SMDs
technology
7 leadet 5 10 15 20 25 30
Surface Density

153. Relative Cost of Technologies
Packaging &
Packaging Performance:
Production Costs
electrical, thermal, mechanical 1
DBC
TF &
Plated Cu
0.8
Performance
IMS
0.6
Circuit cost by
Relative Cost
FR4
change in technology
0.4
Hot
0.2
Embossing 5 10 15 20 25 30
Surface Density

154. Relative Packaging & Production Cost
Relative to 1 in2 of FR4 14 12 a a
leaded auto/10 comp b 10 b
Substr/in2 8
Relative Cost c c 6
Power chip& wire/10 comp 4 d 2
SMD/10 comp d d
Integrated res/10 comp
TTF
TF multilayer
Hot Embossing
FR4 Cu( 2x35um)
FR4 Cu( 4 layer )
IMS (1 layer on Al)
DBC( 0,63 Al2O3 )
Z-strate Cu( 2 layer)
Substrate Technology

155. Relative Production Cost per Technology
120%
100%
80%
Cost/component
60%
40%
20% 0%
Leaded-manual
Leaded-auto
Power chip & wire
SMD-auto
Assembly Technology

156. THE END
Thank you for your interest in DCHopkins & Associates