1

LECTURE 18
Switches and Switch Stress: The Concept of Safe
Operating Area for a Device
I. Ideal Switch Characteristics
A. Block +V with I
OFF
0
B. Pass +I with V
ON
0
C. Zero switching delay and its benefits
D. Power loss due to switches: zero in every way
1. DC Loss: R
ON
= 0, V
ON
= 0
2. Switching Loss: No delays, no device
stored charge
E. No stray L
p
or C
p
for undesired ringing!
F. Real Switches
1. Limited quadrants of operation for real
solid state switches
a. One quadrant and device example
II. Active Switch Stress (S) and Switch
Utilization (U)
A. General Definitions
S(active) ~ V(
sw
off
)
rms I
(
sw
on
) per switch
U
P(load)
S
per switch
B. Case of Flyback Converter
2
V(off) ~ V
g
/D } D
opt
} for
I(on) ~ I D } U
max
C. Table of U
max
and D
opt
for various Converters
The above selection of solid state switches will be
matched to the I(D) through the device and the V(D)
across the device as determined by detailed circuit
analysis in the next few lectures. Analysis of V(D)
and I(D) will follow the same procedure as M(D).
3
LECTURE 18
Switches and Switch Stress: The Concept of Safe
Operating Area for a Device
A. Ideal Switch Characteristics:
There are five characteristics of a SPST ideal
switch. You make think of a semiconductor power
switch as you do of a light switch at home. It
operates with no concern for losses in either the on
or the off state.
i
t
+/-V
1. Block +V with i
t
= 0: No leakage current flows when off
2. V
sw
(ON) 0 at all i
t
, either +i flow allowed
3. Switch on/off or off/on transitions occur with zero delay
or instantaneous response time. Therefore, even for
finite V and I during the switching, the energy E =
(VI)(t) required to switch is near zero because the
switch time is assumed to be zero. Removing this
assumption is the first step to understanding real
switches that operate at f
SW
. Even if the switch
transitions lose only a little energy and with a fast switch
time, they switch at 100kHz to 1 MHz so that over one
second close to a million transitions occur, each adding
to the total lost energy.
4. Power required to drive the switch is negligible. No DC
4
losses nor any dynamic switching loss. Typically, a switch driver
controls 100-1000 times the power it dissipates. However, for
some devices this drive power can approach 10 % of the energy
switched as we will see.
5. No device capacitance is charged or discharged during
V
OFF
or V
ON
states. No charge Q is stored in semiconductor
devices. No inductor current i storage occurs during the on state
due to either real inductors or due to leakage inductances of
cores or due to wiring. Hence no additional stored energy flows to
stress the switch via either I
max
and V
max
. Removing this
assumption is the another big step to truly understanding switch
losses
In summary an ideal switch is able to pass currents bi-directionally
or to block voltages bi-directionally.
Note that SPDT
switches may be
modeled by two
synchronized or
commutated
SPST switches.
One is on while
two is off and
vice versa.
1
2
B
A
+
+
-
-
v
A
v
B
i
A
i
B
i
L
V
g
The ideal switch model now needs to be deconstructed
assumption by assumption to better appreciate switch losses,
which account for 5-15 % of total loss depending on the specific
conditions of the PWM converter involved.
5
B. Real Switches
There are five types of real switches we will predominantly deal
with in power electronics, diodes, bipolar junction transistors(BJT),
gate turn-off thyristors(GTO), field effect transistors(FET), and
insulated gate bipolar transistors(IGBT), each of which has unique
V-I characteristics shown below.
Five types of Real Switches
Switch Property

1. Passes current in one direction, blocks in the other 1. Diode
2. Passes or blocks current in one direction 2. BJT
3. Can pass in one direction or block in both directions 3. GTO
4. Can pass in both directions, but blocks in only one direction 4. FET
5. Passes or blocks current in one direction 5. IGBT
1. Diode
2. BJT
3. GTO
4. FET
5. IGBT
Device
Symbol
(including body diode)
We need to realize that each switch has unique and limited
turn-on and turn-off characteristics. Hence, depending on what is
required in terms of current passing through the device when it is
on and voltage that the device is required to block when it is off,
we may have only one choice for a switch regardless of cost. Of
course some circuit circumstances allow for us to choose from
several switch possibilities, perhaps introducing cost as a final
6
determining factor. Below we examine the diode and transistor
switches briefly in a four quadrant V-I space to introduce the
concept of safe operating area (SOA) for the specific switch .
This is a rule of thumb that is easy to define and provided by each
switch manufacturer to guide the user. For any real
semiconductor switch as opposed to mechanical switches multi-
quadrant operation is never easy unless we employ series and
parallel combinations of our five main switches.
v
i
I
II III
IV
Four quadrant operation is not
possible in single device solid
state switches. To achieve four
quadrants one needs several
devices in various
series/parallel combinations
a. One quadrant switches
v
i
A
switch A
on
v
switch B
on
i
B
i
L i
L
V
g
-V
g
switch B
off
switch A
off
SW A is?
Pnp
bipolar?
MOSFET?
SW B is?
Diode,
other?
Note the difference between bipolar transistors (only one-way
current flow) and MOSFETs, (bi-directional current flow). Also, the
bipolar of a different type can indeed stand off the opposite
polarity but current flow is one-way, the other way. The key to
consider in one quadrant switches is the V
max
(off) and I
max
(on)
levels the switches are subjected to by circuit conditions.
7
+
-
v
B
V
g
i
L
(t)
L
+
-
v
A
i
A
i
B
To help thinking, consider
the specific example of
both switches above in a
buck converter.
Several simple one quadrant switches are as shown below for
both switches
SW A
a)
C
C
1
0
i
+
i
-
v
1
0
+
-
v
v
i
on
off
stands off
V
g
b)
BJT and IGBT symbols (a), and
their idealized switch
characteristics (b).
SW B
1
0
+
-
v
i
v
i
on
off
stands off
V
out
Diode symbol and its ideal
characteristics.
The dynamic behavior during switching from a static off/on state to
the other static state depends both on the switch device and the
circuit. In the figure below we show three curves:
1. Perfect Switch trajectories
2. Typical turn-on V-I trajectory
3. Typical turn-off V-I trajectory
8
Switch current vs. voltage, showing the switching trajectories.
The safe operating area as obtained from the manufacturers
device specifications must lie well outside switching trajectories as
shown below:
Switching trajectories within safe operating area.
Switch manufacturers also provide estimated device switching
times based on operating conditions to better estimate the energy
lost during switching W = V*I dt.
First lets refresh our memory on bipolar diode and transistor DC
characteristics as well as some dynamic issues for these two one
quadrant switches. Hopefully, this is a review for most students. If
not we will cover these devices again when we focus in on the
power electronic versions of these classic devices. All bets are off
that these power electronic versions even vaguely resemble the
microelectronic versions, especially for FETs
9
Static models of diodes involve the following:
R
off
R
on
V
on
For HW #4 from the MUR3040PT diode data book for
practice obtain all three values: R
on
= 0.015, R
off
= 40 m, and
V
on
= 0.94 V.
Static characteristics do not tell the full story of any device. Like
people the dynamic characteristics may reveal new and
unexpected behavior. For example, the V
on
for the diode above
does have a brief voltage overshoot when driven by a constant
current source to turn it on. This needs to be accounted for in any
dynamic model of diode operation as the dynamic I-V is unique.
10
Forward recovery characteristic of a diode.
A static npn bipolar transistor model should involve the following
two values:
R
on
V
on
For HW #4 from the measured transistor characteristics below
for practice obtain: R
on
= 134 m and V
on
= 0.08 V.
11
The discussion of the thyristor, the gate turn-off thyristor and the
insulated gate bipolar transistor will have to await a more full
description of these unique switches. This will occur in later
lectures.
II. Active/Switch Stress and the Concept of
Switch Utilization
We will below introduce quantatation to the concept of switch
stress via a term, S, and also introduce an engineering term, U, to
describe how well we are utilizing the chosen switch capability to
the circuit need of that switch. The former can be compared to the
SOA given by the switch manufacturer to avoid switch failure and
the later can be used as a guide to answer the question of
whether or not we are fully utilizing a chosen switch in the specific
circuit to better access the costs on the bill of lading. What
complicates the matter is that S and U depend on the duty cycle
range employed causing tradeoffs to be made in any PWM
converter design between the power switches and the controller
conditions allowed.
A. General: S(active) only
The cost of many PWM converters is primarily the expensive solid
state actively driven switches as well as the passive diode
switches. For a circuit of K such switches
12
S =
total switch
stress

k
V(
peak
blocking
)I(
on
rms
)

This sum, S, depends both on V blocking levels and current on
levels as well as topology of the circuit and use of isolation
transformers with known turns ratios. Ringing due to leakage
inductances only adds to peak voltage stress.
We want to minimize this stress yet still have maximum power
flow! And we want to use the cheapest switch possible to make
more profit on the PWM converter.
Utilization
P(load)
S(active)
We want to maximize
U although typically it is a value <1 depending on the operating
point, using transformers.
B. Case of Flyback Converter with one transistor
(a) Find peak blocking voltage
C Lm
+
-
D1
V
Q1
Vg
.
.
1:n
(a)
tr V
(
stand
off
) =
g V
+
o V
n
and
o
V
g
V
= n
D
D

o
V
n
=
D
D

g
V

,
tr V
(
stand
off
) =
g V
1 +
D
D
=
g V
D

This is a peak value.

(b) Find the rms current for
on-conditions
I
Q1
(RMS) = I
g
rms
Q
1
I
(rms) = I D
Q1
D'Ts DTs
conducting
devices
ig
D1
Ts
t
I
0
I is
in the
magnetic
13
P
load
V
g
D
S active is the transistor stress!
S( )
V
(off)
I
(rms) = (
g
V
+
o
V
n
) I D
TR Q
1 active
=
g
V
I D / ' D
(c) Utilization
P(load)
S(active)
(1) Lets find P(load) in terms of V
o
I
o
for dc converter. Notice the
three different currents in the flyback:
I
g
from dc input source: I
g
= DI(magnetics)
I
o
into the load = I(magnetics)D/N
I the current in the magnetics
. .
1:D
. .
D':n
R
+
-
Vo Vg
I
Io
o I
= (I)
D
n

I = I
g
/D
P(load) = V
o
I
o
= V
o
I
D
n
2. S(active) should be in same variables as P(load) to get
U(flyback).
S(flyback) =
g
V
I D
D
and
g
V
=
o
V

D
n

1
D

S =
o V
n

I
D
same variables as P(load) for flyback
U =
o
V
I
D
n
o
V
n
I / D
= D (flyback) =
P(loadflyback)
S(activeflyback)

D
To find U
max
(flyback) versus D use dV/dD = 0.
14
dV/dD =

[( ) ]
=
1
0
D D
D
D
-1/2
- 3/2D
1/2
= 0 thus D = 1/3.
D
1/3 VTr(off) ITr(on)
U .385
U
max
= 0.385
at D = 1/3 for
1 transistor flyback
D = 1/3 minimizes the product of V
tr
(off) * I
tr
(on rms). We know a
maximum occurs between D = 0, V = 0 and D = 1 V = 0 for the
flyback. However, U
max
is not the only issue so D @ U
max
is a
starting point all other issues neglected. In design we have some
fixed quantities such as:
V
o
, V
g
and load power (V
o
*I
o
) product.
Then D and n for example are free to choose.
Summarize (transformer/isolated) flyback:
Q I (rms) = I D,
o V
g V
= n
D
D
Q V
(off) =
g V
D
Given:
o V
g V
fixed we can trade off n vs D to achieve
o V
g V
Since
o V
g V
=
nD
D'
two extremes are: (1) D 1.0 n
but this means for fixed V
o
, V
g
which causes V
tr
(off)
(2) D 0 n on secondary but for fixed V
out
this causes
I
tr
(on)
D 1/3 a good first guess.
15
(3) General Table of Switch Utilization for Six Major Converters
Table 6.1. Active switch utilizations of some common dc-dc converters, single operating point.
Converter U(D) max U(D) max U(D)
occurs at D=
1 Buck
D
1 1
2 Boost
D
D
'

0
3 Buck-boost, flyback, nonisolated SEPIC,
isolated SEPIC, nonisolated Cuk, isolated
Cuk
D D '
2
3 3
=0.385 1/3 U for
nonisolated
4 Forward, n
1
= n
2
1
2
D
1
2 2
=0.353
5 Other isolated buck-derived converters
(full- bridge, half-bridge, push-pull)
D
2 2
1
2 2
=0.353 1 isolated have
U
6 Isolated boost-derived converters (full-
bridge, push-pull)
D
D
'
2 1+
1
2
0
Flyback line 3 you sacrifice U to achieve input/output isolation.
Some comments are:
(a) Line 3 flyback results also work for other converters.
(b) Line 2 boost without trf. isolation
C R
+
-
Vo>Vg Vg
L
D = 0: Tr is always off, diode always on, U = P/S then goes to
a large value.
No I
on
stress but V
off
(stress)
D = 0 gives maximum utilization but V
o
= V
g
is limited.
(c) Line 6 only get a maximum switch utilization U= at D=0
Add transformer isolation to boost to get dc isolation BUT
V(off) on switches increases! WIN - WIN Case
(d) Line 5 full-bridge buck-derived
Given: V
g
= 500 V
o
= nDV
g
V
o
= 5
P
o
= kW is a fixed constraint
Choose n 100 because D 1 and at D = 1
U(max) = .35 the best we can hope for.
16
P(load)
S(active)
S(active)@
max U =
kW
.35
= 2.9 kVA = S
max
Compare to line 1 non-isolated buck V
o
= DV
g
and
V
g
= 500, V
o
= 5 I
o
= kW
D must be .01

max U
= D = .1, S =
kW
0.1
= 10kVA
In summary the tradeoffs with extra cost of transformers and core
reset complexity we find:
isolated buck non-isolated
U higher U lower
S lower S higher
A reminder grading in this course is 60% HW (We are doing this
weekly). So far we had
HW#1 Chapter 2 Erickson Pbms 2,3,4,6 and class note questions
HW#2 Chapter 3 Pbms 6,7 of Erickson and class note questions
HW#3 Chapter 6 Pbms 6,7 of Erickson and class note questions
HW#4 Chapter 4 Erickson Pbms2,4,5,6 and class note questions
Note that we reassigned HW pbm. 6.11 which is
now a design problem due for your midterm
exam not in HW # 3. See section 6.4.2 to guide
you.
As a help to Ericksons Chapter 4 HW problems
below we give a converter circuit, analyze its DC states of
operation to find the maximum or worst case DC voltage and
current requirements of the switches employed in that specific
circuit topology. These maximums set by the circuit must be
exceeded by the manufacturers device specifications under all
conditions of operation. There will be both DC maximums and
transient maximums.
17
We will start with the DC levels.
b
a
+
-
V V
g
+
+
+
-
-
-
V
1
C
1
L
1
I
1
L
2
I
2
C
2
R
+
-
V V
g
+
+
+
-
-
-
V
1
C
1
L
1
I
1
L
2
I
2
C
2
R
SWa on/SWb off SWa off/SWb on
V
L1
= V
g
V
L2
= - V
1
+ V
g
- V
I
C1
= I
2
I
C2
= I
2
- V/R
V
L1
= V
1
V
L2
= - V
I
C1
= - I
1
I
C2
= I
2
- V/R
Volt-sec balance on L
1
V
L1
= DV
g
+ DV
1
V
1
= -
D
D

V
g

Volt-sec balance on L
2
V
L2
= D(-V
1
+ V
g
- V) - DV
0 = - DV
1
+ DV
g
- V
=
2
g
g
D V
D
+
DV
- V

=
g
DV
(D+D )
D
- V

V =
g
DV
D
2
V =

18
I-sec balance on C
I
C1
= DI
2
- DI
1
2
I
=
V
R
1
I
=
DV
D R
The above equations give the equations for the DC operating
voltages. Now we consider values of V
off
and I
on
for both switches.
+
-
V V
g
+
+
+
-
-
-
V
1
C
1
L
1
I
1
L
2
I
2
C
2
R
(a)
+ -
(b)
+
-
Maximum DC switch ratings will be functions of both circuit values
and switching duty cycles but will not vary with, f
sw
.
V
OFF
and I
ON
for switch a
i
a
V
g
a
V
=
g
V
-
L1
V
=
g
V
-
g
DV
D (
D
D
g
V
)

V
a
= V
g
a 2 1
I
=
I
+
I
=
V
R
+
D
D
V
R
=
V
D R
V
OFF
and Ion for switch b
i
b
V
g
-V
1
V
b
= V
g
- V
1
= Vg/D
b
I
=
1
I
-
2
I

=
'
-
'

=
DV
D R
V
R
DVg
D R
2
< I
b
0
HW#4 problem 4.6 of Erickson, is a also a one quadrant switch