Switching Regulators for Poets

Application Note 25
September 1987
Switching Regulators for Poets
A Gentle Guide for the Trepidatious
Jim Williams
The above title is not happenstance and was arrived at after
considerable deliberation. As a linear IC manufacturer, it is
our goal to encourage users to design and build switching
regulators. A problem is that while everyone agrees that
working switching regulators are a good thing, everyone
also agrees that they are difficult to get working. Switching
regulators, with their high efficiency and small size, are
increasingly desirable as overall package sizes shrink.
Unfortunately, switching regulators are also one of the
most difficult linear circuits to design. Mysterious modes,
sudden, seemingly inexplicable failures, peculiar regulation characteristics and just plain explosions are common
occurrences. Diodes conduct the wrong way. Things get
hot that shouldn’t. Capacitors act like resistors, fuses
don’t blow and transistors do. The output is at ground, and
the ground terminal shows volts of noise.
Added to this poisonous brew is the regulator’s feedback
loop, sampled in nature and replete with uncertain phase
shifts. Everything, of course, varies with line and load
conditions— and the time of day, or so it seems. In the face
of such menace, what are Everyman and the poets to do?
The classic approach is to seek wisdom. Substantial
expertise exists but is concentrated in a small number of
corporate and academic areas. These resources are not
readily accessed by Everyman and some cynics might
suggest that they are deliberately protected by a selfserving priesthood. A glance through conference proceedings and published literature yields either a storm of
mathematics or absurdly coy and simple little block diagrams that make everything look just so easy. Either way,
Everyman loses. And the poets don’t even get to try.
Something to think about is that most people who want
switching regulators don’t need 98.2% efficiency or
100W/cubic inch. They aren’t trying to get tenure and don’t
care about inventing a new type of circuit. What they want
are concepts directly applicable to construction of working circuits with readily-available parts. Thus equipped,
Everyman can build and sell useful products, presumably
buy more components and everyone’s interests (not incidentally, including ours) are served.
As author, I must confess that I am more poet than
switching regulator designer, and my poetry ain’t very
good. Before this effort, my enthusiasm level for switchers
resided somewhere between trepidation and terror. This
position has changed to one of cautiously respectful
optimism. Several things aided this transformation and
significantly influenced this publication. The “encouragement” of the Captains of this corporation, emphasized
over the last year at increasingly insistent levels, constituted one form of inspiration. Conversations with users
(or people who wanted to be) provided more valuable
perspective and strength in the knowledge that I was not
alone in my difficulties with switchers.
At the circuit level, a significant decision was to employ
standard, off-the-shelf magnetics exclusively.1 This policy
was driven by the observation that the majority of problems encountered with switchers centered around inductive components. This approach almost certainly prevents
precisely-optimized performance and may horrify some
veteran switcher designers. It also eliminates inductor
construction uncertainties, saves time and greatly
increases the likelihood of getting a design running. It’s
NOTE 1: For recommended magnetics supplier, see page 13.
, LTC and LT are registered trademarks of Linear Technology Corporation.
an25fa
AN25-1
Application Note 25
much easier to work with, and get enthusiastic about, a
functional circuit than the smoking carcass of a devastated
breadboard. If standard inductor characteristics aren’t
optimal, it’s easier to see the evidence on a ‘scope than to
guess why you don’t see anything.
Additionally, once the circuit is running, an optimized
version of the standard product can be supplied by the
inductor manufacturer. It’s generally easier for the inductor manufacturer to modify its standard product than to
start from scratch. The process of communicating and
translating circuit performance requirements into inductor construction details is tricky. Using standard product
as a starting point accelerates the dialogue, minimizing the
number of iterations required for satisfactory results.
Often, the standard product suffices for the purpose and
no further effort is required.
to be induced into the area around the inductor. The
amount of energy stored in this field is a function of the
current level, how long current flows, the characteristics
of the inductor and its core material. It is often useful to
think of the inductor as a bucket and analogize current flow
as water pouring into it. The ultimate limit on energy
storage is set by the bucket’s capacity, corresponding to
the inductor’s saturation limitations. The amount of energy that can be put into an inductor in a given time is
limited by the applied voltage and the inductance. The
amount of energy that can be stored without saturating the
inductor is limited by the core characteristics. Size, core
material, operating frequency, voltage and current influence inductor design.
5VIN
22µF
Strictly speaking, it makes more sense to design the
inductor to meet circuit requirements than to fashion a
circuit around a standard inductor. Deliberately ignoring
this consideration considerably complicated the author’s
work, but hopefully will simplify the reader’s (such is the
lot of an application note writer’s life). Those interested in
inductor design theory are commended to LTC Application
Note AN-19, “LT®1070 Design Manual.”
A final aid in achieving my new outlook on switchers was
the LT1070 family. In terms of circuit construction and
ease of use they really are superior switching regulator
ICs. A 75V, 5A (LT1070HV) on-chip power switch,
complete control loop, oscillator and only 5 pins eliminate
a lot of the ambiguity of other devices. Internal details and
operating features of the LT1070 family are detailed in
Appendix A, “Physiology of the LT1070.”
Basic Flyback Regulator
Figure 1 shows a basic flyback regulator using the LT1070.
It converts a 5V input to a 12V output. Figure 2 shows the
voltage (Trace A) and the current (Trace B) waveforms at
the VSWITCH pin. The VSW output is the collector of a
common emitter NPN, so current flows when it is low.
Current is pulled through the 100µH inductor and controlled to a value of which forces the 12V output to be
constant. The circuit’s 40kHz repetition rate is set by the
LT1070’s internal oscillator. During the time VSW is low,
current flow through the inductor causes a magnetic field
100µH
PULSE
ENGINEERING
#51516
+
VIN
GND
VSW
LT1070
MBR735
(MOTOROLA)
10.7k*
FB
VC
1k
+
12V OUTPUT
1.75A
+
1.24k*
470µF
1µF
AN-25 F01
*1% METAL FILM RESISTOR
Figure 1. Flyback-Type Regulator
A = 10V/DIV
B = 1A/DIV
10µs/DIV
Figure 2. Flyback Regulator’s Waveforms at 7W Loading
If the inductor is enclosed in a feedback-enforced loop,
such as Figure 1, the energy put into it will be controlled to
meet circuit output demands. Figure 3 shows what
happens when output demand doubles. In this case duty
an25fa
AN25-2
Application Note 25
samples this output via the 10.7k to 1.24k divider. The
LT1070 compares the feedback pin voltage to its internal
1.24V reference and controls the VSW pin’s duty cycle and
current, closing a loop. Since the LT1070 is trying to force
its feedback pin to 1.24V, output voltage may be set by
varying the 10.7k or 1.24k values.
A = 10V/DIV
B = 1A/DIV
10µs/DIV
Figure 3. Flyback Regulator’s Waveforms at 14W Loading
cycle doesn’t change much but current doubles. This
requires the inductor to store more energy. If it couldn’t
meet the storage requirement, e.g., it saturated and could
not hold any more magnetic flux, it would cease to look
inductive. If this point is reached, current flow is limited
only by the resistance of the wire and rapidly builds to
excessive and destructive values. This behavior is exactly
the opposite of a capacitor, where current diminishes
upon entering saturation. Capacitors can maintain energy
storage with no current flowing; inductors cannot. See
Appendix C, “A Checklist for Switching Regulator
Designs,” for details.
At the end of each inductor charge cycle, current flow in
the inductor decays, and the magnetic field around it
abruptly collapses. The VSW pin is seen to rise rapidly to
a voltage higher than the 5V input. This flyback action
gives the regulator its voltage boost characteristics and its
name. The boost characteristic is caused by the collapsing
magnetic field’s lines of flux cutting across the inductor’s
conductive wire turns. This satisfies the basic requirement
for generation of a current in (and hence, a voltage across)
a conductor. This moving magnetic field deposits energy
into the wire in proportion to how much was stored in the
core during the current charge cycle. It is worth noting that
the operating characteristics shown here are similar to the
Kettering ignition system used in automobiles, explaining
why spark occurs when the points open.2
In this circuit the flyback is seen to clamp to a level just
above the output voltage. This is so because the flyback
pulse is steered through the Schottky diode to the output.
The 470µF capacitor integrates the repetitive flyback events
to DC, providing the circuit’s output. The feedback pin (FB)
All feedback loops require some form of stability compensation (see the appended section of LTC Application Note
AN-18, “The Oscillation Problem—Frequency Compensation Without Tears,” for general discussion). The LT1070
is no exception. Its voltage gain characteristics, combined
with the substantial phase shift of the circuit’s sampled
energy delivery, ensure oscillation if uncompensated.
While the large output capacitor smooths the output to DC,
it also teams up with the sampled energy coming into it to
create phase shift. To complicate matters, the load, which
may vary, also influences phase characteristics. The regulator can only source into the output capacitor. The load
determines the sink time constant, influencing phase
performance and overall stability.
The LT1070’s internals have been designed with all this in
mind and compensation is usually fairly simple. In this
case the 1k to 1µF combination at the compensation pin
(VC) rolls off the circuit, providing stable compensation for
all operating conditions (see Appendix B, “Frequency
Compensation,” for details and suggestions on achieving
stability in switching regulator loops).
As innocent as Figure 1 appears, it’s not too difficult to get
into odd and seemingly inexplicable problems. Note that
the ground connection appears at the ground pin, as
opposed to its customary location at the bottom of the
diagram. This is deliberate and the supply and load return
connections should be made there. The high speed, high
current returns from the output transistor’s emitter (the
“other end” of the VSW pin) should not be allowed to mix
with the small currents of the output divider or the VC pin.
Such mixing can promote poor regulation, unstable operation or outright oscillation. Similarly, the 22µF bypass
capacitor ensures clean local power at the LT1070, even
during the fast, high current drain periods when VSW
comes on. It should have good high frequency characteristics (tantalum or aluminum paralleled by a disc ceramic
type). More discussion of these considerations appears in
Appendix C.
NOTE 2: Back when giants walked the earth, Real Cars used ignition points.
an25fa
AN25-3
Application Note 25
+
3k
1/2W
220Ω
100µF
1N5936
30V
– 48V
68V**
Q1
2N5550
1k
OPTIONAL
LOW DRIFT FEEDBACK
CONNECTION
(SEE TEXT)
FROM
5V OUTPUT
100µH
PULSE
ENGINEERING
#51516
*
+
+
2.2µF
1100µF***
VIN
VSW
GND LT1070HV
3.9k
1%
Q2
2N5401
Q2
2N5401
TO
FB PIN
FB
VC
2k
1.2k
1%
Q3
2N5401
1k
1%
0.22
– 48V
(– 40V TO – 60V)
INPUT
3.01k
1%
5V OUTPUT
4A
TO
– 48V
*
MUR810 (MOTOROLA)
**
1.5KE68A (MOTOROLA)
3.01k
1%
AN-25 F04
*** VPR1127R5E1E (MALLORY)
Figure 4. Nonisolated –48V to 5V Regulator
–48V to 5V Telecom Flyback Regulator
Figure 4’s circuit is operationally similar to Figure 1 but is
intended for telecom applications. The raw telecom supply
is nominally –48V but can vary from –40V to –60V. This
range of voltages is acceptable to the VSW pin but protection is required for the VIN pin (VMAX = 60V). Q1 and the
30V zener diode serve this purpose, dropping VIN’s voltage to acceptable levels under all line conditions.
Here, the “top” of the inductor is at ground and the
LT1070’s ground pin at –48V. The feedback pin senses
with respect to the ground pin, so a level shift is required
for the 5V output. Q2 serves this purpose, introducing only
–2mV/°C drift. This is normally not objectionable in a logic
supply, but can be compensated with the optional appropriately scaled diode-resistor shown.
Frequency compensation is similar to Figure 1, although a
low ESR (equivalent series resistance) capacitor gives
less phase shift, permitting faster loop response with the
reduced compensation time constant. The 68V zener is a
type designed to clamp and absorb excessive line
transients which might otherwise damage the LT1070
(VSW maximum voltage is 75V).
Figure 5 shows operating waveforms at the VSW pin.
Trace A is the voltage and Trace B the current. Switching
characteristics are fast and clean. The ripples in the
current trace are due to nonoptimal breadboard layout
(ground as I say, not as I do). Inductor ringing on turn-off
(Trace A) is characteristic of flyback configurations.
A = 20V/DIV
B = 2A/DIV
500ns/DIV
Figure 5. Nonisolated Regulator’s Waveforms
an25fa
AN25-4
Application Note 25
PULSE
ENGINEERING
#PE-64795
– 48
RETURN
2k 2
5W
0.47µF
250V
10k
MUR860
D
3k
47µF
IRF-830
1N3026A
18V
S
47k
+
MUR120
OPTIONAL
SEE TEXT
+
MBR735
5V
OUTPUT
5A
100k
1/2W
2000pF**
4
50µF
SPRAGUE
TE-1307
6
50Ω
G 50Ω
Q5
+
1
4µH
PULSE ENGINEERING
#52901
10
MUR120
100µF
5
Q1
VSW
X
68V
VIN
LT1071
V2 = 7V
22k
VC
+
NC
FB
39.2k*
22µF
GND
7
Q2
820Ω
220k
Q4
NC
47k
47k
10k
0.47µF
47k
4
100Ω
Q3
+
A1
LT1006
+
–
12.1k*
2k
10k
20Ω
LT1004
1.2V
2M
1µF
10µF
0.1µF
– 48VIN (– 40V TO – 60V) INPUT
* 1% METAL FILM RESISTOR
** QLA202U7R5J1L (MALLORY)
= 1N4148
4N28
"M" PREFIXED DIODES = MOTOROLA
36k
0.47µF
PNP = 2N3906
NPN = 2N3904
X
= 1.5KE68A (MOTOROLA)
AN-25 F06
Figure 6. Fully Isolated –48V to 5V Regulator
Fully-Isolated Telecom Flyback Regulator
Figure 6’s circuit is another telecom regulator. Although it
looks more complex, it’s really a closely related extension
of the previous flyback circuits. The fundamental difference is that the output is fully galvanically isolated from the
input, often a requirement in equipment. This necessitates
a transformer instead of a simple 2-terminal inductor. It
also requires output feedback information to be transmitted to the regulator across a nonconducting path. The
transformer complicates the circuit’s start-up and switching characteristics while the isolated feedback requires
attention to frequency compensation.
In this circuit the VIN pin receives power from a transformer winding. This winding cannot supply power at
start-up because the circuit is nonfunctional. Q1 through
Q4 address this issue. When power is applied, Q5 cannot
conduct because the LT1071 is unpowered. Q1 zener-
connected Q2 and Q3 are off. Under these conditions Q4
is on, pulling the VC pin down and strobing off the LT1071.
The potential at Q1’s emitter slowly rises as the 10k-100µF
combination charges. When Q1’s emitter rises high enough,
it turns on. Zener-connected Q2 conducts when the voltage across it is about 7V, biasing Q3 on. Q1 sees regenerative feedback, turning Q3 on harder. Q3’s turn-on cuts off
Q4, allowing the VC pin to rise and biasing up the LT1071.
The rate of rise is limited by the 10µF diode combination
at the VC pin. This network forces the VC pin to come up
slowly, providing a soft-start characteristic (the 100Ω
diode string discharges the 10µF capacitor when circuit
input power is removed). Because of this sequence, the
LT1071 cannot start up the circuit until the VIN potential is
well established. This prevents start-up at “starved” or
unstable VIN voltages which could cause erratic or destructive modes. When start-up does occur, the transformer feeds the VIN pin with DC via the MUR120 diode.
an25fa
AN25-5
Application Note 25
The 50Ω resistor combines with the 100µF capacitor to
give good ripple and transient filtering. This voltage is
ample to run the LT1071 and reduces the current through
the 10k resistor, saving power. Q1, Q2 and Q3 remain on,
biasing Q4 to allow LT1071 operation.
switching, these capacitances can cause excessive transient voltages to appear. The 18V zener diode insures
against gate-source breakdown (VGSMAX = 20V) and the
diode clamps the VSW pin to the VIN potential. Mention of
these considerations appears in Appendix C.
In the previous flyback circuits, the VSW pin drove the
inductor directly. Here, a power MOSFET is interposed
between the VSW pin and the inductor. In this arrangement
the inductor is a transformer and its flyback characteristics are different from a simple 2-terminal inductor. For the
simple inductor, the flyback energy was clamped by and
dumped directly into the output capacitor. Excessive voltages did not occur. In the transformer case, all the flyback
energy does not end up in the output capacitor. Substantial flyback voltage spikes (>100V) appear across the
transformer primary when the LT1071 driven MOSFET
turns off.
The transformer’s rectified and filtered secondary produces the 5V output. This output is galvanically isolated
from the circuit’s input. To preserve this desired feature,
the feedback path must also be galvanically isolated. A1,
the optoisolator and their associated components serve
this function. A1, powered by the 5V output, compares a
resistively-sampled portion of the output with the LT1004
1.2V reference. Operating at a gain of 200, it drives the
optoisolator’s LED. The optoisolator’s output transistor
biases the LT1071’s VC pin, closing a regulation loop. The
feedback amplifier inside the LT1071 is essentially bypassed by the A1 optoisolator combination and is not
used. Normally, the optoisolator’s drifty transmission
characteristics over time and temperature would result in
unstable feedback. Here, A1’s gain is placed ahead of the
optoisolator. This attenuates these uncertainties,
providing a stable loop. This approach is not too different
from inside-the-loop booster transistors and buffers used
with op amps. Both schemes rely on the op amp’s gain to
eliminate uncertainties and drifts. Returning the
optoisolator to VREF instead of ground forces the op amp
to bias well above ground, minimizing saturation effects
during output transients.
Several measures prevent these spikes from destroying
the circuit. The 0.47µF-2k-diode combination, a damper
network, conducts during the flyback event. This loads the
transformer primary, minimizing flyback amplitude. The
damper values are selected empirically, with the trade-off
being power dissipation in them. Very low values markedly reduce flyback potentials but cause excessive dissipation. High values permit low dissipation but allow
excessive flyback voltages. The damper values should be
selected under fully-loaded output conditions because
flyback energy is proportionate to transformer power
levels. Appendix C contains additional information on
damper network considerations.
Even with the damper network, the flyback voltage is too
high for the LT1071 output transistor. Q5 prevents the
LT1071 from seeing the high voltage. It is connected in
series with the LT1071’s output transistor. This connection, sometimes called a cascode, lets Q5 stand off the
high voltage and the LT1071 operates well within its
breakdown limits. Development and testing of this configuration is detailed in Appendix D. Q5 has large parasitic
capacitances associated with all terminals. During
Frequency compensation is somewhat more involved in
this circuit than the previous examples. A1 is rolled off by
the 0.1µF unit. This keeps gain low at high frequency,
preventing amplified ripple and noise from being fed back
to the LT1071. Local compensation at the LT1071 VC pin
stabilizes the loop. The 100Ω resistor at the 5V output, a
deliberate sink path, allows loop stability at light or no
load. Appendix B discusses frequency compensation.
Additional transformer secondary windings could be added
if desired. The input zener clips transient voltages.
an25fa
AN25-6
Application Note 25
Circuit waveforms appear in Figure 7. Trace A is Q5’s drain
voltage and Trace B the drain current. Trace A shows that
the MOSFET sees about 100V due to flyback effects, but
this is well within its rating. The ringing on turn-off is
normal and is similar to the waveform observed in
Figure 4’s circuit. Trace B shows that the current flow is
fast, clean and controlled. Figure 8 shows transient response for a 1A step on a 2.5A output. When Trace A goes
high the step occurs. Trace B shows that output sag is
corrected in about 8ms. When Trace A returns low the 1A
load is removed and recovery is similar to the positive
step. Broadband output noise, about 75mVP-P, may be
reduced with the optional filter shown.
100W Off-Line Switching Regulator
One of the most desirable switching regulator circuits is
also one of the most difficult to design. Figure 9’s circuit
has many similarities to the previous design but is powered directly from the 115V AC line. This off-line operation
is desirable because it eliminates large, heavy and inefficient 60Hz magnetics and filter capacitors. The circuit
provides an isolated 5V, 20A output as well as isolated
±12V, 1A outputs. Additional features include operation
over a 90V AC to 140V AC input range, AC line surge
suppression, soft-starting and loop stability under all
conditions. Efficiency exceeds 75%.
BEFORE PROCEEDING ANY FURTHER, THE READER IS
WARNED THAT CAUTION MUST BE USED IN THE CONSTRUCTION, TESTING AND USE OF THIS CIRCUIT.
HIGH VOLTAGE, AC LINE-CONNECTED POTENTIALS
ARE PRESENT IN THIS CIRCUIT. EXTREME CAUTION
MUST BE USED IN WORKING WITH AND MAKING
CONNECTIONS TO THIS CIRCUIT. REPEAT: THIS
CIRCUIT CONTAINS DANGEROUS, AC LINE-CONNECTED HIGH VOLTAGE POTENTIALS. USE CAUTION.
A = 50V/DIV
B = 0.5A/DIV
2µs/DIV
Figure 7. Fully Isolated Regulator’s Waveforms
A = 5V/DIV
B = 100mV/DIV
(AC COUPLED
ON 5V OUTPUT)
5ms/DIV
(3kHz BANDPASS)
Figure 8. Fully Isolated Regulator’s Transient
Response for a 1A Change on a 2.5A Load
AC line power is rectified and filtered by the diode bridge–
470µF combination. The MOV device provides surge suppression and the thermistor limits turn-on in-rush current.
Start-up and soft-start circuitry is similar to Figure 6’s
circuit, with some changes necessitated by the higher
input voltage. Erratic operation at extremely low AC line
voltages (70V AC) is prevented by the 220k-1.24k divider.
At very low AC line inputs, this divider forces the LT1071
feedback pin to a low state, shutting down the circuit. The
high input voltage, typically 160V DC, means that the
LT1071’s internal current limit is set too high to protect the
regulator if the circuit’s output is shorted. Q6 and its
associated components provide about 2A limiting. The
LT1071’s GND pin current flows through the 0.3Ω resistor, turning on Q6 if current is too high. The 22k-50pF RC
filters noise, preventing erratic Q6 operation.
an25fa
AN25-7
Application Note 25
Q5, a power MOSFET, is cascoded with the LT1071 for
high voltage switching. Circuit topology is similar to
Figure 6, with Q5’s voltage breakdown increased to 500V.
LT1071 have changed, reflecting this circuit’s different
gain-phase characteristics.
Additionally, the 50Ω resistor combines with the gate
capacitance to slightly slow Q5’s transitions, reducing
high frequency harmonics. This measure eases layout
considerations. The transformer’s damper network borrows from Figure 6, with values reestablished for this
circuit.
ALL WAVEFORM PHOTOGRAPHS WERE TAKEN WITH
AN ISOLATION TRANSFORMER CONNECTED BETWEEN
THE CIRCUIT’S 90V AC-140V AC INPUT AND THE AC
LINE. USERS AND CONSTRUCTORS OF THIS CIRCUIT
MUST OBSERVE THIS PRECAUTION WHEN CONNECTING TEST EQUIPMENT TO THE CIRCUIT TO AVOID
ELECTRIC SHOCK. REPEAT: AN ISOLATION TRANSFORMER MUST BE CONNECTED BETWEEN FIGURE 9’S
CIRCUIT AND THE AC LINE IF ANY TEST EQUIPMENT IS
TO BE CONNECTED.
The A1-optocoupler-enforced feedback loop preserves
the transformer’s galvanic isolation, allowing the regulator output to be ground referenced. The feedback loop is
also similar to Figure 6. Compensation values at A1 and the
4µH
PULSE ENGINEERING
#52901
RT
90V AC
TO
140V AC
INPUT
2A SLOW
BLOW
DANGER! LETHAL POTENTIALS PRESENT
IN SCREENED AREA! DO NOT CONNECT
GROUNDED TEST EQUIPMENT — SEE TEXT
1N4005s
3
470µF
250V
+
0.22µF
250V
47k
1W
220k*
MUR860
MTH7N50
50Ω G
1N3026A
18V
3k
+
100k
1/2W
B
MUR120
3.8k*
7
WINDINGS FOR
OPTIONAL
±12V DC
OUTPUTS
9
D
20µF
+
A
1
50µF
SPRAGUE
TE-1307
5V
OUTPUT
IN5831
3000µF**
4k
10W
+
OPTIONAL
SEE TEXT
PULSE
ENGINEERING
#PE-64780
Q5
12
1.2k*
S
5
MUR120
47k
1/2W
6
Q1
FB
* 1% METAL FILM RESISTOR
+
LT1071
22µF
22k
2k
VC
GND
** QLA302VO10J2L (MALLORY)
7
= 2N3904
220k
NC
Q4
100Ω
2k
4
47k
= V150LA20 (GE)
Q3
820Ω
0.01µF
47k
1/2W
47k
+
A1
LT1006
Q2
= 1N4148
= D504CS
(MIDWEST
COMPONENTS
100Ω
VSW
VIN
+
1.24k*
10µF
0.47µF
–
36k
20Ω
+
10k
1µF
LT1004
1.2V
15k
4.3M
= 2N3906
22k
Q6
0.3Ω
4N28
50pF
AN-25 F09
Figure 9. 100W Off-Line Switching Regulator
DANGER! Lethal Potentials Present—See Text
an25fa
AN25-8
Application Note 25
Figure 10 shows circuit waveforms at 15A output. Trace A,
Q5’s drain, shows the flyback pulse being damped below
300V (for a discussion of the procedures used to design
the damper network and other design techniques in this
circuit, see Appendix D, “Evolution of a Switching Regulator Design”). Trace B, the LT1071’s VSW pin, stays well
within its voltage rating, despite Q5’s high voltage switching. Trace C, Q5’s drain current, shows that transformer
current is well controlled with no saturation effects. Trace␣ D,
damper network current, is active when Q5 goes off.
Figure 11 is a time and amplitude expansion of Q5’s drain
(Trace A) and transformer primary current (Trace B).
Switching is clean, with residual noise due to nonideal
transformer behavior. The damper network clamps the
flyback pulse well below Q5’s 500V rating and the transformer rings off after the flyback interval. The noise on the
current pulse, due to resonances in the transformer, has
no significant effect on circuit operation.
Figure 12 shows output noise with the optional LC filter in
use. Without the filter, noise is about 150mV. Superimposed, residual 120Hz modulation accounts for trace
thickening at the peaks and could be eliminated by increasing the 470µF value.
Figure 13 shows transient response performance. When
Trace A goes high, a 5A transient is added to a 10A steadystate load. Recovery amplitude is low and clean with a first
order response. When Trace A goes low, the transient load
is removed with similar results.
A = 200V/DIV
A = 100V/DIV
B = 20V/DIV
C = 2A/DIV
B = 0.5A/DIV
D = 2A/DIV
5µs/DIV
Figure 10. Off-Line Switcher’s Waveforms
DANGER! Take This Measurement Only With
an Isolation Transformer in Use—See Text
2µs/DIV
Figure 11. Detail of Off-Line Switcher’s Transformer
Primary Voltage and Current Waveforms
DANGER! Take This Measurement Only With
an Isolation Transformer in Use—See Text
A = 10V/DIV
B = 0.05V/DIV
(AC COUPLED)
A = 0.05V/DIV
50µs/DIV
Figure 12. Figure 9’s Output Ripple at 10A Output
with the Optional LC Filter Added—Without the
Filter, Ripple Increases to About 150mVP-P
10ms/DIV
(1kHz BANDPASS)
Figure 13. Figure 9’s Circuit Responding
to a 5A Change on a 10A Output
an25fa
AN25-9
Application Note 25
Figure 14 shows response for shifts in the line. When
Trace A is high, the AC line is at 140V AC. Line voltage
drops to 90V AC with Trace A low. Trace B, the regulator’s
AC-coupled output, shows a clean recovery with small
amplitude error. The ripples in the waveform, 120Hz input
residue, could be reduced by increasing the 470µF
capacitor.
Figure 15 shows the 5V output at start-up into a 20A load.
Response is slightly underdamped and can be modified by
adjusting the frequency compensation. The compensation shown in Figure 9 is a good compromise between
transient response and turn-on characteristics.The delay
on turn-on and the controlled rise time are due to the slowstart circuitry.
Figure 16 plots regulator efficiency. As would be expected,
efficiency is best at high currents, where static losses are
a small percentage of output power.
Switch-Controlled Motor Speed Controller
Voltage regulators are not the only switching power circuits. Figure 17 shows a motor speed regulator. The
LT1070 provides simplicity and switch-mode control
efficiency. Although this circuit controls a motor, it shares
many considerations common to voltage regulators. When
power is applied, the tachometer output is zero and the
feedback pin (FB) is also at zero. This causes the LT1070
to begin pulsing its VSW pin at maximum duty cycle. The
motor turns, forcing tachometer output. When the FB pin
A = HI = 140V AC
LOW = 90V AC
B = 0.01V/DIV
(AC COUPLED)
20ms/DIV
(1kHz BANDPASS)
Figure 14. Figure 9 Responds to a 90V AC—140V AC Line
Change—Loading is 10A—120Hz Residue in Output Could be
Reduced by Increasing the 470µF Input Filter
A = 1V/DIV
25ms/DIV
Figure 15. Start-Up for Figure 9 at 20A Loading—The 10µF
Capacitor at the LT1070’s VC Pin Produces the Slow-Start
Characteristic. If the Small Overshoot is Objectionable,
Modified Frequency Compensation Can Eliminate it at
Some Cost to Transient Response
an25fa
AN25-10
Application Note 25
arrives at the LT1070’s internal voltage reference value
(1.24V), the loop stabilizes. Speed is adjustable with the
25k potentiometer in the feedback string. The MUR120
damps the motor’s flyback spike. The characteristics of
the motor specified permit no current limiting in series
with the diode. Other motors might require this and
damper network optimization should be done for any
specific unit. Similarly, frequency compensation values
will vary with different motor types. The diode at the
tachometer output prevents transient reverse voltages
due to tachometer commutator switching.
100
EIN = 115V AC
90 PQUIESCENT = 2.5W
80
EFFICIENCY (%)
70
60
50
40
30
20
10
0
0
2
6 8 10 12 14 16 18 20
OUTPUT CURRENT (A)
4
AN-25 F16
Figure 16. Figure 9’s Efficiency vs Operating Point
12V
+
47µF
+
MUR120
22µF
+
MOTOR
VSW
VIN
SLOPE
3V/1000RPM
LT1070
GND
VC
+
47µF
+
12V
10k
10µF
1N4002
TACH
–
25k
SPEED SET
(300-9000 RPM)
FB
MOTOR-TACHOMETER IS
TRW-GLOBE 397A120-2
1k
0.47µF
AN-25 F17
Figure 17. A Simple Motor-Tachometer Servo Loop
an25fa
AN25-11
Application Note 25
biases the LT1070’s VC pin. This closes a control loop
around the Peltier cooler, forcing its temperature low
enough to balance the bridge. The 0°C trim adjusts the
servo point to precisely 0°C. A standard RTD should
monitor Peltier temperature when making this trim. Alternately, the sensor specified should be supplied with a
certified 0°C resistance. With the RTD and Peltier cooler
tightly mated, stability is excellent. Figure 19, a plot of
stability over hours in a 25°C ±3°C ambient, shows a
0.15°C baseline.
Switch-Controlled Peltier 0°C Reference
Figure 18 is another switch-mode control circuit. Here, the
LT1070 controls power to a Peltier cooler, providing a 0°C
temperature reference for transducer calibration.
A platinum RTD is thermally mated to the Peltier cooler.
The RTD combines with a bridge network to give a differential output. A1 provides maximum bridge drive without
introducing significant heating in the RTD. The LTC®1043
switched capacitor network converts this output to a
single-ended signal at A2. A2, operating at a gain of 400,
3.16k*
10V
1k*
–
A1
1/2 LT1013
2k
+
10V
10V
4
LT1004
1.2V
1/2 LTC1043
18.2k*
5
1020Ω**
1k**
18.2k*
+
6
A2
1/2 LT1013
50k
0°C TRIM
–
2
1µF
1µF
3
15
RPLATINUM
1k**
** ULTRONIX 105A 0.1%
RPLATINUM = ROSEMOUNT #118ME-1k AT 0°C
16
0.01µF
+
25k
17
22µF
+
PELTIER
COOLER
+
10M
18
10V
* 1% METAL FILM RESISTOR
1µF
(SELECT)
VSW
VIN
LT1070
= PELTIER COOLER = CAMBION #801-2003-01-00-00
1N4148
VC
GND
FB
10k
AN-25 F17
Figure 18. A Peltier-Cooled Switched-Mode 0°C Reference
an25fa
AN25-12
Application Note 25
Acknowledgments
100
80
1°C
60
40
20
0
1 HOUR
AN-25 F19
Figure 19. Stability of Figure 18’s Circuit Over
Many Hours with a 25°C ±3°C Ambient
The author acknowledges Carl Nelson’s abundance of
commentary, some of which was useful, during preparation of this work. Bob Dobkin’s thoughts and patience are
also appreciated. Ron Young made significant contributions towards Figure 6’s circuit. Bill McColey and other
members of the Engineering staff of Pulse Engineering,
Inc.3, supplied invaluable insight and assistance on magnetics issues. As usual, our customers’ requests and
requirements provided the most valuable source of guidance, and they are due a special thanks.
NOTE 3: P.O. Box 12235, San Diego, CA 92112 (619/268-2400)
APPENDIX A
VIN
Physiology of the LT1070
The LT1070 is a current-mode switcher. This means that
switch duty cycle is directly controlled by switch current
rather than by output voltage. Referring to Figure A1, the
switch is turned on at the start of each oscillator cycle. It
is turned off when switch current reaches a predetermined
level. Control of output voltage is obtained by using the
output of a voltage-sensing error amplifier to set current
trip level. This technique has several advantages. First, it
has immediate response to input voltage variations, unlike
ordinary switchers which have notoriously poor line transient response. Second, it reduces the 90° phase shift at
mid-frequencies in the energy storage inductor. This
greatly simplifies closed loop frequency compensation
under widely varying input voltage or output load conditions. Finally, it allows simple pulse-by-pulse current
limiting to provide maximum switch protection under
output overload or short conditions. A low dropout internal regulator provides a 2.3V supply for all internal circuitry on the LT1070. This low dropout design allows input
voltage to vary from 3V to 6V with virtually no change in
device performance. A 40kHz oscillator is the basic clock
for all internal timing. It turns on the output switch via the
logic and driver circuitry. Special adaptive antisat circuitry
detects onset of saturation in the power switch and adjusts
driver current instantaneously to limit switch saturation.
This minimizes driver dissipation and provides very rapid
turn-off of the switch.
16V
2.3V
REG
SWITCH
OUT
FLYBACK
ERROR
AMP
40kHz
OSC
LOGIC
DRIVER
5A, 75V
SWITCH
ANTI-SAT
MODE
SELECT
COMP
–
FB
ERROR
AMP
+
SHUTDOWN
CIRCUIT
1.24V
REF
+
VC
CURRENT
AMP
GAIN ≈ 6
0.02Ω
–
0.15V
AN-25 FA1
Figure A1. LT1070 Internal Details
an25fa
AN25-13
Application Note 25
A 1.2V bandgap reference biases the positive input of the
error amplifier. The negative input is brought out for
output voltage sensing. This feedback pin has a second
function; when pulled low with an external resistor, it
programs the LT1070 to disconnect the main error amplifier output and connects the output of the flyback amplifier
to the comparator input. The LT1070 will then regulate the
value of the flyback pulse with respect to the supply
voltage. This flyback pulse is directly proportional to
output voltage in the traditional transformer-coupled
flyback topology regulator. By regulating the amplitude of
the flyback pulse the output voltage can be regulated with
no direct connection between input and output. The
output is fully floating up to the breakdown voltage of the
transformer windings. Multiple floating outputs are easily
obtained with additional windings. A special delay network inside the LT1070 ignores the leakage inductance
spike at the leading edge of the flyback pulse to improve
output regulation.
The error signal developed at the comparator input is
brought out externally. This pin (VC) has four different
functions. It is used for frequency compensation, current
limit adjustment, soft-starting and total regulator shutdown. During normal regulator operation this pin sits at a
voltage between 0.9V (low output current) and 2.0V (high
output current). The error amplifiers are current output
(gm) types, so this voltage can be externally clamped for
adjusting current limit. Likewise, a capacitor-coupled external clamp will provide soft-start. Switch duty cycle goes
to zero if the VC pin is pulled to ground through a diode,
placing the LT1070 in an idle mode. Pulling the VC pin
below 0.15V causes total regulator shutdown with only
50µA supply current for shutdown circuitry biasing. For
more details, see Linear Technology Application Note
AN-19, pages 4-8.
APPENDIX B
Frequency Compensation
Although the architecture of the LT1070 is simple enough
to lend itself to a mathematical approach to frequency
compensation, the added complications of input and/or
output filters, unknown capacitor ESR, and gross operating point changes with input voltage and load current
variations all suggest a more practical empirical method.
Many hours spent on breadboards have shown that the
simplest way to optimize the frequency compensation of
the LT1070 is to use transient response techniques and an
R/C box to quickly iterate toward the final compensation
network.
There are many ways to inject a transient signal into a
switching regulator, but the suggested method is to use an
AC-coupled output load variation. This technique avoids
problems of injection point loading and is general to all
switching topologies. The only variation required may be
an amplitude adjustment to maintain small signal conditions with adequate signal strength. Figure B1 shows the
setup.
A function of generator with 50Ω output impedance is
coupled through a 50Ω/1000µF series RC network to the
regulator output. Generator frequency is noncritical. A
good starting point is 50Hz. Lower frequencies may cause
a blinking scope display which is annoying to work with.
Higher frequencies may not allow sufficient settling time
for the output transient. Amplitude of the generator output
is typically set at 5VP-P to generate a 100mAP-P load
variation.
For lightly loaded output (IOUT <100mA), this level may be
too high for small signal response. If the positive and
negative transition settling waveforms are significantly
different, amplitude should be reduced. Actual amplitude
is not particularly important because it is the shape of the
resulting regulator output waveform that indicates loop
stability.
A 2-pole oscilloscope filter with f = 10kHz is used to block
switching frequencies. Regulators without added LC
output filters have switching frequency signals at their
an25fa
AN25-14
Application Note 25
the same point as the probe ground clip. Any reading on
channel A indicates a ground loop problem.
outputs which may have much higher amplitude than the
low frequency settling waveform to be studied. The filter
frequency is high enough to pass the settling waveform
with no distortion.
Once the proper setup is made, finding the optimum
values for the frequency compensation network is fairly
straightforward. Initially, C2 is made large (≥2µF), and R3
is made small (≈1kΩ). This nearly always ensures that the
regulator will be stable enough to start iteration. Now, if
the regulator output waveform is single-pole overdamped
(see the waveforms in Figure B2), the value of C2 is
reduced in steps of about 2:1 until the response becomes
slightly underdamped. Next, R3 is increased in steps of 2;1
to introduce a loop “zero.” This will normally improve
damping and allow the value of C2 to be further reduced.
Shifting back and forth between R3 and C2 variations will
now allow one to quickly find optimum values.
Oscilloscope and generator connections should be made
exactly as shown to prevent ground loop errors. The
oscilloscope is synced by connecting the channel B probe
to the generator output, with the ground clip of the second
probe connected to exactly the same place as the channel
A ground. The standard 50Ω BNC sync output of the
generator should not be used because of ground loop
errors. It may also be necessary to isolate either the
generator or oscilloscope from its third wire (earth ground)
connection in the power plug to prevent ground loop
errors in the ‘scope display. These ground loop errors are
checked by connecting the channel A probe tip to exactly
SCOPE FILTER
SWITCHING
REGULATOR*
GND
OUTPUT
1k
VC
0.015
10k
A
1500pF
50Ω
1W
R3
IOUT
VIN
B
SCOPE
GROUND
CLIP
1000µF
(OBSERVE
POLARITY)
50Ω
GENERATOR
f ≈ 50Hz
C2
*ALL INPUT AND OUTPUT FILTERS MUST BE IN PLACE. INPUT SOURCE (VIN) MUST BE
ACTUAL SOURCE USED IN FINAL DESIGN TO ACCOUNT FOR FINITE SOURCE IMPEDANCE.
AN-25 FB1
Figure B1. Testing Loop Stability
GENERATOR OUTPUT
REGULATOR OUTPUT
WITH LARGE C, SMALL R
WITH REDUCED C AND
SMALL R
EFFECT OF INCREASED R
FURTHER REDUCTIONS
IN C MAY BE POSSIBLE
IMPROPER VALUES WILL
CAUSE OSILLATIONS
AN-25 FB2
Figure B2. Output Transient Response
an25fa
AN25-15
Application Note 25
If the regulator response is underdamped with the initial
large value of C, R should be increased immediately before
larger values of C are tried. This will normally bring about
the overdamped starting condition for further␣ iteration.
Just what is meant by “optimum values” for R3 and C2?
This normally means the smallest value for C2 and the
largest value for R3, which still guarantee no loop oscillations, and which result in loop settling that is as rapid as
possible. The reason for this approach is that it minimizes
the variations in output voltage caused by input ripple
voltage and output load transients. A switching regulator
which is grossly overdamped will never oscillate but it may
have unacceptably large output transients following sudden changes in input voltage or output loading . It may also
suffer from excessive overshoot problems on start-up or
short-circuit recovery.
To guarantee acceptable loop stability under all conditions, the initial values chosen for R3 and C2 should be
checked under all conditions of input voltage and load
current. The simplest way to accomplish this is to apply
load currents of minimum, maximum, and several points
in between. At each load current, input voltage is varied
from minimum to maximum while observing the settling
waveform. The additional time spent “worst-casing” in
this manner is definitely necessary. Switching regulators,
unlike linear regulators, have large shifts in loop gain and
phase with operating conditions. If large temperature
variations are expected for the regulator, stability checks
should also be done at the temperature extremes. There
can be significant temperature variations in several key
component parameters which affect stability—in particular, input and output capacitor values and their ESRs and
inductor permeability. The LT1070 parametric variations
also need some consideration. Those which affect loop
stability are error amplifier gm, and the transfer function
of VC pin voltage versus switch current (listed as a
transconductance under electrical specifications.) For
modest temperature variations, conservative overdamping
under worst-case temperature conditions is usually sufficient to guarantee adequate stability at all temperatures.
If external amplifiers or other active devices are included
in the loop (e.g., Figures 6 and 9), their effects must be
included in stabilizing the loop. LTC Application Note
AN-18, pages 12-15, provides commentary that may be
useful in these situations.
APPENDIX C
A Checklist for Switching Regulator Designs
1. The most common problem area in switching designs
is the inductor and the most common difficulty is
saturation. An inductor is saturated when it cannot hold
any more magnetic flux. As an inductor arrives at
saturation it begins to look more resistive and less
inductive. Under these conditions the current flow
through it is limited only by its DC copper resistance
and the source capacity. This is why saturation often
results in destructive failures. Figure C1 demonstrates
saturation effects. The pulse generator drives Q1, forcing current into the inductor. The diode and RC combination form a typical load. Figure C2 shows results. The
voltage at Q1’s collector falls when it turns on (Trace A
is pulse generator output, Trace B is Q1’s collector).
Trace C, the inductor current, ramps in controlled
fashion. When Q1 goes off, current falls and the induc-
tor rings off. In Figure C3, drive pulse width is longer,
allowing more inductor current buildup. This requires
the inductor to store more magnetic flux. Its ramp
waveform is clean and controlled, indicating that it has
the necessary capacity. Figure C4 brings some unpleasant surprises. Drive pulse width has been increased.
Now, the inductor current departs from its ramp characteristic into a nonlinear slope. The nonlinear behavior
starts between the third and fourth vertical divisions.
This curve shows a rapidly increasing current characteristic. These conditions indicate that the inductor is
entering saturation. If pulse width is increased much
more, the current will rise to destructive levels. It is
worth noting that some inductors saturate much more
abruptly than this case.
an25fa
AN25-16
Application Note 25
TYPE 547
OSCILLOSCOPE
TYPE
1A4
30V
RLOAD
PULSE
GENERATOR
Q1
AN-25 FC1
Figure C1. Inductor Saturation Test Circuit
A = 20V/DIV
A = 20V/DIV
B = 50V/DIV
B = 50V/DIV
C = 200mA/DIV
C = 200mA/DIV
50µs/DIV
50µs/DIV
Figure C2. Normal Inductor Operation
Figure C3. Normal Inductor Operation at Increased Current
A = 20V/DIV
B = 50V/DIV
C = 500mA/DIV
50µs/DIV
Figure C4. Inductor Being Driven into Saturation
an25fa
AN25-17
Application Note 25
2. Always consider inductive flyback effects. Are semicon
ductor breakdown ratings adequate to withstand them?
Is a snubber (damper) network required? Consider all
possible voltages and current paths, including the
transient ones via semiconductor junction capacitances,
to avoid evil problems.
3. Think about requirements in capacitors. All operating
conditions should be accounted for. Voltage rating is
the most obvious consideration, but remember to plan
for the effects of equivalent series resistance (ESR) and
inductance. These specifications can have significant
impact on circuit performance. In particular, an output
capacitor with high ESR can make loop compensation
difficult.
4. Layout is vital. Don’t mix signal, frequency compensation, and feedback returns with high current returns.
Arrange the grounding scheme for the best compromise between AC and DC performance. In many cases,
a ground plane may help. Account for possible effects
of stray inductor-generated flux on other components
and plan layout accordingly.
conducting cycle. This stored charge causes the diode to
act as a low impedance conductive element for a short
period of time after reverse drive is applied. Reverse
recovery time is measured by forward biasing the diode
with a specified current, then forcing a second specified
current backwards through the diode. The time required
for the diode to change from a reverse conducting state to
its normal reverse nonconducting state is reverse recovery time. Hard turn-off diodes switch abruptly from one
state to the other following reverse recovery time. They
therefore dissipate very little power even with moderate
reverse recovery times. Soft turn-off diodes have a gradual
turn-off characteristic that can cause considerable diode
dissipation during the turn-off interval. Figure C5 shows
typical current and voltage waveforms for several commercial diode types used in an LT1070 flyback converter
with VIN = 10V, VOUT = 20V, 2A.
20V
2A
REVERSE CURRENT FLOW
CURRENT
“Simple” diodes furnish a good example of how carefully
semiconductor operating conditions must be considered
in switching regulators. Switching diodes have two important transient characteristics—reverse recovery time and
forward turn-on time. Reverse recovery time occurs
because the diode stores charge during its forward
0
DIODE VOLTAGE AND CURRENT
5. Semiconductor breakdown ratings must be thought
through. Account for all conditions. Transient events
usually cause the most trouble, introducing stresses
that are often hard to predict. Things to watch for
include effects of feedthrough via semiconductor junction capacitances (note the clamping of Q5’s gate in
Figures 6 and 9). Such capacitances can allow excessive voltages to occur for brief durations at what is
normally a low voltage node. Study the data sheet
breakdown, current capacity, and switching speed ratings carefully. Were these specifications written under
the same conditions that your circuit is using the device
in? If in doubt, consult the manufacturer.
USD 735C (SCHOTTKY)
VOLTAGE
20V
2A
MUR 415 (ULTRA FAST)
VOLTAGE
CURRENT
REVERSE CURRENT FLOW
0
MR 856 (FAST)
20V
2A
VOLTAGE
CURRENT
REVERSE CURRENT FLOW
0
0
40
80
120
TIME (ns)
160
200
240
AN-25 FC5
Figure C5. Diode Turn-Off Characteristics
an25fa
AN25-18
Application Note 25
Long reverse recovery times can cause significant extra
heating in the diode or the LT1070 switch. Total power
dissipated is given by:
Ptrr = V • f • tRR • IF
V = reverse diode voltage
f = LT1070 switching frequency
tRR = reverse recovery time
IF = diode forward current just prior to turn-off
With the circuit mentioned, IF is 4A, V = 20V, and f = 40kHz.
Note that diode on current is twice output current for this
particular boost configuration. A diode with trr = 300ns
creates a power loss of:
show diode turn-on spikes for three common diode
types—fast, ultrafast, and Schottky. The height of the
spike will be dependent on rate of rise of current and the
final current value, but these graphs emphasize the need
for fast turn-on characteristics in applications which push
the limits of switch voltage.
Fast diodes can be useless if the stray inductance is high
in the diode, output capacitor or LT1070 loop. 20-gauge
hook-up wire has 30nH/inch inductance. The current fall
time of the LT1070 switch is 108A/sec. This generates a
voltage of (108)(30 • 10–9) = 3V per inch in stray wiring.
Keep the diode, capacitor and LT1070 ground/switch lead
lengths short!
Ptrr = (20)(40 • 103)(300 • 10–6)(4) = 0.96W
Diode turn-on time can potentially be more harmful than
reverse turn-off. It is normally assumed that the output
diode clamps to the output voltage and prevents the
inductor or transformer connection from rising higher
than the output. A diode that turns on slowly can have a
very high forward voltage for the duration of turn-on time.
The problem is that this increased voltage appears across
the LT1070 switch. A 20V turn-on spike superimposed on
a 40V flyback mode output pushes switch voltage
perilously close to the 65V limit. The graphs in Figure C6
USD 735C (SCHOTTKY)
3V
–10V
FORWARD SPIKE
DIODE VOLTAGE
–20V
8V
MUR 415 (ULTRA FAST)
0V
–10V
–20V
FORWARD SPIKE
23V
MR 856 (FAST)
0V
–10V
–20V
DIODE CURRENT
If this same diode had a forward voltage of 0.8V at 4A, its
forward loss would be 2A (average current) times 0.8V
equals 1.6W. Reverse recovery losses in this example are
nearly as large as forward losses. It is important to realize
however, that reverse losses may not necessarily increase
diode dissipation significantly. A hard turn-off diode will
shift much of the power dissipation to the LT1070 switch,
which will undergo a high current and high voltage condition during the duration of reverse recovery time. This has
not been shown to be harmful to the LT1070, but the
power loss remains.
FORWARD SPIKE
0V
4A
2A
0
0
100
200
300
TIME (ns)
400
500
AN-25 FC6
Figure C6. Diode Turn-On Spike
an25fa
AN25-19
Application Note 25
APPENDIX D
Evolution of a Switching Regulator Design
A good way to approach designing a switching regulator
is to break the problem into small tasks and then integrate
everything. The combination of inductors, a sampled
feedback loop, and high speed currents and voltages
leaves much room for confusion. The approach used in
Figure 9’s design is illustrated as an example of an iterative
approach in switching regulator design. This off-line
circuit features high power, an isolated feedback loop and
the aforementioned complexities. Any attempt to get
everything working on the first try is beyond risky.
The transformer drive is the most critical part of Figure 9’s
design. Fast switching of over 100W at high voltage
requires care. In particular, two issues must be addressed.
Will the high voltage FET-LT1071 cascode connection
really work? What amplitudes of flyback voltage are going
to occur and what will their effects be?
Figure D1 begins the investigation. This test circuit allows
checking of the high voltage cascode. To start, a resistive
load is used, eliminating the possible (certain!) complications of the inductive load. Figure D2 shows waveforms.
Switching is clean. Trace A is the FET drain, while Trace B
is the LT1071 VSW pin. Drain current appears in Trace C.
Pulse width is kept deliberately low, minimizing load
power dissipation. Everything appears well ordered, and
the LT1071 VSW pin does not see any high voltage excursions. Artifacts of the MOSFET’s high voltage switching
do, however, appear at the LT1071 VSW pin. On the falling
edge, the ringing appears, albeit at lower amplitude. The
rising edge shows a slight peaking. These effects are due
to the high voltage coupling through the MOSFET’s junction capacitances. The diode clamps the source to 10V,
but the effects of the high voltage slewing are still noticeable. This doesn’t cause much trouble with the resistive
load, but what will happen with the inductor’s higher
flyback voltages?
160V
+
RLOAD
200Ω-50W
NONINDUCTIVE
10µF
D
G
10V
+
350V
MTH8N35
S (MOTOROLA)
20µF
8k
2k
DUTY CYCLE
ADJUST
MUR120
(MOTOROLA)
B = 10V/DIV
VSW
C = 1A/DIV
VIN
NC
FB
VC
A = 50V/DIV
LT1071
GND
200ns/DIV
AN-25 FD1
Figure D1. Test Circuit for MOSFET-LT1071
Cascode with Resistive Load
Figure D2. Testing the MOSFET-LT1071
Cascode Switch with a Resistive Load
an25fa
AN25-20
Application Note 25
Figure D3 shows the test circuit rearranged to accommodate the transformer load. The transformer replaces the
resistor. Its terminated secondary allows it to present a
significant load. The fixed 160V supply has been replaced
with a 0V to 200V unit, permitting voltage to be slowly and
cautiously increased1. The 350V transistor is replaced
with a 1000V unit, in preparation for inductive events.
Figure D4 shows waveforms. As expected, the inductive
flyback (Trace A) is significant, even at low supply voltage
(VSUPPLY = 60V in this photo).
Trace C, the drain current, rises with a characteristic
indicating the inductive load. Trace B, the source voltage,
is of greater concern. The flyback event, feeding through
the MOSFET’s capacitances, causes the source (and gate)
to rise above nominal clamped value. At the higher supply
voltages planned, this could cause excessive gate-source
voltages with resultant device destruction. Because of
this, the zener diode in dashed lines is installed, clamping
gate-source voltages to safe values. This component
appears in Figure 9’s final design. With this correction,
behavior at higher supply voltages may be investigated.
0V TO 200V
VARIABLE SUPPLY
+
10µF
A = 100V/DIV
NC
0.2Ω
50W
B = 10V/DIV
G
10V
+
D 1000V
MTH5N100
(MOTOROLA)
S
C = 1A/DIV
MUR120
(MOTOROLA)
20µF
2µs/DIV
18V
8k
VIN
NC
2k
DUTY CYCLE
ADJUST
FB
VC
Figure D4. MOSFET-LT1071 Cascode Switching the
Transformer Primary—Secondary Load is 0.2Ω
VSW
LT1071
GND
AN-25 FD3
Figure D3. Test Circuit for MOSFET-LT1071
Cascode with Transformer Load
NOTE 1: “For fools rush in where angels fear to tread”—An Essay on Criticism, A. Pope.
an25fa
AN25-21
Application Note 25
Figure D5 shows the MOSFET drain at VSUPPLY = 160V. The
load draws about 2.5A. Flyback voltage rises to 400V. At
5A loading this voltage approaches 500V (Figure D6),
while a 10A load (Figure D7) forces almost 900V flyback.
In actual regulator operation, supply voltages, switch ontime and output current can go higher, meaning flyback
potentials will exceed 1000V. This graphically mandates
the need for a damper network. A simple reverse-biased
diode or zener clipper will work, but will suffer from
excessive dissipation. The network shown in Figure 9 is a
good compromise between dissipation and reasonable
flyback voltages.
A = 100V/DIV
A = 100V/DIV
2.5µs/DIV
2.5µs/DIV
Figure D6. Undamped Regulator
Flyback Pulse at 5A Output
Figure D5. Undamped Regulator
Flyback Pulse at 2.5A Output
A = 200V/DIV
2.5µs/DIV
Figure D7. Undamped Regulator
Flyback Pulse at 10A Output
an25fa
AN25-22
Application Note 25
Once the drive-flyback issues are settled, a feedback loop
is closed around the transformer. This allows checking to
see that loop stabilization is possible. Figure D8 diagrams
the loop. In this configuration the regulator will function,
but is unusable. The output is not galvanically isolated
from the input, which ultimately must be directly AC line
driven. After this loop has been successfully closed, the
isolated version is tried (Figure D9). This introduces more
phase shift, but is also found to be stable with appropriate
frequency compensation. Finally, the connection between
the input and output common potentials is broken, achieving the desired galvanic isolation. The start-up, soft-start
and current limit features are then added and optimized.
Testing involves checking performance under various line
and load conditions. Details on circuit operation are
covered in the text associated with Figure 9.
160V
5V
OUTPUT
D
G
10V
S
VIN
VSW
LT1070
GND
VFB
VC
160V
RETURN
AN-25 FD8
Figure D8. Developmental Version of Off-Line Switching Regulator—No Isolation is Included and
the Scheme is Solely Intended to Verify that a Loop Can be Closed Around the Transformer
5V
OUTPUT
160V
D
G
10V
S
+
A1
–
VIN
VSW
LT1070
GND
160V
RETURN
VFB
N/C
VC
AN-25 FD9
Figure D9. Developmental Version of Off-Line Regulator with Isolation— the Circuit Verifies That Loop Stability is Achievable with the
Added Phase Shift of A1 and the Opto-Isolator—Start-Up, Current Limit and Soft-Start Features Must be Added to Complete the Design
an25fa
Information furnished by Linear Technology Corporation is believed to be accurate and reliable.
However, no responsibility is assumed for its use. Linear Technology Corporation makes no representation that the interconnection of its circuits as described herein will not infringe on existing patent rights.
AN25-23
Application Note 25
an25fa
AN25-24
Linear Technology Corporation
LT/TP 0202 1.5K Rev A • PRINTED IN USA
1630 McCarthy Blvd., Milpitas, CA 95035-7417
(408)432-1900 ● FAX: (408) 434-0507 ● www.linear-tech.com
 LINEAR TECHNOLOGY CORPORATION 1988