hat operates in multiple modes that are determined by power demand, where each
mode enhances the efficiency within its power range. A flyback converter is highlighted as the solution;
it is controlled in burst, frequency foldback, discontinuous conduction, and quasi-resonant modes that
are shown to enhance efficiency from no-load to full-load, respectively. In addition to the design steps,
this topic also includes test techniques and performance verification.

I.

G
REEN
-M
ODE
P
OWER
S
YSTEM

Todays converters must operate efficiently
from full-load to no-load. Converter efficiency is
often judged as the average of the efficiency at
25%, 50%, 75% and 100% full rated load,
between the AC line and the output of the
converter.
[1]
An additional no-load AC line power
criterion is further imposed on the converter in
order to judge the conservation merits of the
power converter. Thus, power supply efficiency
and performance at light loads is as important as
at full rated load.
Modern appliances such as battery chargers,
personal computers, monitors, printers and
televisions require a small amount of bias power
so that they can quickly spring to life, perform
their tasks and slip back into their idle state
without operator intervention. Previously, idle
state power was simply eliminated with an AC
line switch; now, many modern systems must
have regulated power available at all times.
During idle states, the systems enter sleep-
modes where they impose micro-watts of load
on the DC converter output. During idle state, the
seemingly small operational and parasitic losses
in the power supply are the most significant loads
that are ultimately imposed on the AC line.
During full load operation, the mode of control
must change in order to efficiently meet the
power demand.
The power supply must have the ability to
change modes of operation based on load in order
to maximize efficiency over a broad load range.
Often, the task of coordinating the different
modes falls upon a primary-side controller, such
as the UCC28600.
Furthermore, the power system architecture
may need a small power supply to coordinate
larger internal power supplies in order to reduce
no-load power. Intuitively, it is easier to make a
50-W power supply have an idle power less than
500 mW than it is to make a 500-W power
supply have an idle power less than 500 mW.

2-2
MAIN SUPPLY
150 W < P
LOAD
ON/OFF
V
OUT _1
V
OUT _N
V
IN
CONTROL
1
2
AUXILIARY
SUPPLY
IDLE
LOGIC
V
OUT _1
V
OUT _K
V
IN
CONTROL
1
2
LOAD
ON/OFF
IDLE REQ
PFC
CONTROL
BIAS
1
PRIMARY
SECONDARY
0 W < P < 75 W

Fig. 1. High-power system architecture for green mode.
V
OUT _1
V
OUT _K
V
IN
CONTROL
1
2
PFC
CONTROL
BIAS
1
PRIMARY
SECONDARY
LOAD
MAIN SUPPLY
75 W < P < 150 W

Fig. 2. Mid-power system architecture for green mode.
2-3
V
OUT _1
V
OUT _K
V
IN
CONTROL
1
2
PRIMARY
SECONDARY
LOAD
MAIN SUPPLY
0 W < P < 75 W

Fig. 3. Low-power system, the main supply has green mode.

Power system architecture for power levels
that have more than 75 W capacity often include
multiple power processors such as, a Power
Factor Corrector (PFC), an isolated auxiliary
supply, an isolated main converter, etc., as shown
in Fig. 1. During idle periods, all of the supplies
are turned-OFF except the isolated auxiliary
supply that maintains the minimum amount of
system functionality in order to properly revive
the power system. There is a mid range of power
supplies, shown in Fig. 2, that have large enough
loads to require PFC, yet they do not merit the
complications of the high-power systems. We can
address the mid-range converters with the power
system architecture that is shown in Fig. 2.
Lower power level systems (below 75 W) can be
implemented with single converter that can
provide true regulation from 0 W to full load,
such as Fig. 3. The isolated auxiliary supply in
high-power systems and the single converter in
low-power systems must both have a feature that
enables them to maintain control at zero load
power while drawing minimal power from the
AC line. We call this feature Green Mode.
One of the typical green-mode specifications
is a maximum no-load power, usually below
500 mW. Unfortunately, it is difficult to predict
the no-load power of a converter because of the
myriad of interdependencies and non-linear
switching losses. As of this writing, it is not
possible to directly use a predefined no-load AC
line power as a direct design parameter for power
supplies; we must simply take the best measures
that we can during the design process and be
prepared to iterate the design.
The recurring element in the power systems
of Fig.1 and Fig. 3 is a small (usually less than 75
W, perhaps as large as 150 W), wide input line
converter. The flyback topology is a popular
solution for this kind of application. It can be less
expensive to add a low power green-mode
converter to a high power system than to try to
make a high power converter meet a small no-
load AC load specification. If the controller
includes a feature that detects when the converter
is operating at extremely light loads and uses that
information to turn-OFF a PFC and it offers
quasi-resonant full-load control, the flyback
topology can be extended to be sufficient for the
75 W-150 W range, in Fig. 2.
2-4
II.

M
ULTI
-M
ODE
F
LYBACK
C
ONTROL

The flyback topology has a minimum number
of components while providing galvanic isolation
for the outputs. Input ripple current, output ripple
current and leakage inductance energy usually
limits the practical power range. The flyback
converter is also popular in the role of an isolated
auxiliary supply due to power level and the low
cost of having multiple outputs. Keep in mind
that a multiple output topology will have poor
cross regulation during idle states due to the
power-saving Discontinuous Conduction Mode
(DCM) operation.
There are a variety of operational modes that
are compatible with the flyback topology; fixed
f
S
Continuous Conduction Mode (CCM), fixed f
S

Discontinuous Conduction Mode (DCM), Quasi-
Resonant (QR) Mode, constant-on time
Frequency Fold-back Mode (FFM) and hysteretic
burst mode. Controllers that are suited to
simultaneous wide-line and wide-load operation
often employ at least three of the aforementioned
modes, if not four of them, so that the converter
can maintain high efficiency at any rated
operating point.
A.

QR and DCM Modes
During heavy load operation, the practical
choices are CCM, DCM or QR modes. Reverse
recovery losses from the output rectifier and
stability issues often eliminate the CCM option.
Operation in DCM is limited by turn-ON loss and
high RMS currents in the primary side. The QR
mode offers the lowest turn-ON loss because the
switching event occurs at the valley of the
resonance that occurs after the flyback
transformer is de-energized, as illustrated in
Fig. 4. In DCM operation, the turn-ON loss can
be as large as ½C
D
(V
1
+ nV
2
)
2
Joules, depending
on line and load conditions. In QR operation, the
turn-ON loss is ½C
D
(V
1
- nV
2
)
2
Joules for valley
switching and 0 Joules for Zero Voltage
Switching (ZVS).
V
GS
I
M
I
PRI
V
DS
t
L
M
C
D
Resonance
V
1
L
L
C
N
Resonance
I
SEC
At Rated Load
½ C
D
V
DS2
C
2
C
N
R
N1
C
D
R
N2
D
N
D
2
L
L
I
PRI
L
M
n 1
I
SEC
I
1
+
-
+
-
+
-
+
-
V
N
V
DS
V
1
V
2
V
GS
+
-

Fig. 4. Flyback converter (top), QR waveforms
(bottom).
2-5
Why should we settle for valley switching;
why not choose a transformer turns ratio that will
always operate in the QR mode with ZVS?
Unfortunately, the MOSFET stress voltage would
be at least twice the input voltage at all line
conditions. Operation at the maximum universal
AC line voltage (265V
AC
) translates to a peak
operational stress voltage of at least 748 V
DC
. As
of this writing, MOSFET switching devices with
sufficient voltage ratings and low R
DS(on)
are not
economical.
Valley switching saves nearly 2 W of loss for
a 65-W test converter with a 600-V MOSFET
and a reflected secondary voltage of 100 V.
Furthermore, the output rectifier commutates-
OFF when it has zero current. Thus, the output
rectifier does not incur reverse recovery losses.
So, QR control with valley switching has merits
over fixed frequency CCM and DCM designs.
The price for saving the MOSFET turn-ON loss
and the output rectifier reverse recovery loss is
that the switching frequency must vary in order
to maintain valley switching.
What happens to the switching frequency of
the QR flyback converter when the input voltage
changes or the load changes? In general, the
switching frequency increases with increasing
input voltage and with decreasing load, as shown
in Fig. 5. Non-linear steady state behavior of the
converter leads to complications in quantitatively
determining the steady state peak current, I
P
, and
switching frequency, f
S
.
0.0
I
PR
I
-
P
eak
P
r
i
m
a
r
y

Cu
rr
en
t -
A
I
PRI
vs f
S
MAP
50 k
f
S
- Frequency - Hz
1.5
2.5
4.0
0.5
1.0
2.0
3.0
3.5
70 k
90 k
110 k
130 k
150 k
P = 5%
40%
Load
75%
Load
95%
Load
235 V
DC
374 V
DC
High Line,
Variable Load
96 V
DC
Low Line,
Variable Load
Constant Load at 100%,
Variable Line

Fig 5. QR flyback operating plane.
The form of the steady state relationship for a
Q