Chapter 2 Conventional Isolated Feedback Network and Previous Researches….7
2.4 Conclusion
The conventional feedback network is widely adopted in isolated offline
switch-mode power supplies in industry owing to the benefits of the simple circuit
structure and low cost. However, the power loss under very light/no-load conditions
makes it become an obstacle for designers to pursue low standby power. Besides,
although the compensator of the conventional feedback network suffices the needs in
most applications, there exists a minimum midband gain limitation on building the
type-2 compensation, which degrades the flexibility in system design.
Previous literature provides some techniques to address the power loss issue.
Primary-side control removes the entire traditional feedback network and features low
standby power and low cost. Nonetheless, a primary-side-control converter has a poor
output regulation and is mainly limited to the discontinuous-conduction mode due to the
difficulty in extracting the output voltage. Furthermore, the necessity of a dummy load
at the output increases the power loss. The output voltage control technique and the
feedback impedance modulation offer controller solutions without any change in the
system circuit. Under very light/no-load conditions, the output voltage control technique
regulates the supply voltage of the controller to a lower value for dropping down the
output voltage. Although ILED and IFB are reduced, the current flowing through RP is as
large as before. Besides, controlling the output voltage indirectly by regulating VCC will
give rise to a poor output regulation. The feedback impedance modulation switches the
feedback impedance to a high value under very light/no-load conditions. The current
flowing through RP is truly reduced, and the output voltage is kept regulated. However,
both the feedback impedance modulation and the output voltage control technique
require a long recovery process to avoid the instability when the load changes from a
light to a heavy one, leading to a long transient response time.
Chapter 3
Low-Standby-Power Output Feedback Scheme
3.1 Introduction
The proposed output feedback scheme with low standby power consumption is
introduced in this chapter. Beginning from the central concept that originated from the
previously described knowledge, the following sections explain the whole thoughts on
the design of the proposed feedback network circuit step by step. After that, the analyses
of the power loss and the compensation method are carried out and compared with those
of the traditional feedback network. Finally, comparisons with prior techniques are
provided to reveal the superiority of this solution.
3.2 Phase Reversal Concept
In Section 2.2.4, we have disclosed that the power loss of the conventional
feedback network increases with the decrease of the output power and it comes to a
maximum value while there is no output load applied. Although the power consumption
can be reduced by existing approaches [55], [56] under very light/no-load conditions, it
is still not minimized and those techniques need to sacrifice the light-to-heavy-load
transient response time for ensuring the system stability. These facts thus motivate us to
think over the possibility of a better solution.
To overcome the difficulty suffered by the conventional feedback topology, we first
review the relationship between VFB and the output power, as the gray line shown in Fig.
3.1. Two major problems can be pointed out. One, the positive correlation between VFB
and the output power is unfavorable for pursuing a low standby power. Two, the current
for supplying the shunt regulator causes a minimum voltage drop across RP and leads to
irremovable losses on both the primary and the secondary sides. In consideration of
these, we come up with a helpful solution to reverse the situation. The fundamental
concept is that, as shown in Fig. 3.1, we flip the conventional VFB curve vertically. The
resulting new VFB curve (black line) suggests that a higher VFB should correspond to a
lighter load, making the power loss due to ILED and IFB decrease with the output power.
That is, the phase relationship between ILED (or IFB) and the output power is reversed. In
this way, the power consumption naturally shrinks to a minimum value when no load is
presented at the output. Besides, to eliminate the voltage drop IFB,min.RP caused by the
supply current of the shunt regulator and in the meanwhile to prevent that current from
causing power loss on the primary side, we should keep the supply current from flowing
through the optocoupler.
3.3 System Architecture
For realizing the proposed concept, the whole feedback circuit should be thought
over. Starting from the circuit on the secondary side to the primary side, the circuit
Fig. 3.1. VFB versus the output power in conventional and proposed networks.
design considerations are fully uncovered in the following subsections.
3.3.1 Secondary Side
To make VFB increase with the decrease of the output power, we should try to
reduce IFB when the output demands less energy such that the output voltage starts
growing. This means the desired correlation between IFB and the output voltage should
be reversed compared to that in the conventional circuit. With this observation, the first
thought coming to our minds is that it can be easily achieved by simply interchanging
the positive and the negative terminals of the operational amplifier in the shunt regulator,
as shown in Fig. 3.2(b). For reference, the conventional circuit is also illustrated in Fig.
3.2(a). However, the configuration in Fig. 3.2(b) makes it not a negative feedback path
from V2 back to V1. The Miller capacitor CC thus can not be placed between the two
nodes to build a dominant pole. A compromise is to connect the capacitor between V3
and V1; however, not only an additional pin is required but also the output current
capability of the operational amplifier should be quite enhanced for ensuring a sufficient
(a) (b)
(c)
Fig. 3.2. (a) Conventional secondary-side feedback circuit and (b)(c) two modifications.
slew rate. These outcomes make this solution not really fascinating at all. Instead of
interchanging the two input terminals of the operational amplifier, another choice is to
replace the npn transistor with a pnp type. In Fig. 3.2(c), the pnp transistor
fundamentally acts as an emitter follower, and now V2, V3, and V1 all have the same
phase polarity. The Miller compensation therefore can not be built between any two of
the three nodes, which discourages the feasibility of this idea.
Through the two previous trials of circuit modification, we get to understand that
the expected relationship between ILED and the output voltage will destroy the negative
feedback path from V2 to V1. The reason is that ILED flows through RLED before through
the current-controlling transistor. With this recognition, we then attempt to change that
sequence and offer two possible structures drawing in Fig. 3.3. In both Fig. 3.3(a) and
(b), VOUT is divided by R1 and R2 to VOF, and an operational amplifier (functionally an
error amplifier) amplifies the difference of VOF and a reference voltage to drive a
current-controlling device. Note that here we use a MOS transistor rather than a bipolar
transistor as the current-controlling device just because we will apply a CMOS process
technology to implement chips. In Fig. 3.3(a), the nMOS MN works as a source follower,
while the pMOS MP in Fig. 3.3(b) is configured as a common-source amplifier. An
optocoupler and a resistor RLED are placed in series between the current-controlling
device (MN or MP) and the ground. Functionally speaking, both of the two possible
structures meet our need for a negative correlation between VOUT and ILED. The major
difference between the two circuits in Fig. 3.3 and the three in Fig. 3.2 is that the
current-controlling devices in Fig. 3.3 directly connect to VOUT to serve as current
sources rather than current sinks. As a result, when VOUT and thus VOF increase, MN in
Fig. 3.3(a) or MP in Fig. 3.3(b) lowers down ILED and V2 also drops. This represents that
the path from V2 back to VOF is a negative feedback where the Miller compensation can
be constructed. Since both of the two circuits in Fig. 3.3 are feasible, what are exactly
(a) (b)
Fig. 3.3. Two practical secondary-side circuits with, respectively, (a) nMOS and (b) pMOS as current-controlling devices.
their merits and drawbacks in comparison with each other? In Fig. 3.3(a), the source
follower provides better noise immunity as VOUT would see a relatively large impedance
toward MN. However, the serious body effect of MN enlarges its threshold voltage badly.
Even though there is a device called isolated nMOS which enables its body and source
to be tied together, it consumes a relatively large area in comparison with a standard
nMOS. In view of this, we eventually make a choice of the pMOS as the
current-controlling device. Note that IQ in Fig. 3.3 is the current for supplying the
internal error amplifier and the voltage reference, and that current will not pass through
the current-controlling device. By separating it from ILED, the minimum values of IFB
and ILED are basically not limited.
3.3.2 Primary Side
Fig. 3.4(a) is the original primary-side feedback circuit. If we do nothing to it but
(a) (b)
(c)
Fig. 3.4. (a) Conventional primary-side feedback circuit and (b)(c) two modifications for the phase-reversal technique.
close the loop with the circuit shown in Fig. 3.3(a) or (b) adopted as the secondary-side
feedback circuit, what results will occur? When the output voltage drops and ILED and
IFB are increased, VFB is pulled down. Then, the modulator narrower the duty cycle of
output pulses, making even less power delivered and the output voltage continuously
drop. Similarly, when the output voltage increases, the modulator outputs even wider
pulses to raise the output voltage even more quickly. In other words, the system forms a
positive feedback loop and in the end leads to a malfunction.
The problem lies in the feedback signal polarity. Because we have reversed the
VOUT. That is, VFB will grow up with VOUT, and vice versa. Therefore, using a typical
PWM modulator to directly modulate this VFB into switching pulses simply causes a
positive feedback loop. A simple solution is to just put an inverting amplifier with a unit
gain before the modulator, as shown in Fig. 3.4(b). It helps reverse again the phase of
VFB before it is used for the following modulation. The output voltage VRFB of the
inverting amplifier will thus vary in the opposite direction of the way that VFB goes. By
doing so, the negative feedback loop is maintained. Fig. 3.4(c) provides an alternative
way where the optocoupler is connected between the supply voltage VCC of the
controller and VFB. Instead of sinking current, the optocoupler in Fig. 3.4(c) sources the
induced current equal to ILED multiplied by CTR to RP. As a result, the phase of VFB is
also reversed compared to that in Fig. 3.4(a), and this VFB can be directly modulated.
Then, which solution of Fig. 3.4(b) and (c) is better? The circuit in Fig. 3.4(c) is merely
a rearrangement of the circuit in Fig. 3.4(a), and it seems there is no more additional
effort should be paid for it. However, we can think that when the output voltage of a
converter is still lower than the desired value, ILED will be at its maximum value which
is mainly determined by the present output voltage and RLED. If ILED is not properly
limited, VFB in Fig. 3.4(c) has a possibility of exceeding the internal supply voltage VLO
which is generated from VCC for supplying low-voltage-rating devices. In view of this
concern, circuits inside the controller that would be connected to VFB should be
designed using high-voltage-rating devices, but the consequence is the largely increased
area consumption. The circuit in Fig. 3.4(b) is free from this issue, although it requires
an extra inverting amplifier and thus causes more power loss. Fortunately, unlike the
operating currents of the optocoupler, this inverting amplifier which is just used to
process an on-chip low-speed signal needs only few tens of microamperes. Hence, we
choose it as the preferred solution.
3.3.3 Overall System
The proposed complete feedback scheme applied to an isolated switch-mode power
supply is shown in Fig. 3.5, and two examples of a flyback and a forward converters are
presented in Fig. 3.6. In these implementations, a secondary-side integrated circuit is
substituted for the traditional shunt regulator. It pulls down ILED according to the
difference between VOF and the built-in reference voltage. The higher VOUT is, the
smaller ILED will be conducted. As its operation is reversed compared to the traditional
shunt regulator which will draw a larger ILED with a larger VOUT, we call it the
reverse-type shunt regulator (RTSR). Note that the supply current IQ will not flow
through MP and is not contained in ILED. On the primary side, the only difference in the
controller is that an inverting amplifier is presented before the PWM modulator. Other
off-chip components, including RLED, CP, RC, and CC, are added for implementing a
frequency compensator, which will be described later.
Fig. 3.5. The proposed complete low-standby-power feedback network.
The proposed feedback network basically performs the same function as the
conventional one does, but the key point is that the phase of the intermediate error
signal for optical coupling is reversed. With this proposed feedback scheme adopted, a
higher VFB, which gives a lower VRFB, will correspond to a higher VOUT, and therefore
losses due to ILED and IFB will automatically reach minimum values under the no-load
condition. Concerns may be aroused that whether or not the additional power losses
caused by the inverting amplifier and IQ surpass the saved power under the no-load
condition. As previously mentioned, the current consumption of the inverting amplifier
can be designed to be only a few tens of microamperes. Also, the supply current of the
shunt regulator is not contained in ILED, and thus the minimum values of ILED and IFB for
operating are essentially not limited and can be designed to be very small. With these
two features, the power loss of the feedback network under the no-load condition can be
(a)
(b)
Fig. 3.6. Proposed feedback network adopted in (a) flyback and (b) forward topologies.
minimized. In the following sections, we will present the power loss analysis as well as
the control loop compensation analysis.
3.4 Power Loss Analysis
As what we have done for the conventional feedback network in Section 2.2.4, we
also want to formulate the power loss that is associated with the proposed feedback
circuit. First, we can recognize from Fig. 3.5 that there are five current branches. The
first one is the current flowing through the voltage divider. The second one is IQ, which
is consumed by the error amplifier and the voltage reference in the shunt regulator. The
third one is ILED, which is conducted by MP and the optocoupler. The fourth and fifth
ones are respectively IFB and the current dissipation of the inverting amplifier. Since
there is only a slight power consumed by the inverting amplifier, we directly denote it as
PIV for convenience. Thus, if we first make an assumption of ideal energy conversion,
the power loss (PL,pro.) of the entire feedback network can be estimated by
IV
From observing (3.3), we see that the second term is the power loss caused by currents
flowing through the optocoupler on the primary and the secondary sides. For a
well-designed power converter, this part of loss will vary with VFB, which is determined
by the present load condition. Equation (3.3) is a simplified general estimation for any
transformer-isolated converter adopting the proposed feedback network. If we solely
consider a flyback converter, as shown in Fig. 3.6(a), equation (3.3) can be further
When operating under the no-load condition, converters generally adopt the burst
mode control to regulate their outputs [39], [40]. As previously mentioned in Section
2.2.4, for a conventional PWM controller, it will start using the burst mode to control
the system when VFB is lower than a threshold voltage [57]. This mechanism is
inappropriate for the proposed feedback topology in which, as shown in Fig. 3.1, VFB
increases with the decrease of the output power. Under this circumstance, the burst
mode threshold voltage VBU should be set close to VLO, and the burst mode control
should be activated when VFB is larger than VBU. Fig. 3.7 illustrates simulated
waveforms of VFB and the gate-driving signal VG in the burst mode under the no-load
condition. In this case, VBU is set 4.5 V while VLO is 5 V. The driving signal VG is
Fig. 3.7. Simulated burst-mode waveforms with proposed feedback topology adopted.
minimum values of ILED and IFB are not limited in the proposed feedback scheme) and
therefore the loop response is very slow under the no-load condition, VFB can be
designed deliberately to touch and stay at VLO in the period between the bursts. In this
way, the optocoupler actually conducts zero currents on both sides in that duration.
Because the only current dissipation at the output node comes from the resistor divider,
the switching-ceased period is relatively long. Thus, under the no-load condition, (3.3)
and (3.4) can be respectively approximated as
IV current of the inverting amplifier is 25 μA. The second and the third terms in (3.8)
together add up to merely 3.4 mW under the condition that VOUT = 12 V, VCC = 10 V,
and VD1-D2 = 0.5 V, making PL,pro. mainly dominated by the power consumption of the
resistor divider only. Recall that, in Section 2.2.4, the conventional feedback network in
a typical flyback converter having the same conditions consumes a power of 23 mW
excluding the part of the resistor divider. Comparing the power losses of the
conventional and the proposed feedback circuits, we can find that a power of 19.6 mW
can be saved by simply applying the proposed feedback scheme. It should be noted that
here we do not consider switching losses for simplicity. The estimated saved power is
thus underestimated, which will be discussed more in Chapter 4.
3.5 Control Loop Analysis
The compensation design in the control loop must be considered for achieving a
stable converter system. The proposed feedback network is provided with a
compensator which is very similar to that in the conventional one. Fig. 3.8 shows the
small-signal equivalent circuit of the proposed feedback network from VOUT to VRFB (the
output voltage of the inverting amplifier). The reverse-type shunt regulator is regarded
as a voltage-controlled current source with a transconductance Gmrv, while the
optocoupler is treated as a current-controlled current source with a current gain of CTR.
Again, the internal pole of the optocoupler is considered by including COPT, and the
dynamic resistance of the light emitting diode is omitted from the following analysis as
it is much smaller than R .
Fig. 3.8. Equivalent circuit for small-signal analysis.
Observing Fig. 3.8, we recognize that
LED
(
P OPT)
we arrive at the final transfer function
(
1)(
12)
These results show that the proposed network still exhibits a two-pole one-zero
characteristic, which can be easily utilized for the type-2 compensation. Moreover, the
magnitude of G0 and positions of the two poles are basically the same (i.e., determined
by the same parameters) as their counterparts in the conventional feedback network.
The only difference is that the zero in the conventional topology is at about 1/R1CC,
whereas in the proposed topology, we need to additionally place RC in series with CC to
intentionally create a negative zero given by (3.17).
RLED and the minimum midband gain issue of the conventional feedback network. Its
RLED and the minimum midband gain issue of the conventional feedback network. Its