Feedback Impedance Modulation

Very recently, in [56], another technique called feedback impedance modulation

was revealed to address the power loss issue of the conventional feedback network.

Similar to the output voltage control described in Section 2.3.2, this technique provides

a controller solution without the need to modify the system circuit. It proposes

increasing the value of the resistor which is connected to the collector of the

phototransistor inside the optocoupler only under very light/no-load conditions, and the

operating current will be reduced to an extent. Fig. 2.16 shows the proposed feedback

circuit in [56]. VFB should still be connected to the phototransistor collector of an

optocoupler. A resistor string composed of R1-RN with each resistor in parallel with a

separate switch S1-SN is in series with a fundamental feedback resistor RP, and each of

the switches is controlled by an individual digital signal from a counter. A comparator

Fig. 2.16. Primary-side feedback circuit with impedance modulation.

with a hysteresis window compares VFB with a threshold voltage VT to monitor the

output load condition, and its output is sent to the counter. VT is also the burst mode

threshold voltage of the controller. An oscillator provides a clock signal for the counter

and the modulator with its frequency fSW controlled by VFB, and the PWM modulator

generates pulses for gate driving.

When a converter adopting the feedback impedance modulation technique operates

under a heavy-load condition, VFB is higher than VT and all the switches S1-SN are closed

to bypass the resistor string. The feedback impedance remains only RP, which is set the

same as that in the conventional feedback network. However, when the load varies to a

very light/no-load condition, the counter resets all its output signals (VS1-VSN) to zero as

soon as VFB drops below VT. Now, the total feedback impedance ZFB becomes

This abruptly increased feedback impedance will likely let VFB fall down to well below

the burst-mode threshold voltage (VT) because ILED has not been changed at that instant.

Then, the output voltage starts dropping a little bit due to the insufficient power delivery.

Eventually, it will settle down to a level where the corresponding VFB is approximately

equal to the burst-mode threshold voltage since the converter basically operates in the

burst mode. In overall, the enlargement of ZFB makes it much easier to drop down VFB

with only little IFB, and the resulting effect is that the output voltage drops a little (but it

is still regulated) to lower down ILED and thus IFB. ZFB given in (2.25) can be chosen

such that the minimum supplying current ILED,min. for the shunt regulator is sufficient for

pulling down VFB. Therefore, the power loss of a flyback converter described in

equation (2.21) under very light/no-load conditions can be estimated as

⎟⎠

which indicates that the second term is minimized under such traditional system circuit

structure. Now, if the output load increases, VOUT drops and VFB rises to exceed VT. The

counter starts gradually bypassing the resistor string as soon as the comparator changes

its output state. Why can not all the resistors R1-N get shorted at a time? Because if we

do that, VFB will rush quite highly and the PWM modulator will widen the pulse duty

cycle rapidly. Then, the resulting a large amount of energy is poured to the output,

which will probably make the controller enter the light-load operation mode again, and

the reciprocating between the light-load and the heavy-load operation modes leads to an

instability phenomenon in the end. In view of this, switches S1-N in Fig. 2.16 will be

closed in sequence.

Fig. 2.17 summarizes the complete operating procedure with practical values.

When VFB is high, the switching frequency fSW is at its maximum value (say, 60 kHz)

and the feedback impedance ZFB is at its minimum value of 5 kΩ. If VFB reduces (but

still larger than VT), fSW may be also decreased with a minimum value of 20 kHz while

ZFB still remains at 5 kΩ. Once VFB drops below the burst-mode threshold voltage VT (a

small hysteresis window exists), the switching stops and ZFB is directly switched to a

maximum value of 50 kΩ. No switching action in the following time makes VFB

recovers. If VFB exceeds VT, the resistor string begins to be gradually bypassed. Unless

VFB falls below VT once again, ZFB will be decreased from 50 kΩ toward 5 kΩ with a step of 1 kΩ in every switching cycle. Fig. 2.18 shows the transient waveforms of VFB

and the gate-driving pulses in the burst mode. In the switching-ceased period, ZFB is 50

kΩ, while within the burst period every switching pulse is accompanied with a 1-kΩ

decrease of ZFB.

Compared with the output voltage control technique [55] which regulates VCC in

the light-load operation mode, a converter adopting this feedback impedance

modulation technique still controls the output voltage rather than other system variables.

Fig. 2.17. The change of ZFB and the relationship between switching frequency fSW and VFB.

Fig. 2.18. Waveforms of VFB and gate-driving signal in burst mode operation.

Therefore, the output regulation is largely improved. Besides, although both of the two

techniques [55], [56] aim at minimizing currents of the optocoupler, the current flows

through RP is in fact not reduced in [55], making the the reduction in power loss not

thorough enough. However, both of the two techniques would cause a risk of instability

when their systems go from light-load to heavy-load conditions. It is therefore that a

slow recovery procedure (i.e., the soft-start process in [55] or the gradual resistors

bypass in [56]) is necessary in both of the techniques, and this shortcoming brings about

the need to trade off the response time required for the transient from a very light to a

much heavier load.