When operating a switching converter from the wall outlet, part of the absorbed energy feeds the load while the rest is lost and dissipated in heat. To get an idea of the loss amount in a given application – assume a set-top box adapter – simply place your hand on the cover while it operates and feel the heat: a warm enclosure implies significant losses while a cold case (no pun intended) characterizes a highly-efficient power supply. If you now press the standby button, you expect the box to cool down after several minutes. However, in some cases, you can still feel a warm enclosure, telling you that losses still exist despite standby.
Losses in switching converters have been the subject of numerous publications. In this short paper, we will only concentrate on the secondary side shown in Figure 1 for a typical converter.

Figure 1 - In most of today’s converters, control is located in the secondary side. Here a TL431 hosting a compound error amp and a reference voltage.
In this application circuit, you recognize a TL431 whose reference pin receives a portion of the variable you want to monitor, Vout. The TL431 cathode current I1 is adjusted in relationship to the load current so as to maintain a constant output voltage. The goal is to modify the primary-side feedback voltage shown in Figure 1’s left side via a current optically transmitted by the optocoupler. This feedback voltage will ultimately set the converter peak current in a current-mode converter. The optocoupler is affected by a current transfer ratio (CTR) defined as
(1)
where IC is the collector current and IF the LED forward current.
A TL431 aggregates 11 bipolar transistors to form an op amp and a 2.5-V reference voltage. The part directly operates from the current absorbed by the cathode. This current must be 1 mA minimum as stated in the component data-sheet. This 1 milliamp comes on top of the feedback current necessary to produce IFB in the primary side. Capitalizing on the LED 1-V forward drop, resistor Rbias builds a cheap current generator with the optocoupler LED. If Rbias is 1 kΩ, then you roughly inject 1 mA. Assuming a primary-side pull-up resistor RFB of 20 kΩ and an internal Vdd of 5 V, the maximum current needed to move the feedback voltage between 5 V and 0.3 V (the optocoupler saturation voltage) is found to be
(2)
Considering a 25% CTR, this collector current translates into a LED forward current of
(3)
Adding the 1-mA extra bias current provided by RFB, you total a current of 1.94 mA, drawn from the output. If this output is 20 V (typical value for a notebook adapter), you permanently dissipate
(4)
Compared to the power delivered by a typical adapter (45 to 65 W), it may not be regarded as a tremendous amount of waste. However, when you struggle to limit the no-load power below the 100-mW barrier, you chase every mW you can save. In (4), part of the result is the 1-mA bias current brought by RFB. This current is important to ensure a proper TL431 bias at full power (high dc gain) but in no-load standby, we could think of a way to remove it.
Figure 2 shows you an ON Semiconductor proprietary circuit that nicely biases the part at high power but as the converter starts operating in light load, the bias gradually decreases to completely disappear in standby.

Figure 2 - A simple peak rectifier helps removing the bias current in a no-load situation.
The peak detector built around D1, C1 and Rbias ensures a full bias at high power where the voltage across C1 equals Vout. When the controller enters skip cycle mode, the peak voltage on C1 remains the same but the valley voltage drops owing to the C1Rbias time constant. The current in Rbias reduces to fully disappear when a deep standby mode is entered: you save 20 mW from (4)’s total.
CTR and Secondary-Side Power
As indicated by (3), the primary-side current reflects to the secondary via the optocoupler CTR. Unfortunately, to reduce losses in standby mode, the feedback resistor RFB is usually kept high (20-40 kΩ) to force a low collector current. In this mode, the optocoupler exhibits a poor CTR, often as low as 25%. The drawback of a low CTR is seen as extra dissipated power in the secondary side, especially at a high Vout as indicated by (4). In this case, why not select higher CTR optocouplers? A popular reference such as SFH615 comes in different CTR grades. For instance, the -2 shows a CTR varying from 63 to 125% at a 10-mA current. If you select a -4, the CTR now varies between 160% and 320%. Even if at a low feedback current you won’t get a 160% CTR, it is likely that the LED current is significantly lower than the -2 version. Measurement on the board with -4, shows a CTR of 43% versus 25% with the -2. Considering the disappearance of the TL431 bias current in standby, the secondary-side power drops
(5)
which is a quarter of the power brought by (4). Higher CTRs can be selected and a possible option is the FOD817S that exhibits a CTR of 75% even at low collector currents. In this case, the LED bias drops to 313 µA, inducing a power loss of
(6)
The selection of high-CTR optocouplers is not trifling. An optocoupler is affected by parasitic elements, in particular a capacitor seen between collector and emitter. This capacitor together with RFB creates a low-frequency pole that can affect the converter phase margin if crossover is high. It is important to characterize this pole and quantify its impact on stability. Its position varies in relationship to the CTR grade. If you want high-bandwidth, select low-CTR optocouplers. Standby power will suffer but the pole position should be less of a problem. On the opposite, high-CTR optocouplers are good for low-power systems but you may have to select a low crossover frequency so that the pole does not hamper stability. In any case, if you decide to replace your optocoupler by a higher-CTR type, make sure to measure the converter’s dynamic response again to check for any possible troubles.
The below numbers show the no-load standby power performance improvement brought by upgrading the optocoupler CTR. The board uses the new low-voltage NCP1256 controller which includes brown-out sensing. All measurements are carried at a 230-V rms input voltage.

Solutions exist to lower a converter standby power when operated in a no-load condition. Increasing the voltage divider sensing network value is one obvious path but reducing the TL431 bias current or improving the optocoupler CTR are simple to implement rewarding options.