Many power supplies, especially offline power supplies, require low standby power. The most cost-effective isolated topology for power levels below 100 W is a flyback, because it requires the fewest components. Flyback converters often generate multiple secondary outputs, which require relatively precise regulation. This article will describe the challenges of achieving well-regulated output voltages while still achieving low standby power.

Low-power AC/DC flyback power supplies are widely used in industrial applications such as motor drives and appliances because they can achieve good voltage regulation and low standby power losses. A typical application for an isolated low-power design often requires more than one secondary output. Figure 1 shows an example of a flyback topology that generates the outputs VOUT1 and VOUT2 from a universal input (85 VAC to 265 VAC). Transformer T1 provides galvanic isolation between the AC power line (mains) and the loads. Auxiliary winding AUX powers the primary-side flyback controller.

Figure 1 The simplified schematic of a multiple-output flyback that provides galvanic isolation between AC power line and loads. Source: Texas Instruments

How to reduce standby power

Let’s briefly review the known techniques to reduce standby power. Standby power mainly depends on the cycle energy, startup circuit, snubber network, and minimum load requirement. Reducing the no-load switching frequency and using active startup circuitry and a Zener snubber network instead of a resistor-capacitor-diode snubber leads to lower standby power. Unfortunately, other circuit properties can increase standby losses as well. So, it’s helpful to develop a strategy in advance that will help keep the standby power low.

One of the main challenges for a power-supply designer is that it’s impossible to build an ideal circuit, as any real board must deal with parasitic capacitances and inductances, as well as with noise in the system.

These challenges become even greater when generating two or more isolated outputs, as shown in Figure 1. Normally, a voltage control loop regulates only one output; the coupling of the transformer windings semi-regulates the other output. Figure 2 shows the regulation of one output. An external error amplifier (U1) connects to output VOUT2 through a resistor divider (Rhigh1, Rlow1). An optocoupler helps transmit the error signal to the primary side.

Figure 2 The schematic of an external error amplifier connected to VOUT2 shows the regulation of one output. Source: Texas Instruments

The other output, VOUT1 (3.3 V), is only semi-regulated because of the coupling of the transformer windings. But what happens during standby mode with a light or no-load condition? To answer this question, consider Figure 3, which shows the secondary winding voltages—also called secondary switch nodes—of VOUT1 (3.3 V) and VOUT2 (12 V).

Figure 3 Overshoot of the secondary-side switch nodes can be a challenge at light or no-load conditions. Source: Texas Instruments

You can easily recognize the overshoot, followed by ringing after the end of the on-time. Basically, the overshoot of the primary switch node reflects to the secondary side. This overshoot can be a challenge at light or no-load conditions, especially for an unregulated output, because it charges the output capacitance through output diodes D1 and D2, as shown in Figure 1. The overshoot can cause the unregulated output voltage to rise to very high values.

What is the main cause of unintended overshoot and ringing? It’s the parasitics of the power stage and board, including the leakage inductance of the transformer. The leakage inductance is caused by the magnetic flux from one winding in a transformer that does not couple to other windings. This energy dissipates externally to the transformer and an overshoot occurs. Figure 4 shows the primary switch-node voltage, which is basically drain-to-source voltage of the metal-oxide semiconductor field-effect transistor (MOSFET).

Figure 4 The primary switch node is MOSFET’s drain-to-source voltage. Source: Texas Instruments

Influence of transformer leakage inductance

Now that you’ve seen how overshoot can have a detrimental effect on cross-regulation for light loads, the question arises: Why don’t you just clamp it strongly? Typically, a snubber clamp circuit limits the overshoot voltage to a certain level. The clamp circuit absorbs energy stored in the transformer’s leakage inductance and, depending on the value of the clamp voltage, will also absorb a fraction of the magnetizing energy. The energy lost in the clamp increases rapidly as the clamp voltage drops.

Because of the high energy losses, you must allow for a certain switch-node voltage overshoot. The minimum overshoot depends mainly on the leakage inductance. With an existing transformer, it’s not possible to clamp the overshoot to every intended level. You have to think about an optimized transformer structure before ordering a custom transformer sample. The goal should be to minimize the ratio of the leakage to the magnetizing inductance.

The leakage inductance depends largely on the physical winding geometry. In general, two changes will reduce the leakage inductance: decreasing the dielectric spacing between the primary and secondary windings and increasing the surface area of overlap between them. Thus, using an interleaving winding structure and a wider bobbin and moving the layers further together will lead to low leakage inductance. Unfortunately, there is a trade-off. These changes usually involve increasing the parasitic interwinding capacitance, which increases common-mode electromagnetic interference. Therefore, you should work closely with a transformer manufacturer right from the start in order to find an optimized transformer structure.

Now, let’s look again at the design generating two outputs: 3.3 V (VOUT1) and 12 V (VOUT2). Some applications require tighter regulation of the lower output voltage because it typically requires a smaller tolerance. Assume that VOUT1 (3.3 V) will be regulated and the higher output voltage VOUT2 (12 V) will remain unregulated. Therefore, VOUT1 is regulated to 3.3 V, while the turns ratio of the transformer winding determines VOUT2. This configuration can work well for a system with low parasitics, including low leakage inductance, even at light loads.

However, if the leakage inductance is large, the coupling of the windings is poor and the overshoots are large, then the cross-regulation is no longer good because the transformer winding voltage ratio is no longer directly proportional to the winding turns ratio. As a result, VOUT2 can rise very quickly, easily becoming twice as large as the intended level or even larger. A resistor or Zener diode would limit the voltage but also significantly increase the standby power. So, you’ll need to consider other possibilities.

Therefore, instead of regulating the lower output voltage, it might be helpful to regulate the higher output voltage VOUT2. If the unregulated output VOUT1 typically does not exceed the value of VOUT2, in principle the low-voltage output could at most reach the level of the high-voltage output. This means that in some cases, it’s advantageous to regulate the higher voltage because doing so will maintain a lower absolute maximum voltage in the system.

As always, there is a trade-off because the regulation of the unregulated output will be worse. A compromise would be to regulate both outputs at the same time, as shown in Figure 5. This works well as long as you do not require isolation between the outputs, but has a disadvantage because it becomes impossible to regulate any output with very high precision.

Figure 5 The schematic shows an external error amplifier connected to VOUT1 and VOUT2. Source: Texas Instruments

Another alternative would be to take the inner loop—which connects to the anode of the optocoupler—from one output and the outer voltage loop from the other output, as shown in Figure 6, to achieve precise regulation of VOUT2 and somewhat improve the regulation of VOUT1.

Figure 6 This is how an external error amplifier is connected to the inner loop and VOUT2. Source: Texas Instruments

Because the final regulation depends heavily on the parasitic capacitances and inductances of the power-stage components and layout, evaluating the alternatives in the lab is recommended.

Modern flyback controllers can achieve very low standby power because a pulse-width modulation algorithm varies both the switching frequency and primary current while maintaining discontinuous conduction mode. This algorithm reduces the switching frequency and peak current for light loads. With a modern flyback controller, it’s even possible to achieve standby power of less than 20 mW for certain applications. When designing a power supply, however, it is essential to avoid causes that increase power dissipation.

To achieve low standby power, it’s essential to reduce the energy taken every cycle from the input by lowering the switching frequency and primary peak current using an active startup circuit and reducing the secondary-side pre-load resistors. A good layout also reduces noise in the system, while suitable snubber networks for the primary and secondary switching nodes can further reduce noise and overshoots. Finally, don’t neglect the transformer; other than the controller, it’s the most important part of the power supply.

Florian Mueller, power systems engineer at Texas Instruments, has authored this Power Tips article #105 for EDN.

Isn’t it better to combine both secondary windings and use a tap for 3.3V ? Then the coupling is for sure 100% between the secondary windings up to the 3.3V tap.

You must Sign in or Register to post a comment.