It appears that there is a more and more push pull converter working in low power application. It seems that those converters are essentially working in open loop, i.e with no output voltage feedback. Here is an example:

enter image description here

It is probably well suited for input voltage already regulated as there is no output voltage feedback. Also the load have to be not really varying. The output voltage regulation is probably poor.

According to an article push-pull converter are recommanded for "medium power" application. I mean for power above 200 W.

enter image description here

What I do not understand is why use a push converter for those low power application rather than a forward converter? I understand that it is more complicate to predict the output voltage of a flyback converter without regulation? But a forward converter is pretty simple to predict? So what are the advantages to use a push-pull converter for those kind of applications?

Thank you,

  • \$\begingroup\$ There is nothing inherently stopping you from using a push-pull at very low power, but how come you can’t use regular feedback from the secondary side and a flyback? \$\endgroup\$
    – winny
    Sep 28, 2023 at 8:45
  • \$\begingroup\$ Thank you for your comment, my question is more why Texas Instrument has choosed to use push pull converter rather than a forward converter at such a low power application. TI has developped an IC for this specific purpose. Analog Device has done the same. \$\endgroup\$
    – Jess
    Sep 28, 2023 at 9:07
  • \$\begingroup\$ Simple at low input voltage and if you need to run it open loop without feedback. Please simulate it to better understand it. \$\endgroup\$
    – winny
    Sep 28, 2023 at 9:16
  • \$\begingroup\$ Push-pull uses the core more effectively because of AC magnetization. The forward works with DC magnetization. \$\endgroup\$ Sep 28, 2023 at 9:30

2 Answers 2


It's not a forward converter, it's a charge pump. Sort of. That's a bit of an abuse of terminology, but there is no closer, specific term for it.

To be clear, it's not a forward converter, because, while conduction occurs on the forward phase, there is no PWM control (or at least, no meaningful control). It's not strictly a charge pump, because there is no "flying" capacitor, and it's transformer-isolated rather than capacitor-coupled. I gravitate towards the "charge pump" side, personally, because the dynamics are closest to that. (It's also a "chopper", but that's generally anything that chops, not that it's used specifically for power conversion with a rectifier at the output. Maybe add the specifier, then: "chopper DC-DC"?)

Anyway, consider its operation. When one transistor is on, 5V is applied to one half of the primary, and presumably about 14V is developed on the secondary. C37 is charged rapidly through the rectifier, current limited by leakage inductance and switch resistance. The switch turns off, opposite side turns on, process repeats. Output voltage rises rapidly until settling at the ratio, and then sits there.

Operation is identical to a regular charge pump, where a capacitor is charged directly by a switch, current limited by its resistance. There's merely a coupling transformer added inline.

These converters are distinguished by poor efficiency under load, particularly startup and fault conditions, which makes them poorly scalable. The complete lack of output regulation (input and output are simply in ratio) is also a big disadvantage.

Regarding scalability, they tend to be small -- hence the SN650x for example. Probably the only large-scale application of such circuits is in automotive audio amplifiers. For example:

enter image description here
Source: Car Amplifier 1 x 120W + SMPS based on KAC-716 | electronica.mk

Typically, a parallel array of MOSFETs drives a toroidal transformer. (One each is shown here, for a relatively low power system, but kW+ amplifiers may use dozens.) Cheap old MOSFETs are used to increase robustness while keeping Rds(on) usefully low. The toroidal transformer is wound with little interleaving, so that leakage inductance dominates, providing the slightest amount of current limiting. Startup is achieved by slowly ramping PWM from an open-loop controller (typically SG3524, TL494 or KA7500); since the primary halves are not well coupled to each other either, most of the leakage inductance is discharged as flyback into the MOSFETs' avalanche capacity, hence the need for "robust" types. (Clamp diodes/snubbers are rarely if ever used. Older MOSFETs are typically rated for repetitive avalanche, IRFZ46N being a common choice.)

Needless to say, such applications utterly lack regulation and fault protection. Regulation isn't really a problem for an audio amplifier with reasonable PSRR, but protection is a problem, when an output transistor inevitably fails shorted from overheating, or a shorted speaker, and the fault cascades through the power supply, until finally the inlet fuse blows, leaving the whole unit a crater of burned semiconductors.

Many of these faults are readily cured by adding an inductor after each rectifier, turning it into a standard push-pull forward converter; preferably current-mode control would be implemented as well, improving AC regulation and fault protection.

(The irony is, a properly designed switching supply (with these basic features) would probably be cheaper: fewer transistors are needed, and frequency can be raised, shrinking the transformer. But that would require a fair amount of design work -- which costs real time and money. Meanwhile, what they have, Just Works(tm), so manufacturers don't care.)

The same can be said for the tiny converter -- control, filtering and protection can all be integrated. Some indeed do: Analog Devices' combined digital isolators and power converters come to mind,
ADuM4470 Isolated Switching Regulator with Quad-Channel Isolators | Analog Devices
which stipulate output filter inductors.

But why go to such lengths, making a complicated circuit with all those control and protection features, when such little power can be easily handled by a single chip?

So, the SN650x (and equivalents like MAX253). They are low power (a watt or two), low voltage, and just supposed to be some dumb standalone thing, simple, no features, little if any protection, it just does one thing and that's it. So they can get away without current-mode operation, without output filter inductor, without PWM control, fully open loop.

When all you need is some isolated voltage, and the risk of failure (or consequence) is small, it's a good choice.

Examples of low-risk or low-consequence applications might be auxiliary outputs, isolated communication ports (RS-232 and RS-485 are common applications), etc. The isolated circuit should have low risk of failing short circuit (thus causing a chain failure); comm interfaces are an excellent example because transient voltages can be limited by TVS diodes, currents by fusing or other protective elements, and current consumption even into a short circuit is well-defined (10s of mA for RS-232, 200mA for RS-485). It's hard to abuse such an interface to the point of complete cascading failure.

A less appropriate application might be a gate driver on a main power inverter: the consequences are high (basic functionality of the whole system is at risk), and the risk of cascading failure is also high (blown transistor --> gate driver --> isolator --> ..). A flyback, forward or resonant converter might be a more judicious choice in that case. Alternately, if board-level repair is of no concern (or actively discouraged..), a cascading failure isn't much of a concern, and such a design [chopper based] might still be chosen.

  • \$\begingroup\$ Thank you Tim for this answer :) I appreciate all your comments. It gives a lot of informations on many aspect :) \$\endgroup\$
    – Jess
    Sep 28, 2023 at 14:20

It appears that there is a more and more push pull converter working in low power application.

I'm unsure what would suggest that there's any power limitation here.

A push-pull converter laid out as you show is just one way to drive a transformer primary with a square wave without using a full bridge.

The whole thing is fundamentally no different than a regular mains transformer using for step-up or step-down. As you know, transformers are used for AC power conversion all the time, from milliwatts in audio signal applications to megawatts in energy production, energy distribution, and industry.

So, as long as you figure out a way to drive the transformer primary with an AC signal, you'll get a voltage proportional to the turns ratio on the output. The transformer will have some voltage sag between minimal load and full load, but that's not a big deal. Every linear power supply deals with it without any trouble!

The citation you provide is silly without context. Push-pull is used in HVDC-to-AC applications at megawatt scale. There's nothing precluding its use at any power level. It may not be the most cost effective solution for a particular application, but it will absolutely work for every application where you need simple voltage step-up, isolation, or step-down - from the smallest to the biggest. It will also be competitively efficient. It may cost way more than some other solution would, and that's a big driver in mass market devices.

What the citation meant, probably, is very specific: In the mass market of mains-to-low-voltage switching supplies, push-pull converters make economic sense in a given power range. That's all that's being said here.

Over consecutive cycles, the flux density in the core accumulates to higher and higher levels, eventually driving the core into saturation.

These imbalances only exist with half-bridge drive, and even then there are simple solutions to them in the design of the transformer. No need for current mode control as long as the transformer was designed for this push-pull application.

This concern applies to a niche where two slightly unbalanced windings are driven from a half-bridge, and the transformer is built for maximum efficiency and minimal cost so that this flux will have any chance of accumulating at all.

If you drive one winding with an AC waveform that has zero DC component, this won't ever be the case. Otherwise our entire electric distribution system wouldn't work :) It's all push-pull except rotating electric machinery is doing the push-pulling, not DC voltage and H-bridges.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.