11
\$\begingroup\$

Looking at several different isolated converter topologies, flyback looks like it's the simplest at first glance. There's only one switch, so there's only one driver, which (all other things being equal) should reduce the cost. However, at high power levels (5kW+) flyback seems to generally not be considered practical. I asked why early in my career, and the answers I got were vague.

I met one person who was commonly winding his own flyback transformers; he said he got 500W out of one once, but just barely, and with lots of rewinding to optimize the transformer. The commercial manufacturers I talked to went silent, or asked what insane thing I was doing to want a flyback transformer that big.

An old book I came across said that flyback transformers need to be operated at high frequencies, and the available switches couldn't survive the stresses of a flyback converter at those power levels. However, it wasn't clear on why those stresses were worse than other single-switch topologies, like boost converters. Nor was it clear on why the frequencies needed to be so high. I suspect it's because exceptionally tight coupling is needed across the transformer/coupled inductor, which limits the choice of core materials and sizes, dictating frequency choice, further dictating switch selection. But that's just a guess.

So what's the real deal? What's the effective power limit of the flyback topology, and why?

\$\endgroup\$
  • 1
    \$\begingroup\$ flyback converters run at high frequencies for the same reason non-isolated SMPSs do: it allows a smaller core. Keeping the switching periods short means lower peak energy stored in the inductor, and lower peak current. Thus a smaller core is possible: because a high saturation current isn't necessary. If it weren't for switching losses, switching frequency would probably tend towards infinity. \$\endgroup\$ – Phil Frost Jan 22 '14 at 13:28
7
\$\begingroup\$

There is no hard limit to the output power from a flyback topology. It's a matter of which is best for a given situation. One could create a 1kW flyback, but it would not likely be economical. This is a business where they have blood-on-the-carpet meetings over 3-cent diodes and recognize that it is cheaper to hire another full-time engineer than to put an extra few pennies of cost into their product- so not picking the best topology for the requirements could foreshorten one's career.

The flyback converter uses the core less efficiently (means more money, size and weight for a core, which matters more as power levels go up). As Russell points out, the flyback stores the transferred energy in the inductor, and releases it to the output, as opposed to most other types that transfer energy when the switch is on. That means necessarily the current stress must be higher, since all the energy is being transferred by a single switch, and it can only be on a part of the time. (Keep in mind that some losses are proportional to the square of the current, so 10A for 33% of the time vs. 3A for 100% of the time represent the same load power, but the resistive losses in the low duty cycle switch are 3.7 times higher.

The voltage stress on the switch in a flyback is far higher (double input voltage) compared to a two-switch forward converter (just the input voltage). This makes the switch more expensive, especially for MOSFETs, where chip size (and therefore cost) rapidly rises with voltage rating, all other things being equal. Switches that are less sensitive to voltage (in cost) tend to be rather slow (BJTs and IGBTs), so again less suitable for flyback converters because they would require a bigger core.

Flyback converters have a number of advantages (potential simplicity because of the single switch, no output inductors required because the leakage inductance works for you, wide input voltage range), but those advantages mostly dominate at lower power levels.

That's why you'll almost always see flyback converters used in AC adapters, and you'll never see it in a 250W+ PC power supply-- both applications where any excess cost that is safe to squeeze out has been squeezed out (sometimes more that that!).

\$\endgroup\$
  • \$\begingroup\$ I would expect that for transformerless designs, the efficiency disadvantages of a flyback topology would be minimized when the ratio between input and output voltage is large (regardless of which is higher); if one is stepping 6 volts to 9 volts, a non-flyback boost configuration could pass 2/3 of the power through "directly", and the inductor would only have to "handle" 1/3 of it. A coil in a flyback configuration would thus have to handle three times as much power. If converting 5 volts to 50, however... \$\endgroup\$ – supercat Oct 13 '14 at 15:44
  • \$\begingroup\$ ...the coil in a boost configuration would have to handle 90% of the power, so the fact that a coil in a flyback configuration would have to handle 100% wouldn't affect much. On the other hand, one of the main advantages of flyback configurations is that they don't need to distinguish between step-up and step-down scenarios, and the cases where one would need something that can handle both step-up and step-down equally are those where the input and output voltages would generally be similar--precisely those where flyback operation has the greatest efficiency cost. \$\endgroup\$ – supercat Oct 13 '14 at 15:46
4
\$\begingroup\$

Past bedtime - so short answer. All are happy :-).

You differentiate 'flyback' and boost' - which can mean the same thing, but may not.

Flyback's most unique feature is that the energy to be transferred is stored entirely in the inductor when the switch is on, and transferred to the output by the collapsing magnetic field when the switch is off. Some thought will reveal that in an air-gapped core (or one in which air gaps are distributed throughout the inductor) the energy is in fact stored mainly in the 'air' in the gap - a statement that will attract 'robust contrary comment'. Regardless of the exact storage location, energy is stored in the magnetic field, and increased power requires an increased core size.

Converters which transfer power during the switch's on state do not rely principally on core and field for storage of energy.

To transfer more power in a flyback system you must increase the energy transferred per cycle and/or the number of cycles per second. For a fully 'discharged' inductor:

  • \$E\$ = Energy stored in inductor = \$\frac{1}{2}LI^2\$

  • Power = Energy transfer rate per second = \$f\cdot\frac{1}{2}LI^2\$

Where:
\$f\$ = cycles of discharge per second
\$I\$ = peak current
\$L\$ = inductance

For a given system voltage, to get more power in a given available inductor charge time you must DECREASE \$L\$, as \$I = V\cdot t/L\$, and \$t\$ and \$V\$ are fixed.

Because energy transfer = \$f\cdot\frac{1}{2}LI^2\$, doubling \$I\$ in isolation would increase energy transfer rate by a factor of 4, BUT as \$L\$ must decrease to do this, in fact \$E\$ rises approximately linearly with increasing current.

The only remaining "free" variable is frequency. \$t_{charge}\$ needs to be \$<\$ to \$<<\$ \$1/f\$, but as charge and discharge times are inversely proportional to voltage, as output rises, \$t_{off}\$ falls leaving more time for \$t_{on}\$ and inductor charging.

Early MOSFETs were extremely limited in cutoff frequency. Modern FETs are far more capable BUT for high speed high voltage switching IGBTs are often advantageous.

So ... you are unlikely to see flyback converters at more than a few hundred Watts, and usually less.

More later maybe.

\$\endgroup\$
  • 1
    \$\begingroup\$ "Flyback's most unique feature is that the energy to be transferred is stored entirely in the inductor when the switch is on, and transferred to the output by the collapsing magnetic field when the switch is off" wait...what? Is this not also true of non-isolated boost or buck converters? \$\endgroup\$ – Phil Frost Jan 22 '14 at 14:05
  • 1
    \$\begingroup\$ It's certainly not true of transformer based converters. And for buck converters, much of the power is passed straight from input to output while the switch is on. \$\endgroup\$ – Brian Drummond Jan 22 '14 at 15:11
  • \$\begingroup\$ @PhilFrost As I noted above - the term "boost" and "flyback" may or may not be used synonymously. "Flyback" is specific in meaning that energy is stored in an inductor during and "on" cycle and transferred from the magnetic field collapses. "Boost" means at a basic level that Vout > Vin. Topology used can vary. A simple +ve output single winding flyback converter with the inductor returned to Vin_+ is a boost converter (Vout > Vin) and provides energy from the inductor "standing on" and so adding to Vin. A two winding flyback with Output winding returned to gnd delivers only inductor ... \$\endgroup\$ – Russell McMahon Jan 22 '14 at 21:01
  • \$\begingroup\$ ... energy, so is less efficient, and is also non isolated - but "safer" or less or more useful" deep-ending on what you are seeking to achieve, as Vin now no longer can send current to Vout when the switch is off. | SO one really needs to be specific about topology to make really clear statements about where energy is sourced from. ... \$\endgroup\$ – Russell McMahon Jan 22 '14 at 21:07
  • \$\begingroup\$ ... || A buck converter stores only a portion of the energy in the magnetics (that portion which is delivered during the switch-is-off part of the cycle. This asymptotes to zero as Vout rises towards Vin - which is one reason for the high efficiency of buck converters in low delta V applications. | So yes/no/maybe/depends :-). Specific topology needs to be described. \$\endgroup\$ – Russell McMahon Jan 22 '14 at 21:07
2
\$\begingroup\$

Energy is lost in each closure of switch capacitance.

This makes ever increasing frequency an impractical answer to a flycore with greater energy storage gap at the cost of lower inductance.

You can have a big core with lots of turns, but then you are losing more in copper.

SIC, GAN, and Silicon Superjunction mosfets all have much less capacitance than the best devices a decade ago. Higher power hard switching flybacks are possible.

The best techniques use resonance to remove some or all of the charge stored on the switch before turning it on.

\$\endgroup\$
1
\$\begingroup\$

Switch peak currents and peak voltages limit practical power outputs, BUT the Semiconductors are getting much better. For example a SiC 1200 Volt 100m ohm Mosfet could turn off 30 amps peak . Hence one could think about 1Kw off line . Although these modern switches have low switching losses there is the energy trapped in the transformer leakage inductance that doesn't get to the load which when you use orthodox transformer technology you will find is worse than any prospective switching losses when running at normal frequencies . SO active clamp or anything that addresses leakage is the passport to high power with low losses.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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