How to Determine Appropriate Frequency of Misc. SMPS Transformer Empirically?

[Mechatrommer]

we all know and love the big bulky 50/60Hz mains power transformer. 1 winding in primary, and 1 winding in secondary. we just connect the primary to mains directly, add a full bridge rectifier on secondary, and we are good to go, no hassle of electronics logic ICs nor transistor whatsoever. fast forward to today's age, the weight, size and its low frequency have become its limitation, so here comes the "flyback" or smps transformer in variety of lightweightness, secondary windings count and operating frequencies...

tldr skip the intro.. lets say i have an unknown flyback/smps power transformer. all i care is its 1 major primary winding and 1 major secondary winding and i want to drive it like the old bulky power transformer like in the picture. ie current can go in both directions and open loop without any feedback to tell to stop it switching, it just run "full banana" as someone i've heard termed it. i can only vary the switching frequency of the half h-bridge drive. the question is... how do i make a test setup rig/circuit to determine the ideal operating frequency of the unknown flyback/smps transformer?

ps: i know the common variety of driving flyback transformer, but they are usually by avoiding current saturation which can easily be determined by adding a shunt resistor on the low side, controlled by an smps IC but they are usually only current in one direction. i want to explore the possibility of my question. the other thing i wonder, a mystery to me is what parameter/equation/physical properties that make our old bulky transformer are so efficient without blowing up even connected directly to mains line, it must something to do with flux/core saturation of magnetization right? if i know the physics, maybe i should know how to apply it in the flyback transformer question above. thanks.

[planet12]

I'd start with this as a crash course into the physics, using modern SI units (which are less crazy-making than the old ones by orders of magnitude):

http://ludens.cl/Electron/Magnet.html

the other thing i wonder, a mystery to me is what parameter/equation/physical properties that make our old bulky transformer are so efficient without blowing up even connected directly to mains line, it must something to do with flux/core saturation of magnetization right?

The key parameter for this is the "magnetising inductance" - this is the inductance measured on the primary, with the secondary open-circuit. This appears as a reactance to the mains input, with an impedance of 2 * pi * f * L. With a big iron core and lots of primary windings, you can easily get an inductance of several henries (10's to 100's even). If you assume L=10H and f=50Hz, the impedance will be about 3.1kohm - which with 230V RMS mains, gives you a magnetising current of ~73mA RMS.

If you're wanting to run it "full banana" you want to be clear whether it's a flyback (effectively coupled inductors) or a true transformer; although a flyback transformer can in theory be used directly as a transformer, it'll be a pain to drive and have high losses, as the magnetising inductance will be very low, resulting in high currents even at no load.

If you know the circuit it came from, you can often determine this by looking at the output - if it's direct into a rectifier then capacitor with no output inductor, chances are it's a flyback. Other ways are by measuring the primary inductance - flybacks will typically be in the low microhenries, transformers in the millihenries.

The "ideal" operating frequency is harder to determine, as a particular transformer design is the result of a set of compromises between core losses (flux density, frequency), copper losses (current, skin effect), core size, winding area, current/voltage requirements, isolation requirements, and so forth.

You can make guesses based on the core material - since you're talking SMPS transformers, you'll be looking at a form of ferrite - however even these can range from 20KHz up to megahertz range operation. If you're lucky and it's stamped with a material code, you can find the datasheet for the material, then probably the datasheet for the core. You won't often be lucky though.

How I've approached it in the past when I've wanted to reuse a core is to disassemble it (using either solvents or hot air to soften the glue to get the core apart), then take measurements of the core and rewind it for my purposes. With the core apart you can also clearly tell if it's gapped for use as a flyback, or ungapped and therefore suitable for forward conversion operations.

• Measure physical size of central leg of core so I can later work out flux density limits
• Add a known number of turns, eg. 10, and measure the inductance, allowing me to work out the A_L value
• Use my inductor saturation checker (mine is based on http://elm-chan.org/works/lchk/report.html) to determine what current the primary saturates at and what shape the saturation takes (is it a hard edge, or a softer rolloff? This helps identify between painted ferrite and iron powder torroid cores for example)
• Make some guesses about what flux density to run at based on material type and frequency I want to use (some guidance on this is in the first link above)
• Calculate and wind a new set of windings, keeping a good safety margin since without getting really detailed I won't have full information about the core

Hope this helps.

[Mechatrommer]

The "ideal" operating frequency is harder to determine, as a particular transformer design is the result of a set of compromises between core losses (flux density, frequency), copper losses (current, skin effect), core size, winding area, current/voltage requirements, isolation requirements, and so forth.

i got an idea of measuring various load to get output power (at secondary) vs input power (primary), and various frequencies test setup. then plot graphs and find the maximum efficiency. thats the point of interest.

You can make guesses based on the core material - since you're talking SMPS transformers, you'll be looking at a form of ferrite - however even these can range from 20KHz up to megahertz range operation. If you're lucky and it's stamped with a material code, you can find the datasheet for the material, then probably the datasheet for the core. You won't often be lucky though.

i think larger smps transformer will be operated at kHz range of frequency, MHz range usually very small transformer. i dont plan to unwind the existing (salvaged) transformer, i just want to reuse for hobby/learning purpose in quick and most efficient way. if i want to design my own trasnformer, i'll read transformer design book and ask in different thread.

thanks for your info, i now know a little bit of important term such as magnetising inductance so later i can google up. and reactance of 2 * pi * f * L which should be pretty basic. whats magic about transformer is when load is applied at secondary, reactance seen at primary will also changed (reduced), so its more complicated math involved. cheers

[magic]

This circuit is potentially prone to core saturation and magic smoke if Vdc+ and Vdc- aren't perfectly balanced. For example, during startup/shutdown or if there is DC on mains or failure of one rectifier. Maybe even normal rectifier mismatch could be enough to cause problems, but I'm not sure of that last one. Also, the FETs need to deal with over 600V.

A more common configuration is to fully rectify mains and drive one end of the primary with a half-bridge and ground the other end through a capacitor.

I'm not 100% sure, but I think I have seen ATX PSU where the main primary capacitor was made of two lower voltage caps in series and the primary winding was connected between the capacitors. This way the primary always sees approximately 50% of full positive rail, which automatically tracks ripple and mains variations.

It is probably a good idea to look up schematics of existing half-bridge PSUs to avoid stupid problems that have been solved half century ago

[planet12]

This circuit is potentially prone to core saturation and magic smoke if Vdc+ and Vdc- aren't perfectly balanced. For example, during startup/shutdown or if there is DC on mains or failure of one rectifier.

Quite right, I had in my head to mention "flux walking" in my initial response. The volt-seconds needs to be perfectly balanced, and anything - such as slight differences in transistors or gate drive - can cause it to walk if you don't protect yourself from it. How you do that depends on the converter topology.

I'm not 100% sure, but I think I have seen ATX PSU where the main primary capacitor was made of two lower voltage caps in series and the primary winding was connected between the capacitors. This way the primary always sees approximately 50% of full positive rail, which automatically tracks ripple and mains variations.

This helps with flux balance, but also was used to allow 110/220V supply switching prior to universal supplies - by changing the rectifier setup with a switch you could end up with the same Vdc over the totem pole of capacitors.

For OP - another thing I forgot to mention - a common design constraint is to have magnetising current be less than 10% of the full capacity primary current (although 5% or lower is probably more common). Once you know the primary inductance, and presumably a good guess at the expected primary voltage, you can work backwards from that to get an estimate of the frequency it was previously run at in its original circuit ("at what frequency does this inductance provide an impedance that translates to 5% of the full load primary current")

[nenea dani]

   A ferrite core transformer does not have a specific frequency, but it can be said that it works with high efficiency on a frequency band. The cores are manufactured on very wide frequency bands, for example 50-250Khz. You have to search for various photos on the internet to find out the manufacturer and then the core type, possibly the code. From my experience I say that some transformers that were built for 65Khz work very well at 100Khz without overheating.

[Mechatrommer]

From my experience I say that some transformers that were built for 65Khz work very well at 100Khz without overheating.

this is my thought that i should start at higher frequency first before lowering down to find "sweet spot". running a transformer at lower frequency we have higher risk of saturating the core. @planet12 thanks you introduced me to some new issue such as "flux walking" at first i thought the flux that we use to solder. reading material in google, the old transformer is auto flux zeroing due to primary winding's resistance. thats explained why they dont blow up easily.

[T3sl4co1l]

I never liked the term "flux walking" and I suggest people move away from using it.  It simply means unbalanced DC.  It doesn't matter that it goes up incrementally over many cycles, of course it does, it's a low frequency transient, what else would it be?

It's easily avoided in half-bridge configurations by simply cap-coupling the output.  Most ATX PSUs for example have a 0.47-3.3uF cap in series with the winding, which may be in series with the inverter, or returned to DC link +, center tap (for the voltage-doubler input type) or -.  A pair of caps in series can also be used (this is preferred when a center tap is absent, as in single-supply or active PFC converters; basically, making the center-tap with a cap divider).

It's least easily avoided in push-pull (CT primary) configurations, which cannot be cap-coupled.  In that case, magnetizing current can affect the fall time or voltage level during the dead time between pulses.  It may be helpful to reduce inductance (gap the core) to force this to occur faster, which is to say, the flux imbalance needs to drive an impedance (whatever is driving the transformer when the switches are off -- in a forward converter, this includes the secondary rectifier, and primary (switch) capacitance), and it should do so without saturating.  Minimizing flux imbalance in the first place is best, which can be done with toggle-type controllers like TL494.

A transformer can be measured for DC resistance, magnetizing and leakage inductance, turns ratio, saturation current, and other parameters that may be of interest (stray capacitance, bandwidth, ringing, etc.) in the usual ways.  Inductance with a suitable meter (beware that crude methods, or at too low frequency, likely give erroneous results), magnetizing inductance with all windings open, leakage with all others shorted.  Saturation is easiest with a pulsed method, measuring the current ramp in a boost converter type configuration.  Operate at a low duty cycle, and increase pulse width until the current ramp is seen to steepen.  Mind that DC resistance (which slows dI/dt) may mask saturation, in which case you would need to test at higher voltage (--> shorter pulse widths) or with less resistance somehow (say if the switch and current shunt are dominating loop resistance, as is often the case for small value, high current inductors).

Saturation is the upper limit of peak current, or flux (flux = applied voltage (square pulse) * pulse width), that can be applied to a winding while having a linear enough response.  Typically, operation is somewhat below saturation due to core losses.  The easiest way to test core losses is simply to wire it up for an application and see how hot it gets.  If it gets too hot, reduce the applied voltage, power level, or increase frequency (may not be feasible e.g. that would shift a flyback into CCM operation), and see if that helps.

Typical ferrite materials have reasonable core losses in the 50-200kHz range, for flux density around 0.1-0.3T (and saturation 0.3-0.45T).  Larger cores should run lower density, in part because of surface area to volume ratio (effectively less cooling).  Higher performance materials may be suitable at 50-100mT at 500kHz or more.

There are broadly two kinds of transformer: coupled inductor (purpose is to store and release energy within a cycle, or over several cycles) and transformer (purpose is to transform power instantaneously, less a minor delay due to leakage).  Flyback transformers are the former case, as well as SEPIC/Cuk inductors.  Forward converters are the latter case, though as mentioned above, they may intentionally store some energy, but for purposes other than supplying output power (i.e., some energy transfer (from/to the primary side alone) may be required to achieve flux balance without saturating).  Note then, CMCs are also a type of transformer, merely used sideways (instead of transforming power, they are used to block noise power).  Resonant transformers may be an inbetween case, or also may have lower coupling factor (relatively high leakage inductance), in which case "coupled inductors" is the more meaningful description but notice the three-element (Lpri, Lsec, Lleak) equivalent circuit, in other words the general non-ideal transformer case.  (They're all NI transformers of course; the distinction is more about how it's used, which correlates strongly but not perfectly with the parameters.)

So, you will need to identify the type/application as well.  A forward converter transformer might be usable as a flyback coupled inductor, but at much lower power levels (because saturation will be relatively low).  Type is also easy to recognize from the circuit the component is found in.

Tim