Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

How difficult is it to design a broadband transformer?

  1. Nov 2, 2012 #1
    Hello everyone,

    I am working on a project in which I need to make a broadband/pulse transformer.
    The transformer will operate in the 20-80kHz range.

    The transformer will use about 20 watts of power.
    I need a turns ratio of 1:3.
    20V input 60V output
    It's just for hobby use. I don't have the money to pay an engineer to design it and it does not have to be perfect, I just need to get a good square wave on the secondary side.

    I have tried to make this type of transformer in the past but I find it very difficult to get a good square wave on the secondary side. I usually end up getting AC on the secondary side. I know this could be due to a number of factors such as coil capacitance, core material, or leakage inductance.

    So, my question is, how difficult is it to make a broadband/pulse transformer when one only has basic electronics knowledge?

    What general rules should I go by?

    Could anyone here offer any suggestions or advise...Is it even worth my time to try to figure out?
  2. jcsd
  3. Nov 2, 2012 #2


    User Avatar
    Gold Member

    Transformers operate with AC voltages, normally sinusoidal waves pass through them. This means the voltages/currents change gradually and smoothly. A square wave changes instantaneously (or nearly) from low to high state and then the reverse. Transformer primary and secondary coils are big inductors with iron cores and will not pass a square wave from input (primary)side to output (secondary) side. Inductors, by their nature, resist any instantaneous change in current. Apply a square wave to the primary, expect a sine wave (or, at least not square) wave out. Your own results already show this.

    You must re-think your design.
  4. Nov 3, 2012 #3
    High voltage pulse transformers are difficult to design. Low voltage pulse transformers are usually easier to design.
    If you would describe what you are trying to do, someone may be able to help you.

    Pulse transformers if correctly designed will give a nearly square or rectangular output.

    Are you sure that you require a pulse transformer? Pulse transformers are usually specified as pulses per second (PPS) and have a pulse width, input and output voltage and current, load resistance, and a maximum droop.
  5. Nov 3, 2012 #4
    The transformer I need requires both a square wave input and output. It is different than a pulse transformer in the fact that it operates at a much lower frequency (20-80kHz).

    What I need is coil that will output a decent square wave.
    The coil will be used to charge a capacitor so I need a square wave that will not swing negative.
    I am going to add a resistor to the secondary side to limit charging current.

    So, input voltage and current
    20V @ 1A max
    Output voltage and current
    60V @ 300mA max
    Turns ratio 1:3

    My first question is, how do I determine the type of core material I need?, and how do I determine the secondary inductance I need?
  6. Nov 3, 2012 #5
    This is false. Once a voltage is applied across a transformer winding, magnetic flux starts increasing in the entire core, producing voltage on all other windings. Voltage is the time derivative of magnetic flux divided by number of turns. Since materials are not perfect, their do not possess infinite permeability. Thus, a small magnetizing current exists in the core with non-zero magnetic flux.

    Isolated converter topologies such as flyback, half-bridge, full-bridge, etc. are all based on square-voltage waveforms.

    What you need to do is to calculate the positive volt-seconds during one cycle of the primary windings, select a core size, calculate N1 of turns based on allowable flux density, determine other turns, and allocate window area to different windings.

    Make sure you use a high-frequency magnetic material such as 3c94.

    Transformer size is limited only by the required winding current.

    Last edited: Nov 3, 2012
  7. Nov 3, 2012 #6
    What's difficult about pulse transformer design? The only problem is the usual high required turns ratio that allows quite large core flux leakage and thus leakage inductance proportional to magnetizing inductance.

    Also, 80 kHz is a standard frequency for power converters.
  8. Nov 3, 2012 #7
    Why do you require a square wave?

    Using resistors when charging capacitors reduces efficiency.

    There are many circuits that place DC on a transformer secondary. But a simple transformer will only have AC on it's secondary.

    At these power levels, a flyback transformer is a good approach.
    Flyback transformers are fairly complicated to design. If you cannot find instructions on designing them, I will look through my files to see if there is any information on designing flyback transformers in them.

    Google may have some useful information if you search for "battery charger ic"

    Ferrite is usually used for cores at this frequency.
  9. Nov 3, 2012 #8

    I need a square wave so I can charge the capacitor fully before it discharges.

    Also, your right about resistive charging.

    I wonder if it's possible to add the current limiting inductor to the same transformer core. This would reduce the needed secondary turns since the indcutor would double the output voltage and act as a secondary coil as well. Then I would not have to make a seperate inductor.

    Does anyone know if you could add an additional coil on the same transformer core to act as a current limiting inductor?
  10. Nov 3, 2012 #9
    The volt-second balance must be maintained on all windings. Otherwise, the voltage component would have a DC component and your transformer will quickly saturate.

    What you can do is to use a rectifier secondary topology for non-flyback converter:

    1) Half-wave
    2) Full-wave
    3) Center-tapped
    4) Current doubler

    Flyback converter simply needs a rectifying diode on the secondary side that block voltage during the primary device on-time.

    To charge a battery you could use any of those. However, the current is not limited in any of those topologies (they contain an inductor/inductors that allow current-mode control) so a resistor might be your only choice to reasonably limit transformer/device current.
  11. Nov 3, 2012 #10
    There are ways of adding inductive element on the same core; however, those are not easy to implement.

    An inductor will limit the current only in current-control mode. In such a mode the duty ratio is dynamically changed to maintain constant input/output current.
  12. Nov 3, 2012 #11

    Thanks for the comments thus far, I really appreciate them.

    So, if I used a 50% duty cycle could I get a constant output current?

    Are there any examples I could look up where an inductive element is used on the same core as the transformer? That is something I have had interest in for a while but had no luck finding any information on. Even if I don't utilize it or it's too complicated for me I would like to know understand how it works.
    Last edited: Nov 4, 2012
  13. Nov 4, 2012 #12
    Can you explain this comment further? Why would a square wave fully charge a capacitor better than any other waveform?

    Why would the duty cycle determine whether the output current is constant or not?

    Trying to obtain a square wave output will reduce your efficiency significantly. I can think of two relatively easy ways of getting a square wave but they both are rather inefficient.

    1. Put a relatively high value resistor in both the primary and secondary circuits. This reduces the effect of the inductance on the bandwidth thus increasing it.

    2. Drive the core into saturation which will result in a clipped output waveform.

    In both cases I think you would be better off using a sine wave. Perhaps you can explain why you don't like this solution.
  14. Nov 4, 2012 #13
    A transformer always gives zero as a mean DC output. To charge a capacitor, what you need is a diode.

    Adding a series inductor to a transformer is very easy. Just separate the primary from the secondary winding and you get a "leakage inductance" which has been used in the past to produce a constant current from a transformer.

    BUT people try hard to REDUCE this inductance! If you have one, you can forget your square wave, as this one will be in series with the load.

    In fact, a difficulty of broadband transformers is to minimize the leakage inductance be putting the windings close (which means: split the secondary to wind the primary between the half-secondaries, or split them even further) AND minimize the stray capacitance between them, which increases when you put them closer.

    The product of the stray cpacitance and the leakage inductance is limited by the winding's size, the speed of light, the relative permittivity.

    As a core material at 100kHz, it will be ferrite.
  15. Nov 4, 2012 #14


    User Avatar
    Science Advisor
    Homework Helper

    I think you are misunderstanding something here. If you connect a capacitor and an inductor, you have made a tuned (resonant) circuit, which will defeat any attempt to drive it with a square wave.

    But without seeing the whole circuit, we are guessing what you are realliy trying to do.
  16. Nov 4, 2012 #15
    Broadband xmfr1.png

    Here's a diagram of what I have so far. The diode will prevent any AC from forming between L and C.

    So far from the feedback I have recieved here (Thanks SunnyBoy) I need to:
    -Calculate volt-seconds/Webers of primary coil during one cycle
    -Select a core size and material
    -Calculate Primary turns based on allowable core flux density
    -Determine secondary turns and window area
    -Ensure volt-second balance is maintained on all windings

    I definately have some studying and math to do, but it seems possible. I always like a new challenge to learn and build something!

    I still would like to have more information on adding the current limiting inductor to the same core, but I have not found any instances where it's utilized.

    Is there anything else I should consider in the design?
    Last edited: Nov 4, 2012
  17. Nov 4, 2012 #16
    A couple of things to do in case you would like to proceed with a flyback converter:

    * You can ditch the diode in series with the BJT.
    * Replace the BJT with a MOSFET (100V / 10A should suffice)
    * Output diode must be fast-recovery diode. SiC carbide diode should do wonders.
    * The primary winding should not be shorted by the diode but have an RCD snubber instead. Place a resistor in parallel with a capacitor between the cathode of your diode and the upper winding node. R & C depend on leakage inductance and switching frequency.
    * A flyback converter does not need an output inductor. Its output impedance is large enough not to cause excessive currents when in current-control mode.
    * You can search for a flyback driver chip. Its datasheet should contain a detailed description on how to measure and limit the input current, which is a scaled version of output current.

    With regards to the current limiting inductor: I suggest you take this learning experience one step at a time. Only once you have designed a functional prototype of a current-limited flyback converter, then you should proceed to a different kind of converter. Converters with inductors on the secondary (or primary) are called direct converters and are generally more complex than indirect converters such as flyback converter.

    Btw. the flyback converter transformer is not really a transformer as much as a coupled inductor because the core stores converter energy. An air-gap will be needed.

  18. Nov 5, 2012 #17
    I agree with everything that SunnyBoyNY says.
    The only thing that might be added is that polarity markings should be added to the flyback transformer. If the flyback transformer is connected with incorrect polarity, the circuit will not work.
  19. Nov 8, 2012 #18
    A flyback xfmr is indeed a xfmr. It is also a pair of coupled inductors as well. A transformer is merely 2 inductors with a high degree of coupling. They are 1 & the same thing, so I don't know how one can say it can be 1 but not the other, since xfmr & "coupled inductor" are the same.

    The only exception is multiply wound inductors for ripple current filtering where the flux is a small ac ripple plus a large dc pedestal. Here we have a coupled inductor since several windings share the same core & filter several power supply outputs. But since most of the flux is dc (non-varying wrt time) there is little to no xfmr action, since xfmr action cannot happen with static flux.

    As far as energy storage goes, that itself does not define xfmr behavior. A forward converter transformer does not store much energy, but transfers it to the secondary while the primary is conducting. Some energy is stored in the form of leakage inductance, an undesirable loss element. But a forward xfmr is gapped, since operation is limited to 1st quadrant, & w/o gap the core would saturate.

    So a forward xfmr has the capability to store substantial energy, but does not do so because the forward network allows transfer to take place. When the primary power switch is on, the secondary rectifier allows secondary current, so that energy transfer takes place with little stored in the xfmr.

    With flyback topology, the core is gapped just as in the forward case. This gap prevents core saturation since flybcaks also operate in 1st quadrant. But the flyback topology does not allow primary & secondary conduction at the same time. When the primary power switch is closed, the primary current builds up, increasing flux, & stored energy in the gap. But the secondary cannot conduct since the rectifier is reverse biased.

    When primary power switch turns off, the secondary voltage inverts polarity, diode conducts, & energy is transferred. Both cases involve gapped cores which possess energy storage capability. Whether the magnetic element stores energy then releases it next half cycle, vs. instant transfer, is determined by the circuit topology, flyback or forward.

    The flyback xfmr obeys Faraday's law, just like in the forward case. It also obeys Ampere's law when averaged over a cycle. It also provides isolation just as good as a forward converter xfmr. It also exhibits core loss which varies with ac flux swing just like its forward counterpart. The flyback xfmr is a xfmr in every sense of the word.

    Whether the gapped core stores energy & releases a moment later, vs. transferring instantly is determined exclusively by the external circuit topology. A xfmr from a forward & that from a flyback can be exchanged. Both are gapped to prevent saturation. Which stores energy & which transfers it instantly is purely a function of the circuit, not the core/winding assembly.

    For well over half a century, flyback magnetic elements have been called "transformers". Because they are just that. Just because it stores substantial energy does not mean it cannot be a xfmr. A forward xfmr has the same capabiltiy to store energy, it having a gap. The forward circuit does not allow storage. The flyback circuit does not allow instant transfer.

    The circuit topology determines which mode it operates in. Flyback transformers are indeed true transformers as are forward xfmrs. However, multiple windings on a common core to filter ac ripple from several dc outputs are not a true transformer because since the flux is predominantly dc, transformer induction action cannot occur at dc. This is truly a coupled inductor which cannot be called a transformer.

  20. Nov 8, 2012 #19

    While I am not refusing your reasoning and I do not want to bicker over pedantry, I could not help myself to write at least the following paragraph. In case I am wrong, I will be happy to learn something new or correct my mistakes.

    Xfrm action can happen with a flux DC offset just fine.

    Exactly which forward converter operates only in the first quadrant? Why would you operate a forward converter only in the first quadrant - the effective number of turns is doubled as opposed to the converter working in the 1st and 3rd quadrants.

    Standard HF XFRMs have ferrite cores and certainly do not have a gap.

    XFRMs for forward converters do store very little energy in comparison to gapped cores. See above. Forward converters such as full-bridge ZVT operate in the 1st and 3rd quadrant. Magnetizing current is reset to zero every two switching cycles.

    Flyback operating in the CCM certainly has a large DC flux offset. Btw. if inductor had static flux, it would be just a piece of wire. Thus, inductor also has AC flux.

    A flyback converter working in the continuous conduction mode has a constant DC flux offset, just as a series inductor.
  21. Nov 8, 2012 #20
    My replies to the issues you raised, which I numbered.

    1) Ok fine.

    2) Of course xfmr operation can happen in the presence of dc flux offset. Reread what I stated. Only the ac component of flux can have xfmr action. The dc component of flux cannot induce a seconday emf/mmf.

    3) Single switch forward converter. Some variations operate in 1/3 quadrant, but single switch must have an air gap. Check your smps references & it will be affirmed.

    4) When you use an acronym like "HF", please define it once so I know what you mean. Is "HF" to mean "half forward"?

    5) I already stated that forward converters store negligible energy in the xfmr. I stated it clearly. But being gapped, said xfmr has storage capability. But the forward topology does not allow for storage because of polarity. When primary power FET is on, secondary diode id conducting & energy is transferred instantly. No storage followed by release as in flyback network.

    6) I've already acknowledged that flyback xfmr has a dc flux offset, hence gap is needed to prevent saturation in addition to energy storage.

    7) Yes it does. But a series inductor does not transfer energy from 1 winding to another. The flyback xfmr transfers energy from primary coil to secondary via magnetic flux coupling. The flyback xfmr is consistent with laws of Faraday & Ampere just like forward xfmr. The fact that forward xfmr transfers energy w/o storage vs. the flyback which stores energy w/o transfer until next half cycle, is strictly a characteristic of the network driving it.

    The flyback xfmr obeys the volts per turn relation per Faraday just like a forward, i.e. Vp/Np = Vs/Ns.

    It obeys Ampere's law of balancing amp-turns when averaged over a full cycle, just like a forward, i.e. Np*Ip = Ns*Is.

    It provides galvanic isolation between primary & secondary just like a forward xfmr. Again for a 1 quadrant forward, the xfmr is gapped. A flyback xfmr is gapped. If you have 2 converters, 1 forward & 1 flyback, designed for the same xfmr spec, you could interchange xfmrs. The xfmr in the forward converter is forced to transfer energy, not store it. The xfmr in the flyback converter is forced to store energy, transferring it later on the next half cycle.

    Both devices can store an transfer. How they do it depends only on the network. Both obey the volts per turn relation, as well as the amp turns relation. Both provide galvanic isolation. Both are 2 inductors coupled on a common core. As I said, "coupled inductor", & "transformer", are one & the same thing. You cannot differentiate them. The operating mode regarding energy transfer vs. storage is a circuit characteristic, not the magnetic device characteristic.

    With 2 or more inductor windings wound on a common core, where the flux is a small ripple plus a large dc offset, here is the scoop. The ac ripple exhibits xfmr action, since the ac flux couples the other winding. But the dc flux in 1 winding DOES NOT INDUCE a dc flux in the other. So this arrangement can rightfully be called a coupled inductor that is NOT a xfmr.

    Make sense. Have I clarified it well? Feel free to ask for elaboration if something isn't clear. BR.

Share this great discussion with others via Reddit, Google+, Twitter, or Facebook