Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Off-the shelf transformer for push pull converter

Status
Not open for further replies.

stenzer

Advanced Member level 3
Joined
Oct 1, 2012
Messages
864
Helped
122
Reputation
248
Reaction score
185
Trophy points
1,323
Activity points
8,925
Hi,

I'm aiming to design a push-pull converter for isolation purposes. Therefor, I intend to use a primary and secundary center tapped transformer, which enables a full wave rectification with two diods (circuitry shown in [1]). For simplicity I'm neglecting the diode voltage drop as well as the voltage drop at the switching MOSFET at this point, thus assuming a transformer turn ratio of Ns/Np = 1 (I'm only interested in the isolation).

My design parameters are:

Vin = 5 V (regulated; MOSFET voltage drop is neglected)
Vout = 5 V (diode voltage drop is neglected)
Pout,max = 5W
f_switch = 250 kHz
D = 50 % (duty cycle)

The designed converter should be used for a low product quantity (~100 pieces), thus an off the shelf transformer should be used, and that's the point where I'm start to struggle.

A couple of datasheets of PWM controllers explain how to choose a suitable transformer by means of the ET product (sometimes also called VT product). Which accomplished by choosing a transforme which fullfills

ET >= Vin / (2 * f_switch) = 5 V / (2 * 250E3 Hz) = 10 V*µs

So if I'm using a transformer which has an ET larger (or equal) than 10 V*µs everything should work fine for me. E.g. [2] (has a turns ratio of 1:1.1) with an ET product of 27 V*µs. To my knowledge for a center tapped application this value has to be divided by 2 (as at each switching period only the half primary inductance is applied to Vin).

Although the ET product method seems to be a fast and easy way to select a transformer I have the feeling I'm missing something i.e. transformer current rating (interestingly [2] does not include a current rating) and power. I know the magnetic flux B is proportional to Vin * ton [3, page 14]. Further, B = µ0 * µr * H, where the magnetic field strength H is a function of the coil geometry i.e. number of coil turns, length and diameter as well as current. So the current rating might be covered by the ET product, but it is not obious for me.

Further, there are also off the shelf transformers available, which are not stating the ET product e.g. the PA3964.002NL [4]. The voltage rating and turn ratio of [4] do no fit with my design considerations, this reference is only given to show there are transformers without stating an ET product.

So my overall questions are:

- If I'm finding a transformer which has an ET product larger than that calculated above, I'm fine? (what't about the power and current?)
- How can I identify if a fransformer can be used, not stating an ET product?

Maybe someone can resolve my uncertainties. Unfortunatly, electromagnetism is not my strength :-?.

BR & thx


[1] https://cdn.eeweb.com/articles/articles/Isolated-DC-DC-Converter_5-1429519249.png
[2] https://www.we-online.de/katalog/datasheet/750315090.pdf
[3] **broken link removed**
[4] https://products.pulseelex.com/files/product_files/P730.pdf
 

I'm pretty sure I've seen Coilcraft transformers that
they say are for push-pull.

One issue with push-pull is the potential for "flux
walk", accumulated duty cycle imbalance on the
A/B phases can eventually push you into core
saturation over time. Some controllers attempt
to manage that, some not so much. But it indicates
to me that some "core margin" beyond raw
volt-seconds is wanted?

I'd bet Wurth Elektronik (sp?) has something to
offer as well, and maybe a "product chooser"?
 

Hi,

yes there are a couple push pull "rated" transformes from Coilcraft, Wurth [5], TDK and Pulse. I usually use the parametric search function from Digikey to get an overview of possible manufacturers.
Yes, as far I know a push pull topology is prone to imbalances of the MOSFET switching singnals. If the two switching phase signals are not symmetrical a core-current is built up und drives the core into saturation. Please correct me if I'm wrong.

What is meant by core margin? A high ET popduct? In what kind would that be beneficial (core does not get saturated so easy/fast)?

Yes, it is possible to choose power magnetics for DCDC converters [6] at the Wurth homepage, but there is no ET product parameter to select. However, it is possible to choose output voltage and current, and turn ratios.

BR

[5] https://www.we-online.com/web/en/el..._pbcm/product_spotlight/max13253_max13256.php
[6] https://www.we-online.de/katalog/en/pbs/power_magnetics/transformers_for_dcdc_converter/
 

Most transformers on the market, e.g. from WE, have maximum flux specification in the datasheet.

Dedicated flyback transformers, e.g. the PoE WE transformers in your second link usually have a saturation current specification which can be easily converted to maximum flux.

Application specific transformers as in the first link can alternatively have a voltage + switching frequency specification.
 

A simulator makes it easy to experiment with component values.
My simulation provides insight about the primary winding inductance value. (Although not all questions are answered.)

In order to clarify cause and effect in the waveforms, I made a simple transformer (no center taps). As the simulation ran, primary inductance was increased as labelled.

transformer 1_5x step-up increas primary induc supply 5VAC square 5 ohm load.png

The goal is to obtain most efficient use of the transformer. This appears to be at the middle value. (The low value requires a greater saturation rating. The high value inhibits throughput.)

I chose 1.5x step-up ratio because losses are easily under-estimated, such as: parasitic resistance, diode drops, transformer inefficiency, etc.
 

@ FvM:
Indeed, most manufacturers specify the magnetic flux Φ by means of the ET product ([ET] = [Φ] = Wb = V . s).

Ok, can you tell me how the saturation current can be converted into a corresponding magnetc flux?

I also noticed that there are a couple of transformers which recommend a switching ferquency and rate their input voltage range. I assume the compliance of the stated input voltage is criticel due to the maximum current rating of the copper wire. I also assume that an undercuting of the stated input voltage is not recommended, if so, why (too small magnetic field change)?



@BradtheRad:
Thank you for the attached simulation result, which well ilustrates the effect of the primary inductance (BTW, which program have you used?). According to the simulation result it seems to me not only the ET product have to be kept in mind when choosing an transformer, but also its primary inductance.


But my main question is still unanswered: What have to be considered, to provide the desired power from the primary to the secondary side. By that I mean, transformer prameters (ET, W and Iout if stated, seem to be plausible :-D). Further, how large is the power conversion influenced by the chosen switching frequency (of course, changing f_switch results in an other ET)?

BR
 

To carry a desired spec Watt level, the transformer needs to have:

sufficient metal
thick enough wire gauge
correct number of turns per winding

Insufficient metal results in low inductance and tendency to overheat. By experience formulas have been developed telling how large core dimensions are needed for how many Watts.

There are tables telling what size wire can carry how many Amperes.

A lower frequency needs a longer time constant, therefore more inductance and more turns.

Once you have a transformer in front of you, for 1.5x step-up and 1.5A, you can sweep the frequency faster or slower, until you get maximum voltage to your load. With square wave power source you'll see similar waveforms as in my simulation. With sinewave input you'll get sinewave output (and some amount of power factor error).

I use Falstad's animated interactive simulator. Free to download and use at (Java must be installed on your computer):

falstad.com/circuit

Its component menu includes a transformer with center tap on the secondary. (It can be made to work in reverse.) The menu has no transformer with center tap on both windings. A more sophisticated simulator would have that.
 

@ BradtheRad: Thank your for your tips! Maybe I should try to make an transformer on my own, at least once, to get a feeling what's important :wink:.

@ treez: Thank you for the link, I have read this thread before I start this one, as it has not clarified all of my questions.

@ FvM: Thank you, I had that in mind in the form: I . L = ∫V . dt = ET. So, it is a little bit more obvious (for me) how the (satuartion) current and the ET product are related. Thanks :thumbsup:.
 

Although the ET product method seems to be a fast and easy way to select a transformer I have the feeling I'm missing something i.e. transformer current rating (interestingly [2] does not include a current rating) and power.

That is interesting. Please take a look at the datasheet. **broken link removed**

This is the one you are using. Consider, for simplicity, an ideal transformer. It has a zero resistance but finite inductance. For this core the inductance is reported at 10 kHz. Although for an ideal transformer the inductance is independent of freq, but I am not sure whether this spec will be there at 250 kHz. This is one point.

Also: VT is same as J/Amp; So the relation to power and current is only indirect. So if you select a given VT for one application, you can increase the power by increasing the current (but that means you can increase the frequency and reduce the inductance).

- If I'm finding a transformer which has an ET product larger than that calculated above, I'm fine? (what't about the power and current?)

Your power and current will be locked by the voltage and frequency. Confusing, but true.
 

Your power and current will be locked by the voltage and frequency. Confusing, but true.

That's obvious for me for the current as the following holds, I . L = ∫V . dt. But how does the appropriate mathematical equation regarding your statement look like?
 

Theres no “equation” that you can use to tell if any given transformer core set is ok for your particular application.
Things like “has it got a bobbin with enough pins on it” , also come into it….."are the bobbin pins robust enough for me to terminate to them without them coming off"…etc etc.
"Is the bobbin going to be deep enough for my turns…and for all the extra low power secondaries that I want"…etc etc….

"Is the bobbin footprint small enough for my pcb layout"...etc etc

Basically, you pick a core, you wind it up…..making sure you are not saturating…ie your b(max) should be less than 300mT …….then maybe when you’ve wound it up and done the calcs for it…you see that your Bmax was just 0.01T…...and you had loads of depth in the bobbin for your turns…..so you pick a smaller bobbin and core…..then go again from there……
Its quite quick when you get your calcs going in an excel spreadsheet….but this iterative way is the way to go…………………..there is no magic formula that will give you a result…and then you can type that result into digikey and a load of acceptable cores come up…it wont work like that.

Pick one, do it...see where you are...then go from there...you'll find after youve done a few you get a knack for picking a good one in the starting place....then maybe you only have to re-iterate once.
All the maths you need is in here...
https://massey276.wixsite.com/maths
 
Last edited by a moderator:

@ treeze: Of corse that's true, but as the title of this thread indicates, I'm interested in choosing an off-the shelf transformer with appropriate power handling capability.

Unfortunatly the added link doesn't work (Secured connection failed) :-?.
 

ok in that case its even easier for you...pick a core, taking a general note of its size, cross sectional area, in relation to your power , and go through the PP design with it....find your Bmax and your conductor losses....assess whether or not they are ok....ensure the core is litz wound.

Even with OTS txfmr, there is no magic figure......pick one, calc it through...if its not great pick another....do it again....compare them....choose.

After a while you will get into it and find yourself picking better each time.

If you sit thinking and thinking, you will never get started... do it ...make the mistakes we all make...then do it again...but better the 2nd time, 3rd time etc etc

The link works for me and others have got it too

- - - Updated - - -

You are one of a large number of people on this site and others seeking the magic formula...there isnt one.
 

Ok, thank you for your advice! I will evaluate a couple of off-the shelf transformers by means a prototype.
 

Whats your likely best one.?..(part num?)....tell it here and ill design it through for you and post it.

- - - Updated - - -

i think Kashcke (kaschke? kashke?) do OTS txfmrs aswell
 
Last edited by a moderator:

I would probably go with the Wurth 750315090 [1]. Uasually I would have tested it on my own, by determing the saturation current as a function of applied voltage and frequency, and converted power by varying of the load. But unfortunatly due to the current pandemic, I have no access to propper equipment (home office). That's the reason why I'm trying to make a most adequate analytical choice.

[1] https://www.we-online.de/katalog/datasheet/750315090.pdf
 

your spec is 5vin 5vout 5wout 250khz.
That transformer is 1:1.......so you wont be able to get 5vout from 5vin with a pushpull with split secondary coil......because this would need more than max duty cycle can give you...........there are diode drops in there too making it worse.

why dont you just do a flyback....there are loads of OTS flyback txfmrs..

- - - Updated - - -

Hang on though...lets look into it by using the output coils as a single coil in a "current doubler" configuration...bak soon.
 

That's obvious for me for the current as the following holds, I . L = ∫V . dt. But how does the appropriate mathematical equation regarding your statement look like?

More formally, L. I(t)= integral V(tau) d(tau)

Yes, it can get messy.

My suggestions:

1. Make sure that the core you have selected (post #1; ref #2) can work at 250 kHz

2. Use the supplied inductance and estimate the power the core can handle at 250 kHz

3. If the VT is not specified, use the max_flux (of the core) and inductance and the frequency to estimate the suitability.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top