[SOLVED] 1-Turn inductor (transformer's primary)

[mrinalmani]

Velkarn, No I dont think that at all.
Primary leakage is independent of secondary leakage. For a symmetrically designed transformer, primary and secondary leakage do not differ much. But I meant something else.

The primary leakage when referred to the secondary gets transfered in the ratio squared. Consider the attachment below
Since the turn ratio is 1:30, 1uH leakage in the primary gives a reactance of 0.6 ohm. This when referred to the secondary turns out to be 0.6X900 = 540 ohm. Which is very high compared to the load resistance at full load.

 

[red_alert]

Um, just to clarify things:

I was talking about a 24V to 350V full bridge DC-DC converter. The ferrite core used for the step-up HF transformer is a ETD59 type (3C90).

The turn ratio will be 1:9.

So, the primary winding will have to withstand 160 Amps (24V) and the secondary windings 12 Amps (350V).

The ETD core is kind of E-shaped, so no toroid discussion here. My latest approach is to use two identical transformers (half rated) connected in parallel/series (primary/secondary).

For the S1-4 switches, I've choose IRFP7530 (Vds=60V, Imax=280A, Rds=2mOhms). D1-4 it's not a critical choice (any 600V/15A fast recovery diodes).

Seems like the most feasible solution is to put 3-4 copper foils in parallel for the primary windings then using litz wire for the secondary windings.

 

[FvM]

The turn ratio will be 1:9

Seems to fit neither the claimed current nor voltage ratio.

Seems like the most feasible solution is to put 3-4 copper foils in parallel for the primary windings then using litz wire for the secondary windings.

Only suitable if foils and secondary windings are interlevead.

 

[red_alert]

Seems to fit neither the claimed current nor voltage ratio.

Sorry, the 1:9 turn ratio was for the latest solution (two smaller ETD59 transformers: 24V (actually 20V, for the worst battery output) to 175V) . Anyway, the initial turn ratio was 1:15 not 1:30 as mentioned eralier.

Only suitable if foils and secondary windings are interlevead.

I need 2 turns for primary windings and 18 turns for the secondary one.

So it's OK to put:

9 turns (secondary winding: litz wire) -> 2 turns (primary winding: triple layer copper foil) -> 9 turns (secondary winding: litz wire)?

 

[mrinalmani]

I dont know why leakage is not bothering you?? In fact you have not even thought of it and are completely overlooking it!! Serious trouble ahead...

 

[red_alert]

With all due respect, I just have doubts about those calculations. To get only 132W from a 4kW transformer it's the worst engineer's nighmare! And it seems like there's nothing you can do about it.. then how (real) things work anyway??

Now serious.. could you point me out to some online resources to further read about it? Thank you anyway for your concerns.

PS: By the way, the primary inductance is about 20uH and I'm using an E-shape (not toroid) ferrite core - maybe this help a little.

 

[Velkarn]

9 turns (secondary winding: litz wire) -> 2 turns (primary winding: triple layer copper foil) -> 9 turns (secondary winding: litz wire)?

i think, two layers is better. due to proximity effect in internal layer there is no current, i suppose

 

[mrinalmani]

No using E shape makes leakages even worse. Toroids have the lowest leakage. If you want to learn more about how practical high frequency, high power transformers work, search on line for "resonant converters". In fact at such high powers, only resonant topology can work. Have you ever seen such a high power transformer that is non resonant based? I dont think any one has. How fancy are you thinking your leakage inductance to be? Below 100nH?!! This is an extremely state of the art figure. Look at the datasheets of pulse transformers of top manufacturers. The leakage is in the order of 0.3% to 1.5% of the magnetization, with turn ratio of 1:1.

Let us assume, that you have made an excellent transformer that has a leakage of only 250nH! (do u think u can actually achieve this!)
Let's carry out a quick calculation.
Reactance1 = wL = 0.15 ohm
Refer this to the secondary with a turn ratio of 20
Reactance2 = 0.15*400 = 60 ohm!

Even with a 250nH leakage we are getting a terrible impedance. And trust me 250nH is not an easy figure. Therefore the only solution, as far as I know, is to add a resonance tank and use the resonant topology. Hope you realize this at an early stage!

 

[red_alert]

Ok, you managed to scare me real good!

So you think one couldn't even make a 2kW ferrite transformer?

Well, lucky me - I haven't bought the ferrite cores yet. Because I've just decided to make FOUR identical ferrite transformers (1kW, 24V/40A in, 90V/12A out) and put their primary windings in parallel and their secondary windings in series.

This way, I have a 4kW distributed ferrite transformer.

By the way, I'll use copper foil (0.2 * 35mm single layer) for both primary and secondary windings. Is that ok? Is it E55 (or E65) a good option for ferrite core? (working frequency: 75-100kHz)

 

[mrinalmani]

The only reason not to use a toroid core for high power application is due to the expensive winding procedure during mass production. If you do not intend to manufacture on a mass scale, toroids are the ideal core with minimum guaranteed leakage.

Even with toroids, when working with frequencies above 25kHz, leakage becomes the main problem. Do not expect your leakage to be below 300nH. (the photo attached in previous replies has a leakage of 900nH) And by the way as far as your question of not being able to make a transformer of even 2KW, yes it is true. Using PWM you cannot, but using resonance you may.

You must realize that 2KW is a lot of power in case of HF transformer. In fact it is so much that you would generally not find a documentation online as to how to go about it.

- - - Updated - - -

And wait a bit. Did u just say putting primaries in parallel. For HF transformers it may not be all that simple! Are you aware of 'flux walking' ? Unless you have a dedicated control system to guard flux walking for every core, I am certain you are going to saturate one or more of them!

 

[red_alert]

If the transformers are (almost) identical and are driven by the same H-bridge, isn't the flux walking simultaneously controlled for all of them (by modifying the H-bridge signals)?

Afaik, the flux walking has something to do with asimetrical command signals of H-bridge MOSFETs.

 

[mrinalmani]

Sounds good theoretically. And the only reason that it may fail will be due to asymmetrical response of cores to opposite voltages. Anyway that is a later chapter.
As for now what have you thought about the leakage?

 

[red_alert]

For each of the 4 transformers, the primary inductance will be 6uH and the turn ratio will be 5.

PS: By the way, the transformer's output DC bias it's a good indicator of flux walking?

 

[mrinalmani]

Output cannot have a DC bias. It can only have an average DC component. Note the difference between DC Bias and average DC.
With primary magnetization as low as 6uH you will saturate your core. With 5 turns your leakage reactance per x-mer will be approx 3 ohm. Therefore a total of 12 ohm. Which is almost as much as your load resistance. Use a series capacitor to nullify this.

 

[red_alert]

Average DC, sorry! So do I have to measure the output's average DC for flux walking feedback circuit?

The latest calculation results show 25uH for primary (100mT, 100Khz, 3A/mm2). Could you elaborate on nullifying capacitor, please?

 

[mrinalmani]

25 uH is a practical value, good.
There are two techniques for monitoring flux walking.
1. Voltage measurement. (easy but inferior)
2. Differential current measurement (difficult but superior)
Average DC measurement does not seem practical to me coz high resolution integration will be required.

If you add a capacitor to the secondary, it will come in series with the leakage. Now for a particular frequency resonance will occour and the effect of leakage reactance will be nullified. The value of capacitor is chosen such that resonance occours at the switching frequency.
Consider the following links. It may be helpful.

- - - Updated - - -

https://www.edaboard.com/threads/315362/

 

[red_alert]

Have you came to any conclusion yet?

..

Speaking of inverter's topologies.. do you think that using a LF output transformer is the only (diy) way for high power inverters? This seems to be the most used solution in commercially available UPS anyway but I thought it belongs to the past.

 

[mrinalmani]

Yes I have come to a partial conclusion.
1. I will optimize my winding to reduce the leakage inductance from 900nH to about 500nH.
2. I will then use a 1.1nF polypropylene capacitor from the market or will build one out od a 150uM thick PP film myself.
3. Use a closed loop control to drive the circuit in resonance.

 

[red_alert]

@mrinalmani, could you explain me why these kind of problems (leakage inductance) arise only on SMPS output power > 1 kW?

I mean, if you can make an efficient (95%) low power SMPS (300W) using a small ferrite core (E32, by example), aren't the leakage losses supposed to be proportional (same percentage) if you're using an identical type but larger size ferrite core (E80, for a 3kW SMPS), following the same design/schematics?

What core/inductor/magnetic parameters are not following a linear characteristic in this situation?

 

[mrinalmani]

Its not about linearity but about the ratio of internal to external impedance.
For a 300W load, the load resistance would be 175 ohm. If you have optimized your transformer and managed to reduce the leakage to approximately 0.5uH then your internal impedance would be approximately 120 ohm. Since the internal impedance is not many times larger than the load, a simple over-winding factor of 1.6 to 1.7 would compensate for the internal drop. (Although voltage regulation would be poor)
Notice that if your load was of 4KW then the load resistance would be 15 ohm only, which is much smaller than the internal impedance. And by the way for frequencies as high as 100kHz we generally do not use PWM for loads as high as 300W also. Resonant would be preferred.