My half-bridge smps transformer seems awfully large and expensive?

[grizedale]

Hi,

I am designing an offline Half-Bridge SMPS at 350W and Vin = 90-265VAC.
(there is no PFC stage)

I am finding that I need to use an ETD44 core , but this seems awfully large.
-I am wondering if it is normal to allow the core to go into saturation (i.e. above 300mT during transients etc?)

In selecting the core, I simply picked the smallest one that would NOT go above 300mT with the maximum possible primary volt.seconds.

Is this the usual way?

ETD44 Core datasheet.
**broken link removed**

(I am using N87 with no gap.)

The switching frequency (Bridge frequency) is 111KHz.
(obviously each fet switches at half of this)

The transformer must be sized such that flux density does not exceed 300mT.

Now the maximum voltage that will appear across the primary is 375/2 = 187.5 Volts, (since the rail-splitting capacitors divide the DC Bus down by two.)

The maximum time that any FET can be on for is = D(max) x T
= 0.8 * 9us = 7.2us.

Therefore the maximum magnetising current in the primary is di = (V.dt)/L = 241mA
(where: L(pri) = 5.6mH, V=187.5V, dt = 7.2us)

The saturation current (based on saturation = 300mT) is = I(sat) = B.A.N./L = 368mA

= 0.3 x 172e-6 x 40 / 0.0056
(where: Amin = 172mm^3 and N(pri) = 40 )

Therefore, this shows that ETD44 will not saturate at the maximum possible volt.seconds seen by the primary.

The next size down (ETD39) goes over 300mT at the maximum volt.seconds on the primary so I avoided it.

Am I being a bit over-conservative here?……is it normal to choose a cheaper , smaller core and just let it fly off into saturation during transients etc?

 

[goldsmith]

Dear Friend!
Hi
I think your calculations are not optimum because in this frequency and this low power , your core will become small. i designed a 430w power supply with Pq3230 .!!!
Goldsmith

 

[grizedale]

Hi,

I ran my calculations with Ferroxcube PQ32/30 core and it would be above 300mT at my maximum volt.seconds.........even with any of the gapped varieties.

PQ32/30 Ferrite core
https://www.ferroxcube.com/prod/assets/pq3230.pdf

I admit that maximum duty cycle does not occur in steady state when at maximum Vin....however, during start-up, or transient situations, it will go to maximum duty cycle.

I would be very grateful if you could possibly provide the .................

-maximum possible duty cycle of your pwm controller
-Topology
-Vin
-Switching frequency

Half-bridge transformers, as you know, do not saturate on the power current, but saturate on the magnetising current


Another reason that i cannot use PQ3230 is that it only has six pins on the one side.

I am doing a half-bridge with 3 secondaries , so therefore i need 9 pins on the secondary side.

-If i was using a full-wave rectifier on each secondary, then i could manage with six pins, however, that would mean using 12 secondary power diodes......way too many and too expensive.

 

[goldsmith]

Dear friend
Again Hi
my topology was forward. and my out put voltage was 30volts and current was 14.4 amperes. and my frequency was 300KHZ. and i used gap space for my core (between each E). and my input voltage was 311 volts DC unregulated. and my maximum duty cycle was 0.65 . and the ratio of my turns was step down . and my feed back was with tuned feed back. and its efficiency was good. and i designed a push pull topology with Pq4040 core. its power was 650VA . pq 4040 has more pins.
Best Lucks
Goldsmith

 

[mtwieg]

Why are you defining your magnetizing inductance so high and your magnetizing current so low? By my calculations, you could use a ungapped ETD39 with N=36 and you would get dB=300mT. Your inductance would be down to 1.376mH, and your magnetizing current would be 0.98A peak. If you want higher inductance and lower Imag, just increase the turns. You should be using up pretty much the full window of the core with copper.

 

[grizedale]

The magnetising inductance is 5.6mH because thats what you get when you do two primary layers of 2x0.56mm enammelled copper wire (its 2 wires to reduce skin effect)

We should always wind two primary layers.....each layer being a full layer....these two layers then sanwich the secondaries to reduce leakage inductance.

So, yes, , 5.6mH may sound high, but that is neither here nor there....its a half bridge and the power current wont even see the 5.6mH.........it will simply run out of the secondary.

Having a large primary inductance is not my goal.....but it is useful in that it reduces magnetising current....who wants loads of magnetising current sloshing around?


sorry mtwieg but i disagree with your calculation for ETD39.

I did ETD39 with N = 34, and this gives a L(pri) of 3.47mH with 3C90 => AL = 3000.

So with ETD39 i end uP with the worst case magnetising current peaking at 390mA.......and 300mT means 361mA...SO THEREFORE IT GOES OVER 300mT and therefore is in saturation.

(apologies for capitals....i am not going back to correct it because this window keeps jumping)

Anyway. i am wondering whether i should allow excursions above 300mT for short transient incidences?

because , even though its ungapped.....i bet its not cliff-edge saturation..........so maybe i can get away with transient peaks of flux density up to 500mT?

 

[svhb]

Indeed, as you state, saturation is not a steep effect. It is not a yes/no. But be careful, saturation is lower at higher temperatures. At 100°C it looks like you can use the material up to 350mH.

Why are you designing against saturation only? There is another parameter involved: core loss. If you look careully to the datasheet you will see it is only speced up to 200mT for the core loss figures. (https://www.epcos.com/web/generator...F/PDF__N87,property=Data__en.pdf;/PDF_N87.pdf). So probably, for nominal continuous operation this shoudl set the maximum.

I suppose you should design for core (and copper loss of course), at maximum continuous operation (with the correct duty cycles for minimumand maximum input voltage) atmaximum ambient temperature, and then check the saturation levels for transient behavior. But maybe you did this already. Also an interesting design criteriumis the desired efficiency of the supply. You mentioned it must be on the cheap side, but a more efficient design will run cooler, and will need less money on heatsinks, smaller space requirements, ... It is not easy to tell what the cheapest option is, a better converter design or an efficient one that needs more space.

You can also try to limit the maximum duty cycle and/or use a soft-start circuit.

If you have a core available, it is quite interesting to see how steep the saturation is when you do some windings on it to create a coil, drive it with a Mosfet and a pulse generator (don't forget to frewheel the solenoid...). If you look at the current, you can decide for yourself it it is acceptable behavior or not...

Stefaan

 

[mtwieg]

The magnetising inductance is 5.6mH because thats what you get when you do two primary layers of 2x0.56mm enammelled copper wire (its 2 wires to reduce skin effect)

Why do you think that having two wires in parallel vs one wire with the same area will be less lossy? I've never heard that before. I'm hoping you don't think that doubling your surface area will halve your resistance (that's wrong for conductors in close proximity).

We should always wind two primary layers.....each layer being a full layer....these two layers then sanwich the secondaries to reduce leakage inductance.

Makes sense, I suppose. But keep in mind that following that rule might make you end up using a lot more copper than you need.

sorry mtwieg but i disagree with your calculation for ETD39.

I did ETD39 with N = 34, and this gives a L(pri) of 3.47mH with 3C90 => AL = 3000.

First of all, you said you were using N87, not C90 (though C90 is probably better if you're able to use it). Second, for some reason in my calculations I used the AL for a 0.1mm gap for some reason. My mistake, for an ungapped ETD39 core of N87 with 36 turns, you'll get 2.075mH of Lp.

So with ETD39 i end uP with the worst case magnetising current peaking at 390mA.......and 300mT means 361mA...SO THEREFORE IT GOES OVER 300mT and therefore is in saturation.

Yes, because you chose N=34 turns (for some reason). As I said, you need at least N=36 to prevent saturation.

Anyway. i am wondering whether i should allow excursions above 300mT for short transient incidences?

because , even though its ungapped.....i bet its not cliff-edge saturation..........so maybe i can get away with transient peaks of flux density up to 500mT?

If you want to push it beyond 300mT, you should be prepared to do extremely stringent stress testing on it. Meaning you give it its worst case transient line/load conditions at maximum operating temperature. If you can't do that, then don't push your luck. Put a hard limit on the primary volt time product and call it a day.

But to your original point, I still don't see why you think a smaller core wouldn't work. Your limit on core size should be justified not just by saturation, but by the core window area, and transformer losses, and I haven't seen you address either of the latter two.