"Sandwich" wound transformer secondary is sandwiched between unequal windings

[zenerbjt]

Hi,
In interleave wound transformers (where a secondary is “sandwiched” between two series primary halves), what is the effect on coupling of having the two primary halves of unequal turns?

For example, DER368 by power integrations shows a 12W offline flyback, and a 180W offline forward converter, each with a secondary sandwiched between unequal primary sections.
The flyback (page 29) has the primary split into 27 Turns and 60 Turns. The Two transistor forward (page 240 has the primary split into 22 turns and 41 turns.
Surely the Ampere turns should best be equal either side of the sandwiched secondary? (to get best coupling)

DER368

DER-368 - 190 W Continuous, 280 W Peak DC-DC Forward Converter with Standby

ac-dc.power.com

 

[zenerbjt]

Hi,
In interleave wound transformers (where a secondary is “sandwiched” between two series primary halves), what is the effect on coupling of having the two primary halves made up of unequal numbers of turns?
For example, DER368 by power integrations shows a 12W offline flyback SMPS, and a 180W offline forward converter, each with a secondary sandwiched between unequal primary sections.
The flyback (page 29) has the primary split into 27 Turns and 60 Turns. The Two transistor forward (page 240 has the primary split into 22 turns and 41 turns.
Surely the Ampere turns should best be equal either side of the sandwiched secondary? (to get best coupling)
Presumably, the uneven winding is purely done to make the cost of winding less?

DER368:
https://ac-dc.power.com/design-supp...es/der-368-190-w-continuous-280-w-peak-dc-dc/

 

[Easy peasy]

To get "best" coupling, aka better coupling, go to 5 layers, e.g. 3 primaries ( in series or not ) and 2 sec's ( series better here ) ....

 

[zenerbjt]

To get "best" coupling, aka better coupling, go to 5 layers, e.g. 3 primaries ( in series or not ) and 2 sec's ( series better here ) ....

Thanks, so lets say we have the three series primaries (triple sandwiching the two secondaries), then to get best coupling, each of the three series sections should have the same number of turns?

 

[Easy peasy]

exact same is not necessary ...

 

[zenerbjt]

Thanks, and in the flyback of post #2 above, would it not have given better coupling between pri and sec if the primary had been divided into say one "half" of 43 turns, and the other "half" of 44 turns? (ie, more equal halves).
In the DER368 doc, the Flyback turns are 27 for one pri "half", and 60 turns for the other pri "half"...that surely would mean poor coupling compared to having more equal halves?
Though i wonder, perhaps if the coil "halves" extend fully and evenly across the bobbin then perhaps it really doesnt matter that the sandwiching "halves" are not equal in turns?

 

[Velkarn]

To get "best" coupling, aka better coupling, go to 5 layers, e.g. 3 primaries ( in series or not ) and 2 sec's ( series better here ) ....

yes, but the capacitance pri to sec will be high...

i tryed this kind of trans construction ang get the best efficiency, but the worst EMI

 

[Easy peasy]

There are ways to lower the capacitance and make multi-layers very effective ...

 

[Velkarn]

There are ways to lower the capacitance and make multi-layers very effective ...

what ways, can u tell please?

 

[zenerbjt]

[@Velkarn...I guess its things like using TIW and more layers of tape, to increase the distance between windings, as well as things like shield windings.]

...But i am amazed nobody is saying a "43/sec/44" sandwich is worse for coupling than a "27/sec/60" sandwich.
...i mean lets take it to extremes, and think of a "2/sec/85" sandwich...now we all surely agree that unequal sandwich winding is worse.!
...so why does power integrations tout their "27/sec/60" sandwich.?
Where is the point where the unequalness of sandwiching significantly worsens coupling between pri and sec?

 

[Easy peasy]

Did they mean 0 - 27 on the first and 28 - 60 on the 2nd ... ?

 

[zenerbjt]

Did they mean 0 - 27 on the first and 28 - 60 on the 2nd ... ?

Yes

 

[Easy peasy]

Well - that is not that unbalanced then - is it ... ?

 

[zenerbjt]

The ratio is 33% to 67%...id say that was very unbalanced.....its nowhere near 50:50

 

[Velkarn]

Did they mean 0 - 27 on the first and 28 - 60 on the 2nd ... ?

no

--- Updated Dec 29, 2020 ---

now i ged it! when take a look at design

they do it to make integer number of layers

 

[zenerbjt]

now i ged it! when take a look at design
they do it to make integer number of layers

Yes i agree, but as you know its not neccesary, and as you know it could have been organised to be done 43/sec/44.
(even if that meant using a different bobbin/core set)
Our mystery is why they have chosen to do 27/sec/60?
We all agree that pri-sec coupling is worse with such uneven "sandwiching".
In the extreme we all agree that eg 2/sec/85 would give even worse coupling.....even if the "2" turns was done as tape windings which completely filled the bobbin winding width.

 

[Easy peasy]

it's basically been forced on them to make the wire fit the Tx - to meet marketing size requirement ....