SMPS quadfilar transformer construction

[fourtytwo42]

Hello all, I am designing an ETD59 running at 25Khz with two secondary's that may be connected in series or parallel. As the primary will fit in a single layer I propose to split the secondary's either side of it. To cover the current density and skin depth the secondary's require two wires each.

Now comes the conundrum, I am concerned if the two secondary's are wound as two wires (bifilar) in two layers there may be circulating eddy currents when they are connected in parallel so I am thinking of winding them together in quadfilar (split across the two layers) however that is going to be really hard to do especially on the outer layer wound over a lumpy primary.

Am I being to concerned about eddy currents circulating in parallel connected secondary's, has anybody designed a similar transformer and how did they deal with the issue ?

[sandalcandal]

To clarify, the wire you are using is single core enamelled copper wire (magnet wire)?

I'm not sure what eddy currents in a parallel configuration you are thinking of. Are you concerned current at the termination of the parallel pair will flow back into one of the secondaries? A counter example to your situation would be Litz wire where there are sometimes essentially hundreds of "secondaries" connected in parallel. Also I'm not sure spelling out the maths/physics would make it particularly clearer but I believe a physics analysis should show matched windings are simply parallel sources as one would expect and any mismatch between windings will result in "shorting" extra turns to the lowest effective turns in the paralleled windings.

[fcb]

Never worried about or had issues with eddy currents or unexpected losses with parallel secondaries, even past 100KHz.

ETD59 is quite big, what sort of diameter is your secondary - are you expecting uneven windings?

[fourtytwo42]

Hi Sandical

Yup, magnet wire, so two secondary's of two parallel wires each, each secondary needs a whole layer.

So build-up 1 : Layer 1  secondary 1, Layer 2 Primary, Layer 3 Secondary 2. This is the one where I worry about connecting the two secondary's in parallel.

Build-up 2 : Layer 1 1/2 of secondary 1 & 2 in quadfilar, Layer 2 Primary, Layer 3  1/2 of secondary 1 & 2 in quadfilar.

I am sorry I did not do a very good job of explaining it in the original post. Build-up 2 is exactly akin to using Litz and I have no worries about it electrically but physically I think it will be a pig to (hand) wind.

However I am concerned build-up 1 will have unequalness in some way and result in nasty circulating currents when the two secondary's are connected in parallel.

Or am I just overthinking this ?

[fourtytwo42]

Never worried about or had issues with eddy currents or unexpected losses with parallel secondaries, even past 100KHz.

ETD59 is quite big, what sort of diameter is your secondary - are you expecting uneven windings?

Hi Fcb, Layer 3 will be about 29mm diameter and the problem I foresee is a bit lumpy due to the primary underneath whereas layer 1 (1st secondary) will be on the smooth former). Thank you for your comment, how were your sections/layers arranged ?

[sandalcandal]

However I am concerned build-up 1 will have unequalness in some way and result in nasty circulating currents when the two secondary's are connected in parallel.

Or am I just overthinking this ?

From a theoretical perspective I can't think of any way that should cause problems. In theory, the outer winding will have greater diameter thus longer wire length thus higher ESR and will take a slightly lower share of the current under load but unless the primary is particularly thick I think it will be negligible. The underlying EMF/open circuit voltage should stay the same however and as long as (effective) number of turns is the same there will be no "leakage"/mismatch current. (penfold below makes a point to be careful of accidental half-turns)

That being said I haven't actually built and/or fully analysed a transformer of your description so I can't say with 100% certainty. There might be something I've missed that may cause issues. Hopefully some more experienced people will chime in.

[penfold]

In terms of any significant errors that could cause problems, i.e. whole numbers of turns missing, I'd be concerned about the production process if this happened with 5 turns, and at 500 turns the voltage/turn to total resistance of the winding would minimise the problem itself. So, wouldn't usually worry in that regard.

However, it is worth being a little careful about which pins you route the windings out to avoid half-turns worth of errors, can add a little extra leakage inductance you'd be better without but probably wouldn't notice, but would ever so slightly imabalnce the current sharing

[fourtytwo42]

Thank you all for your very helpful replies, the primary is about 2.5mm thick (0.8mm*7), the secondary's are 29 turns each.
I like the comment about bobbin pins, easily overlooked, thank you. I feel considerably relieved 
I should mention I have never built a ferrite transformer with parallel secondary's before 

[fcb]

Never worried about or had issues with eddy currents or unexpected losses with parallel secondaries, even past 100KHz.

ETD59 is quite big, what sort of diameter is your secondary - are you expecting uneven windings?

Hi Fcb, Layer 3 will be about 29mm diameter and the problem I foresee is a bit lumpy due to the primary underneath whereas layer 1 (1st secondary) will be on the smooth former). Thank you for your comment, how were your sections/layers arranged ?

We tried winding both secondaries at the same time - worked but was awkward, the we tried winding the second ontop of the first - also worked.  There was a small but measurable difference in secondary ESR, but not enough to effect the overall design, but we weren't really stuffing the bobbin.  The biggest risk - as Penfold says - is unequal secondary winding numbers.
25KHz is fairly low - assuming you have some sort of performance/efficiency target to hit? Do you have the budget for real Litz?

[fourtytwo42]

Never worried about or had issues with eddy currents or unexpected losses with parallel secondaries, even past 100KHz.

ETD59 is quite big, what sort of diameter is your secondary - are you expecting uneven windings?

Hi Fcb, Layer 3 will be about 29mm diameter and the problem I foresee is a bit lumpy due to the primary underneath whereas layer 1 (1st secondary) will be on the smooth former). Thank you for your comment, how were your sections/layers arranged ?

We tried winding both secondaries at the same time - worked but was awkward, the we tried winding the second ontop of the first - also worked.  There was a small but measurable difference in secondary ESR, but not enough to effect the overall design, but we weren't really stuffing the bobbin.  The biggest risk - as Penfold says - is unequal secondary winding numbers.
25KHz is fairly low - assuming you have some sort of performance/efficiency target to hit? Do you have the budget for real Litz?

Fairly relaxed 0.6 fill factor on this one 

Ahh REAL LITZ  Something I find myself drooling over in catalogues but sadly not available in prototype quantities otherwise I would instantly increase frequency, so I just spend my humble life designing without it 

[fcb]

http://www.osco.uk.com/products/cable-litz-wire/litz-wire-winding-wire
https://www.hi-wire.co.uk/acatalog/litz_wire.html

Or if you just need to prove it and don't mind the hifi prices: https://www.wires.co.uk/acatalog/litzwire.html

[fourtytwo42]

http://www.osco.uk.com/products/cable-litz-wire/litz-wire-winding-wire
https://www.hi-wire.co.uk/acatalog/litz_wire.html

Or if you just need to prove it and don't mind the hifi prices: https://www.wires.co.uk/acatalog/litzwire.html

Thanks for the links, I have looked before a few years ago but perhaps they are more flexible (pun) now! Any thoughts on a Litz equivalent to TIW again in small volumes (I am trying to work out if the outer servings do the job) ?

[T3sl4co1l]

I wouldn't worry about it. The mismatch between two wires over that many turns will be small.  I would be more worried for a shorter winding -- which would also then have problems with leakage so I'd be looking at a foil construction instead.

If you have the winding area, you can always use twisted pairs (or quads or whatever).  (That is to say, the fill factor will be lower.)  Downside, you have to ohm them out at the end(s).  Which eh, it's fine for one-offs?

Tim

[fourtytwo42]

I wouldn't worry about it. The mismatch between two wires over that many turns will be small.  I would be more worried for a shorter winding -- which would also then have problems with leakage so I'd be looking at a foil construction instead.

If you have the winding area, you can always use twisted pairs (or quads or whatever).  (That is to say, the fill factor will be lower.)  Downside, you have to ohm them out at the end(s).  Which eh, it's fine for one-offs?

Tim

Thanks Tim, I am by nature a worrier  but I also have the determination to get off the pot  I thought of the twisted solution but the window utilisation was getting poor and I was not to keen on an even number (4) in the twist so I am very happy you guy's all helped me understand what I was worrying about  Yes I also figured the buzz it out bit that I will now not have to do!

I have been musing about foil too as a solution to the frequency problem and lack of Litz but typically my voltages are to high and consequentially need quite a few turns, the thoughts of long lengths of copper foil and mylar insulation running amok during the winding process, picking up bits of sawdust from the floor and bench put me off

[fcb]

Get to it fourtytwo42! Stop procrastinating and wind yourself a bloody prototype. Your next posting should be a picture of the transformer.

[fourtytwo42]

  Ok the bobbin should arrive this week but I am clean out of 0.8mm wire so that's a few weeks and I am waiting for TIW quotes, loif in the countryside  , nothing happens quick around here but to be sure I will keep you posted, now back to sleep while I wait 

[fourtytwo42]

Two and a half months later we finally have a transformer 

The major holdup was the TIW, one lot arrived and seemed fine until actually in the wind I discovered an unexpectedly small number of turns per layer and on close inspection discovered the insulation was 0.125mm thick instead of 0.1mm 
The cores were a good laugh too, I had to wait for three shipments altogether with much supplier discussion as they kept arriving smashed 

So another zzzzzzzzzz time later with the correct TIW winding began again, a layer of tape over the 1st primary layer helped to protect it from the heavy secondary wind. 2 layers of tape were required over this to prevent the next secondary layer dropping between the primary turns.

The build-up is a little weird as although the primary is a single layer both secondaries are a layer and a half each (sods law in action) 
So layer 1 is S1, layer 2 is primary, 3 is S2 and 4 is the remaining halves of both S1 & S2. Resulting leakage 450nH.

This is an ETD59, gone up a core size as it was to tight on an ETD54. The two secondaries work very happily in parallel that if anybody can remember was the original question  I also chose to use flying lead terminations as this is a chassis mount transformer rather then pcb and the primary is to big for the little bobbin pins anyway.

Ohh I should mention I sweated the slot (pin) numbers carefully to ensure the two secondaries anded up with exactly the same number of fractional turns.

Lot's of pictures ...........

[fourtytwo42]

The rest

[Terry Bites]

Lisztomania wire?

[fcb]

Nice job.

Have you calculated/measured the difference in resistance between the two secondaries. Presumably you've got plenty of margin to cope with more current contribution from the inner secondary (they are in parallel if I remember?).

Although if the inner secondary gets warmer it's resistance will go up, might balance out nicely.

[fourtytwo42]

Lisztomania wire?

Hand-made  Primary is 7 strands of 0.8mm ecw using electric drill & bench vice.

[fourtytwo42]

Nice job.

Have you calculated/measured the difference in resistance between the two secondaries. Presumably you've got plenty of margin to cope with more current contribution from the inner secondary (they are in parallel if I remember?).

Although if the inner secondary gets warmer it's resistance will go up, might balance out nicely.

Thank you 
I must confess I have not attempted to measure them, I could probably calculate it from the length differance BUT I prefer your last point, the advantage of copper having a positive temperature co-efficient vs silicon, no thermal runaway here. 
Only running <5A/mm2 so plenty of margin indeed 

[sandalcandal]

Beautifully done!
Would be interested in seeing full performance numbers but it looks quite good already.

[fourtytwo42]

Beautifully done!
Would be interested in seeing full performance numbers but it looks quite good already.

Thank you kind Sir  but I dont think as neat as some of yours!
It's OK on the bench but thats only capable of ~200W so have to wait till its outside again for a full power test (it's a freaky kind of solar converter), but it's not just that transformer that needed upgrading so I have some more work to do on cooling (uprating again) and a new APU to power the uprated cooling hohoho don't we like lots of work to do 

Picture is primary ends at ~100W unfortunatly using ubiqutios long ground lead probes! I decide to put up with a bit of ringing to reduce snubber power, the EMC filters stop it getting outside the box anyway.

[sandalcandal]

I decide to put up with a bit of ringing to reduce snubber power, the EMC filters stop it getting outside the box anyway.

Careful now. Ringing can result in more losses the semiconductor switches and other parts of the circuit with greater total loss. Even though snubbers "lose" energy there can be overall efficiency gains by reducing ringing.