Electronics > Projects, Designs, and Technical Stuff

7kwt LLC 45 khz transformer design issues

<< < (4/6) > >>

T3sl4co1l:
I haven't done exactly this.  I've made foil-based transformers in other sizes, and made converters in comparable power levels.  (In particular was a 5kW inverter for induction heater service; it ran at 20-400kHz depending on configuration, and used a T107/65/25-3F3 core with Litz for primary and secondary.  The Litz cable was easy, no problems...)

Last one I did with foil, was... this:



(don't mind the notes about finding the right gap ;) ), for a little converter,



120-800V DC output, about 50-100W capacity (12V input), dual output flyback topology.  (The outputs are wired in series, so that doesn't actually amount to anything in the end; it does make the waveform far cleaner, though -- the secondary is effectively a balanced circuit.)

The windup is in order: 1 turn primary, 15 secondary, 1 pri, 15 sec, 1 pri.  Don't have any pics of the transformer under construction, but what I did was, I used copper foil tape for the turns themselves, and made connections by soldering 20AWG bare copper wire onto the ends.  Secondary is 28AWG.  This took up very little space on the bobbin, only needing ~1kV functional isolation (since the output is common ground).  I happen to have a bunch of these core sets, so it's rather oversized, or, the windup is rather disproportionate.  It doesn't seem to heat up much under normal load conditions though.  Just a simple module for lab use, no big deal.

The obvious downside to this construction, is the thin strip of wire joining together what's otherwise a beastly thick foil turn.  Preferably, you'd keep the foil going, by folding it 90° so it goes up and out of the bobbin, so the next section can be wound, and so on.

That would still concentrate current around the foil edges, because you're trying to make a 90 degree turn -- the current has to bunch up somewhere.  In this case, it's bunching up under the primary, because of image currents, gathering on the fat outside corner of the bend, and getting carried up the terminal edge and out of the bobbin.  But this at least happens only at the ends of the turn, not over its whole length.)

It also doesn't help that, every time you start/end a section with foil turned sideways, the available winding area is reduced by the foil thickness.  Preferably, you have a wide winding area.  But that implies you need a lot of room to turn the foil through 90°, so the core cross section should be rectangular as well, so the turn can be made on a wide face of the bobbin.

The next-last foil winding I made, I did actually stack two foil layers, but I haven't tested that one yet at full load, so I'm not sure how good its losses are.  (That was a necessary step, because the two turns are the ends of a CT winding -- as with your case.  Overlapping foil gives low leakage between ends of the CT winding, necessary to minimize snubbing of those switches.  Which in your case is the synchronous rectifier; in mine, it's a PP primary, same thing.)

Arguably, this should still be acceptable, as the two halves of the CT winding draw current alternately, so there shouldn't be proximity effect where both turns (on top of each other) are active at the same time -- except only for the currents due to leakage and snubbing (if applicable), which are a small part of the total.

Actually, that may be a much better example; unfortunately, I don't have any photos or drawings handy.  Perhaps I can describe it.  The windup was thus:
- Start with T87 toroid
- At 90° intervals around the toroid, apply single turns of copper foil tape.  Insulate with polyimide tape, leaving the ends of each turn exposed for connection.  The connections shall be exposed on one flat face of the toroid.  (So, not the inner or outer cylindrical surface, the face between them.)
- Apply single turns of copper foil tape on top of the existing turns.  Cut the ends a little bit shorter, so they do not overlap the previous layer's connections.
- Using copper tape, solder and polyimide tape, construct a 90 degree turn, so that the four connections from each pair of turns becomes a 4-layer laminated (flexible) connection, standing up perpendicular from the toroid plane, and aligned radially.
- Wind 24AWG over top the foil windings, and secure with tape.
- Unfold the laminated connections, and connect the foil turns in series-parallel to make a 2+2CT primary winding.  Each half of the CT winding is made of the two stacked foil turns in series; two of these are further connected in series to make a 2+2CT primary half.  Two of these are connected in parallel to give the full primary winding.
- The primary wiring connections are made in laminated foil and tape construction again; primary leads is made with a 3-layer flexible connection.

This is...probably a nightmare of a description if you have aphantasia. :scared: Ah well...

Anyway, the important part, applicable to your case, is this: to get low leakage between ends of the secondary, you should really use two foil strips stacked (instead of just one), and wrap that around for two turns.  Join the end of one with the start of the other, and that's your CT.  Except, because this would have about as much current crowding as it is now: instead of making 2+2 turns in a single section, make a single 1+1 turn pair for each section, and use two sections, with primary sections around them (PSPSP if possible, but as you said, you don't have much room to do that; SPS may actually be better here, then?).  Then wire the individual turns in series, to construct a 1+1:1+1 CT winding, where the 1's from opposite sides are closely coupled.  That is, if we number the turns a, b, c and d, with a+b being one section and c+d being the other, connect them as a+c:b+d.

So the full windup would be:
[a b] [16 turns pri] [c d]



I would also be tempted to suggest a planar transformer.  This should be quite reasonable at the low impedance here, but cooling may be a challenge.  I've seen designs advertised in this power range before, so it's definitely possible.  Downside, lots of NRE and fab time, so if it turns out it sucks anyway, or you end up needing to change the ratio or something...

Tim

axizep:
Tim! its very interesting what you do! please tell me have you been experiencing problems with an air gap fringing field? how can one estimate its influence on the winding impedance?

T3sl4co1l:
Heh, I've managed to avoid fringing effects most of the time...

Estimate... how would you estimate it, anyway?  The gold standard would of course be simulation.  This is perfectly doable, but requires specialized tools.

I think I would... model the gap by transforming the corner of the core into a pair of complementary current loops carrying about as many amp-turns as the windings.  These currents are positioned at the corners of the core piece, pointing in opposite directions around the core, so that they cancel at a distance, but the field is quite intense up close.  Then, say we put a sheet of metal near the gap: the induced eddy current is equivalent to a differential microstrip over ground plane situation, so we can think in terms of transmission lines and couplings, which are easy to calculate (say for arbitrary cases, use ATLC2).  When the plane is closer than the distance between wires (the gap length), coupling is stronger to the plane, and therefore most of the current is induced there.  Say you have 100 amp-turns on the core: at a distance equal to the gap separation, that might be around 50At induced in the plane (or maybe it's half that distance, I don't know), which implies you need a pretty stout piece of metal to have low losses (in particular, something which magnetic fields can permeate: Litz), and if not, you'll get that much current flowing on its surface, probably having big losses as a result.

So we can, at this point, also consider it as an application of induction heating, noting that the induced current is a shorted turn against some fraction of the total amp-turns, and we can calculate the L and R of this based on the geometry and material properties.

Hmm, now... amp-turns isn't necessarily the actual winding total amp-turns.  You'll get intense fringe heating in a transformer that has net ~0 amp-turns (i.e., primary N1*I1 = secondary N2*I2).  The trick is to also find the amount of flux that is fringing, and convert that to amp-turns in free space.  So this depends on the gap width to height ratio, and flux density is converted to amp-turns per length with the magnetic constant µ0.

This is a very rough and off-the-cuff approach but I think it could be developed to give intuitive results within a useful ballpark, say 20% accuracy on eddy current losses.  I'm not sure how practical it is, in terms of "intuitive", while needing so many steps.  You'd want to do a bit of analysis and see if it can be simplified down to a transformer equivalent model, and what variables are needed to fill that in (namely, a transformer with an effective shorted turn, and what coupling factor and resistance that turn has), and how to derive those from geometry.

In the mean time, if you can keep the gap thinner (bigger core, fewer turns -- also helps with windup if you can cut the turns in half, say), or keep the windings more than, say, 2 gap lengths away from the gap, that will help greatly.

Increasing frequency also helps with volts/turn and core size.  Downside, you might have to use SiC MOSFETs instead of IGBTs (I assume you're using IGBTs), which is a much bigger change than the transformer alone.  No idea if this might be prohibitive for development or cost reasons.

Tim

axizep:
Hey,Tim,Thank you for your thoughts!
i will comment starting from the bottom thesis.
yeah,i have started thinking about Sic fets and rising frequency.that would,of course,lead to designing completely new gate drivers,with optoisolators instead of gate transformes,and dedicated isolated power supply with +15v -5v outputs. but i thought that if iam experiencing such overheat presumably due to the fringing/proximity effect and other hf effects which maybe iam not aware of - problems on higher frequencies will worsen all those problems and make overheat problem more significant.
yeah,i guess its possible for me to avoid fringing field from a gap,if its enough 2 lengths of air gap to avoid those effects.

Also i wanted to ask you about a connections to the copper foil. i was thinking about a cross section of transistors for example. they have 0.5mm2 leads,but datasheet claims 150A drain current.i guess its possible to handle this kind of current density,maybe its not that bad idea to have thin copper stripes soldered to the foil on the edge, 90 degrees angle.alowing for a connection to leave the transformer,and outside litz will be soldered to it right on exit. do you think its acceptable if on few mm length there will be reduced copper cross section?
iam asking,because i have doubts about my method of connection to foil,as it could raise impedance of the whole secondary,because of thickness of those connections (field will not penetrate such thick layers with those solder tabs and will build up at those areas and lead for heating)?
Also wanted to ask you if you can say something on paralleling primary's inside and outside of secondary.can eddy currents circulate on such contour of closed loop of wire and build up excessive heat? or is it preferable to divide primary into series connected portions?

jmelson:

--- Quote from: axizep on April 16, 2019, 07:32:55 am ---with no load condition there is no heating in the coils at all!transformer runs cool. Ferrite runs cool also at all situation :with full load and no load.
i have computed dc current density. at most it was 7.5A/mm2

--- End quote ---
OK, that helps a lot.  So, it seems that eddy current is NOT the issue.  Now, I think the obvious problem is just primary resistance.  Note that Litz wire has a lot less copper area than plain wire, so you have to increase the wire size a lot.  I think that is what you are running into, here.

Possibly, since the ferrite stays cool, you have too many turns.  So, you could also try reducing the turns of both primary and secondary by a factor of two.
That will make the core run hotter, but will cut primary resistance by about half.

Jon

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod