SMPS Transformer - Practical Limitations

[archnemesis]

I'm developing a push-pull SMPS power supply, 12VDC to 180VDC, and my power target is 500W. The design of the rest of the system is all fine and good, but when it comes to actually building the transformer, I'm at a loss of how to proceed.

I purchased a core and bobbin as per my design, and the design also calls for a 20mil copper foil, 6 turns center-tapped. My problem is that this is proving to be quite cumbersome, the first issue being attaching wires to the foil that are large enough to carry the requisite current, but also fit within the bobbin and can be soldered to the legs. Then I thought why not just cut some copper tabs, and have them exposed and wired directly to board. However this is all bringing up a question I've been having, which is, should I temper my expectations with regard to the kind of power I will be able to get out of a home-constructed transformer? I would love to have one made (or find one with the right turns ratio, inductance, etc) but I already got two "thanks, no thanks" replies from Coilcraft and someone else, my project is probably too small for them to want to bother with. So that leaves me stuck either re-designing for smaller power loads, or trying something else entirely. But I've seen people build much larger transformers successfully, I just can't seem to find much practical information out there with my specific situation.

I guess what I'm asking for is practical advice for building these things. How would the pros tackle something like this? My eventual goal is to be a power electronics engineer so this is above all a learning experience. Additionally, are there small shops out there that someone can recommend who would build something like this for me?

Thanks!

[xygor]

One way is to use a bunch of smaller transformers.  Parallel the primaries and series the secondaries.

[jbb]

I'm sure you'll learn a lot from this project.

For connecting the 12V windings, I suggest that you don't bother with the legs - terminate the secondary straight onto the board (e.g. using bolted connections).  I guess you're looking at around 70 A RMS for each half of the primary winding?

About terminating the foil windings: this is tricky.  I've not done it myself, but have seen two major methods.

• Fold the foil over to make a 90 degree bend and have that pop out the side of the winding pack (may not be practical, depends on foil geometry).
• Solder on a collecting wire (e.g. thin copper bar) at right angles to the foil, and have that pop out the side of the winding pack.

When you're making the transformer, I suggest you:

• Wear gloves.  Copper foils can be very sharp (think angry angry paper cuts).
• Be very careful to smooth the edges of the foil (maybe use fine grit sandpaper). A small rough edge will rub through insulating tape.
• Try to have an exact integer number of turns, i.e. 6 on the dot, not 6 1/2 (unless the design calls for it).
• Observe carefully all the requirements for the insulation.

Have you built a converter of this power or voltage level before?

[NiHaoMike]

One way is to use a bunch of smaller transformers.  Parallel the primaries and series the secondaries.

If you tear down power inverters, you'll find that most of them 400W and up use multiple transformers, each transformer being driven from a separate bank of MOSFETs. That's a good reason for that. If each transformer is rectified separately, it's possible to shut off some phases in order to boost efficiency at part load.

If isolation is not needed, rectifying the ends of the primary will give you roughly 13% of the output voltage for "free" while giving you a regenerative snubber.

[archnemesis]

Thanks for the responses so far, all good things to consider. This is indeed my first try at something this powerful, I did a 100W 12V to 24V supply as a test and mostly to get over the hurdle of stability, which I finally somewhat understand enough to implement something.

So about paralleling transformers, one suggestion was to parallel the primaries and put the secondaries in series, obviously this does not let you turn off individual transformers to gain efficiency at low load. Does each transformer need its own set of FETs and controller, or can you simply place them in parallel? In the case that you parallel both the primary and secondaries, I'm assuming you would need a controller for each transformer, would that be correct? I know some controllers support parallel operation, but I thought you could just parallel the outputs and each stage runs pretty much independently? I seem to recall something regarding current sharing resistors being used in this application?

I've been steeped in theory for the last couple of years but you quickly realize that practical knowledge is far harder to obtain.

[T3sl4co1l]

Far too much for a single transformer -- a single foil winding of multiple turns has multiple downsides.

A single turn foil winding is practicable, so if you can run at higher frequency, or handle a larger core, that's a big win.

Definitely a case for push-pull on the primary side, which you've correctly chosen.

As for custom transformers, don't expect a commercial company to necessarily do better -- they're geared for winding wire, and sure you can get prototypes, but you may not get something as good as you can pull off yourself.  What you might get, for example: using a wad of wire in parallel, instead of foil.

Which, in and of itself, isn't bad either, because wire allows magnetic fields around it, that foil doesn't.  Foil, the current flow is pushed to the edges.  That's why multiple turns of foil is bad.

Tim

[fourtytwo42]

Some more information would be helpful particuarly operating frequency ? and is this intended to meet re-enforced insulation requirements ?

IMOP using multiple transformers is a bad idea especially hand wound as it will be very difficult to ensure power balance between them. A single transformer for 500W is a doddle, I am using an ETD54 for 1500W@25Khz and it is nowhere near the maximium power it could acheive at that frequency.

Are you sure you need foil ? it is hard to handle and if your frequency is moderate litz may be easier to use, in fact you can also make your own sucsesfully if you dont need to many strands. Think about your primary terminations and pcb layout, you dont want bulges in the winding caused by large wires crossing at 90degrees to reach a termination.

Don't forget to split your secondary with the primary in the middle to reduce leakage inductance.

[ConKbot]

I refurbished a 12v "3kw" inverter where someone had done naughty things to the output transistors. (insides were well enough put together justify refurbing, single PCB, not a spaghetti mess like a different brand 1kw inverter I got at the same time)  the ends of primaries of the two push pull transformers were multiple foil layers brought out and screwed down to the heatsinks of the mosfets, a clamp bar spread the force from the screws, this also provided the connection to the mosfet drains.

Not sure how the center tap was connected to the mounting pins like they had it, but if you can't spot weld copper to copper (hardest kind of spot welding to do) I'd definitely consider just folding it 90 degrees, having it come out the bottom of the transformer and get bolted/soldered to the input bus bar, then running it back in for the rest of the windings.

By no means an expert on this, just tore apart and fixed an inverter once. . Luckily, with the board being neat and tidy, that left room for me to add my own spaghetti mess adding in a chunky common mode choke.

[T3sl4co1l]

I did this once (for a 5V 100A PS), wind the primary on the bobbin, then loop the "foil" (copper plate in this case) through the core.  I used only two turns (center tapped, because PP output with synchronous rectification -- basically the inverse of what you're doing here, but lower voltage), so it was fairly easy to pull off.

The transformer was mounted horizontally, so the secondary pieces were bent simply, soldered into slots in the PCB.

Pic:



As you can see the output chokes were terrifically oversized, too.

Tim

[ahbushnell]

One way is to use a bunch of smaller transformers.  Parallel the primaries and series the secondaries.

If you tear down power inverters, you'll find that most of them 400W and up use multiple transformers, each transformer being driven from a separate bank of MOSFETs.

That might be true for old designs but new designs don't need that.  Much larger MOSFET's are available. 

[archnemesis]

My switching frequency is 100kHz, and I'm using an ETD 44/22/15. I'm shooting for current density < 400A/cm^2. I had chosen foil because the alternative was 40 parallel strands of mag wire (to keep penetration ratio close to 2) and that seemed a bit harder to work with. I'm not really concerned at this point about isolation requirements, obviously I want it to be safe but I'm not quite at the level where I know exactly what I need to specify with regards to isolation and safety, that is probably one of my next steps.

To reduce leakage I am placing the primary between the series-split secondary. I think the idea of using tabs may work the best here, since the orientation of the coil is good for just running the tabs down to the board and soldering/screwing them in place.

[fourtytwo42]

My switching frequency is 100kHz, and I'm using an ETD 44/22/15. I'm shooting for current density < 400A/cm^2. I had chosen foil because the alternative was 40 parallel strands of mag wire (to keep penetration ratio close to 2) and that seemed a bit harder to work with. I'm not really concerned at this point about isolation requirements, obviously I want it to be safe but I'm not quite at the level where I know exactly what I need to specify with regards to isolation and safety, that is probably one of my next steps.

To reduce leakage I am placing the primary between the series-split secondary. I think the idea of using tabs may work the best here, since the orientation of the coil is good for just running the tabs down to the board and soldering/screwing them in place.

Ahh 100Khz ok that makes a lot of difference with regard to foil, I have long since reduced operating frequency simply to reduce transformer construction problems like exactly this but I must admit that at only 12V input for 500W it's a bit of a beast! as Tim pointed out looking at low voltage output smps's gives some good clues and I must admit they universally use foil although professionally built. Have you tried the design at 25Khz, although you will need a comiseratly larger core it might help in other ways (switching losses) ? I am also interested in Tim's construction of secondary over the top, definetly simpler but I would be worried about leakage OTH good if your going for resonant switching I always find the tradeoffs around the transformer/inductor the most entertaining I should mention my construction is limited to the tools in my garage, vice for keeping things taught, electric drill for makeing Litz and a lot of patience Also you could push your 400A/Cm^2 if it helps but beware core heating at 100Khz, what material are you using ?

[PChi]

For the lead outs on a 24 V 30 A output supply on the secondary I used copper sheet insulated with kapton tape and used braided copper soldered to the copper sheet turns at right angles with a large soldering iron. A cheap alternative is used desoldering braid.

[archnemesis]

what material are you using ?

I'm using N97, chosen based on core loss calculations.

For the lead outs on a 24 V 30 A output supply on the secondary I used copper sheet insulated with kapton tape and used braided copper soldered to the copper sheet turns at right angles with a large soldering iron. A cheap alternative is used desoldering braid.

I've considered braided copper wire or even flat braid copper (like the desoldering braid) but I would be worried about the skin effect here. Some of the transformers I've taken apart, I noticed that the wires connecting the winding to the board/legs was often smaller than the windings themselves (for instance, a rather large foil connected by only 3-4 strands of small diameter magnet wire). Can I derate those conductors a bit due to their relatively short length?

[fourtytwo42]

what material are you using ?

I'm using N97, chosen based on core loss calculations.

For the lead outs on a 24 V 30 A output supply on the secondary I used copper sheet insulated with kapton tape and used braided copper soldered to the copper sheet turns at right angles with a large soldering iron. A cheap alternative is used desoldering braid. N97 cool! BTW are you using the russian guys trafo software ?

I've considered braided copper wire or even flat braid copper (like the desoldering braid) but I would be worried about the skin effect here. Some of the transformers I've taken apart, I noticed that the wires connecting the winding to the board/legs was often smaller than the windings themselves (for instance, a rather large foil connected by only 3-4 strands of small diameter magnet wire). Can I derate those conductors a bit due to their relatively short length?

I must admit I was ammused by the solder braid concept, the stuff I use is very thin for SMT work and it will wick solder and become solid very easely. Yes you can derate for short lengths but remember the heat generated will have to go somewhere, pcb & transformer not very good news! Better to either fold the foil 90degrees, do something like Tim or solder a short lump of Litz to the foil. BTW are you using the russian guy's software ?

[archnemesis]

Russian guy's software?

[fourtytwo42]

Russian guy's software?

Yahh Excellentit7100 it's good for toying around different ideas but a bit impenetarable at times! If you want to try it be very carefull finding downloads there are plenty with nasty little hitch-hikers included.

So what software do you use or do you do your own custom spreadsheets like I do ?

[archnemesis]

I am using my own spreadsheets yes, based on a few books, the main one being Switching Power Supply Design & Optimization by Maniktala.

[archnemesis]

I think what I'll do is switch to using a core with a square bobbin. I didn't account for having to bend the foil at 90 degrees for a tab and also curve it around a round core. The flats should give me plenty of space for the tabs as well. I'll have the center tap come out as a bundle of mag wire/litz wire then the two legs will attach to the board either through slots or pads.

[fourtytwo42]

I am using my own spreadsheets yes, based on a few books, the main one being Switching Power Supply Design & Optimization by Maniktala.

Excellent I always think my own calculations are better than anyone else's as I can customise them to my own limitations etc and many of those are derived from practical implementation experience, it's kinda tune as you go 180V is quite dangerious perhaps you are building an inverter/GTI and it's not my place to enquire particuarly but please don't leave insulation as an afterthought, it can make a big and sometimes annoying differance in transformer construction, you may even need to increase core size to accomodate it but nothing is more important. There are to many DIY designs and sites that ignore this I am afraid. There was a bid to make a high-voltage section on this site and I think it was decided against on the balance of encouregemnt vs safety. I am sorry to sound like a crazed preacher, just be safe ok

[T3sl4co1l]

I've considered braided copper wire or even flat braid copper (like the desoldering braid) but I would be worried about the skin effect here.

Braid alone isn't enough, but you're on the right track -- if you look closely enough you can find braided enameled wire (flat Litz), however, which is excellent, much better than foil!

Necked down conductors are perfectly fine, as long as you can handle the increased temperature, or have enough thickness nearby to sink the heat away with.

BTW, simply using large Litz isn't that great, because the magnetic field penetrates through the bundle -- all that field means more leakage inductance than with a solid wire of the same size.

Leakage inductance is easily calculated (well, to the extent that these things are calculable at all..) by assuming one winding is solid metal (shorted turns) and the other is working against that shield.  For example, a single layer secondary, with a foil primary around it, is equivalent to a round wire microstrip transmission line structure, which has around 0.3uH per meter of wire length.  Wire length directly matters, which is another reason to err on the side of larger cores.

Tim

[archnemesis]

180V is quite dangerious perhaps you are building an inverter/GTI and it's not my place to enquire particuarly but please don't leave insulation as an afterthought, it can make a big and sometimes annoying differance in transformer construction, you may even need to increase core size to accomodate it but nothing is more important. There are to many DIY designs and sites that ignore this I am afraid. There was a bid to make a high-voltage section on this site and I think it was decided against on the balance of encouregemnt vs safety. I am sorry to sound like a crazed preacher, just be safe ok

I totally understand, I am taking a few precautions like using margin tape, 3-4 layers of mylar tape between layers, etc. However I'm just not at a point where I could say how much isolation that gives me, so I'm trying to use an abundance of caution.

[strawberry]

Bridge inverter will give 20A and more efficient transformer

[David Hess]

My solution for this sort of problem is to use a larger than necessary transformer core to make the winding and construction easier or at least feasible.

I have seen designs which used two smaller transformers (or inductors) in parallel to allow a symmetrical layout for easy current balancing.

[phenol]

i have done a couple of beasties of similar nature with different transformer designs. They all boost 12v to 440V (one to 600v) in current-fed PP topology. My first attempt involved a toroidal transformer with a flat multistrand assembly of enameled wire on top of the secondary. An awkward variation of this described on some Russian forums splits the center-tapped multistrand primary into multiple single-strand primaries tightly wound together in a single layer around the entire circumference of the core. Then they are connected in parallel and used as one composite primary giving lower leakage inductance.
I haven’t tried that magical recipe and went for copper foil and E cores, which gave the lowest leakage. I used copper straps or paralleled strands of enameled wire (flat, so it doesn’t bulge) to connect to the foil windings .