Electronics > Projects, Designs, and Technical Stuff
Wind a transformer like a roll of tape? Pros and cons?
(1/3) > >>
cur8xgo:
My project is currently using a transformer as follows:

Pri: 6 turns
Sec: 16 turns
Core: 0.014" laminated M6, 24 sq cm. core area
Wire gauge: 2 awg solid for both windings (about 30 feet total). Heat shrink tubing for insulation.

Operating: 350A in 125A out (roughly)

Things are working well. However, the transformer could be improved. As it is, its operating at only around 0.25T, and the windings are much bigger than they need to be because they were hand wound and the insulation is very thick.

But besides fixing those things, I think I could go down to a significantly smaller core, like an EI-162. And to make that happen, I'd replace the 2AWG solid circular with solid strip of the same area.

I have an easy source for copper strip of that dimension, and it _should_ fit into an EI-162 maintaining the same number of turns and copper area. However, winding with strip would be much more efficient with almost zero wasted space.

Think tape roll winding, not on-edge. So the entire winding in one plane, overlapping itself. Can tuck the end of the first wind under at 90 degrees so it can get out.

My question is, what is the downside of this? I think the winding thickness overall will end up sticking out from the core around 0.9" with 16 turns. Right now 16 turns of 2 AWG solid secondary on TOP of the primary is sticking out much further than that.

To insulate I could put a single layer of kapton tape on one side of the copper strip. Or if thats a bit too weak, something marginally thicker like paper or whatever is currently the rage.

So it seems to me leakage inductance would go down, even though this method of winding pushes each turn further away from the core.

Any other issues with this? Seems really win win.

I know this is done with foil wound transformers. But I'm just a little hesitant, not sure this is apples to apples. My "foil" would be identical except maybe 0.056" thick. And the pri and sec would not be wound on top of each other, they would be side by side.


 


T3sl4co1l:
I take it, this is at 50/60Hz?

You don't want to put down an entire winding all by itself, because the magnetic field can't escape from the pile of turns.  Within a section, the field from turns underneath are shielded by the overlaying turns, while adding on top of the field from those turns, vastly increasing eddy current losses (or skin effect losses, or whatever; this is broadly termed "proximity effect").  With copper of this size, I think even at 50Hz, you'll have issues with this, and I don't recommend it.

Wire-wound windings are normally done this way, i.e. in single sections, because the wire is smaller (up to, oh say, 10 AWG?).  This is around where copper wire starts getting unfavorable due to proximity effect.

Another mitigation for proximity effect is to simply not have windings so tall -- reduce the number of layers that are stacked upon each other.  If you're limited to "wasteless" style stampings, there's probably not much room to do this; but if you have the option to use a wider (and less tall) winding area, that would help.

A toroidal transformer may be attractive for this reason.  The winding area is quite wide (the circumference of the core!).  Rather difficult to use foil windings with!  Multi-filar windings would be more attractive (i.e., using a ribbon of finer wires in parallel for the same total cross section).

Multi-filar also helps with proximity effect (you're making flat Litz, in a sense), which is another option.  You may get a better fill factor with smaller enameled wire, than whole round wire.


Back to foil windings.  What you can do, is wind both primary and secondary at the same time -- much as you would wind a capacitor with two plates and two dielectrics, this keeps the current balanced within the winding, so that currents aren't piling up.  There's always one forward current (primary) beside an opposing current (secondary), alternating.

In this case, you might wind the primary and two secondaries, getting 6 and 6+6 turns respectively, then add another 4 turns (2 each at the top and bottom) to bring the total up to the desired 6:16.  The additional turns aren't paired, so do exhibit proximity effect, but are a small part of the total so it's an acceptable compromise.  Other divisions are possible but this is a simple one, and wouldn't be too bad.

In summary, that would be:
Pri, sec: full width foil; ampacity attained by selecting foil thickness appropriately.
Windup:
A: 2 turns sec
B: 6 turns of triple layer: sec-pri-sec (call these B1, B2, B3)
C: 2 turns sec
Wiring:
Secondary start = A start
Connect A end to B1 start
Connect B1 end to B3 start
Connect B3 end to C start
Secondary end = C end
Primary start = B2 start
Primary end = B2 end

BTW, if you have to neck down the foil, to get it out of the bobbin, that's fine.  At the start/end of a winding, you'll do a 45° fold, to make a 90° bend, to escape vertically from the bobbin (as you already noted :) ).  If this is too wide (the bobbin winding area width, and therefore foil width, may be wider than the slot in the bobbin cheek), the vertical lead can be cut narrower (increasing current density, but just for a short distance -- no big deal).  Or actually, it could be folded in half, lengthwise, or thirds or whatever for that matter.  Crush it down with pliers or hammer, it's going to get pretty thick pretty quickly...

You'll end up with a bunch of foil stubs sticking out of the bobbin, which need to be wired together.  Just resolve these however you can.  Expect to take up some space in the process.  You'll probably fold over a few against the bobbin and solder them (if they're folded or rolled, make sure solder gets into the layers), or for that matter, drill holes and make bolted connections.

Or you can do some simplification right away, and, I think, assemble one single strip, with the other layers positioned on it exactly as needed.  Start with a 10-turn length of secondary, and assemble the 6-turn lengths of primary and additional secondary, onto it, at the 2-turn mark.  Now you only have six terminals to resolve.  You need to get the lengths and positions just right, however!

By the way, regarding insulation: do use heavy plastic, or vulcanized paper ("fish paper").  Don't want that foil edge chafing through a weak layer of film tape or something!  I would guess 10 mil PET should be effective here, or maybe two layers of 3-5 mil.  Or anything equivalent in strength but higher rated in temperature, as needed.


There's actually kind of a downside to this construction, in that it may be too good -- the characteristic impedance will be much lower than most mains-frequency transformers.  If you were expecting some leakage inductance in the process (as might be beneficial in the below-mentioned example, converting MIG to stick -- the inductance improves arc stability), you'll get very little with such a tight windup, and may need to add an extra reactor (AC inductor) in series.  In other words, the available short-circuit current will be very high, or the regulation will be very good.  (These are also attributes that toroidal transformers share, and these are precisely the reasons why. :) )


Somewhat more out-there ideas: is a transformer really the best solution?  This sounds like a pretty unusual application, and maybe there is a better overall approach.  The numbers roughly sound like, changing a 350A MIG welder to a 125A TIG/stick welder?  Or, stepping up a UPS/inverter's primary (~24VAC) to some oddball voltage (UPS'd 1kW audio amp??).

In short, just to say: there may be a better (easier, cheaper), less clever alternative, e.g. selling the 350A MIG welder and buying a 125A welder.

Or you could design a switching supply which performs the transformation, preferably DC to DC, but AC to AC is also possible; but a switcher is considerably harder to pull off well, so I'm not going to recommend it without more background details.

No idea, these numbers are really oddball -- hence the custom transformer -- but when this happens, it's often a case of a missed assumption, and the reason no one uses this low-level solution, is because there is a better high-level alternative.

Tim
cur8xgo:
Thanks for the detailed answer Tim.

The operating frequency right now is 625Hz. Its a full bridge switching DC into the transformer. The output of the transformer is rectified and used as DC.

My perspective right now is that the prototype is working as designed (with some anomalies but nothing serious). The existing transformer, although hand wound somewhat uneven, is working and does not appear to be breaking a sweat during operation.

Transformer cost is critical here, so I'm pretty sure a stamped wasteless core is the only way I can go. (also why I can't use a toroid) An EI-162 would in theory fit 16 turns of 0.056" thick copper strip with the primary at 6 turns right above it, and room left for a bobbin, the 45 degree bend,and inter-winding gaps and insulation. All while being much smaller than the working prototype core, and using about 1/3rd less copper. Lots of gains to be had both in cost, weight, size, and performance.

If eddy currents will go up by stacking 16 turns on top of each other, it matters very critically, by how much. I think in this situation there is definitely a budget to afford that, so I need to quantify it. Any idea how I can do that besides just building the new transformer and having at it? Even just a ballpark.  EDIT: I dont think this applies to conductor-to-conductor effects, especially at idle.->At the moment the transformer draws 18Arms at idle, so I have to think proximity losses aren't that significant at least at the current prototype config. Do they go up proportionally with turns or exponentially? I mean, the copper area/resistance wont change (EDIT: but the geometry will, significantly, and thats very important for this effect), just the number of turns on top of each other. So should be able to ballpark this right?

The savings here are so big and the construction of the transformer would be so significantly simplified that its a road I have to go down unless there is a clear dead end. At the moment the prototype has 3 layers of windings on top of each other (1 pri, 2 sec).

Your idea of using wider windings is something I could look into. If I use the entire core length and wind the pri and sec together like you suggested, maybe we end up with the same core, but an much improved proximity effect situation. Its not out of the question to jump to an EI-175 or bigger, allowing even wider windings and making pri/sec simultaneous winding more do-able. The cost of the core for these stamped laminations is not too prohibitive yet, I dont think it will go up very fast from EI-162 to jump to the next size.

I'm going to have to stare at your summary of winding arrangements..I think I need a diagram. I will look up winding strategies like that so I can know what you are saying.

Also something to consider are losses and expense of joining several foil stubs together. With the single winding of thick strip, I can weld it to wherever it goes, and possibly make the stub long enough to reach where it needs to reach in the chassis in an inexpensive and joint-free way. But with many thin foil strips, I need to come up with a reliable mechanism to join them all without creating significant drops. Even a 0.2V drop is significant here on the primary side. I've worked hard to cut out 0.0015 ohm drops from the various pathways, especially on the primary side.

This application doesn't need the leakage inductance, thankfully, at least, in my analysis. It might benefit in inductance after the bridge but at the moment thats not showing up as a must have.

This is a project from scratch, not running on mains.

I'm seeing some good resources on calculating proximity effect but I'm going to have to learn a bit to be able to take a stab at it. Its not easy for me to do experiments without disassembling the prototype core and I'd like to avoid that if possible, its an epoxied-together situation that works.

Its too bad I can't wind the strip on-edge as a normal helix. Would get the same savings. I just don't think I can do that without special equipment. Whereas with strip wound in one plane I can do that by hand for prototyping and just as simply for making more than one.





cur8xgo:
I'm seeing lots of good papers on calculating proximity effect..just not the easiest reading

https://ecee.colorado.edu/~ecen5797/course_material/Lecture35slides.pdf

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.916.5488&rep=rep1&type=pdf

cur8xgo:
Wikipedia shows the "dowell" method to calculate Rac/Rdc for proximity effect in rectangular wires. I think that I may be able to ballpark with this... maybe

https://en.wikipedia.org/wiki/Proximity_effect_(electromagnetism)

Why is "b = width of winding window" in there?

Hmm Dowells curves. Seems like this is the ticket. Should at least be able to compare current design to proposed design and ballpark increase in losses.

Navigation
Message Index
Next page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod