Electronics > Projects, Designs, and Technical Stuff

Wind a transformer like a roll of tape? Pros and cons?

<< < (3/3)

cur8xgo:

--- Quote from: Kleinstein on June 04, 2019, 06:07:28 am ---There is not much wrong wit using the tape like winding. It adds to the winding capacitance - this could be a problem at high frequency, but usually is not at 60 Hz or 600 Hz.

Why use the same gauge for both windings ? normally it is a similar amount of copper for both windings, so a thinner wire for more turns.
In theory the optimum winding uses thicker wires more to the outside,  reducing the current density invers proportional to the length per turn.

It usually is attractive to really make use of the possible magnetization. So going from 0.25 T to 1 T would only need 1/4 the size of the core and correspondingly shorter turns. The downside is slightly higher no load loss, and less effective cooling, but this is usually more than compensated by less loss.

--- End quote ---

I don't have a reason for using the same gauge for both windings. Its a blind spot I haven't examined yet. The only rationale I have at the moment is that the working prototype also uses the same gauge for both windings and so to keep things equivalent I'd do the same with an improved transformer design, if not just to maintain or improve on the DCR.

If the flux density goes up significantly, I think there will be some pretty crazy core hysteresis loss once things go up near the knee, at 625Hz. I submitted for a quote on a similar design and the engineer there suggested the core would melt at 1.6T at 600Hz, and that 0.5T was the "absolute max" for M6 at that frequency (1 min on 1 min off duty with forced air). That seems a little extreme to me but I have not done the calcs so I will use it as a guideline.

So my proposed improved design will operate at 0.5T and see what happens. If my calcs are correct this puts the new design in a cost, size, and weight area that looks great for this application. Basically an EI-162 stamped core with about 18 feet of 0.056" copper strip to make both windings.

MagicSmoker:
Yeah, neither proximity loss (which increases with the number of layers) nor distributed capacitance are really relevant at <1kHz. However, M6 seems a bit thick for operation at 625Hz, and the engineer advising you that losses will go through the roof if you come anywhere close to its saturation limit is precisely right. You might find a more favorable induction limit vs. core loss vs. cost with the thinner laminations like M3 and M4 even though they are higher in cost.

Kind of a weird frequency which is way too low for ferrite, but a bit too high for most electrical steels, even with really thin laminations.

cur8xgo:
I selected operating frequency in an early stage of prototyping, and its survived until now, although could probably be revisited to get some not-insignificant gains.

At that time,  I believe the factors were:

-make it high enough to reduce idle losses (magnetizing current)
-make it low enough to theoretically leave me some breathing room with frequency dependent effects which I didn't understand well (skin, proximity)
-make it "reasonable" for M6 laminations..this is very hand wavy I admit

Now, further down the road, I have learned:

Power input bus transients occur at twice the operating frequency, and I have been unable to snub them entirely into a capacitor. They avalanche through the fets (and presumably into the bus cap) at 45V for about 5 us per switch event. Luckily, I selected some fets that are fully avalanche rated and they are heavily heat sinked with forced air already, so they don't seem to be breaking a sweat. I'm not sure how much energy is being snubbed by the fets and how much by the bus cap. But if it was all into the fets it would be about 98W per H-bridge leg (625 * 2 * 45V * 350A * 5us).

That said, increasing frequency will increase the power dissipated by snubbing the bus transients more often. And although nothing seems to be running hot, and losses should be proportional to frequency, I would prefer to avoid touching frequency unless I have no choice. But if my thinking is right here, maybe I could double it? I would really have to put thermistors on the fets and know exactly where things stand temperature wise before I could do that though.

And there are other switch-event-frequency related energy absorbing events:

-primary snubber
-bottom mosfet gate dv/dt transient
-possibly the full bridge output cap ripple
-mosfet linear region dissipation

So, 625Hz was originally somewhat arbitrary, but turned out I got a little lucky and my intuition was also a little right, I think.

Full-bridge was selected based on advice given in many books and online for this power level. And, because it seemed like something I could get my head around as being new to SMPS. Although now, I am very tempted to experiment with a boost design at some point, at least in a simulator, to see what would be what.

Isolation, unfortunately, is probably required in this application. There may be a twist on this device that could allow it to use an autotransformer or non-isolated topology. But I don't think I should jump into that since things appear to be going well with this full bridge isolated design.

Now, since the prototype works, my pragmatic choice would seem to be turning my somewhat-realistic prototype transformer into a much improved, producable transformer.

To do that:

EI-162 M6 0.014" lams with a core area of about 3.1" for a flux density of about 0.5T (twice now)..about $24 in lams (about 125 pairs per core) from Tempel in small qty
0.056" thick copper strip (because that thickness is easily available to me and coincidentally fits into an EI-162 with the sames turns as prototype) about $25 retail in very small qty for 18 feet (both windings)
Custom bobbin: lets just say $5
Varnish, paper/kapton, hardware, etc..etc.. lets say $10
An hour to wind and assemble? $15?

Haha so $80 for a 3kw strip wound custom transformer! Hmmm that can't POSSIBLY be true can it?

To prototype it the lams are the expensive part..looking at a minimum order of $124 from Tempel, but I can will call it so no shipping. So we are looking at around $200 to try this out.

As far as literally having someone else make these...I've gotten not too unreasonable quotes from Pacific Transformer..about $220 in single quantity for something similar (not identical though) to this. Thats much higher than the budget for this though. I need this transformer to be under $100 or I will feel like I haven't done my job.

This just in: its occured to me that my original excel spreadsheet showed me 0.25T at the LOADED input voltage..not the unloaded input voltage. And, I have been talking about 0.5T for the new design at the UNLOADED input voltage. So that means I could probably further reduce the core size to get to where I thought I was going, and use even less core and less copper, than the EI-162 idea. But..to be conservative I will probably just ignore this to leave some slack, since I still don't know what will happen with the smaller core with the same number of turns. Nice to know there is some breathing room though.





Navigation

[0] Message Index

[*] Previous page

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