Transformer design: how to size a core?

[Buk]

I want step up Li-ion battery voltage to 100-125V @ 0.5A and ~145kHz.

I'm thinking of using a toroidal transformer to get the voltage boost.

The first thing I need to know is will a suitable size toroid physically fit into the space I have available.

To that end, are there any rules of thumb for sizing a (toriodal) core to handle 50W of power at 145kHz?

[T3sl4co1l]

What input voltage?

How much experience do you have with switching power supplies, and Li-ion batteries?

Why such a particular frequency?

If it's just a few cells in series, a push-pull converter is a fine way to go, as typically used in automotive equipment.  Matter of fact, you might get away with a 100W automotive inverter, just saw off the AC "modified sine" chopper circuit.

Edit: oh, right!
https://www.eevblog.com/forum/projects/heat-generation-using-piezo-elements/

Tim

[Buk]

Edit: oh, right!
https://www.eevblog.com/forum/projects/heat-generation-using-piezo-elements/

Did that answer all your questions, or should I supply further information?

[WattsThat]

Probably the simplest, most practical explanation of DIY magnetics you’ll ever find:

https://ludens.cl/Electron/Magnet.html

[T3sl4co1l]

I don't remember you mentioning what battery pack you were going to use so let's start with that

Tim

[Buk]

I don't remember you mentioning what battery pack you were going to use so let's start with that

Tim

I'm hoping to get away with a single 21700 4200mAh cell with a (claimed) maximum discharge rate of 45A. spec sheet (pdf) They are from Moli One Energy Corporation of Taiwan and have a good reputation.

At a pinch I might consider a parallel pair of 18650s, say Sony VTC5As with a 25A rating, but I'd rather avoid the extra weight. It's not just the 30g for the batteries, but the extra weight in the bigger case.

[Thomas]

http://schmidt-walter-schaltnetzteile.de/smps_e/smps_e.html#smps

[T3sl4co1l]

What will you use to hold the cell?

My experience with snap-in holders has been, they have quite high resistance, topping out at not many amperes.

If you'll have a spot-welded lead frame, or just something better than what I've used, it should be feasible.

Still, the very high current seems unnecessary.  How about a small 3S pouch or something?

In any case, you will probably need a Baxandall (or "ZVS", often erroneously "Royer") oscillator with a pair of MOSFETs:
https://adammunich.com/zvs-driver/
At the low voltage, the top circuit is actually sufficient; I would add series gate resistors, and maybe clamp zeners to protect them some more.

If plain on-off control is sufficient, then also a switch to supply that, along with logic to manage the battery (basically cutting out below 3V, while not drawing many µA in the process).  Maybe an LED too?  If you want adjustable power, the switch can be PWM'd (don't forget the catch diode, you're making a buck converter with the oscillator's series supply inductor), varying the amount of current into the oscillator (down to a minimum, as it needs some voltage to begin oscillating).

The oscillator will run at whatever resonant frequency it's tuned to, partly a function of the transformer inductance, load capacitance, and the load itself (piezo stack).  This is probably the best option to start out with.

The next step up from an oscillator, is basically a free-running oscillator with PLL and some means of sensing the load tuning (current or voltage phase usually), or some means of forcing it to the operating frequency (selectable capacitors?) -- quite a bit more complicated, as you can see.  So if the oscillator is good enough or can be made to work, it's the preferable option.

Tim

[Buk]

(Not sure what this has to do with the OP question, but thanks for the interest.

The plan for the cell contacts looks like this:



The copper colored component is an 8mm diameter copper rod with a 10mm x 1mm contact face (all one piece). The white thing is a silicone insulator. At the base, there is a screw in lid, with a 10mm raised face that connects negative to the case. The circuit will have reverse polarity, under voltage etc. detection.

The current plan for the circuit is a VCO chip driving a mosfet (or H-bridge pair if a tapped primary is beneficial) connected to the primary side of a toroidal transformer; in the form of a flyback converter. A second inductor in parallel with the secondary coil provides the energy storage. The piezo is connect in series with the secondary coil via a diode. Probably a limiting resistor, and maybe a cap is needed there also.

I've determined that a 20x10x10 toroid can handle the load, maybe a 16x8x8 will also depending on its permeability curve.

[WattsThat]

For one offs, you’ll find pot or E cores with a plastic bobbins are far easier to DIY than toroidal cores. Obviously it depends on the number of turns required and the gauge of the wires. The two piece type cores are also easier to deal with when you need to create air gaps if saturation becomes a problem.

[T3sl4co1l]

So like a flashlight?

I wouldn't recommend making your own.  Like I said, spot welding is probably the only way for that kind of current.  (But, others who have more experience with cells, do chime in!)

50W at 150kHz will probably need more like a 25mm OD toroid, depends.  In that ballpark anyway, yeah.

I would recommend a nice clean cylindrical bobbin style, like E, EP, ETD, and use copper foil for the primary.  Push-pull (as in the ZVS circuit) is preferred to reduce RMS current per transistor.  The two halves of the primary should be tightly coupled together, which is done by using two strips of foil together, and connecting them in series outside the winding.  (Or put down two turns but solder another foil connection in the middle, same idea.  Probably about the same bother to assemble, honestly.)

You can use enameled wire just fine, but you'll need to use many strands in parallel, both to handle the current (ala litz cable), and to interleave enough strands to get reasonable performance.  (Pro tip: make pairs or quads of wires for both halves of the primary, then wire a bunch of those in parallel for ampacity.)  The secondary probably doesn't need to be too tightly coupled to the primary (leakage inductance is acceptable, to some extent) so can just be a single block.

No diodes or anything, maybe some combination of inductor and capacitor to match the secondary winding to the piezo stack.  (A diode would just... charge it up to DC, and, sit there doing nothing?)

Tim

[Buk]

So like a flashlight?

I wouldn't recommend making your own.  Like I said, spot welding is probably the only way for that kind of current. 

The battery must be removeable, so spot welding isn't an option.

For my prototype, I'll be making the body out of copper. The tube, which is one part of the full case, will be made from 22mm heating pipe. The rest will be made from copper sheet silver soldered together.

If it ever got to being made commercially -- extremely doubtful -- then probably a Zamak (Zinc alloy) die casting would be the solution.

50W at 150kHz will probably need more like a 25mm OD toroid, depends.

I've finally got a FEA simulation to run at 145kHz, and a 16x8x8 core gets to 0.47T; which unless someone tells me I am wrong, it what I will be using.

[Buk]

For one offs, you’ll find pot or E cores with a plastic bobbins are far easier to DIY than toroidal cores. Obviously it depends on the number of turns required and the gauge of the wires. The two piece type cores are also easier to deal with when you need to create air gaps if saturation becomes a problem.

Easier, but they take more space; which is the big limitation.

[sandalcandal]

(Not sure what this has to do with the OP question, but thanks for the interest.

The plan for the cell contacts looks like this:

Word of caution with the cell holder design. Slightly less of an issue with flashlights/torches but highly enclosed designs like this circumvent cell venting safety features and turn the cell+holder into a pipebomb/shotgun. There are a numerous deaths due to lithium ion battery explosions with vapes; DIY, modded, misused or poorly designed. For this reason, most companies don't like selling their cells (directly) for vape applications and official material forbids their use in vapes/e-cigs. If you must use this form factor design you need to add pressure relief/venting.

Edit: Even with pressure relief, there is still significant danger for any highly enclosed design.

[T3sl4co1l]

0.47T, with what ferrite?

0.2T may be more comfortable, with respect to core loss at that frequency.  Maybe less.  Definitely look for a low loss type like 3C95 or better.

Tim

[Buk]

0.47T, with what ferrite?

0.2T may be more comfortable, with respect to core loss at that frequency.  Maybe less.  Definitely look for a low loss type like 3C95 or better.

Tim

I believe that is true for EI  and pot transformers, less so for toroidal.

At a pinch -- and considerably more money -- I could opt for an Nanocrystalline amorphous metal core that can run upto 1.56T and 1/5th the losses; but I'd like to avoid that cost if I can.

[Buk]

Word of caution with the cell holder design. Slightly less of an issue with flashlights/torches but highly enclosed designs like this circumvent cell venting safety features and turn the cell+holder into a pipebomb/shotgun.

I'm aware of the requirement; but the warning is appreciated.

Edit: Even with pressure relief, there is still significant danger for any highly enclosed design.

The way to deal with it is the same way electrolytics caps and aircraft :

[uer166]

You're not going to go far with a nanocristalline core at that frequency, the losses would be massive. I'm not sure where you got the 1/5 loss number from, but I doubt you can run it at any more than 10s of kHz.

[Buk]

You're not going to go far with a nanocristalline core at that frequency, the losses would be massive. I'm not sure where you got the 1/5 loss number from, but I doubt you can run it at any more than 10s of kHz.

FINEMET® FT-3K50T and FT-8K50D

These are brand new materials produced by applying a controlled magnetic field during annealing to industry leading thin nano crystalline ribbon. This provides a material that satisfies both high saturation magnetic flux density and high permeability. Other standard features: Low core loss, Low magnetostriction, Excellent temperature characteristics and small aging effects, Excellent high frequency characteristics and the Flexibility to control magnetic properties " B-H curve shape " during annealing process.

[Buk]

You're not going to go far with a nanocristalline core at that frequency, the losses would be massive. I'm not sure where you got the 1/5 loss number from, but I doubt you can run it at any more than 10s of kHz.

More info:

[uer166]

Seems like you're right, I did some math for a few cores, e.g. MP2010:

• 100kHz
• Vin=3.6, Vout=125, Iout=0.5
• Ipk = 22.5A, Iripple=10A
• Suggested L = 3.6uH
• Number of turns = 7
• Bpeak = 0.3T, Bripple=0.14T
• Inductance at Bpeak = 90% of zero, Core loss is ~2W at 0.14 Bpeak (about 3% efficiency lost in core losses)

It's a little surprising to have such low core loss for laminated metal, so it would be great to see real life results. At such high currents and frequency you'd do well to use some form of litz as well, although a 1-layer windup here helps massively.

Edit: above is for a straight boost which at this ratio would be impractical. Also these cores are not suitable for transformer due to very low permeability (they're gapped), so you'd be better off finding ungapped Metlgas cores, or regular low loss ferrite.

[T3sl4co1l]

I believe that is true for EI  and pot transformers, less so for toroidal.

At a pinch -- and considerably more money -- I could opt for an Nanocrystalline amorphous metal core that can run upto 1.56T and 1/5th the losses; but I'd like to avoid that cost if I can.

Those are mains frequency transformers; the core losses are already low.  Nanocrystalline materials are promoted in that space simply to achieve even lower idle losses.  (When the replacement schedule is half a century as can be the case for distribution (pole/pad) transformers, it's worth spending a few thousand bucks up front, to save a few cents a day of electricity over that time frame!)

Core materials fall off at high frequency.  Typically there is a cutoff frequency where the impedance changes, from largely inductive (impedance proportional to frequency, ~90° phase shift) to resistive (impedance constant, no phase shift), or something inbetween.  As it happens, sheet metals have a diffusion characteristic (impedance ~ sqrt(f), 45° phase), which takes over when skin depth equals sheet thickness, more or less.  (At still higher frequencies, the capacitance between layers dominates, and impedance falls with increasing frequency.)

For most nanocrystalline materials, this cutoff is in the 50kHz range.  Which is not to say it's automatically useless up there -- but it is a hint that losses won't be as low as you might've hoped, and it's worth considering other materials too.

Ferrites drop off in the 100k to 10MHz range, depending on material and size.

Tim

[sandalcandal]

Word of caution with the cell holder design. Slightly less of an issue with flashlights/torches but highly enclosed designs like this circumvent cell venting safety features and turn the cell+holder into a pipebomb/shotgun.

I'm aware of the requirement; but the warning is appreciated.

Edit: Even with pressure relief, there is still significant danger for any highly enclosed design.

The way to deal with it is the same way electrolytics caps and aircraft :

Good to know you are aware.

[Buk]

For most nanocrystalline materials, this cutoff is in the 50kHz range.

I may have linked the wrong page or part of the page. That website has very long pages and lot of different things per page and many of the #targets are vague.

And I can only go by the specifications provided by the Metglas/Hitachi. I have B-H curves for several samples of products from  both of them -- now co-owned -- and  if you look at the
core load v frequency curve for 2506SA1, they go up to 150kHz.

And I am led to believe that the newer FINEMET® FT-3K50T and FT-8K50D products, used correctly, can produce even lower losses on small toriods at higher frequencies. But I am no expert, and can offer no first-hand experience here. I do know that most peope do not realise quite how different amorphous metal -- Metallic Glass -- is from regular transformer steel.

For a start, it is an order of magnitude thinner -- 0.015mm versus ~ 0.18mm -- than the best silicon steel sheets you can get. Add to that a surface texture that means that even when wound together without additional interleaved insulation, virtually no current can flow between layers. And the final piece of the puzzle appears to be the application of magnetic fields during annealing above the Curie temp. that freezes magnetic domains into the cooled product in much the same way they do with NeFeB magnets.

But now I'm just quoting stuff from spec. sheets, and quite possibly mixing stuff from several different material specs; but I did do some (FEA) testing of various MetGlas/FineMet specs a few years ago for a different project -- a transverse flux motor design -- and I am convinced that for my application of a no-gap toriod, size for size versus a ferrite (FeNiZnV), at 145kHz, I can get 1/5th the losses, or use half the turns, or some balance between the two. If only I was prepared to pay the cost. For now, I do not think I need to.

This is a 16 od x 8 id x 6 ferrite core with a 5 turn/15A primary , 155 turn/0.5A secondary @145kHz.


The calculated losses are so low that I will need to  recalibrate and re-run the simulation using individual wires -- this is a 10/330 --


 rather than block level simulations, before I am convinced the results are correct.; but it is looking feasible.

[T3sl4co1l]

Truly it's amazing stuff, amorphous/nanocrystalline materials; not only low loss, but some offer permeability rivaling supermalloy.  The now-obsolete ISDN transformers, using the stuff, are just about unmatched in bandwidth.  The cores are still around fortunately, so you can still make your own, just not off-the-shelf ready made transformers.

That's quite extreme flux density for ferrite, are you sure your loss model is correct?  Also, how exactly are you going to get 150 total amp-turns through a 1/4" hole?

I'd like to see core loss more like 0.5W for a core that size, at least for continuous operation, let alone copper losses as well.

Never heard of FeNiZnV, what's that?

Tim