Inductor Core Saturation and Losses

[samovar]

I want to make an inductor on toroid ferrite core. I know all the dimentions of the core and permeability - 4200. I have a formula to calculate inductance based on this information and number of turns.

I want to run a high current through it. As I understand, as I apply voltage, the current will be increasing slowly (based on inductance value), but at some point the core will saturate and current will increase dramatically.

How do I calculate the current at which the saturation happens?

 

[johansen]

use my spreadsheet.
**broken link removed**

you need open office to run it, whenever i convert it to excel the formulas don't work.

btw, the air gap value is the magnetic air gap, a physical airgap will be half of that value.

 

[samovar]

btw, the air gap value is the magnetic air gap, a physical airgap will be half of that value.

Thank you! I do not have open office, but I'll try to get it.

 

[samovar]

I looked at DigiKey and they have only ungapped ferrite cores. This gives me good inductance. I can use their larger 102mm ferrite core to get 10mH. I only need 42 turns, which seems quite feasible with #10 wire.

However, when I try to calculate B I get u * I * N/l = 0.00529*100*42/0.255 = 87 T. Wikipedia says that ferrite saturates at 0.2-0.5 T, so I am hopelessly out of range.

Even for 1mH I'll need 13 turns, which leads to B of 27 T. Still too much.

I also tried to do calculations with air core, but this leads to a huge several pound inductor with tremendious heat losses. I can cut heat losses by increasing wire size, but then it becomes so heavy that I probably won't be able to lift it. Not exactly what I want.

I would need something around 5mH, preferably 10mH, which would carry 100A DC without saturation. Is there any way I can build that?

 

[MrAl]

Hi,

You are asking for a lot here because 10mH requires many turns and for every turn the DC current causes the core to saturate more easily. So the more turns the more easily it saturates. Adding a gap makes it take more DC current to saturate, but when you add a gap you lower the effective permeability which lowers the inductance value so you have to add turns to make up for that so it ends up being a trade off. Luckily the inductance value per turn goes up more than the ease of saturation per turn goes up, so eventually you reach a point where it works. Just where this point is though could mean adding a lot more turns after adding the air gap.

If you are going to experiment with this a toroid core will be hard to deal with because they are hard to gap if not impossible unless you have expensive equipment or get it gapped at the factory, and getting it gapped at the factory means knowing precisely what size gap you need. But then again a 100 amp DC inductor is probably not something a hobbyist does.

There are other factors too like wire size vs skin effect.

Depending on what frequency you intend to operate at and what it is going to be used for, you may get away with a different kind of core.

A 10mH 100 amp inductor will be quite large however, no matter how you do it. What is this going to be used for, perhaps there are other ideas too?

 

[ronsimpson]

What are you building?
100A and 10mH. That is a large inductor.

 

[samovar]

I'm thinking about building high power low voltage constant current power supply. That's for anodizing aluminum. I need at least 100A @ (8-20V). I wanted to go with 200A first, but I don't think I can source parts for that. I try to go for low ripple and high efficiency. Design is very simple - bridge rectifier with relatively small capacitor, followed by a buck SMPS - that's what I need an inductor for.

I understand the inductor is not going to be very small, but I would want to make it inexpensive.

So, what is the best way to go? Is there a place on the Internet that sells gapped cores with known parameters. Or should I try to buy an old transformer, unwind it, and saw in a gap?

 

[johansen]

the inductor you need for a buck converter is a trade off between energy stored, efficiency and cost.

I can push 250 watts through a ferrite core with a .8cm ^2 cross section, 16? turns of 14 awg wire, and the core is 34mm by ~50 iirc, at 50khz.
the inductor losses were negligible. at 50Khz and 500 watts, i would have to quadruple the copper losses, and if i wanted to try and push 1 kilowatt through the core at 100Khz it would need a fan, but it would not be impossible. (this was a buck converter operating at 50 volt input, and 25 volt output, for 50% duty cycle)
**broken link removed**

what you're probably not anticipating is the inductor for a buck converter is the easy part. the low inductance and low ESR capacitors needed for the input are not negligible.

if you use forced synchronous rectification, you don't even need capacitors on the output for electrolysis type applications

 

[johansen]

btw, for iron core inductors, figure a pound of copper and a pound of steel per joule of energy stored for reasonable losses at 120-360 hz applications.
copper magnet wire costs 9$ per pound and electrical steel is on the order of 3$ per pound...

a 100 amp 10mh inductor stores 50 joules

 

[johansen]

**broken link removed**
another design idea for you.
you could probably get 2 KW from something like that, without heatsinking, depending on how much money you want to throw at the mosfets. irfz46 fets are good to about 5 amps without a heatsink. provided you have some airflow to cool them off. --but they make 60 volt fets with one fourth of the on resistance these days, for pretty cheap.

i would not bother trying to remove the 120 hz ripple for electrolysis operations.
if you do insist on removing that ripple, you have to store that energy in something.
There are ways you can fake a 3 phase supply with a 3 phase induction motor, if you have some of those laying around then you would not need a big stack of 50 volt capacitors..
as the 360hz ripple from a 3 phase transformer (or 3 single phase transformers) is negligible, because the buck converter can maintain constant current from a 3 phase supply provided it does not run higher than 80% duty cycle.

btw.. i intended for that board to be a 5 phase buck converter, but i never finished that project.
if i tried to run 100 amps through that board as a single phase buck converter, the capacitors would explode within a minute, i'm sure.

 

[MrAl]

Hi there johansen and samovar,

johansen:
I took a look at your circuit boards and i have to say that's pretty cool. I like your ad-hack method of making the circuit boards which looks like maybe a Dremel and cutter bit is all that is needed to yield a quick circuit board with plenty of copper for the higher current apps. Very nice. Too often i see fancy dancy pc boards which are over complicated for what they do. This is nice and anyone can make them that way without too much trouble. So thanks for posting the pics.

samovar:
Did you mention yet what your buck operating frequency was going to be? The higher the frequency the smaller the inductor, but im not sure how small you can get with that 100 amp requirement. The 10mH requirement can be brought down considerably at higher frequencies. But im not sure if you intended to use that as an input filter for the rectifiers. If so then it can not be reduced, but if for the buck you can reduce that value.

 

[johansen]

I use kicad and a milling machine now
**broken link removed**
**broken link removed**

to get from kicad to the mill i use pcb2gcode, which is free.
for someone who is new to this i would recommend on linuxcnc, which is free.
but you could use any pcb designer to output gerber files. then go from gerber to pcb2gcode, to gcode, and from the same machine run the cnc machine to cut the board.
most folks use windows but i'm not aware of any free software to go from gerber to gcode on windows.

practical limits for diy milling in my opinion are 0804 components and minimum of 15 mills gap between pins.. which means .03" pin spacing or bigger.

 

[samovar]

Thank you.

I don't have a set frequency. I was going to switch on when I get current 10% less than set value and off when it is 10% higher. By my calculations at 100V difference between the source and the load, 10mH will lead to about 1kHz switching frequency, and 5mH - 2kHz. Above that, switching losses get high.

Probably I need to re-think something because 100 pound inductor just doesn't seem right.

For example, I can get a transformer to reduce source voltage to about 20V. Then 1mH inductor will do and switching losses go down, so I probably can easily switch around 10kHz and inductor size goes down to 100μH, but 2kVA 60Hz transformer is not an easy thing neither. This also will bring into picture 100A diodes for the rectifier, instead of 20A with high voltage. I'm not sure such diodes ever exist.

 

[johansen]

yeah i was assuming you wanted something on the order of less than 50 vdc to buck down to 8-20 volts.
I'm not familiar with most electrolysis set-ups other than trying to make sodium from table salt at 800C, or making hydrogen and oxygen from car batteries at 18 volts and 30 amps per battery...
but 100 amps seems insane, seems to me that much current would be used to anodize an entire 4 by 8 foot sheet at one time.

and as far as i know once the initial annodization is done, you need a few hundred volts at mere milliamps to get that nice hard thick layer of aluminium oxide.. and then you take the part out and quickly dump it in the dye solution so the crystals of Al2O3 suck up the colourizing agent...

in anycase, 1Khz was on the slow side 30 years ago. bipolar darlingtons can get you reasonable losses at up to 10Khz, and mosfets are good to 500Khz today.


my calculations at 100 volts in, 10 volts out and 100 amps at 1Khz and 1 volt drop across the switch, and 1 volt drop across the diode mean you're going to be running at about 10.2% duty cycle and for 10 amps of peak to peak ripple current means you need 505uH inductor. and its going to store 3 joules of energy.
the output capacitor is only going to see 5.7 amps of ripple current, and you're going to need 25,000 uF
the input capacitor is going to see 33.389 amps of ripple current and you need a minimum of 3000 uF


oh, with a buck regulator you're not getting something for nothing either...if you're worried about the diode losses i would recommend just making your own 100 amp 10-20 volt transformer and running that into a stack of 30 amp 45 volt Schottky diodes which only have a .5 volt forward voltage under rated load... well heatsinked (you can parallel them, provided they are from the same manufacturer)
run the transformer from a variac.

 

[samovar]

High current anodizing requires 20 amps/sf. Since this is on both sides, 100 amps will only give me 2.5 sf - I would want more. I've never tried it though. The voltage is supposed to start at about 10 volts and I assume will go to 20 volts at the end, but I'm not really sure, so my idea was to go with 120V, so that I could go to 30-40 volts or more if needed. Switching high voltage is more difficult. Hence low frequency. I didn't expect that it would be so difficult to get 10mH inductor.

Anyway, back to inductors. I couldn't find any that would handle 100A, so I will need to make my own. I see now that 10mH is practically impossible. The question is how big can I practically get? My idea was to buy a core and wind a wire around it - looks very easy. But even for 100 μH I still need a big core. Is it possible to buy a core somewhere? If not, can I make it? Is it a good idea to buy inductors designed for lower current and then get the core out and re-wind it with thicker wire?

 

[ronsimpson]

The size of the inductor is related to 1/frequency.

10mH will lead to about 1kHz switching frequency, and 5mH - 2kHz

I am switching at 1mhz with uHs.

Schottky diodes

At high current I some times use sync-diodes. (mosfet used as a diode) Synchronous rectification

It looks like you want to come off the power line, 4-diodes, caps, buck, 20 volts out.
I think you need isolation. It is dangerous to come off the power line without isolation.

 

[samovar]

my calculations at 100 volts in, 10 volts out and 100 amps at 1Khz and 1 volt drop across the switch, and 1 volt drop across the diode mean you're going to be running at about 10.2% duty cycle and for 10 amps of peak to peak ripple current means you need 505uH inductor. and its going to store 3 joules of energy.

Thank you. I totally missed this. It is only going to be on during 10% of a cycle. If that is 1 ms then the frequency is only 100 Hz. For 1 kHz, the on time is only 100 μs. And I need much smaller inductor! Thank you very much!!!

The only question remains, what is the best way to make 100 amps 0.5mH inductor?

 

[samovar]

I think you need isolation. It is dangerous to come off the power line without isolation.

I know My idea is to test the design without isolation, then, perhaps, rent (borrow) a 1:1 transformer to do actual work.

 

[MrAl]

Hi,

I would think you could get to 10kHz at least, maybe 50kHz.

Remember also that two equal inductors in parallel provides only 1/2 the inductance of one unit but twice the operating current. So two 1mH 50A inductors can make one 500uH 100A inductor. And another way to go is to parallel complete power supplies.

I ran into this problem too when i went to build an offline power supply for a heavy duty DC motor that was going to draw about half this current level. I had problems finding an inductor that did not cost a million dollars. Another idea is to simply buy a ready made power supply from a manufacturer that specializes in power supplies. That could actually end up being much cheaper too.

The largest inductor(s) i ever designed was for use for testing a high power three phase synthesized sine power supply. Three of them, each one weighed 150 pounds and used 1/4 inch square copper wire. These used standard EI laminations, if i remember right 3 x 3 inch square center leg.

 

[johansen]

it is not dangerous to come off the line without isolation, but without which you're going to have 170 vdc to deal with not 100 volts. which is totally impractical.

my opinion is samovar you should make a 100 amp power supply from re-wound MOT transformers to find out if your electrolysis bath will even accept 100 amps.