Looking for manufacturer of huge inductor

[GK]

With high enough AC current amplitude and frequency however, core losses would dominate over winding losses if the core is used all the way to its saturation flux density. In this case it is likely possible to achieve lower total loss by increasing the gap length and number of turns (keeping the same core and inductance) as this would decrease the AC flux density. This is likely to be needed in for example converters operating in discontinuous mode from 100 kHz or so and up (for ferrite cores) and the resonant inductors in quasi-resonant converters.

I'd like to see a worked example of this, backed up by measurement. Assessing / quantifying and actually measuring the effects of core / gap losses isn't as trivial as some make out with their simplified formula and theory, and especially so if these phenomenon are to be accurately juggled for optimal efficiency.

Incidentally, when designing an inductor for typical buck/boost SMPS I select my gap on the DC current plus the peak ripple current. That generally works out to be a good "rule of thumb" compromise between minimising copper losses and having some safety factor to keep away from saturation.
 

[johansen]

With high enough AC current amplitude and frequency however, core losses would dominate over winding losses if the core is used all the way to its saturation flux density. In this case it is likely possible to achieve lower total loss by increasing the gap length and number of turns (keeping the same core and inductance) as this would decrease the AC flux density. This is likely to be needed in for example converters operating in discontinuous mode from 100 kHz or so and up (for ferrite cores) and the resonant inductors in quasi-resonant converters.

I'd like to see a worked example of this, backed up by measurement. Assessing / quantifying and actually measuring the effects of core / gap losses isn't as trivial as some make out with their simplified formula and theory, and especially so if these phenomenon are to be accurately juggled for optimal efficiency.
Incidentally, when designing an inductor for typical buck/boost SMPS I select my gap on the DC current plus the peak ripple current. That generally works out to be a good compromise between minimising copper losses and having some safety factor to keep away from saturation.

megajocke.. what if instead of using a lower flux density to lower core losses... what if you just... used a smaller core?
what you said is true, core losses dominate over winding losses when the core is used at high flux levels.. but you can always decrease the permeability of the core to increase copper losses to arbitrarily high values --this is a big difference between inductors and transformers, transformers ideally store no energy.

Somewhere on the internet is account and photos of someone's 12kw boost converter using 12 awg solid wires and he didn't even put the I on the E core, left the E core out on its own... sure you can't saturate the core without melting the wires.. but with the right sized gap he could have cut the volume of the inductor at least in half, and it would probably be twice the "Q"

theoretical examples such as the above thought experiment are hard to accurately determine because they cover too many variables.

What it really comes down to is acceptable power density, why spend twice as much if its only 10% lower losses...where does that stop.
Some people spend 3$ per watt for a 99% efficient dc-dc widget to manage power from 1$ per watt solar panels. this is illogical unless you can't buy more solar panels right?. same analogy goes for the inductor at the beginning of this thread, just different variables


i also suggest reading these papers.
http://power.thayer.dartmouth.edu/shapeopt_intro.html

the second paper linked there, and here: http://thayer.dartmouth.edu/inductor/papers/cooscost.pdf
mentions a Q of 300 for a 19mH 100khz half amp inductor.. which is fantastic.
--the available window area decreases quickly with frequency.

[megajocke]

If you design for continuous mode with typical ripple current amounts (say below +-20 % of the DC current) and typical frequencies (say below 200 kHz), you're unlikely to see much core loss with ferrites so just keeping out of saturation in the way you do by following the energy storage diagrams is certainly enough.

Bulk core losses are nothing strange, and those are the ones I am talking about. If you try to build something like a discontinuous mode flyback converter operating at 200 kHz with flux density swinging between zero and 300 mT each cycle you have almost certainly blown more than all of your loss budget just on core loss. It would overheat even without any help from copper losses. On for example a 3C85 (not very high performance) material ETD49 core, that would be like 25 W of core loss, which is way too high for any reasonable cooling scheme.

Extra copper losses from gap fringing fields and other "strange" losses are not so trivial to reason about, but in a case such as the one I just described, a gap just large enough to satisfy the energy storage need wouldn't do.

[GK]

Extra copper losses from gap fringing fields and other "strange" losses are not so trivial to reason about, but in a case such as the one I just described, a gap just large enough to satisfy the energy storage need wouldn't do.

Well that's a pretty extreme example that I would say is a fair departure from inductor design problem that started this thread.

But considering your example, why would you even choose to make such a thing with a "not very high performance" ferrite core with a (possibly) bloody huge gap and who knows what kind of EMR (and associated issues) due to all that fringing flux? To me that problem just screams out for a high (frequency) performance powered iron toroid.

[megajocke]

megajocke.. what if instead of using a lower flux density to lower core losses... what if you just... used a smaller core?

...

i also suggest reading these papers.
http://power.thayer.dartmouth.edu/shapeopt_intro.html

the second paper linked there, and here: http://thayer.dartmouth.edu/inductor/papers/cooscost.pdf
mentions a Q of 300 for a 19mH 100khz half amp inductor.. which is fantastic.
--the available window area decreases quickly with frequency.

Well, I was assuming the inductor was limited by heat dissipation rather than some arbitrary loss goal. Using a smaller core if you can should be done, and because the surface area grows slower with increased size than volume does, smaller cores can support higher loss densities and therefore higher flux densities in a core loss limited design. I was assuming you had already found the smaller core to be unworkable, or that you would evaluate the design as the last step and change core size if it seemed feasible. If you start with some core and optimize the design and end up with something that has much lower temperature rise (or loss) than you were aiming for (or what is safe) the core could be made smaller.

And thanks for the links! I have read some of those papers before, but I don't think I had seen that 19 mH inductor. Neat.

[GK]

Quote from: GK on November 08, 2014, 11:49:02 AM

    Well, I've always approached DC-carrying inductor design with the idea that  core gaping is a necessary evil to prevent saturation. If you are concerned with efficiency, then you only gap the necessary amount and no more. The suggestion that (for a desired inductance value and DC current) the core gap (and presumably the number of turns) can be adjusted for a happy or optimal medium between "core" and copper losses doesn't make any general sense to me.

this is the commonly taught and fundamentaly backwards way of looking at the matter.

Umm, it's an entirely practical one given the constraints that I either loosely defined or thought were obvious. So now in this thread we have leaped from the specific topic of "DC-carrying inductor design" to flyback transformer design.............

 

[megajocke]

Well that's a pretty extreme example that I would say is a fair departure from inductor design problem that started this thread.

But considering your example, why would you even choose to make such a thing with a "not very high performance" ferrite core with a (possibly) bloody huge gap and who knows what kind of EMR (and associated issues) due to all that fringing flux? To me that problem just screams out for a high (frequency) performance powered iron toroid.

Yes, quite different. Now I forgot what my point was. Maybe it was that both your method, the other one seen in this thread, and the one I like to use really are the same thing but starting from different ends of the problem. And that the application discussed here is going to be saturation limited, in which case the manufacturers LI^2 graphs do apply, as opposed to my somewhat contrived example. Though, where is the dividing line between DC-carrying inductors and high-ripple stuff?

At 60 kHz an iron powder based inductor may very well be core-loss limited which means the number of turns and inductance (for a certain core) will by necessity be high. If using a ferrite core, it might be beneficial to use a much lower inductance if the higher inductor ripple current is acceptable for the other parts than what would have been practical for an iron-powder design. I've once used a ferrite based 10 µH inductor for a 60 V input 30 V 20 A output buck converter at 100 kHz (the output has lots of low-ESR capacitors) which works quite nicely, and I actually needed the high ripple and low inductance to achieve high current slew rate.

For my discontinuous mode example, about 0 - 120 mT flux density swing would probably be about right (with about 100 mW/cm³ of core loss) if using 3C85. More modern materials could take more flux swing for the same loss. Whether the gap would have to be huge or not depends on what output power one is aiming for, which wasn't a fixed parameter.

Iron powder toroids are nice in that there is no gap with fringing fields that could cause excessive winding losses, but on the other hand even the lowest-loss iron powder materials have higher core loss than quite high-loss ferrites at typical SMPS frequencies at the same flux density. So you'd need more turns for the same cross-sectional area for the same core loss density. Tradeoffs everywhere!

If you are operating at low frequency and/or low ripple current (which is quite typical) the high saturation flux density of iron powder cores is a very nice feature. With low ripple (unlike my example) you don't need good core loss performance anyways. For the application discussed at the start of the thread, an inductor based on a powdered iron core or similar is probably a good fit.

[johansen]

For the application discussed at the start of the thread, an inductor based on a powdered iron core or similar is probably a good fit.

I have shoved 250 watts through a single T106 green/blue powdered iron toroid pulled from a mystery atx supply. --at 60Khz.
it got very warm.. but if i recall correctly, it was a 50v in, 24 volt out buck converter at about 10 amps out. losing half a volt in the switches, and about the same in the inductor.. so that's 5 watts for both the inductor and the switches. epoxying the inductor to a heatsink would have worked.

[johansen]

Quote from: GK on November 08, 2014, 11:49:02 AM

    Well, I've always approached DC-carrying inductor design with the idea that  core gaping is a necessary evil to prevent saturation. If you are concerned with efficiency, then you only gap the necessary amount and no more. The suggestion that (for a desired inductance value and DC current) the core gap (and presumably the number of turns) can be adjusted for a happy or optimal medium between "core" and copper losses doesn't make any general sense to me.

this is the commonly taught and fundamentaly backwards way of looking at the matter.

Umm, it's an entirely practical one given the constraints that I either loosely defined or thought were obvious. So now in this thread we have leaped from the specific topic of "DC-carrying inductor design" to flyback transformer design.............

I mentioned transformers because they do not store energy.

flyback transformers are coupled inductors, they make crappy transformers.

the fundamental difference between myself and your perspective is looking at the gap as a necessary evil.
the gap or permeability sets the maximum energy storage.
energy storage is proportional to copper losses squared.
the number of turns is unimportant, and the last matter to worry about.

Well, I was assuming the inductor was limited by heat dissipation rather than some arbitrary loss goal. Using a smaller core if you can should be done, and because the surface area grows slower with increased size than volume does, smaller cores can support higher loss densities and therefore higher flux densities in a core loss limited design. I was assuming you had already found the smaller core to be unworkable, or that you would evaluate the design as the last step and change core size if it seemed feasible. If you start with some core and optimize the design and end up with something that has much lower temperature rise (or loss) than you were aiming for (or what is safe) the core could be made smaller.

well no offense but all those hidden assumptions are why so many people find inductor optimization to be black magic.

consider this theoretical thought experiment.
given a T106 mix 26 core
find the optimum frequency and ripple current % to shove 300 watts through it with the least losses. assume a buck or boost converter.
assume that you are limited to 70% fill factor with 26 awg wire.
edit:forgot to mention, assume 50% duty cycle for worst case senario.

[GK]

the fundamental difference between myself and your perspective is looking at the gap as a necessary evil.

In a typical ferrite design of which this thread topic started core losses are generally negligible, maximum efficiency is had by minimising copper losses and and core gapping therefore IS a "necessary evil" from a practical perspective. Have you ever built a SMPS with a 50+ mJ (let alone the 100+ mJ the OP wants) gapped ferrite inductor next to any sensitive control circuitry or tried to test it for electromagnetic interference? Huh?
 

[johansen]

In a typical ferrite design of which this thread topic started core losses are generally negligible, maximum efficiency is had by minimising copper losses and and core gapping therefore IS a "necessary evil" from a practical perspective. Have you ever built a SMPS with a 50+ mJ (let alone the 100+ mJ the OP wants) gapped ferrite inductor next to any sensitive control circuitry or tried to test it for electromagnetic interference? Huh?

I still don't understand how you see it that way.
Grind the gap in the center leg, 3-d print a custom bobbin to keep the wires away from the gap, those things are fantastic for that purpose, and wrap the core with copper foil.
There is no other optimization to make. if OP can handle the heat from three or four cores side by side, then he could probably manage with 2 cores and copper strip wound on edge, with a slow stream of air flowing through the coil.

in both cases the gap isn't evil.. its what makes it an inductor.

[GK]

 

[T3sl4co1l]

I don't get why you're getting so frustrated...

Did you learn to design inductors *one way only*, without understanding the underlying theory enough to form your own, critical yet open, opinion on the subject?

As has been indicated time and time by the experts* in this thread, there are a great many variables which are inter-related, complex to analyze, poorly quantified, and generally not worth much concern, given that the resulting error is in the sub-10% "good enough" range.

*For lack of a better word; meaning, those who have presented theoretical or empirical evidence (which includes you).

This approximation of approximations means that, in the space of solutions which are all "about good enough", one can have very differing opinions which are nonetheless valid.  This should be, not a cause for concern, but an understanding that, in all disciplines, technical or otherwise, there is no absolute certainty, indeed only absolute uncertainty, and one must accept and understand that, in order to work best with it.  This is not the hard logic of mathematics, this is the art of engineering!

I'm not trying to be condescending, I would just rather have you understand, than be frustrated and angry..

Tim

[GK]

I don't get why you're getting so frustrated...

Did you learn to design inductors *one way only*, without understanding the underlying theory enough to form your own, critical yet open, opinion on the subject?.

The goal posts keep getting shifted around and everyone, at best, keeps competing to say the same things in different words. That gets frustrating. I presented one very simple method to design the inductor because that method is directly applicable to the design problem at hand - and having an immediate, intuitive grasp of that fact doesn't betray a general, underlying lack of understanding.

As has been indicated time and time by the experts* in this thread, there are a great many variables which are inter-related, complex to analyze, poorly quantified, and generally not worth much concern, given that the resulting error is in the sub-10% "good enough" range.

Where did I say anything that contradicts this?

This approximation of approximations means that, in the space of solutions which are all "about good enough", one can have very differing opinions which are nonetheless valid.  This should be, not a cause for concern, but an understanding that, in all disciplines, technical or otherwise, there is no absolute certainty, indeed only absolute uncertainty, and one must accept and understand that, in order to work best with it.  This is not the hard logic of mathematics, this is the art of engineering!

I'm not trying to be condescending, I would just rather have you understand, than be frustrated and angry..

[GK]

in both cases the gap isn't evil.. its what makes it an inductor.

The gap makes an inductor that can maintain its inductance with a DC current flowing. The operating conditions, for example, of a flyback transformer are completely different from those of an inductor carrying a large DC current with a comparably small ac ripple component.

Please explain to me how a general example of the latter case can benefit from a gap significantly larger than that necessary to avoid saturation.




 

[johansen]

When a smaller core + larger gap is cheaper than what you suggested; I made no such general rule apart.such.
You keep adding unknown varariables, I explains from the beginning how to calculate watts per $ and provided a spreadsheet designed for that purpose.

I covered flyback transformers in reply #63, that has no bearing here

[GK]

When a smaller core + larger gap is cheaper than what you suggested;

That's not even a logical answer to the question. If the inductor can be made with a cheaper, smaller core then question still begs - why should that core be given a gap larger than necessary to avoid saturation? And what exactly did I suggest? Right towards the start of this thread I simply gave a brief worked example to put some kind of perspective to the (physical) scale of the inductor the OP was asking for. I even said that it wasn't supposed to be a practical solution. Is that what you are talking about?... wait.... forget that I asked. I think that I am through with this thread. There are much better things I can do with my time. In fact I have a freshly gapped core for a 50mJ inductor ready to wind right here.....


 

[johansen]

In fact I have a freshly gapped core for a 50mJ inductor ready to wind right here.....

fantastic..
i heated the copper you see here to 500C at least twice to get it to fit.
--90% fill factor btw.
http://johansense.com/bulk/P1050464.JPG

When a smaller core + larger gap is cheaper than what you suggested;

That's not even a logical answer to the question.

You have had other hidden unknown variables since the beginning.
answer my question.
what is the optimal frequency for which to use a T106 mix 26 core for a 300 watt buck converter at 50% duty cycle.

you should know that to find that answer requires a shit load of data... for which the manufacturers will not provide.

[johansen]

I hardly ever use bobbins.
It is not hard to wrap some slippery paper around a wood block carved out to the right size and then slip the coil off the block and onto the core.

Concerning ferrite cores, you can easily grind them yourself with most any grinder.
If you want to grind swing chokes, tapered gaps, step gaps or whatever, that is also very easy.
find a fine grit diamond tile saw blade, 4 inch diameter, 1/16th inch thick, no slots.
the ferrite is soft enough you don't need to bother with water, but you could easily get a cheap tile saw from harbor freight to do all the work.

[GK]

When a smaller core + larger gap is cheaper than what you suggested;

That's not even a logical answer to the question.

You have had other hidden unknown variables since the beginning.
answer my question.
what is the optimal frequency for which to use a T106 mix 26 core for a 300 watt buck converter at 50% duty cycle.

you should know that to find that answer requires a shit load of data... for which the manufacturers will not provide.

WTF? You can't answer the question (or rather admit that answering makes your original claim of contention moot) so you just completely change the topic?

[T3sl4co1l]

His point is, the missing variables to answer both questions are unavailable...

Tim

[GK]

His point is, the missing variables to answer both questions are unavailable...

Tim

Baloney! So far only one person posting on this topic has actually understood the point I have made:

Well, I've always approached DC-carrying inductor design with the idea that  core gaping is a necessary evil to prevent saturation. If you are concerned with efficiency, then you only gap the necessary amount and no more. The suggestion that (for a desired inductance value and DC current) the core gap (and presumably the number of turns) can be adjusted for a happy or optimal medium between "core" and copper losses doesn't make any general sense to me.

Yes, if the current is pure DC enough, core losses will be negligible even in a minimum gap/minimum turns design like you say. In that case minimum gap is certainly the configuration which will give minimum total loss. Many authors call this a "saturation limited desgin" or something to that matter. If you have for example +-20% of ripple current and design for a peak (AC+DC) flux density of 300 mT you'll only have +-50 mT of AC flux density. (300 mT / 1.20 gives 250 mT of average DC flux density) Even the crappiest power ferrites have quite low loss up to 200 kHz or so for this ripple level.

Going back to the specifications actually listed by the OP, the desired inductance was in the range of 150 - 200 uH and the switching frequency 30 - 60 kHz, a current output of 25A and Vout up to 15VDC with Vin up to 35VDC.

At Vin = 30V and Vout = 15V for a 50% duty cycle and picking the lowest f and smallest L (30 kHz and 150 uH) the ripple current works out to 1.7A peak-to-peak. For a DC load current of 25A, that's well under +/-5 % ripple.